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 neural fate commitment of pluripotent stem cells requires the repression of extrinsic inhibitory signals and the activation of intrinsic positive transcription factors . However , how these two events are integrated to ensure appropriate neural conversion remains unclear . In this study , we showed that Pou3f1 is essential for the neural differentiation of mouse embryonic stem cells ( ESCs ) , specifically during the transition from epiblast stem cells ( EpiSCs ) to neural progenitor cells ( NPCs ) . Chimeric analysis showed that Pou3f1 knockdown leads to a markedly decreased incorporation of ESCs in the neuroectoderm . By contrast , Pou3f1-overexpressing ESC derivatives preferentially contribute to the neuroectoderm . Genome-wide ChIP-seq and RNA-seq analyses indicated that Pou3f1 is an upstream activator of neural lineage genes , and also is a repressor of BMP and Wnt signaling . Our results established that Pou3f1 promotes the neural fate commitment of pluripotent stem cells through a dual role , activating internal neural induction programs and antagonizing extrinsic neural inhibitory signals .
Early vertebrate development is the process by which unrestricted pluripotent stem cells progressively make lineage fate choices . Central to cell allocation is gastrulation , during which the epiblast responds to secreted signals and generates three primary germ layers ( Lu et al . , 2001 ) . In early mouse embryos , gastrulation initiates at embryonic day ( E ) 6 . 5 . Posterior epiblast cells ingress through the primitive streak to form the mesoderm and endoderm , whereas the cells that remain in the anterior part of the epiblast form the ectoderm ( Tam and Loebel , 2007 ) . Then , a portion of the anterior ectoderm is specified to adopt the neural fate and subsequently , develops into the neuroectoderm , forming a plate-shaped structure called the neural plate at approximately E7 . 5 ( Tam and Zhou , 1996 ) . Previous studies have indicated that neural fate specification from embryonic ectoderm occurs autonomously in the absence of inhibitory signals such as bone morphogenetic proteins ( BMPs ) and Wnts ( Munoz-Sanjuan and Brivanlou , 2002; Stern , 2005b ) . In early Xenopus , chick , and mouse embryos , BMP and Wnt signals prevent neural conversion and contribute to non-neural fates such as epidermal differentiation and primitive streak formation ( Winnier et al . , 1995; Hemmati-Brivanlou and Melton , 1997; Liu et al . , 1999; Wilson et al . , 2001 ) . BMP and Wnt inhibition in the prospective neural ectoderm is essential for proper neural development ( Wilson and Edlund , 2001 ) . In mouse embryonic stem cells ( ESCs ) , BMP and Wnt signals are required for self-renewal and readily repress neural differentiation partially through their targets , such as Id1 , Id2 , and Myc ( ten Berge et al . , 2011; Varlakhanova et al . , 2010; Ying et al . , 2003; Zhang et al . , 2010a ) . BMP and Wnt antagonists have been utilized to generate neural lineage cells in mouse or human ESCs ( Blauwkamp et al . , 2012; Chambers et al . , 2009; Gratsch and O'Shea , 2002; Watanabe et al . , 2005 ) . In addition to extrinsic signaling pathways , neuroectoderm specification is also controlled by the sequential activation of intrinsic neural fate-promoting factors . Sox2 , which is an ESC pluripotency-maintenance factor , plays an important role in ESC neural differentiation , indicating that Sox2 is a neural lineage-poised factor ( Thomson et al . , 2011 ) . Zic2 and Otx2 are also involved in epiblast stem cell ( EpiSC ) neural conversion ( Iwafuchi-Doi et al . , 2012 ) . Recently , Zfp521 was identified as an intrinsic factor that promotes the progression of early neural development ( Kamiya et al . , 2011 ) . Studies concerning these neural fate-promoting factors have partially revealed the internal mechanism of early neural development . However , how these neural factors are activated during neural fate commitment remains unclear . Moreover , considering the importance of the effect of extrinsic signals on the neural fate decision , it remains unclear whether the inhibition of extrinsic signals and activation of internal factors are regulated separately or are integrated by a single determinant . POU family transcription factors play important roles in the development of the nervous system ( Veenstra et al . , 1997 ) . Pou3f1 ( also known as Oct6 , Tst1 , or as SCIP ) has been reported as the earliest expressed POU III family member in mouse embryo development ( He et al . , 1989; Monuki et al . , 1989; Meijer et al . , 1990; Suzuki et al . , 1990 ) . During gastrulation , Pou3f1 expression is observed in the chorion and in the anterior epiblast ( Zwart et al . , 1996 ) . As embryonic development proceeds , Pou3f1 expression becomes restricted to central nervous tissues and is detectable in the midbrain and in the forebrain ( He et al . , 1989; Zwart et al . , 1996 ) . Pou3f1 has also been documented as a crucial regulator of the myelination of Schwann cells in the peripheral nervous system ( Bermingham et al . , 1996; Jaegle et al . , 1996 ) . In vitro , the rapid increase of Pou3f1 mRNA in retinoic acid-induced neural differentiation of P19 cells suggests that Pou3f1 may be functionally associated with neural fate commitment ( Meijer et al . , 1990 ) . Recent reports have proposed that Pou3f1 might be a potential regulator associated with early neural development ( Kamiya et al . , 2011; Iwafuchi-Doi et al . , 2012; Yasuhara et al . , 2013 ) . However , whether Pou3f1 is involved in the neural initiation of pluripotent stem cells remains elusive , and the underlying mechanism requires further investigation . In this study , we show that Pou3f1 is necessary and sufficient for the neural fate commitment of ESCs and of EpiSCs . In chimeric mice , Pou3f1-knockdown cells display suppressed neuroectoderm distribution . Conversely , ESCs with Pou3f1 overexpression preferentially contribute to the neuroectoderm but not to other lineages . We further demonstrate that Pou3f1 promotes the neural fate commitment of pluripotent stem cells through the activation of intrinsic neural lineage genes and through the inhibition of extrinsic BMP and Wnt signals .
We previously established an efficient system to induce ESC neural differentiation in serum-free medium ( Zhang et al . , 2010a ) . To investigate neural conversion mechanisms , we performed a microarray-based screening and identified Pou3f1 as one of the genes significantly up-regulated during pluripotent stem cell neural differentiation . Pou3f1 was moderately expressed in ESCs . The highest levels were observed from days 2–4 upon neural differentiation , and then the expression of Pou3f1 declined ( Figure 1A , Figure 1—figure supplement 1A ) . Gene expression profiling indicated that the Pou3f1 expression peak occurred between the epiblast marker Fgf5 and the neural stem cell marker Sox1 ( Figure 1A , Figure 1—figure supplement 1A ) . This result suggests that Pou3f1 might play a role in ESC neural differentiation . 10 . 7554/eLife . 02224 . 003Figure 1 . Pou3f1 is essential for ESC neural differentiation . ( A ) Schematic expression profiles of Pou3f1 and of several key marker genes during ESC neural differentiation in serum-free medium . Rex1 , ESC marker; Fgf5 , EpiSC marker; Sox1 , NPC marker; Tuj1 , neuron marker . Detection of Pou3f1 protein expression during ESC neural differentiation by Western blotting . ( B ) Gene expression levels in control-ESCs ( Ctrl ) and in Pou3f1-knockdown ESCs ( Pou3f1-KD1 , Pou3f1-KD3 ) at neural differentiation day 4 were determined by Q-PCR . Three independent experiments were performed . ( C ) Immunocytochemical assays of Sox/Oct4 , Pax6 , and Tuj1 in day 4 EBs described in B . DNA is stained with DAPI . Scale bars: 50 μm . ( D ) Statistical analysis of Sox+/Oct4− , Pax6+ , and Tuj1+ cells in C . ( E ) Gene expression levels in control-ESCs and inducible Pou3f1-overexpressing ( Pou3f1-OE ) ESCs at unbiased differentiation ( 10%FBS ) day 8 were determined by Q-PCR . Dox ( 2 μg/ml ) was added for 8 days . ( F ) Immunocytochemical assays of Sox/Oct4 , Pax6 , Nestin , and of Tuj1 in day 8 EBs described in E . Scale bars , 50 μm . ( G ) Statistical analysis of Sox+/Oct4− , Pax6+ , and Tuj1+ cells in F . ( H ) Pou3f1-knockdown ESCs were transfected with control or with Pou3f1-overexpressing lentiviruses . Gene expression levels at neural differentiation day 4 were determined by Q-PCR . The values represent the mean ± SD for B , D , E , G , and for H . ( *p<0 . 05; **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 00310 . 7554/eLife . 02224 . 004Figure 1—figure supplement 1 . Pou3f1-knockdown ESCs could differentiate into non-neural cell lineages . ( A ) Expression profiling of Pou3f1 and of several key marker genes during ESC neural differentiation in serum-free medium , was determined by Q-PCR . ( B ) a , Knockdown efficiency of Pou3f1 with control-shRNA and Pou3f1-KD1/2/3 lentivirus-transfected ESCs was determined by Q-PCR . b , Knockdown of Pou3f1 protein by Pou3f1-shRNAs . ( C ) Gene expression levels in control and Pou3f1-knockdown ESCs were determined by Q-PCR . ( D ) Expression levels of germ layer genes in control and Pou3f1-knockdown ESCs at unbiased differentiation day 8 were determined by Q-PCR . The values represent the mean ± SD . ( *p<0 . 05; **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 00410 . 7554/eLife . 02224 . 005Figure 1—figure supplement 2 . Brn2 could compensate for the Pou3f1 depletion during ESC neural fate commitment . ( A ) Expression levels of the POU III family members Pou3f1 , Brn1 , and Brn2 during ESC neural differentiation in serum-free medium . ( B ) Brn1 and Brn2 expression levels in control and Pou3f1-knockdown ESCs were determined by Q-PCR . ( C ) Expression levels of POUIII family members in control , Pou3f1-knockdown and Pou3f1/Brn2-knockdown ESCs undergoing differentiation for 4 days in serum-free medium . ( D ) Expression levels of neural marker genes in control , Pou3f1-knockdown , and Pou3f1/Brn2-knockdown ESCs undergoing differentiation for 4 days in serum-free medium . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 00510 . 7554/eLife . 02224 . 006Figure 1—figure supplement 3 . Overexpression of Pou3f1 accelerates ESC neural differentiation in serum-free condition . ( A ) Expression levels of neural marker genes in control and Pou3f1-stable-overexpression ESCs differentiated in serum-free medium from days 0 to 6 . ( B ) Immunocytochemical assays of Sox/Oct4 , Nestin , and Tuj1 in day 4 EBs described in A . Cells in day 4 EBs were replated in N2 medium for 2 days . Immunostaining of Tuj1 ( red ) was performed ( g and h ) . DNA is stained with DAPI . Scale bars: 50 μm . ( C ) Statistical analysis of Sox+/Oct4− , Nestin+ , and Tuj1+ cells in EBs and percentages of Tuj1+ cells in adherent culture during neural differentiation from days 0 to 6 in A . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 00610 . 7554/eLife . 02224 . 007Figure 1—figure supplement 4 . Pou3f1 promotes neural differentiation in a cell-autonomous manner . ( A ) a , Q-PCR and b , Western blotting analysis of induced Pou3f1 overexpression . ESCs in adherent cultures were treated with Dox for 48 hr . ( B ) a , Immunocytochemical assays for Tuj1 ( red ) and GFP ( green ) using the co-cultured EBs . Wt ESCs ( GFP− ) were co-cultured with control ESCs ( GFP+ ) or with Pou3f1-overexpressing ESCs ( GFP+ ) in serum-free medium for 6 days . b , Cells in ‘a’ were immunostained by the Tuj1 antibody and then analyzed by fluorescence-activated cell sorting . The values represent the mean ± SD . ( *p<0 . 05; **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 007 To test this hypothesis , a lentivirus-mediated knockdown strategy was utilized to diminish Pou3f1 expression . Two shRNAs ( KD1 and KD3 ) targeting the Pou3f1 3′ UTR region efficiently decreased Pou3f1 expression in ESCs to approximately 50% and 30% , respectively ( Figure 1—figure supplement 1B ) . The control ( Ctrl ) and Pou3f1-KD1/3 ESCs exhibited comparable expression levels of the pluripotency markers ( Figure 1—figure supplement 1C ) . After differentiating these ESC lines in serum-free medium , the transcripts of the neural markers Sox1 , Pax6 , and Tuj1 were reduced in Pou3f1-KD1/3 cells ( Figure 1B ) . Immunocytochemical assays confirmed the reduced percentage of Sox+/Oct4− , Pax6+ NPCs , and Tuj1+ neurons from Pou3f1-KD1/3 ESCs ( Figure 1C , D ) . Moreover , unbiased ESC differentiation in serum-containing medium revealed that the expression of mesoderm ( T and Flk1 ) , endoderm ( Gata4 and Gata6 ) , and epidermal ( Ck18 and Ck19 ) markers was unaltered after Pou3f1 knockdown ( Figure 1—figure supplement 1D ) . These results suggest that Pou3f1 is selectively required for the neural differentiation of ESCs . Because most POU III proteins exhibit extensive functional equivalence ( Andersen and Rosenfeld , 2001; Jaegle et al . , 2003; Friedrich et al . , 2005 ) , we wanted to determine whether other POU III proteins , such as Brn1 and Brn2 , are similarly involved in ESC neural fate commitment . We examined the Brn1 and Brn2 expression profiles , and determined that these proteins are up-regulated in ESC serum-free neural differentiation after day 5 , following the peak of Pou3f1 expression ( Figure 1—figure supplement 2A ) . Interestingly , compared with the control , Brn2 , but not Brn1 , expression was enhanced in Pou3f1-KD1/3 cells ( Figure 1—figure supplement 2B ) . When the expression of Pou3f1 and Brn2 was simultaneously reduced by lentivirus-mediated shRNAs , the expression of the neural marker genes Sox1 , Pax6 , and Nestin decreased more dramatically ( Figure 1—figure supplement 2D ) , although Brn1 expression was not affected ( Figure 1—figure supplement 2C ) . Together , these results suggest that Brn2 , which is a POU III family member , compensates for Pou3f1 depletion . To determine whether Pou3f1 is sufficient to promote the neural differentiation of ESCs , stable Pou3f1-overexpressing ESCs were generated . Compared with the control , the constitutive expression of Pou3f1 notably enhanced the expression of NPC and the neuron markers during serum-free differentiation , particularly at day 4 ( Figure 1—figure supplement 3 ) . Single cell suspensions from EBs at various days were replated in N2 medium for neuronal differentiation . Many Tuj1+ neurons emerged from stable Pou3f1-overexpression ESCs at day 4 , 2 days earlier than the control ESCs ( Figure 1—figure supplement 3C , d ) . These results demonstrate that neural differentiation was accelerated by Pou3f1 overexpression under serum-free conditions . To exclude the influences of Pou3f1 overexpression on the ESC state , doxycycline ( Dox ) -inducible Pou3f1-overexpressing ESCs were generated ( Figure 1—figure supplement 4A ) . As expected , the Dox-induced overexpression of Pou3f1 strongly enhanced ESC neural differentiation in serum-containing medium , which was accompanied by the increased expression of the neural markers Sox1 , Pax6 , Nestin , and Tuj1 in both quantitative polymerase chain reaction ( Q-PCR ) and immunostaining assays ( Figure 1E–G ) . Moreover , the decreased neural marker expression in Pou3f1-depleted ESCs was restored by the overexpression of a Pou3f1 coding sequence ( CDS ) lacking the 3′ UTR ( Figure 1H ) . Cell aggregation assays were performed by co-culturing wild-type ESCs with either GFP-labeled control or Pou3f1-overexpressing ESCs in serum-free medium . The neural differentiation of wild-type ESCs was not affected by Pou3f1-overexpressing ESCs in the culture system ( Figure 1—figure supplement 4B ) , indicating that Pou3f1 promoted neural differentiation cell-autonomously . Taken together , these results suggest that Pou3f1 is both necessary and sufficient for the intrinsic neural conversion of ESCs . Our previous study showed that ESC neural differentiation could be divided into two stages: ESCs to EpiSCs and EpiSCs to NPCs ( Zhang et al . , 2010a ) . Therefore , we investigated which stage of neural differentiation is regulated by Pou3f1 . To address this question , we performed ESC-derived EpiSC ( ESD-EpiSC ) colony formation assays ( Zhang et al . , 2010a ) using day 2 ESC aggregates in serum-free medium . The results demonstrated that both control and Pou3f1-overexpressing ESCs generated similar numbers of homogeneous compact monolayer EpiSC-like colonies that displayed weak alkaline phosphatase activity ( AKP ) and similar levels of Oct4 expression ( Figure 2A , B ) . Furthermore , both types of EpiSC-like colonies expressed comparably high levels of the pluripotency markers Oct4 and Nanog , and of the epiblast marker Fgf5 , with the absence of the expression of the ESC-specific gene Rex1 ( Figure 2D ) . Consistently , Pou3f1 knockdown did not affect the formation and markers' expression of EpiSC-like colonies ( Figure 2C , E ) . These results suggest that Pou3f1 may not be involved in the first stage of ESC neural differentiation . 10 . 7554/eLife . 02224 . 008Figure 2 . Pou3f1 promotes the neural differentiation from EpiSCs to NPCs . ( A ) Inducible Pou3f1-overexpressing ESCs were cultured as EBs for 2 days in the medium with or without Dox and then subjected to the ESD-EpiSC colony formation assay for 6 days in Dox-free CDM/AF medium . EpiSC-like colony cellular morphology , alkaline phosphatase activity ( AKP ) ( purple ) , and Oct4 immunostaining ( red ) are presented . Scale bars , 100 μm . ( B ) Statistical analysis of EpiSC-like colonies in A . ( C ) Statistical analysis of EpiSC-like colonies from the control-ESCs and from Pou3f1-knockdown ESCs ( Pou3f1-KD1 , Pou3f1-KD3 ) in the ESD-EpiSC colony formation assay . ( D ) Gene expression levels in ESCs and in EpiSC-like colonies formed in A . ( E ) Gene expression levels in ESCs and in EpiSC-like colonies formed in C . ( F ) EpiSC-like colonies from control-ESCs ( −Dox ) , short-term Pou3f1-overexpressing ESCs ( +Dox 0–2 ) , and from long-term Pou3f1-overexpressing ESCs ( +Dox 0–6 ) in the ESD-EpiSC colony formation assay . Cellular morphology , AKP activity , and immunostaining for Oct4 , Nestin , or for Tuj1 with DAPI are presented . Scale bars , 100 μm . ( G ) Statistical analysis of EpiSC-like colony numbers described in F . ( H ) Gene expression levels of ESCs and of the EpiSC-like colonies described in F . ( I ) Expression profiling of Pou3f1 and Sox1 during EpiSC neural differentiation in serum-free medium . ( J ) Gene expression levels of control and Pou3f1-knockdown EpiSCs in serum-free medium at differentiation day 2 were determined by Q-PCR . ( K ) Gene expression levels of control and Pou3f1-overexpressing EpiSCs at unbiased EBs differentiation day 2 were determined by Q-PCR . The values represent the mean ± SD for B–E and for G–K . ( *p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 008 To determine whether Pou3f1 plays a role at the second stage of ESC neural differentiation , we used Dox to induce Pou3f1 overexpression during various periods in the ESD–EpiSC colony formation assay . The short-term overexpression of Pou3f1 was achieved by adding Dox for the first 2 days ( +Dox 0–2 ) , whereas the long-term overexpression was achieved by adding Dox for 6 days ( +Dox 0–6 ) ( Figure 2F ) . The number and morphology of EpiSC-like colonies from the short-term treated ESCs were similar to those characteristics of untreated control ESCs ( −Dox ) . Additionally , AKP and Oct4 expression levels were also similar to those levels in the controls ( Figure 2F , G ) . However , the number of EpiSC-like colonies from the long-term treated ESCs was significantly reduced , as was the expression of AKP and Oct4 , whereas the expression of neural makers , such as Nestin and Tuj1 , increased ( Figure 2F , G ) . Moreover , the enhanced expression of Sox1 , Pax6 , and Nestin was also confirmed by Q-PCR ( Figure 2H ) . Therefore , these results suggest that Pou3f1 may function during the second stage of ESC neural differentiation , from EpiSCs to NPCs . To validate this finding , EpiSCs derived from early mouse embryos were differentiated in serum-free medium for 4 days . Gene expression profiling revealed that Pou3f1 transcripts peaked at differentiation day 1 and subsequently declined with the onset of Sox1 expression ( Figure 2I ) . In Pou3f1-knockdown EpiSCs at neural differentiation day 2 , Sox1 , Pax6 , and Nestin expression was reduced ( Figure 2J ) , whereas Sox1 , Pax6 , and Nestin expression was increased in Pou3f1-overexpressing EpiSCs at unbiased differentiation day 2 ( Figure 2K ) . These results suggest that Pou3f1 facilitates the neural differentiation of EpiSCs . Together , these data indicate that Pou3f1 promotes pluripotent stem cell neural differentiation during the transition from EpiSCs to NPCs . To explore the function of Pou3f1 in vivo , first , we verified Pou3f1 expression patterns in early mouse embryos by in situ hybridization . Pou3f1 transcripts were detected in the whole epiblast and in the extraembryonic region of mouse embryos at E5 . 5 ( Figures 3A , a , g ) . Then , Pou3f1 expression was gradually restricted to the anterior part of the embryos from E6 . 5 to E7 . 0 ( Figure 3A , c , d ) . Transverse sections of embryos revealed that Pou3f1 expression was exclusively localized to the anterior region of the inner epiblast , which would prospectively undergo neuroectoderm fate ( Figures 3A , b–d , h , i ) . During the neural initiation stage at E7 . 5 and at E8 . 0 , Pou3f1 expression was further restricted to the anterior neuroectoderm ( Figure 3A , e , f , j , k ) , suggesting a causal correlation with embryonic neural differentiation . 10 . 7554/eLife . 02224 . 009Figure 3 . Pou3f1 promotes neural fate commitment in vivo . ( A ) Whole-mount in situ hybridization of Pou3f1 in early mouse embryos ( E5 . 5–E8 . 0 ) . The arrowhead marks the position-plane of the transverse section of the corresponding embryo below . Scale bars , 100 μm . ( B ) Contribution of injected GFP-labeled control ( Ctrl ) , Pou3f1-knockdown ( Pou3f1-KD ) , and inducible Pou3f1-overexpressing ( Pou3f1-OE ) ESCs to different germ lineages in chimeric embryos . NE , neuroectoderm; M , mesenchyme; and S , somite . Scale bars , 50 μm . ( C ) Statistical analysis of GFP-positive cell distribution in the various germ layer lineages in the ESC blastocyst injection study . The values represent the mean ± SD for C . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 00910 . 7554/eLife . 02224 . 010Figure 3—figure supplement 1 . Information of chimeric mice generated from Pou3f1-overexpressing or knockdown ESCs . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 010 Next , we performed a blastocyst injection study using manipulated ESCs . GFP-labeled control , Pou3f1-knockdown ( Pou3f1-KD ) , and Pou3f1-overexpressing ( Pou3f1-OE ) ESCs were injected into E2 . 5 blastocysts and transferred into pseudopregnant mice , respectively . The developmental potentials of these cells were examined after 7 days post-transplantation ( at E9 . 0−E9 . 5 ) . Chimeras were generated from these three ES cell lines ( Figure 3—figure supplement 1 ) . The number of GFP-positive cells in various tissues was ascertained in sections of chimeric embryos . The control ESCs contributed to a wide range of germ layer lineages , including neuroectoderm ( NE ) , mesenchyme ( M ) , somite ( S ) , heart , gut , and extraembryonic ectoderm , at similar percentages ( ∼60% ) ( Figure 3B , C ) . Surprisingly , only Pou3f1-KD ESCs failed to contribute to the neuroectoderm , but were widely identified in non-neural lineages ( Figure 3B , C ) . By contrast , Pou3f1-OE ESCs preferred incorporation into the neuroectoderm and displayed a considerably reduced contribution to non-neural tissues ( Figure 3B , C ) . These results indicate that Pou3f1 promotes the neural fate commitment of pluripotent stem cells in vivo . To investigate the regulatory mechanism of Pou3f1 at the global level , we performed RNA-seq assays to identify Pou3f1-regulated genes during ESC differentiation . Pou3f1-overexpressing ESCs were differentiated in unbiased medium , and total RNAs were collected from EBs at days 2 , 4 , and 6 for mRNA sequencing . The RNA-seq analysis revealed that the global transcriptome changed dramatically from day 2 to day 6 ( Figure 4A ) . Because day 4 EBs were at the transition state from the epiblast-like stage at day 2 to the NPC-like stage at day 6 ( Zhang et al . , 2010a ) , we focused on the transcriptome data from day 4 . To validate the deep-sequencing data , we examined the expression levels of approximately 30 genes by Q-PCR and found that these expression levels were consistent with the sequencing data ( Figure 4—figure supplement 1A ) . Of the 11 , 356 genes expressed ( rpkm > 1 ) , 768 genes were up-regulated , and 202 genes were down-regulated ( Cuffdiff , FDR < 0 . 05 ) . 10 . 7554/eLife . 02224 . 011Figure 4 . RNA-seq and ChIP-seq analysis of Pou3f1 downstream targets . ( A ) RNA-seq gene expression heat map of control and of inducible Pou3f1-overexpressing ESCs with Dox-treatment for 6 days . Heat-map colors ( red , up-regulation; blue , down-regulation ) indicate gene expression in units of standard deviation from the mean of all samples . ( B ) Analysis of Pou3f1-enriched regions in the ChIP-seq assay . Pie chart showing the percentage distribution of Pou3f1-binding peaks in each category . The ChIP-seq assay was performed with Pou3f1-overexpressing ESCs at differentiation day 4 . ( C ) Venn diagram depicting the overlap ( purple ) of Pou3f1-bound genes ( blue ) and genes with significantly altered expression upon Pou3f1 overexpression ( pink ) at differentiation day 4 . Statistical significance was estimated by Fisher's exact test ( p<4 . 71e−75 ) . ( D ) GO analysis of biological processes of the overlap genes described in C . Many genes involved in neural differentiation processes were up-regulated , whereas a few genes related to pattern specification were down-regulated . Log p value was used to rank the enrichment . ( E ) Genome browser view of the distribution of the ChIP-seq and RNA-seq reads of represented genes . The upper panels show the Pou3f1-binding regions identified by ChIP-seq ( black , input; red , Pou3f1-binding site at genomic loci ) , and the lower panels depict the RNA-seq reads of the represented genes in control ESCs ( gray ) and in Pou3f1-overexpressing ESCs ( green ) at differentiation day 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 01110 . 7554/eLife . 02224 . 012Figure 4—figure supplement 1 . Pou3f1 is enriched in the loci of multiple downstream target genes . ( A ) Correlation between Q-PCR and RNA-Seq data . Approximately , 30 genes were chosen to confirm the RNA-Seq results . ( B ) ChIP-qPCR verification of the ChIP-seq data represented in Figure 4E . Pou3f1 enrichment at identified binding sites of each gene was normalized to corresponding coding regions . ( C ) Genome browser view of the distribution of Pou3f1 binding on the loci of representative genes . The values represent the mean ± SD for B . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 012 To identify genes directly regulated by Pou3f1 on a genome-wide scale , ChIP-seq assays were performed with day 4 EBs . Interestingly , a large percentage of Pou3f1-binding sites ( 47% ) were located in distal regions more than 50 kb away from known or predicted transcription start sites ( TSS ) . Only a small percentage of Pou3f1 binding sites resided in 5′ proximal regions ( 0–1 kb and 1–5 kb ) , reflecting the property of Pou3f1 to control transcription primarily through distal enhancers . To investigate whether Pou3f1 binding to the genomic regions exerts functional consequences through regulating targeted gene expression , we integrated the ChIP-seq data with the RNA-seq data . Among the 4674 Pou3f1-binding genes , 430 genes were modulated significantly ( Figure 4C ) . Gene Ontology term enrichment analysis revealed that genes up-regulated by Pou3f1 were primarily involved in neural differentiation processes , such as neuron differentiation , neuron development , and axonogenesis , whereas Pou3f1-down-regulated targets were highly enriched in pattern specification and in embryonic morphogenesis ( Figure 4D ) . Detailed ChIP-seq and RNA-seq analyses showed that the genomic region of neural development-related genes , such as Pax6 , Sox2 , and Zfp521 , was bound by Pou3f1 and that their expression was up-regulated by Pou3f1 overexpression . Intriguingly , the downstream targets of important morphogens , such as Gata4 in the BMP pathway as well as Myc and Dkk1 in the Wnt pathway , were also bound by Pou3f1 . However , the expression of these genes was down-regulated by Pou3f1 overexpression ( Figure 4E ) . Pou3f1 genomic binding was confirmed by ChIP-qPCR ( Figure 4—figure supplement 1B ) . We also found that Pou3f1 could bind to the genomic regions of Zic1 and of Zic2 , which are related to neural development , and of the BMP and Wnt signaling targets Id1 and Axin2 ( Figure 4—figure supplement 1C ) . Together , these results suggest that Pou3f1 might promote ESC neural fate commitment through regulating the expression of multiple genes . Genome-wide ChIP-seq and RNA-seq assays revealed that Pou3f1 might regulate a group of genes related to neural development , such as Sox2 , Zfp521 , Zic1 , and Zic2 ( Figure 4E , Figure 4—figure supplement 1C ) . Q-PCR confirmed that expression of these neural fate-promoting factors was decreased by Pou3f1 knockdown and increased by Pou3f1 overexpression ( Figure 5A , B ) . Next , we investigated how Pou3f1 regulates the expression of these target genes . As a transcription factor , Pou3f1 contains three domains: the amino-terminal region , the POU domain , and the HOMEO domain . The POU domain and HOMEO domains mediate protein interactions and DNA binding ( Levavasseur et al . , 1998 ) . Among serial deletion mutants ( Figure 5—figure supplement 1A ) , the HOMEO domain deleted mutant ( ΔHOMEO ) exclusively failed to promote ESC neural differentiation ( Figure 5—figure supplement 1B and data not shown ) . This result suggests that the HOMEO domain is essential for the Pou3f1-mediated promotion of the neural fate . 10 . 7554/eLife . 02224 . 013Figure 5 . Pou3f1 increases neural lineage-specifier expression . ( A ) Gene expression levels in control and in Pou3f1-knockdown ESCs differentiated in serum-free medium for 4 days . ( B ) Gene expression levels in control and in inducible Pou3f1-overexpressing ESCs at unbiased differentiation day 8 . ( C ) Luciferase assays using the Sox2N2-luc enhancer in control , Pou3f1-full length , or in Pou3f1-ΔHOMEO vector-transfected HEK293 cells . ( D ) ChIP assay in control , Pou3f1-full length , or in Pou3f1-ΔHOMEO lentivirus-transfected P19 cells . A Pou3f1-specific antibody was used , and Pou3f1 enrichment at Sox2N2 and Sox2N1 enhancer regions was normalized to the Sox2 coding region . ( E ) Whole-mount in situ hybridization of cPou3f1 ( a–h ) and cSox2 ( i–p ) in early chick embryos from HH stage 3+ to HH stage 10 . ( F ) Pou3f1 overexpression induces cSox2 expression ectopically . IRES-GFP ( control vector , a and a′ ) or Pou3f1-IRES-GFP ( b and b′ ) was electroporated into the epiblast layer of the chick embryos . cSox2 ( blue ) expression was examined by in situ hybridization ( a , b , a′ , b′ ) . GFP expression ( brown ) indicating the electroporated field was detected by immunohistochemical assays ( a′ and b′ ) . The arrowhead marks the position-plane of the corresponding embryo transverse section below ( i and ii ) . NC , notochord . The values represent the mean ± SD for A–D . ( *p<0 . 05; **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 01310 . 7554/eLife . 02224 . 014Figure 5—figure supplement 1 . HOMEO domain is essential for the neural-promoting effect of Pou3f1 . ( A ) Schematic structure of full-length and domain-deleted mutant Pou3f1 proteins . Row 1 , full-length Pou3f1; Row 2 , Pou3f1 without the N-terminus ( ΔN , missing 1–244 amino acids ) ; Row 3 , Pou3f1 without the POU domain ( ΔPOU , missing 245–324 amino acids ) ; Row 4 , Pou3f1 without the HOMEO domain ( ΔHOMEO , missing 325–499 amino acids ) . ( B ) Gene expression levels in ESCs transfected with control , Pou3f1-full length or with Pou3f1-ΔHOMEO lentiviruses at differentiation day 4 in serum-free medium . The values represent the mean ± SD for B . ( *p<0 . 05; **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 014 The expression of Sox2 , which is an important neural induction gene , is regulated by different enhancers . For example , the N2 enhancer regulates Sox2 expression in the anterior neural plate , and the N1 enhancer regulates Sox2 expression in the posterior neural plate ( Uchikawa et al . , 2003; Takemoto et al . , 2011 ) . Our ChIP-seq data revealed that Pou3f1 binds to the N2 enhancer region of the Sox2 gene and promotes Sox2 expression ( Figure 4E ) . To further confirm this regulation , luciferase assays were conducted using a reporter construct driven by the Sox2N2 enhancer . Wild-type Pou3f1 enhanced luciferase activity; however , the ΔHOMEO mutant did not enhance this activity ( Figure 5C ) . Similarly , ChIP assays revealed that wild-type Pou3f1 , but not Pou3f1-ΔHOMEO , bound to the Sox2N2 enhancer , and neither of them bound to the Sox2N1 enhancer ( Figure 5D ) . Thus , Pou3f1 regulates Sox2 expression by binding to the N2 enhancer , and this activity is mediated by the HOMEO domain . Chick embryos have been widely used as an in vivo model to study early neural development ( Stern , 2005a ) . In early chick embryos , chick Pou3f1 ( cPou3f1 ) was initially expressed at the anterior portion of the primitive streak at HH stage 3+ ( Figure 5E , a ) . Then , the territory of cPou3f1 expanded to the prospective neural plate , where the earliest expression of cSox2 was detectable at HH stage 4 ( Figure 5E , b , i , j ) . From HH stage 5 onward , cPou3f1 expression highly overlapped with cSox2 in the anterior neural plate ( Figure 5E , c–h , k–p ) . These results demonstrated that cPou3f1 was expressed earlier than cSox2 in the prospective neural plate in early chick embryos , suggesting that cPou3f1 activates cSox2 expression in chick embryos . These results are similar to our findings concerning ESC neural differentiation . To determine the function of Pou3f1 in early chick embryos , either the control vector or Pou3f1 was electroporated into the epiblast layer of HH stage 3 chick embryos as a line extending outwards from the prospective neural plate ( Linker et al . , 2009 ) , and the expression of cSox2 was analyzed 12 hr later . The ectopic expression of Pou3f1 induced the lateral expansion of cSox2 expression ( 7/9 ) , whereas the control vector did not ( 0/9 ) ( Figure 5F ) . Taken together , these results suggest that Pou3f1 promotes neural fate commitment by directly activating the expression of neural development-related genes . In addition to the direct regulation of intrinsic factors , Pou3f1 might also interfere with the activities of extrinsic inhibitory signals , such as the BMP and Wnt pathways , in ESC neural differentiation ( Figure 4E , Figure 4—figure supplement 1C ) . Indeed , during ESC neural differentiation , Pou3f1 knockdown increased the expression of the BMP targets Id1 , Id2 , Msx1 , and Msx2 ( Figure 6A ) , whereas Pou3f1 overexpression generated the opposite effect ( Figure 6B ) . In vivo electroporation studies also revealed that the ectopic expression of Pou3f1 reduced the expression of the BMP target gene cId1 ( 6/10 ) at the edge of the chick anterior peripheral ectoderm , whereas the control vector did not ( 0/11 ) ( Figure 6—figure supplement 1A ) . 10 . 7554/eLife . 02224 . 015Figure 6 . Pou3f1 represses BMP and Wnt signaling at the transcriptional level . ( A ) Expression levels of BMP signaling target genes in control and Pou3f1-knockdown ESCs differentiated in serum-free medium . ( B ) Expression levels of BMP signaling target genes in control and Pou3f1-overexpressing ESCs in unbiased differentiation . ( C ) Luciferase assays using BRE-luc in control and Pou3f1-shRNA vector-transfected ESCs with or without BMP4 treatment in N2B27 medium . ( D ) Luciferase assays using BRE-luc in control and Pou3f1-expressing vector-transfected ESCs with or without BMP4 treatment in N2B27 medium . ( E ) Pou3f1 ChIP assays in control , Pou3f1-full length , or in Pou3f1-ΔHOMEO lentivirus-transfected P19 cells . Pou3f1 enrichment at the Id1-BRE was normalized to the Id1 3′ UTR region . ( F ) pSmad1 ChIP assay in control and Pou3f1-full length lentivirus-transfected P19 cells with or without BMP4 treatment . A pSmad1/5/8-specific antibody was used in the assay . pSmad1 enrichment at the Id1-BRE and control 3′ UTR region were analyzed . ( G ) Dose-dependent inhibitory effect of Pou3f1 on the BRE-luc reporter activities . P19 cells were transfected with increasing amounts of Pou3f1-expressing vector and treated with or without BMP4 in N2B27 medium . ( H ) Expression levels of Wnt signaling target genes in control and Pou3f1-knockdown ESCs differentiated in serum-free medium . ( I ) Expression levels of Wnt signaling target genes in control and Pou3f1-overexpressing ESCs in unbiased differentiation . ( J ) Luciferase assays using TOPflash in control and Pou3f1-shRNA vector-transfected ESCs with or without stimulation of Wnt3a in N2B27 medium . ( K ) Luciferase assays using TOPflash in control and Pou3f1-expressing vector-transfected ESCs with or without stimulation of Wnt3a in N2B27 medium . ( L ) Dose-dependent inhibitory effect of Pou3f1 on the TOPflash luciferase reporter activities . P19 cells were transfected with increasing amounts of Pou3f1-expressing vector and treated with or without CHIR99021 in N2B27 medium . The values represent the mean ± SD . ( *p<0 . 05; **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 01510 . 7554/eLife . 02224 . 016Figure 6—figure supplement 1 . Pou3f1 interferes with BMP and Wnt signaling pathways at the transcriptional level . ( A ) In situ hybridization of cId1 expression ( blue ) in chick embryos that were electroporated with IRES-GFP ( control vector , a and a′ ) or with Pou3f1-IRES-GFP ( b and b′ ) . GFP expression ( brown ) was detected in a′ and b′ by immunohistochemical assay . ( B ) Luciferase assays using BRE-luc in P19 cells that were transfected with control , Pou3f1-full length or with each of the Pou3f1-deletion mutant vectors shown in Figure 5—figure supplement 1 with or without BMP4 stimulation in N2B27 medium . ( C ) Luciferase assays using TOPflash-luc in P19 cells that were transfected with control , Pou3f1-full length expression or with each Pou3f1-deletion mutant vector shown in Figure 5—figure supplement 1 with or without Wnt3a stimulation in N2B27 medium . The values represent the mean ± SD for B and C . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 016 Then , we explored the mechanism underlying the Pou3f1-mediated inhibition of BMP targets . Luciferase assays were performed with a four-repeat BMP responsive element ( BRE ) -driven reporter ( Katagiri et al . , 2002 ) to examine BMP activity in ESCs and in P19 cells . Pou3f1 knockdown increased BRE activity in ESCs with or without BMP4 stimulation ( Figure 6C ) , whereas Pou3f1 overexpression partially inhibited BRE-luc activity ( Figure 6D ) . Furthermore , among the several known functional domains , only the HOMEO domain was necessary to maintain the inhibitory effect of Pou3f1 on BMP signaling ( Figure 6—figure supplement 1B ) . ChIP assays using a Pou3f1 antibody were performed , and we found that the binding of wild-type Pou3f1 , rather than Pou3f1-ΔHOMEO , was specifically enriched at the BRE region of the Id1 gene promoter ( Figure 6E ) . We also performed ChIP assays using a pSmad1 antibody and found that pSmad1 bound to the BRE locus of the Id1 promoter , but not to the 3′ UTR region , in the presence of BMP4 ( Figure 6F , open column ) . Interestingly , Pou3f1 interfered with the binding of pSmad1 to the BRE locus of the Id1 promoter ( Figure 6F , filled column ) . Moreover , Pou3f1 repressed BMP-induced luciferase activity in a dose-dependent manner ( Figure 6G ) . We also observed that Pou3f1 did not affect the stimulation , degradation , dephosphorylation , or intracellular translocation of pSmad1 ( data not shown ) , excluding the fact that Pou3f1 regulates the BMP pathway through a signaling cascade . Together , these results suggest that Pou3f1 may inhibit BMP signaling by interfering with pSmad1 binding to the regulatory elements and then repressing the transcription of target genes . Similar to the BMP pathway , Wnt3a , Axin2 , Dkk1 , and Myc in Wnt signaling were regulated by Pou3f1 during ESC neural fate commitment ( Figure 6H , I ) . ChIP-seq data revealed that Pou3f1 directly binds to the promoter regions of these Wnt signaling targets ( Figure 4E , Figure 4—figure supplement 1C ) . In luciferase-based TOPflash ( TCF optimal promoter ) Wnt reporter assays ( Korinek et al . , 1997 ) , TOPflash-luc activity was enhanced by Pou3f1 knockdown ( Figure 6J ) , and Wnt3a-induced luciferase activity was partially reduced with Pou3f1 overexpression ( Figure 6K ) . We also found that the HOMEO domain is crucial for sustaining the inhibitory effect of Pou3f1 on TOPflash-luc activity ( Figure 6—figure supplement 1C ) . Pou3f1 also inhibited Wnt agonist CHIR99021-induced TOPflash-luc activity in a dose-dependent manner ( Figure 6L ) . Together , these results suggest that Pou3f1 interferes with the BMP and Wnt signaling pathways by directly inhibiting the transcription of their target genes . The BMP and Wnt signaling pathways have strong inhibitory effects on ESC neural differentiation ( Haegele et al . , 2003; Ying et al . , 2003 ) , and the above data suggest that Pou3f1 inhibits BMP and Wnt transcriptional activities . Thus , we investigated whether Pou3f1 could attenuate their inhibitory effects . ESCs were differentiated in serum-free medium with or without BMP4 for 48 hr from day 2 to day 4 , and Dox was simultaneously added to induce Pou3f1 overexpression ( Figure 7A ) . Consistent with our previous observation ( Zhang et al . , 2010a ) , BMP4 inhibited the expression of the neural markers Sox1 , Pax6 , Nestin , and Tuj1 at both the mRNA and protein levels ( Ctrl BMP4+ compared with Ctrl BMP4− in Figure 7A–C ) . As expected , Pou3f1 overexpression fully restored the expression of these markers in ESC neural differentiation ( Pou3f1 BMP4+ compared with Ctrl BMP4+ in Figure 7A–C ) . Furthermore , Pou3f1 overexpression also fully rescued the neural inhibitory effects of Wnt3a ( Figure 7D ) . To test whether Pou3f1 relieves the neural inhibition mediated by the BMP signaling pathway in vivo , we co-electroporated Xenopus BMP4 ( xBMP4 ) with a control vector or with Pou3f1 into the pre-neural plate region of chick embryos at HH stage 3 . cSox2 expression was completely repressed by xBMP4 ( 16/19 ) , whereas the forced expression of Pou3f1 partially recovered cSox2 expression ( 10/19 ) ( Figure 7E ) . Together , these results suggest that Pou3f1 alleviates the inhibitory activities of both BMP and Wnt signals during neural fate commitment . 10 . 7554/eLife . 02224 . 017Figure 7 . Pou3f1 alleviates the inhibitory effects of BMP4 and Wnt3a on neural fate commitment . ( A ) Inducible Pou3f1-overexpressing ESCs were cultured as EBs in serum-free medium for 4 days with or without BMP4/Dox treatment from day 2 to day 4 . Gene expression levels were detected by Q-PCR . ( B ) Immunocytochemical assays using day 4 EBs described in A . The EBs were stained with Sox ( red ) and with Oct4 ( green ) . Scale bars , 100 μm . ( C ) Statistical analysis of results from the immunocytochemical assay of Sox+/Oct4− , Nestin+ , and Tuj1+ cells in EBs and of Tuj1+ replated cells . ( D ) Pou3f1-overexpressing ESCs were cultured as EBs in serum-free medium for 4 days with or without Wnt3a/Dox addition from day 2 to day 4 . Gene expression levels were detected by Q-PCR . ( E ) Pou3f1 partially rescues the inhibitory effects of xBMP4 on cSox2 . In situ hybridization of cSox2 ( blue ) in chick embryos that were co-electroporated with xBMP4 plus IRES-GFP ( control vector , a and a′ ) or Pou3f1-IRES-GFP ( b and b′ ) , respectively . GFP expression ( brown ) was detected in a′ and b′ by immunohistochemistry . The values represent the mean ± SD for A , C , and D . ( *p<0 . 05; **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02224 . 017
In the past decades , studies on early neural development have mainly focused on the role of extrinsic signals . Recent works have provided new insights concerning the intracellular programs involved in early neural fate commitment in the absence of extracellular signals ( Kamiya et al . , 2011; Iwafuchi-Doi et al . , 2012 ) . However , how the intrinsic and extrinsic regulatory networks are orchestrated to ensure the appropriate initiation of neural differentiation remains largely unclear . Our in vitro and in vivo data indicate that Pou3f1 is crucial for ESC neural fate commitment and promotes the transition from EpiSCs to neural progenitor cells . Furthermore , Pou3f1 functions as an intrinsic regulator of both intracellular transcription factors and extracellular inhibitory signals during neural fate commitment . Pou3f1 was previously reported to be a transcription factor that participates in Schwann cell development and myelination ( Bermingham et al . , 1996; Jaegle et al . , 1996 ) . The Pou3f1 gene expression profiles in mouse embryos in vivo ( Figure 3; Zwart et al . , 1996 ) and of ESC differentiation in vitro ( Figure 1 , Figure 1—figure supplement 1 ) imply that Pou3f1 may also participate in early neural development . Indeed , the shRNA-mediated knockdown of Pou3f1 in ESCs results in the reduced expression of the neural markers Sox1 , Pax6 , and Tuj1 in serum-free medium ( Figure 1 ) . However , the compensation of the POU III member Brn2 may be one of the reasons for the mild effects observed during ESC neural differentiation after Pou3f1 depletion ( Figure 1 , Figure 1—figure supplement 2 ) . Brn2 compensation and the different ESC lines and culture system used potentially explain why the Pou3f1 knockdown effects are not reported in Iwafuchi-Doi's study ( Iwafuchi-Doi et al . , 2012 ) . On the other hand , our results are consistent with their results indicating that the forced expression of Pou3f1 promotes the expression of neural markers ( Figure 1 , Iwafuchi-Doi et al . , 2012 ) . Clearly , Pou3f1 is necessary and sufficient for ESC neural differentiation . Pou3f1-overexpressing or Pou3f1-knockdown ESCs generate EpiSC-like colonies that are similar to the control ESCs . However , the neural differentiation of Pou3f1-overexpressing or Pou3f1-knockdown EpiSCs is markedly different from the control EpiSCs , suggesting that Pou3f1 functions specifically during the neural transition from the epiblast to neural progenitor cells ( Figure 2 ) . Furthermore , in our blastocyst injection study , the contribution of Pou3f1-knockdown ESCs to the neuroectoderm was severely impaired ( Figure 3 ) , indicating that Pou3f1 most likely functions cell-autonomously during the neural fate commitment of pluripotent stem cells in vivo . Our findings revealed that Pou3f1 is an essential transcription factor required for the intrinsic neural differentiation of pluripotent stem cells . Cell fate determination is regulated in a step-wise fashion via the activation or inhibition of lineage specification factors ( Pfister et al . , 2007 ) . Several transcription factors , including Pax6 , Sox2 , Zfp521 , Zic1 , and Zic2 , promote neural gene expression and play roles in the derivation of the anterior neural plate ( Iwafuchi-Doi et al . , 2012; Kamiya et al . , 2011; Zhang et al . , 2010b ) . Zfp521 and Zic1/2 are important for neural fate consolidation rather than initiation ( Aruga , 2004; Kamiya et al . , 2011; Iwafuchi-Doi et al . , 2012 ) . To date , the intrinsic modulators essential for the early neural initiation event have not been identified . In this study , the combination of RNA-seq and ChIP-seq enabled us to investigate the underlying molecular mechanisms governing Pou3f1-mediated neural fate commitment in ESCs at the genome-wide level and to determine whether Pou3f1 is involved in the initiation of neural differentiation . Our results indicate that Pax6 , Sox2 , Zfp521 , and dozens of other known neural fate-promoting genes are enhanced by Pou3f1 overexpression during ESC differentiation ( Figures 4 and 5 ) . Furthermore , ChIP-seq data reveal that Pou3f1 is enriched at the regulatory regions of Pax6 , Sox2 , Zfp521 , Zic1 , and Zic2 genomic loci ( Figure 4 , Figure 4—figure supplement 1 ) , indicating that Pou3f1 directly activates these neural fate-promoting genes . Surprisingly , Pou3f1 did not bind the Sox2N1 enhancer , which controls Sox2 posterior neural plate expression; Pou3f1 preferentially binds to the Sox2N2 enhancer , which drives Sox2 anterior neural plate expression ( Figures 4 and 5 ) . This result is consistent with the in vivo Pou3f1 and Sox2 overlapping expression patterns during neural fate commitment . Our results are also consistent with the notion that the anterior-most portion of the epiblast constitutes the primitive neural identity following neural induction ( Andoniadou and Martinez-Barbera , 2013; Li et al . , 2013 ) . Moreover , our observations confirm the hypothesis proposed in a recent study ( Kamiya et al . , 2011 ) that Pou3f1 functions upstream of Zfp521 during ESC neural differentiation ( Figures 4 and 5 , Figure 4—figure supplement 1 ) . Taken together , these findings demonstrate that Pou3f1 is most likely an intrinsic neural initiation factor that participates in the transition of pluripotent stem cells to NPCs by directly activating a group of key neural fate-promoting genes . In addition to intrinsic factors , several extrinsic signals involved in early neural fate commitment have been intensively studied , including BMPs and Wnts . However , how BMP/Wnt inhibitory activities are alleviated to secure neural fate commitment has not been fully elucidated . BMP and Wnt signals function partially through their downstream genes ( ten Berge et al . , 2011; Varlakhanova et al . , 2010; Ying et al . , 2003; Zhang et al . , 2010a ) . Unlike Zfp521 , which did not affect BMP signaling ( Kamiya et al . , 2011 ) , the expression of a few genes related to BMP and Wnt pathways was regulated by Pou3f1 knockdown or by overexpressing in EBs at day 4 ( Figure 6 ) . However , this regulation was not evident in ESCs or in EBs at day 2 ( data not shown ) . This finding suggests that Pou3f1 interferes with the BMP/Wnt signaling pathways during the process of neural conversion from epiblast to NPCs . Moreover , Pou3f1 is recruited to the genomic loci of many downstream targets of BMP and Wnt signals , such as Id1 , Id2 , Myc , and Axin2 ( Figure 4 , Figure 4—figure supplement 1 ) . We also found that Pou3f1 represses the transcriptional activation of a BMP responsive element ( BRE ) by BMP4 and of a TCF optimal promoter ( TOP ) by Wnt3a ( Figure 6 , Figure 6—figure supplement 1 ) . Our data further suggest that the binding of pSmad1 to the BRE locus is potentially compromised in the presence of Pou3f1 , which results in the repression of BMP signaling pathway activity ( Figure 6 ) . However , other possibilities , such as the recruitment of repressing cofactors by Pou3f1 , could not be excluded by the present study . Notably , Pou3f1 overexpression enables neural differentiation even in the presence of BMP4 or Wnt3a ( Figure 7 ) . We propose that the Pou3f1-dependent repression of the BMP and Wnt signaling pathways and the activation of intrinsic neural lineage genes together are involved in the neural fate-promoting activity of Pou3f1 . In summary , our study establishes Pou3f1 as a critical dual-regulator of intrinsic transcription factors and extrinsic signals to promote neural fate commitment . This study provides a better understanding of the internal mechanism of neural initiation . Nonetheless , many questions concerning this process remain unanswered , such as whether the dual regulatory mechanism of Pou3f1 is also utilized to initiate the mouse neural program in vivo , whether this two-way modulating processes occurs simultaneously or in a sequential , temporal manner , and how the controversial activation/inhibition activities of the Pou3f1 transcription factor is achieved . All these unanswered questions lay the foundation for exciting future work concerning the interplay between the extrinsic and intrinsic cues during early embryonic neural fate commitment .
Mouse ESCs ( R1 and R1/E ) were maintained on feeders in standard medium . ESC serum-free neural differentiation ( 8% knockout serum replacement medium ) and EB replating were performed as described previously ( Watanabe et al . , 2005; Zhang et al . , 2010a ) . ESC unbiased differentiation in serum-containing medium ( 10% FBS ) was performed as described previously ( Zhang et al . , 2013 ) . EpiSCs were cultured on FBS-coated dishes in a chemically defined medium ( CDM ) supplemented with 20 ng/ml activin A ( R&D Systems , Minneapolis , MN ) and with 12 ng/ml bFGF ( Invitrogen , Carlsbad , CA ) ( CDM/AF ) as described previously ( Brons et al . , 2007; Zhang et al . , 2010a ) . To generate ESD-EpiSCs ( ESC-derived epiblast stem cells ) , ESCs or cell aggregates were dissociated into single cells after treatment with 0 . 05% Trypsin-EDTA at 37°C for 2 min . Individual cells were seeded at a density of 2 . 0 × 105 cells per 35-mm dish in CDM/AF . After 6 days , the surviving cells formed large compact colonies . P19 cells were cultured as described previously ( Jin et al . , 2009 ) . Factors and inhibitors , including BMP4 ( 10 ng/ml , R&D Systems , Minneapolis , MN ) , Wnt3a ( 100 ng/ml , R&D Systems , Minneapolis , MN ) , and CHIR99021 ( 3 μM , Stemgent , Cambridge , MA ) , were used . For Pou3f1 knockdown in ESCs , the lentiviral vector pLentiLox 3 . 7 , which expresses shRNA and GFP , was used . A reference shRNA sequence ( Huang et al . , 2010b ) was used as a negative control . The control and Pou3f1 shRNA sequences are shown in Supplementary file 1 . Lentiviral packaging and cell transfection were performed as described ( Tiscornia et al . , 2006 ) . GFP-positive cells were sorted using a FACS-Aria cell sorter ( BD Biosciences , San Jose , CA ) and propagated . For stable overexpression , Pou3f1 was cloned into the lentiviral expression vector pFUGW-IRES-EGFP ( Naldini et al . , 1996 ) . The PCR primers used in the cloning are listed in Supplementary file 1 . The empty vector pFUGW-GFP was used as a negative control . For Pou3f1-inducible overexpression , the Pou3f1-IRES-EGFP fragment was constructed and inserted into the lentiviral vector pLVX-Tight-Puro ( Clontech , Mountain View , CA ) . After co-transfection of pLVX-Tight-Puro-Pou3f1-IRES-EGFP and rtTA lentiviruses for 48 hr , the stable transfection was selected by puromycin ( 2 μg/ml , Sigma ) . The culture medium supplemented with Dox ( 2 μg/ml , Sigma-Aldrich , St . Louis , MO ) was used for inducing the overexpression of Pou3f1 , and Dox was not added to the control group . Immunocytochemistry was performed as described previously ( Xia et al . , 2007 ) . The mouse monoclonal antibodies included anti-Oct4 ( Santa Cruz Biotechnology , Santa Cruz , CA ) , and anti-Tuj1 ( Covance , San Diego , CA ) . The rabbit polyclonal antibodies included anti-Nestin ( Upstate Biotech , Lake Placid , NY ) , anti-Pax6 ( Covance , San Diego , CA ) , and an anti-Sox1/ ( 2 ) /3 that preferentially recognize Sox1 and Sox3 over Sox2 ( Okada et al . , 2004; Tanaka et al . , 2004 ) . Cy3 and Cy5 ( Jackson Immunoresearch Laboratories , West Grove , PA ) secondary antibodies were used in this study . Fluorescence detection and imaging were performed on a Leica confocal microscope or on an Olympus fluorescence microscope . Total RNA was extracted from cells using TRIzol reagent ( Invitrogen , Carlsbad , CA ) . Reverse transcription and Q-PCR analysis were performed using an Eppendorf Realplex2 ( Peng et al . , 2009 ) . Primers for Q-PCR analysis are listed in Supplementary file 1 . Whole-mount in situ hybridizations were performed as described previously ( Huang et al . , 2010a ) . The following probes were used: mPou3f1 ( 3′ UTR of mouse Pou3f1 mRNA , PCR-amplified from cDNA ) , cPou3f1 , cSox2 , and cId1 . R1 ESCs constitutively expressing pFUGW-IRES-EGFP were used as the control for visualizing the contribution of the injected cells in vivo . To obtain chimeric embryos , GFP-labeled Pou3f1-KD , Pou3f1-OE , or control ESCs were injected into E2 . 5 mouse blastocysts respectively , and the cells were then transferred into the uteri of day 2 . 5 pseudopregnant ICR female mice . For the inducible Pou3f1-overexpresing ESCs , the recipient ICR female mice were fed with Dox ( 2 mg/ml ) in water after blastocyst injection . Mouse embryos were collected at E8 . 5 to E9 . 0 . After transverse section , the fluorescent signals of embryos were detected by confocal microscope . Our animal experiments are conducted with the highest ethical standards . Fertilized eggs ( Shanghai Academy of Agricultural Sciences , Shanghai , China ) were incubated at 38°C to HH stage 3/3+ ( Hamburger and Hamilton , 1992 ) . Gene electroporation and new culture were performed as described previously ( New , 1955; Voiculescu et al . , 2008 ) . The control vector pCAGGS-IRES-GFP and the Pou3f1 expression construct pCAGGS-mPou3f1-IRES-GFP were used . Whole-mount immunostaining of GFP was performed as described previously ( Huang et al . , 2010a ) . The luciferase assay was described previously ( Jin et al . , 2009 ) . Plasmids were co-transfected in ESCs or in P19 cells in N2B27 medium for 24 hr . f Factor treatment was applied for 10 hr , and then the luciferase activities were measured using a Dual-Luciferase Reporter Assay system ( Promega , Madison , WI ) with a Turner Design 2020 luminometer . ChIP assays were performed according to the manufacturer's protocol ( Protein A/G Agarose/Salmon Sperm DNA [Upstate Biotech , Lake Placid , NY] and Dynabeads Protein A/G [Invitrogen , Carlsbad , CA] ) , and detailed procedures were described previously ( Jin et al . , 2009 ) . ChIP was performed with 2 μg antibody against phosphorylated Smad1/5/8 ( Cell Signaling ) or Pou3f1 ( Santa Cruz Biotechnology , Santa Cruz , CA ) . Normal IgG was used as negative control . Q-PCR was used to amplify various regions of the target gene genome , and primers for ChIP-qPCR are listed in Supplementary file 1 . The high-throughput sequencing was performed by the Computational Biology Omics Core , PICB , Shanghai . The SOAP version 2 . 20 alignment tool was used to align ChIP-Seq reads to the mouse genome build mm9 ( Li et al . , 2009 ) . Only reads with less than two mismatches that uniquely mapped to the genome were used in subsequent analyses . Using FindPeaks Homer software , Pou3f1 binding peaks with fourfold greater normalized tags were identified in ChIP experiments compared with the control ( Heinz et al . , 2010 ) . We calculated the distance from the peak centers to the annotated transcription start sites ( TSS ) and then defined the nearest genes as peak-related genes . Raw reads were mapped to mm9 using the TopHat version 1 . 4 . 1 program ( Trapnell et al . , 2009 ) . We assigned FPKM ( fragment per kilo base per million ) as an expression value for each gene using Cufflinks version 1 . 3 . 0 software ( Trapnell et al . , 2010 ) . Then , Cuffdiff software was used to identify differentially expressed genes between treatment and control samples ( Trapnell et al . , 2013 ) . Differentially expressed gene heat maps were clustered by k-means clustering using the Euclidean distance as the distance and visualized using Java TreeView software ( Saldanha , 2004 ) . To investigate the functions of genes with Pou3f1 binding sites and differentially expressed after Pou3f1 perturbation , functional enrichment analyses were performed using the Database for Annotation , Visualization , and Integrated Discovery ( DAVID ) . Each experiment was performed at least three times , and similar results were obtained . The data are presented as the mean ± SD . Student's t test was used to compare the effects of all treatments . Statistically significant differences are indicated as follows: * for p<0 . 05 and ** for p<0 . 01 . | After an egg has been fertilized , it undergoes a series of divisions to produce a ball of cells known as a blastocyst . The cells within the blastocyst are pluripotent stem cells , which have the potential to become many different types of cell . After a few days , the stem cells organize into three layers—an innermost layer called the endoderm , a middle layer of mesoderm , and an outer layer of ectoderm—that ultimately give rise to different types of tissues . The brain and nervous system are formed from cells in the neuroectoderm , which is part of the ectoderm . Now , Zhu et al . have shown that a transcription factor called Pou3f1 triggers stem cells within a region of the ectoderm to turn into neural progenitor cells , thereby generating the neuroectoderm . These neural progenitor cells then go on to become neurons and glial cells that make up the brain and nervous system . Using a virus to reduce levels of Pou3f1 in embryonic stem cells grown in a dish led to a drop in the number of stem cells that committed to neural progenitor cells . Overexpressing Pou3f1 in the stem cells restored the number of neural progenitor cells . Together these results showed that Pou3f1 is both necessary and sufficient for the conversion of embryonic stem cells into future neurons and glia . The same result was seen when embryonic stem cells containing either reduced or elevated levels of Pou3f1 were injected into 2 . 5-day-old mouse blastocysts , which were then implanted into surrogate females . The resulting embryos comprised some cells with normal levels of Pou3f1 , and others with either too little or too much . Cells with elevated Pou3f1 mostly became neural progenitors , whereas those with reduced levels rarely did so . Gene expression studies revealed that Pou3f1 promoted the formation of neural progenitor cells by activating the expression of pro-neuronal genes inside the stem cells , and by blocking anti-neuronal pathways called Wnt/BMP signaling cascades initiated outside the cells . By revealing the two roles of Pou3f1 , Zhu et al . have increased our understanding of one of the earliest stages of nervous system development . Further work is required to determine exactly how Pou3f1 exerts its effects and , in particular , whether it performs its two roles simultaneously or in sequence . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2014 | The transcription factor Pou3f1 promotes neural fate commitment via activation of neural lineage genes and inhibition of external signaling pathways |
Sperm-packaged DNA must undergo extensive reorganization to ensure its timely participation in embryonic mitosis . Whereas maternal control over this remodeling is well described , paternal contributions are virtually unknown . In this study , we show that Drosophila melanogaster males lacking Heterochromatin Protein 1E ( HP1E ) sire inviable embryos that undergo catastrophic mitosis . In these embryos , the paternal genome fails to condense and resolve into sister chromatids in synchrony with the maternal genome . This delay leads to a failure of paternal chromosomes , particularly the heterochromatin-rich sex chromosomes , to separate on the first mitotic spindle . Remarkably , HP1E is not inherited on mature sperm chromatin . Instead , HP1E primes paternal chromosomes during spermatogenesis to ensure faithful segregation post-fertilization . This transgenerational effect suggests that maternal control is necessary but not sufficient for transforming sperm DNA into a mitotically competent pronucleus . Instead , paternal action during spermiogenesis exerts post-fertilization control to ensure faithful chromosome segregation in the embryo .
Faithful chromosome segregation requires careful orchestration of chromosomal condensation , alignment , and movement of mitotic chromosomes during every eukaryotic cell division ( Rhind and Russell , 2012 ) . The very first embryonic mitosis in animals requires additional synchronization . Paternally and maternally inherited genomes undergo independent chromatin reorganization and replication prior to mitotic entry . For instance , maternal chromosomes must complete meiosis ( Sen et al . , 2013 ) and then transition from a meiotic conformation to an interphase-like state in preparation for replication . The sperm-deposited , paternal chromosomes must undergo an even more radical transition from a highly compact , protamine-rich state to a decondensed , histone-rich state before DNA replication ( Braun , 2001; Miller et al . , 2010 ) . Despite these divergent requirements to achieve replication- and mitotic-competency , maternal and paternal genomes synchronously enter the first mitosis . Failure to carry out paternal chromosome remodeling in a timely fashion results in paternal genome loss and embryonic inviability ( Loppin et al . , 2001; McLay and Clarke , 2003; Landmann et al . , 2009 ) . The transition from a protamine-rich sperm nucleus to a competent paternal pronucleus requires the action of numerous maternally deposited proteins in the egg ( McLay and Clarke , 2003 ) . For instance , paternal genome decondensation post-fertilization requires the integration of histone H3 . 3 , a histone variant deposited by the maternal proteins HIRA ( Loppin et al . , 2005a ) , CHD1 ( Konev et al . , 2007 ) , and Yemanuclein ( Orsi et al . , 2013 ) . Similarly , maternally-deposited MH/Spartan protein localizes exclusively to the replicating paternal genome and is required for faithful paternal chromosome segregation during the first embryonic division ( Delabaere et al . , 2014 ) . These and other studies demonstrate the essential role of maternally-deposited machinery in rendering competent sperm-deposited DNA and ultimately , ensuring faithful paternal genome inheritance . Is paternal control also necessary for the extensive decondensation and re-condensation of the post-fertilization paternal genome ? If so , disruption of such control would manifest as paternal effect lethality ( PEL ) . Unlike male sterility mutants that lack motile sperm , PEL mutants make abundant motile sperm that fertilize eggs efficiently . However , embryos ‘fathered’ by PEL mutants are inviable . Only a handful of PEL genes have been characterized in animals ( Browning and Strome , 1996; Fitch and Wakimoto , 1998; Fitch et al . , 1998; Loppin et al . , 2005b; Smith and Wakimoto , 2007; Gao et al . , 2011; Seidel et al . , 2011 ) . These encode proteins that mediate sperm release of paternal DNA , sperm centriole inheritance , and paternal chromosome segregation . Only one of these PEL proteins directly localizes to paternal chromosomes; the sperm-inherited K81 protein localizes exclusively to paternal chromosome termini and ensures telomere integrity ( Dubruille et al . , 2010; Gao et al . , 2011 ) . The maintenance of telomeric epigenetic identity joins a growing list of examples of sperm-to-embryo information transmission via protein or RNA inheritance ( e . g . , diet: [Ost et al . , 2014] , stress: [Rodgers et al . , 2013] , embryonic patterning: [Bayer et al . , 2009] , transcriptional competency: [Hammoud et al . , 2010; Rando , 2012; Ihara et al . , 2014] ) . Despite our new appreciation of paternal control over epigenetic information transfer , there are no reports of paternal control over the global chromatin reorganization required for synchronous mitosis across paternally and maternally inherited genomes . Indeed , in the absence of any known paternal protein-directed genome remodeling , a model has emerged that maternal proteins might be sufficient for transforming tightly packaged sperm DNA into a fully competent paternal pronucleus . The notion that maternal control is sufficient to accomplish paternal genome remodeling is challenged by recent findings from the intracellular Wolbachia bacterium that infects more than 50% of insect species ( Hilgenboecker et al . , 2008 ) . Wolbachia-infected Drosophila males mated to uninfected females father embryos that arrest soon after the first zygotic mitosis ( Lassy and Karr , 1996 ) . Embryonic arrest occurs because paternal genomes enter the first mitosis with unresolved sister chromatids that fail to separate on the mitotic spindle ( Callaini et al . , 1997; Landmann et al . , 2009 ) . Although the identity of the host factor ( s ) manipulated by Wolbachia to mediate this transgenerational effect is still unknown , what is clear is that pre-fertilization , Wolbachia subverts the paternal germline machinery that helps direct global genome remodeling of paternal chromosomes in the embryo . Wolbachia action during spermiogenesis leads to paternal-maternal genome asynchrony and ultimately , failure of paternal chromosomes to separate on the first mitotic spindle ( Callaini et al . , 1997; Landmann et al . , 2009 ) . Despite decades of interest , the molecular basis of paternal control has remained elusive . To investigate the potential for paternal control over sperm genome remodeling post-fertilization , we took a candidate gene approach , focusing on the Heterochromatin Protein 1 ( HP1 ) proteins that orchestrate genome-wide chromosomal organization in plants , animals , fungi , and some protists ( Lomberk et al . , 2006 ) . HP1 proteins are defined as such by a combination of two domains—a chromodomain that mediates protein-histone interactions and a chromoshadow domain that mediates protein–protein interactions ( Aasland and Stewart , 1995; Eissenberg and Elgin , 2000 ) . The biochemical properties of HP1 members ( Canzio et al . , 2014 ) support a diversity of chromatin-dependent processes in the soma , including DNA replication ( Pak et al . , 1997; Schwaiger et al . , 2010 ) , telomere integrity ( Fanti et al . , 1998 ) , and chromosome condensation ( Kellum et al . , 1995 ) . Recently , we carried out a detailed phylogenomic analysis of the HP1 gene family in Drosophila that revealed numerous testis-restricted HP1 proteins ( Levine et al . , 2012 ) . Given the established roles of HP1 proteins ( Eissenberg and Elgin , 2000; Lomberk et al . , 2006; Vermaak and Malik , 2009; Canzio et al . , 2014 ) , we posited that these newly discovered male-specific HP1 genes might represent excellent candidates for encoding chromatin functions specialized for paternal genome organization and remodeling in the early embryo . Using detailed genetic and cytological analyses , here we show that one of these testis-specific HP1 proteins , Heterochromatin Protein 1E ( HP1E ) , is essential for priming the paternal genome to enter embryonic mitosis in synchrony with the maternal genome in D . melanogaster . Intriguingly , HP1E is able to mediate this priming function transgenerationally i . e . , the HP1E protein itself is not epigenetically inherited . We further show that absence of HP1E especially imperils mitotic fidelity of the heterochromatin-rich , paternal sex chromosomes . Thus , our study firmly establishes that both maternal and paternal control are necessary for paternal genome remodeling in the early Drosophila embryo .
The HP1E gene is a testis-restricted Drosophila HP1 paralog born more than 60 million years ago ( Levine et al . , 2012 ) . To investigate the possibility that HP1E acts during chromatin reorganization prior to sperm maturation , we generated transgenic flies that encoded a Flag- or YFP-tagged HP1E fusion protein , driven by the native HP1E promoter . In addition , we raised a highly specific polyclonal antibody against HP1E ( Figure 1—figure supplement 1 ) . All three reagents revealed that HP1E localizes to developing spermatids subsequent to the completion of meiosis II ( Figure 1A–C , Figure 1—figure supplements 2 , 3 , 4 ) , ruling out a role for HP1E during the pre-meiotic or meiotic phases of spermatogenesis . HP1E signal was generally diffuse across a subset of the chromatin throughout spermiogenesis ( Figure 1A–C ) but disappeared completely at sperm maturation ( Figure 1D ) . Native expression of the HP1E-YFP fusion transgene confirmed our immunofluorescence results ( Figure 1—figure supplement 4 ) , implying that HP1E disappearance in late spermiogenesis is not due to antibody inaccessibility in the highly condensed sperm head . These data demonstrate that HP1E localizes to paternal chromosomes only during the radical reorganization of histone-rich chromatin into protamine-rich sperm DNA and disappears once sperm mature . 10 . 7554/eLife . 07378 . 003Figure 1 . HP1E localization in D . melanogaster spermatogenesis . We highlight four stages of spermatogenesis in D . melanogaster testes: ( A ) round spermatids , ( B , C ) elongated spermatids , and ( D ) mature sperm . HP1E localization was visualized with a Flag epitope-tagged HP1E transgene driven by a native promoter . We find that Flag-HP1E protein ( green ) localizes to the DNA ( blue ) of post-meiotic spermatids and persists until sperm maturation but is not present on mature sperm . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 00310 . 7554/eLife . 07378 . 004Figure 1—figure supplement 1 . HP1E antibody is specific . HP1E antibody recognizes heat-shock induced HP1E expression in cell culture ( S2 cells ) . Predicted sizes: native HP1E 20 . 3 kD , Flag:HP1E , 24 kD , YFP:HP1E 46 kD . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 00410 . 7554/eLife . 07378 . 005Figure 1—figure supplement 2 . HP1E protein localizes to post-meiotic spermatids . HP1E-stained whole-mount testis ( upper panel ) . HP1E localizes to spermatids after meiosis II . HP1E first appears during spermatogenesis on round spermatids with a prominent ‘protein body’ or ‘pseudonucleus’ ( arrow ) , which corresponds to the region of HP1E exclusion . A protein body is diagnostic of a post-meiotic stage . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 00510 . 7554/eLife . 07378 . 006Figure 1—figure supplement 3 . Anti-3xFlag ( M2 ) exhibits no localization to spermatids in a non-Flag tagged HP1E genetic background . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 00610 . 7554/eLife . 07378 . 007Figure 1—figure supplement 4 . Fixed testis expressing HP1E-YFP driven by a native promoter recapitulates immunofluorescence results . HP1E localizes to DNA of round and elongated spermatids . We observed no evidence of HP1E localization to DNA of mature sperm . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 007 Many characterized HP1 proteins serve critical heterochromatin organization roles and are often used as markers of canonical heterochromatin in somatic cells ( Eissenberg and Elgin , 2000 ) . However , heterochromatin organization is poorly defined in post-meiotic developing spermatids ( Dubruille et al . , 2010; Hennig and Weyrich , 2013 ) , precluding our ability to ask if HP1E co-localizes with classic bulk heterochromatin markers . We were also unable to directly ascertain HP1E localization to specific heterochromatin loci using chromatin immunoprecipitation based methods ( e . g . , ChIP-seq ) . Instead , we adopted an orthogonal approach . We conducted RNA-seq on control and HP1E-depleted testes , reasoning that loss of heterochromatin organizing protein would uniquely perturb the global transcriptional readout from this specialized genome compartment . We found that HP1E knockdown during this narrow developmental stage affects the expression of over 700 genes ( fdr < 0 . 05 , Figure 2—source data 1 ) . Of those 700 genes significantly misregulated upon knockdown , there were very few genes that encode known chromatin-modifying or chromosome-bound proteins; of these , most represent uncharacterized genes . However , one intriguing pattern that does emerge from this dataset is that 100% of significantly misregulated heterochromatin-embedded genes are upregulated when HP1E is depleted ( Figure 2 ) . In comparison , less than 60% of significantly misregulated euchromatin-embedded genes are upregulated . This dichotomy between the effects of HP1E depletion on the euchromatic and heterochromatic compartment is highly significant ( p < 0 . 00001 ) and suggests a direct action of HP1E in the heterochromatic compartment , akin to its closest HP1 relative , HP1A ( Kellum et al . , 1995; Levine et al . , 2012 ) . Alternatively , one or more of the 700 mis-regulated genes could be responsible for modifying paternal chromatin . However , no obvious candidate genes involved in chromatin modification or binding emerged from the list of mis-regulated genes ( Figure 2—source data 1 ) . In the absence of direct evidence of heterochromatin localization via cytology or ChIP-seq , we can only tentatively conclude that HP1E acts directly on this genome compartment . In contrast , HP1E unambiguously localizes to chromatin during sperm development . 10 . 7554/eLife . 07378 . 008Figure 2 . Heterochromatin-embedded genes are globally perturbed upon HP1E-depletion . HP1E depletion in testis results directly or indirectly in mis-regulation of hundreds of genes . Volcano plot illustrates the fold up- and down-regulation of euchromatin-embedded genes ( gray points ) and heterochromatin-embedded genes ( black points ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 00810 . 7554/eLife . 07378 . 009Figure 2—source data 1 . Results of RNA-seq comparisons between testes of wild-type vs HP1E-depleted males , rank-ordered by the false discovery rate . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 009 To investigate HP1E's role in male fertility , we generated HP1E-depleted fathers by driving a UAS promoter-hairpin homologous to the D . melanogaster HP1E transcript with either an actin5C-Gal4 driver ( ubiquitous expression ) or vasa-Gal4 driver ( male germline expression ) . Both drivers efficiently knocked down HP1E expression ( Figure 3—figure supplement 1 ) and both resulted in highly penetrant male sterility ( Figure 3A ) . To rule out off-target effects of RNAi , we engineered a recoded version of HP1E in which all synonymous sites were changed but the amino acid sequence was preserved . This recoded , RNAi-resistant HP1E transgene ( Figure 3A , Figure 3—figure supplement 2 ) fully rescued fertility . In parallel , we also generated an HP1E-null allele using a TAL-effector nuclease ( Figure 3—figure supplement 3 ) . We found that HP1E knockout males are also completely sterile , and this sterility is also fully reversed by the recoded HP1E transgene ( Figure 3—figure supplement 4 ) . Thus , HP1E is required for male fertility in D . melanogaster . 10 . 7554/eLife . 07378 . 010Figure 3 . HP1E is a paternal effect lethal in D . melanogaster . ( A ) HP1E knockdown via simultaneous presence of both a UAS-driven HP1E dsRNA gene and a Gal4 driver results in highly penetrant male sterility . Fertility can be fully restored by an HP1E transgene recoded at all synonymous sites , driven by a native HP1E promoter ( ‘rescue transgene’ ) . Please refer to Figure 3—source data 1 . ( B ) HP1E-depleted males produce abundant motile sperm ( seminal vesicle ) , which are efficiently transferred to females ( seminal receptacle ) , and fertilize the egg ( embryo ) . We visualized sperm tails using the ‘don juan-GFP’ transgene ( 29 ) . ( C ) Unlike wild-type embryos ( gray ) , embryos fathered by HP1E-depleted males ( black ) arrest after 3–4 rounds of nuclear divisions ( Mann–Whitney U: p < 0 . 0001 ) . Embryos were collected in the 5–70 min window post-fertilization . Please refer to Figure 3—source data 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 01010 . 7554/eLife . 07378 . 011Figure 3—source data 1 . Number of progeny fathered by males encoding both the UAS-HP1E hairpin and the Gal4 driver ( 24196/A5C ) compared to fathers encoding the Gal4 transgene alone ( w1118/A5C ) , hairpin alone ( 24196/CyO ) , or both plus the native promoter-driven , HP1E recoded transgene ( 24196/A5C + transgene ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 01110 . 7554/eLife . 07378 . 012Figure 3—source data 2 . Mitotic cycle number ( 0 , 1 , 2 etc ) of embryos fathered by wild-type males ( 24196/TM6 ) or PEL embryos fathered by HP1E-depleted males ( 24196/A5C ) collected during the 75-min window post-oviposition . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 01210 . 7554/eLife . 07378 . 013Figure 3—figure supplement 1 . HP1E knockdown using multiple drivers is efficient . RT-PCR on testis cDNA prepared from testis encoding a UAS-HP1E hairpin alone ( ‘HP1Ei’ ) or HP1E hairpin + Gal4 driver . Ubiquitous ( ‘A5CGal4’ ) and testis-restricted ( ‘vasaGal4’ ) drivers of the UAS_HP1E hairpin knockdown HP1E expression . Rp49 is the positive control . ‘CyO’ and ‘SM1’ refer to balancer chromosomes and represent wild-type HP1E expression levels . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 01310 . 7554/eLife . 07378 . 014Figure 3—figure supplement 2 . DNA sequence of recoded HP1E transgene . All third position sites were changed to either the preferred codon ( if un-preferred was encoded ) or the next preferred codon ( if the preferred codon was encoded ) following the Drosophila melanogaster preferred codon usage table . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 01410 . 7554/eLife . 07378 . 015Figure 3—figure supplement 3 . Nucleotide and amino acid sequence of the HP1E mutant . An HP1E-targeted TALEN deleted 7 bp starting at coding sequence base pair 58 ( above ) . Translation of coding sequence encoding the lesion . ‘*’ = stop codon ( below ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 01510 . 7554/eLife . 07378 . 016Figure 3—figure supplement 4 . HP1E mutant recapitulates male fertility defect . ‘pin’ refers to a marked second chromosome , triangle refers to the native promoter driven HP1E transgene inserted on the second chromosome . ‘***’ refers to a p-value < 0 . 0001 in a Mann–Whitney U test . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 016 Unlike the vast majority of male sterility mutants , HP1E-depleted fathers produce abundant motile sperm that transfer to females , fertilize eggs , and initiate embryogenesis ( Figure 3B ) . However , embryos sired by HP1E-depleted fathers failed to hatch ( 0 . 5% hatch rate ) . These data demonstrate that HP1E-depletion results in PEL i . e . , zygotic viability is dependent on father's genotype . Embryos fathered by HP1E-depleted males ( ‘PEL embryos’ hereafter ) arrested after only a few rounds of zygotic mitosis ( Figure 3C ) and exhibited a chromatin bridge in the very first mitotic division ( Figure 4A ) . Aged embryos exhibit increasingly asynchronous nuclear cycling and acute mitotic catastrophe ( Figure 4B ) . This gross chromosome segregation defect ultimately results in highly penetrant embryonic lethality . 10 . 7554/eLife . 07378 . 017Figure 4 . HP1E depletion in testis results in failed first embryonic mitosis and later mitotic catastrophe . ( A ) We observed a chromatin bridge ( arrow ) in the first zygotic telophase in PEL embryos fathered by HP1E-depleted , but not wildtype , males . ( B ) Embryos aged beyond first mitosis exhibit increasingly aberrant nuclear morphology and asynchrony across nuclei ( p = prophase , a = anaphase , t = telophase ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 017 To gain insight into the mechanism of paternal effect lethality in the PEL embryos , we tracked maternal and paternal DNA dynamics prior to the first telophase . We stained fixed , 0–20 min-old embryos with DAPI and an acetylated histone 4 ( AcH4 ) antibody ( Figure 5A ) . AcH4 accumulates preferentially on paternal chromatin prior to and during embryonic mitotic cycle 1 , allowing us to distinguish paternal from maternal DNA ( Adenot et al . , 1997 ) . PEL embryos revealed no gross defects in maternal chromatin dynamics ( Figure 5A ) and exhibited stereotypical centrosome and spindle morphology ( Figure 5—figure supplement 1 ) . Furthermore , sperm DNA in PEL embryos underwent the protamine-to-histone transition ( Figure 5—figure supplement 2 ) , decondensed , migrated toward the maternal pronucleus , and entered into the first mitosis just like in wild-type embryos ( Figure 5A ) . Using antibodies against a replication protein ( PCNA ) and a kinetochore protein ( Cenp-C ) , we also found that both pronuclei recruit replication machinery and initiate kinetochore assembly in PEL embryos ( Figure 5—figure supplements 3 , 4 ) . 10 . 7554/eLife . 07378 . 018Figure 5 . HP1E depletion in fathers results in mitotic arrest due to a paternal chromatin defect . ( A ) Paternal chromatin ( marked by anti-AcH4 ( red ) ) morphology mirrors maternal chromatin in wild-type embryos but differs in PEL embryos . In both wild-type and PEL embryos , the female and male pronuclei ‘migrate’ toward each other , ‘appose’ , and then enter mitosis . However , in PEL embryos , metaphase appears asynchronous between maternal and paternal chromatin , an AcH4-enriched chromatin bridge appears in anaphase ( arrowhead ) and persists at telophase ( arrow ) . ( B ) We calculated a ‘circularity ratio’ ( 1 = perfect circle , 0 = starfish ) for the first metaphase in wild-type and PEL embryos . We found that the paternal chromatin was significantly more circular i . e . , less condensed than maternal chromatin in PEL ( red dots ) but not wild-type embryos ( black dots ) , ( Mann Whitney-U test , p < 0 . 0001 ) . Dotted lines refer to sample means . A circularity ratio of 1 ( gray solid line ) refers to paternal and maternal chromatin with equivalent circularity . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 01810 . 7554/eLife . 07378 . 019Figure 5—source data 1 . Independent measurements of ‘circularity’ of maternal to paternal nuclei at first metaphase in embryos fathered by either wild-type ( ‘wt’ ) or HP1E mutant males ( ‘HP1E’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 01910 . 7554/eLife . 07378 . 020Figure 5—figure supplement 1 . In embryos fathered by both wild-type and HP1E knockdown males , metaphase centrosomes ( red , left panel ) and spindle ( red , right panel ) are indistinguishable . Genotype of father appears on the left . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 02010 . 7554/eLife . 07378 . 021Figure 5—figure supplement 2 . Embryos fathered by HP1E knockout males eject protamines . The protamine-GFP packaged sperm head is the only phase post-sperm entry when GFP signal was observed . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 02110 . 7554/eLife . 07378 . 022Figure 5—figure supplement 3 . PCNA ( replication factor ) is recruited to both maternal and paternal pronuclei at apposition in wild-type and PEL embryos . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 02210 . 7554/eLife . 07378 . 023Figure 5—figure supplement 4 . HP1E PEL embryos initiate kinetochore assembly . Kinetochores of embryos fathered by HP1E knockout males visualized with anti-Cenp-C . In the left panel , the paternal DNA appears in the lower half of the chromosome conformation . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 023 The first visible sign of defects in PEL embryos was observed at metaphase . We found that paternal DNA failed to condense synchronously with the maternal chromosomes in PEL embryos ( Figure 5A , B ) . To quantify this asymmetry , we adopted a measure of ‘circularity’ ( see ‘Materials and methods’ ) of maternal and paternal components across wild-type and PEL embryos ( Figure 5B ) . Condensed chromosomes appear as finger-like projections and therefore exhibit circularity close to 0 . In contrast , interphase chromosomes appear close to a perfect circle and exhibit circularity close to 1 . In wild-type embryos , we found that the ratio of paternal to maternal circularity was equal to one , suggesting that both pronuclei undergo synchronous condensation . In contrast , paternal DNA had twice the circularity of maternal DNA in PEL embryos , suggesting asynchronous condensation due to failure of the paternal genome to compact into resolved chromatids in a timely fashion . Immediately following this asynchronous metaphase in PEL embryos , we found that the paternal DNA failed to separate on the mitotic spindle . Specifically , a prominent chromatin bridge enriched in paternal chromatin-specific AcH4 appeared in PEL embryos at the first anaphase ( Figure 5A , arrowhead , n = 20/20 ) . We also observed AcH4 at the poles , suggesting that only a fraction of the paternal genome mis-segregates in PEL embryos . Based on these findings , we conclude that defects in paternal chromatin organization are the primary source of mitotic arrest in embryos fathered by HP1E-depleted males . However , our analyses could not formally rule out the possibility that mitotic defects additionally result from maternal chromatin defects , the loss of essential mitotic machinery normally contributed by wild-type fathers ( e . g . , the centriole ) , or the deposition of a mitotic ‘poison’ by the PEL fathers . To test whether paternal chromatin alone was sufficient to trigger failed mitosis in PEL embryos , we adopted a genetic approach that took advantage of the D . melanogaster maternal effect lethal sesame185b ( Loppin et al . , 2000 ) . In eggs laid by homozygous mutant sesame females , the paternal DNA completely fails to de-condense and so does not participate in zygotic mitosis . Instead , the haploid maternal chromosomes undergo mitotic cycling like wild-type diploid embryos until late embryogenesis—long after the PEL mitotic arrest observed in embryos sired by HP1E-depleted males ( Figure 3C ) . We crossed the HP1E-depleted males to sesame females to ask whether bypassing paternal chromatin was sufficient to rescue mitotic cycling . We observed full rescue of nuclear divisions in these crosses ( Figure 6A ) ; the resulting embryos cycled maternal haploid nuclei identically to those ‘fathered’ by a wild-type male beyond embryonic cycle 12 ( 85% and 82% , respectively , p > 0 . 2 , Figure 6A ) . We therefore conclude that defects in paternal chromatin dynamics are both necessary and sufficient to explain the mitotic arrest in PEL embryos sired by HP1E-depleted males . 10 . 7554/eLife . 07378 . 024Figure 6 . Embryonic mitosis can be rescued by excluding paternal chromatin but not by ectopic embryonic deposition of the HP1E protein itself . ( A ) HP1E-knockdown males crossed to wild-type mothers father embryos that undergo early arrest ( also see Figure 3C ) . However , both HP1E-depleted and wild-type males crossed to sesame ( ssm ) mothers father maternal haploid embryos that surpass mitotic cycle 12 . Black and gray circles refer to the paternal and maternal DNA contributions , respectively , to the zygotic nuclei . Embryos were imaged after fixation and DAPI staining . Please refer to Figure 6—source data 1 . ( B ) We observe no evidence of rescue when HP1E is deposited ectopically into the egg ( ‘+’ ) prior to fertilization ( ‘***’ refers to a p-value < 0 . 0001 in a Mann–Whitney U test , ‘n . s . ’ = not signficant ) . Western blot probed with the HP1E antibody shows an absence of native HP1E in embryos of wild-type mothers and HP1E deposition into early embryos of the experimental females . Like wild-type females ( ‘−’ on the x-axis ) , these experimental females ( ‘+’ on the x-axis ) fail to mother viable progeny when crossed to HP1E-depleted ( black bar ) males . Please refer to Figure 6—source data 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 02410 . 7554/eLife . 07378 . 025Figure 6—source data 1 . Number of embryos generated by ssm- females that arrested earlier than cycle 3 ( ‘ARREST’ ) or after cycle 7 ( ‘NOarrest’ ) fathered by wild-type males ( 24196/TM6 ) or HP1E-depleted males ( 24196/A5C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 02510 . 7554/eLife . 07378 . 026Figure 6—source data 2 . Number of progeny generated from crosses between mothers encoding a Gal4 driver alone ( MTD/CyO ) or Gal4 driver plus UAS-HP1E construct and males heterozygous ( ‘HP1E/TM6’ ) or homozygous ( HP1E- ) for the HP1E mutant chromosome . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 026 When does this paternal chromatin defect arise ? We failed to detect HP1E protein cytologically in both wild-type mature sperm head ( Figure 1D , Figure 1—figure supplement 3 ) and wild-type early embryos ( data not shown ) . Nevertheless , we wanted to formally consider the possibility that low levels of sperm-inherited HP1E might act during early embryogenesis to ensure proper mitosis . If this was the case ( as with spe-11 [Browning and Strome , 1996] ) , the defects we observed arising in PEL embryos could be due to missing HP1E protein in the embryo itself . To test this possibility , we used ectopic over-expression of HP1E during oogenesis to maternally deposit HP1E protein into the embryo . This over-expression strategy resulted in robustly detectable HP1E levels in the embryo ( Figure 6B ) , whereas embryos laid by control females harbor no detectable HP1E . However , maternally deposited HP1E was unable to rescue the PEL defect associated with sperm from HP1E-depleted males ( Figure 6B ) . Consistent with the testis cytology presented in Figure 1 , these data suggest that it is HP1E action during spermiogenesis , rather than in early embryogenesis , which is responsible for ensuring proper embryonic mitosis . Combined with the successful rescue of mitotic cycling by sesame- mothers , we conclude that HP1E primes the paternal genome during spermiogenesis i . e . , pre-fertilization , to ensure proper remodeling of the paternal genome in embryos post-fertilization . Our analysis of the anaphase bridges in PEL embryos revealed that only a fraction of the paternal genome appears to be affected by the HP1E depletion ( Figure 5A ) . This unusual observation suggested the possibility that all five D . melanogaster chromosomes might not be equally dependent on HP1E function . Based on our previous findings that HP1E-depletion led to a global overexpression of heterochromatin-embedded genes ( Figure 2 ) , we speculated that the paternal chromosomes that encode the longest tracts of heterochromatin might be especially sensitive to HP1E depletion . In the D . melanogaster genome , heterochromatic DNA is most abundant on the sex chromosomes ( Celniker and Rubin , 2003 ) . To test the possibility that specific chromosomes are enriched in the chromatin bridge , we performed fluorescent in situ hybridization ( FISH ) analysis on wild-type and PEL embryos using chromosome-specific satellite probes to all five D . melanogaster chromosomes ( Dernburg , 2011 ) ( Figure 7A , C ) . We found that the paternal Y was trapped as a bridge between nuclei in 94% of the male PEL embryos ( Figure 7B ) , whereas the maternal X was never found in the bridge ( Figure 7B ) . In female embryos , we found that the ( inferred ) maternal X-chromosome segregated faithfully while the ( inferred ) paternal X-chromosome was trapped as a bridge in 60% of embryos . We found that the large autosomes—chromosomes II and III—mis-segregated at only 4% and 15% frequency in male and female embryos , respectively . The dichotomy between sex chromosomes and the large autosomes is highly significant ( p < 0 . 0001 , Figure 7D ) . Homology between the small autosomal fourth ( ‘dot’ ) and Y-chromosomes precluded us from inferring fourth chromosome mis-segregation frequencies in male embryos . However , in female embryos , we found that the paternal fourth chromosome mis-segregated at 25% frequency , intermediate between the autosomes and sex chromosomes . 10 . 7554/eLife . 07378 . 027Figure 7 . Sex chromosomes are especially vulnerable to HP1E depletion in D . melanogaster . Representative images of fluorescent in situ hybridization ( FISH ) analyses of first zygotic telophase in ( A ) wild-type and ( B ) PEL embryos using chromosome-specific satellite probes ( C ) , which recognize chromosome-specific repetitive elements ( Dernburg , 2011 ) . FISH probes against the Y chromosome were tested together with probes against either the chr . X probe ( left ) or chr . 2 + 3 probe ( middle ) or chr . 4 probe ( right ) . ( D ) Using at least 20 images per probe pair per sex , we find that sex chromosomes are statistically enriched in the telophase bridge of PEL embryos . Quantification of chromosomal element appearance in the first telophase bridge in male and female embryos ( PEL embryos ) fathered by HP1E-depleted males . Hybridization of the fourth chromosome probe to the Y chromosome precluded data collection for this probe in male embryos . Data are reported as ‘obs/total/ ( % ) ’ where ‘obs’ = number of embryos observed with the probe appearing in the telophase bridge , ‘total’ = total number of embryos sampled per probe , and ‘%’ = obs × 100/total . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 027 Our discovery of sex chromosome enrichment in the telophase bridge suggests that a heterochromatic locus common to the X and Y chromosomes may underlie PEL . The only known repetitive locus exclusive to the X and Y in D . melanogaster is the multigene cluster of rDNA genes , which encode the ribosomal RNAs . We hybridized labeled rDNA probes ( targeting the IGS sequence ) in combination with the Y-satellite probe to wild-type ( Figure 8A ) and PEL embryos ( Figure 8B ) . In male PEL embryos , we discovered the Y-linked rDNA in the bridge at only 50% frequency ( Figure 8B , Figure 8D ) compared to 95% frequency of the AATAC satellite probe ( p < 0 . 0001 ) . Thus , it is the AATAC satellite DNA or an immediately proximal satellite DNA cluster that is in the telophase bridge in male embryos . In contrast , the X-linked rDNA locus in female PEL embryos occurs in the bridge at 85% frequency ( Figure 7D , Figure 8B , D ) compared to 60% for the 359 bp repeat ( p < 0 . 02 ) . These data suggest that the DNA present in the telophase bridge is more likely to be proximal to the X-rDNA cluster ( although not the rDNA itself ) than the 359 bp satellite repeats . It is currently unclear whether any of our targeted sequences are responsible for PEL . Nevertheless , our discovery of different frequencies for different probes on the X- and Y-chromosomes implicates discrete loci rather than entire chromosomes underlying mitotic failure , as was discovered for D . melanogaster-Drosophila simulans hybrid embryos at nuclear cycles 10–13 ( Ferree and Barbash , 2009 ) . However , unlike the hybrid case where the 359 bp probe signal appears stretched across the chromatin bridge at these later stage embryos , we observed mostly condensed foci in the bridge of the very first mitosis . 10 . 7554/eLife . 07378 . 028Figure 8 . Localization of the paternal X- and Y-linked rDNA locus to the telophase bridge in female and male embryos fathered by the HP1E mutant ( PEL embryos ) . Representative images of FISH analyses of first zygotic telophase in ( A ) wild-type and ( B ) PEL embryos using ( C ) probes that recognize the Y-specific satellite AATAC ( to determine sex of embryos ) and rDNA ( the intergenic spacer ‘IGS’ sequence ) . ( D ) Quantification of FISH signal in the first telophase bridge in male and female PEL embryos . Data are reported as ‘obs/total/ ( % ) ’ where ‘obs’ = number of embryos observed with the probe appearing in the telophase bridge , ‘total’ = total number of embryos sampled per probe , and ‘ ( % ) ’ = obs × 100/total . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 028 Based on these findings , we conclude that HP1E action is required during sperm development to prime the paternal genome for embryonic chromosome segregation . This priming function is especially critical for faithful segregation of paternal sex chromosomes , which appear to be most vulnerable to HP1E depletion . Even though only a fraction of the paternal genome suffers these consequences , the resulting embryonic mitosis is catastrophic and results in highly penetrant developmental arrest . Thus , paternal contributions laid down during the spermiogenesis program play an essential role in ensuring synchrony of paternal and maternal genomes for the first embryonic mitosis .
Properly coordinated chromosome segregation during virtually all mitotic divisions relies on the function of multiple cell cycle checkpoint proteins ( Lara-Gonzalez et al . , 2012; Iyer and Rhind , 2013; Yasutis and Kozminski , 2013 ) . No such cell cycle checkpoint proteins have been identified to act in the very first embryonic mitotic cycle ( O'Farrell et al . , 2004 ) , which must nevertheless accomplish the difficult task of synchronizing maternal and paternal chromosomes that were inherited in very different chromatin states . To investigate the paternal contributions that ensure timely participation of the paternal genome in early embryogenesis , we carried out a detailed functional analysis of the testis-restricted HP1E gene in D . melanogaster . We found that HP1E encodes a novel function that ensures paternal genome stability in the embryo . Our cytological and transcriptome analysis revealed that HP1E is developmentally restricted within the male germline , where it contributes to heterochromatin integrity . HP1E depletion during sperm development results in a highly penetrant PEL phenotype in which paternal chromosomes , especially the paternal sex chromosomes , fail to condense in synchrony with the maternal chromosomes and ultimately cause mitotic catastrophe . We further showed that the PEL embryonic phenotype could not be rescued by egg-supplied HP1E but could be rescued if the paternal DNA was excluded from participating in embryonic mitosis . These observations support a model ( Figure 9A ) under which HP1E acts pre-fertilization to ensure proper chromosome condensation and segregation of paternal chromosomes post-fertilization . 10 . 7554/eLife . 07378 . 029Figure 9 . Proposed model for HP1E ‘hit and run’ priming of the paternal genome for timely entry into embryonic mitosis . ( A ) HP1E localization to post-meiotic paternal chromatin directly or indirectly results in an epigenetic mark transferred to the embryo on sperm chromatin . This mark ensures synchronous paternal and maternal entry into the first embryonic mitosis . The absence of HP1E during postmeiotic sperm maturation leads directly or indirectly to the loss of an epigenetic mark ( designated by the absence of the flag ) . Paternal chromatids fail to resolve and mitotic catastrophe ensues . ( B ) The loss of HP1E in the obscura group of Drosophila dates to the same 7 million-year long branch as a major karyotype innovation involving the sex chromosomes , including the birth of a neo-Y chromosome ( Carvalho and Clark , 2005; Larracuente et al . , 2010; Levine et al . , 2012 ) . For clarity , only the dynamic subset of the chromosomal elements is presented . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 02910 . 7554/eLife . 07378 . 030Figure 9—figure supplement 1 . HP1E is present for more than 60 million years of Drosophila evolution but was lost at least three times over the Drosophila phylogeny . D . willistoni and D . grimshawii each encode other testis-restrcited HP1 genes whereas D . pseuodobscura and its close relatives do not . DOI: http://dx . doi . org/10 . 7554/eLife . 07378 . 030 The ‘hit and run’ priming function clearly distinguishes HP1E from all other previously characterized paternal effect lethal genes , which encode proteins that are transmitted to the embryo via sperm ( Browning and Strome , 1996; Fitch and Wakimoto , 1998; Fitch et al . , 1998; Loppin et al . , 2005b; Smith and Wakimoto , 2007; Gao et al . , 2011; Seidel et al . , 2011 ) . These include the D . melanogaster paternal chromatin-associated PEL , k81 , which encodes a protein that persists on paternal telomeres from late spermatogenesis to the first embryonic mitosis ( Dubruille et al . , 2010; Gao et al . , 2011 ) . The HP1E-depletion phenotype is instead reminiscent of Drosophila fathers infected with Wolbachia bacteria crossed to uninfected females ( Serbus et al . , 2008 ) . Embryonic lethality induced by Wolbachia testis infection is also caused by a pre-fertilization modification to the paternal genome that results in paternal-maternal chromatin asynchrony and mis-segregation at the very first zygotic mitosis . However , Wolbachia-associated PEL results in mis-segregation of the entire paternal genome ( Landmann et al . , 2009 ) rather than just the heterochromatin-rich chromosomes observed in HP1E—PEL ( Figure 5A , Figure 7B ) . Moreover , the HP1E PEL defect is completely independent of Wolbachia ( we find that PEL phenotype persists for Wolbachia-free males and females ) . We therefore conclude that HP1E supports a novel chromatin requirement to prime paternally inherited genomes for synchronous and successful embryonic mitosis . How does HP1E ensure timely mitotic entry ? It is formally possible that the PEL phenotype is the consequence of a dysregulated spermatid transcriptome that is , up- or down-regulation of a downstream gene . However , our finding that HP1E depletion results in the global up-regulation of heterochromatin-embedded genes , together with our observation that the heterochromatin-rich paternal sex chromosomes are most vulnerable to HP1E depletion , lead us to favor the alternate model that HP1E functions as a canonical HP1 protein during spermiogenesis . Based on antibody localization ( Figure 5—figure supplements 3 , 4 ) and chromatin bridge morphology ( Figure 5A ) , we found no evidence for defects in kinetochore assembly or replication machinery engagement in PEL embryos . Instead , our observation that the lethality phenotype first manifests as decondensed paternal chromosomes relative to maternal chromosomes implicates condensation delay of the heterochromatin-rich sex chromosomes . This delay could be the consequence of incomplete replication ( Landmann et al . , 2009 ) . Indeed , large stretches of uninterrupted heterochromatic DNA , as found on the Drosophila sex chromosomes , pose a unique challenge to replication ( Leach et al . , 2000 ) ( Pryor et al . , 1980; Collins et al . , 2002 ) . Alternatively , the mitotic delay may be the result of inadequate condensin protein recruitment , which is required for timely resolution of sister chromatids post-replication ( Steffensen et al . , 2001; Dej et al . , 2004; Savvidou et al . , 2005; Cobbe et al . , 2006; Hirano , 2012 ) . Previous studies have shown that heterochromatin can also impair chromosome condensation ( Peng and Karpen , 2007 ) . Timely completion of replication and condensation requires the action of HP1E's closest relative , HP1A , in somatic cells ( Kellum et al . , 1995; Schwaiger et al . , 2010; Li et al . , 2011 ) . However , in developing spermatids , HP1A localizes to telomeres ( Dubruille et al . , 2010 ) rather than broadly to heterochromatin as observed in virtually all other cell types . We posit that HP1E adopts a global , HP1A-like chromatin function during this highly specialized developmental stage and ensures the recruitment or retention of either replication or condensin proteins that are required post-fertilization . Previous studies have shown that HP1A is essential for embryo viability ( Eissenberg et al . , 1992 ) . We show here that paternally-acting HP1E is also essential for embryogenesis . Both HP1A and HP1E evolve under purifying selection ( Levine et al . , 2012 ) . However , unlike HP1A ( encoded by Su ( var ) 205 ) , HP1E has an unusually dynamic evolutionary history . Despite ancient origins , HP1E has been recurrently lost over evolutionary time . HP1E has been apparently replaced by younger , testis restricted HP1 paralogs on at least two occasions during Drosophila evolution ( Levine et al . , 2012 ) ( Figure 9—figure supplement 1 ) . Curiously , Drosophila pseudoobscura and related species encode neither HP1E nor a putative replacement testis-specific HP1 gene . How do we reconcile the paradox of HP1E essentiality in D . melanogaster with its loss in D . pseudoobscura ? We previously found that HP1E loss along in D . pseudoobscura-related species occurred during the same 7-million evolutionary period as a major sex chromosome rearrangement event ( Levine et al . , 2012 ) , in which the ancestral Y was lost , a neo-Y chromosome was born , and the ancestral X fused to an autosome ( Carvalho and Clark , 2005; Larracuente et al . , 2010 ) ( Figure 8B ) . Our finding that the D . melanogaster sex chromosomes are especially vulnerable to HP1E depletion , combined with the emergence of novel sex chromosome arrangements along the same narrow branch as HP1E pseudogenization ( Figure 9B ) , suggests a model under which rearrangements of heterochromatin-rich sex chromosomes in the obscura group rendered HP1E non-essential . Such karyotypic changes can bring distal heterochromatin into closer proximity to euchromatin and be sufficient to alter heterochromatin packaging ( Spofford , 1976 ) , replication timing ( Abramov et al . , 2005 ) or even delete blocks of satellite repeats ( Garagna et al . , 1995 ) . Thus , heterochromatin evolution via chromosomal rearrangements may have obviated maintenance of HP1E's essential heterochromatin function , leading to its degeneration in D . pseudoobscura . Our finding that HP1E is essential in D . melanogaster yet lost in the obscura group highlights the lineage-restricted essential requirements of chromatin genes . Intriguingly , the only other characterized PEL gene that supports paternal chromatin function in Drosophila embryos , k81 , is similarly lineage-restricted despite being essential for paternal telomere function ( Dubruille et al . , 2010; Gao et al . , 2011 ) . In contrast , maternally deposited proteins required for paternal chromatin reorganization following fertilization are generally conserved from fly to human ( e . g . , Loppin et al . , 2005a; Konev et al . , 2007; Delabaere et al . , 2014 ) . This dichotomy is striking . It specifically suggests that even though the essential functions of paternal control of DNA deposition and chromatin remodeling for embryonic mitosis are likely to be conserved in most animals , whereas the identity of those genes is not . PEL chromatin genes like HP1E and k81 thus challenge the dogma that ancient , conserved genes always encode essential conserved functions . Not only can young genes rapidly acquire essential chromatin functions due to dynamic chromatin evolution ( Chen et al . , 2010; Ross et al . , 2013 ) , but chromatin changes , such as those driven by karyotype evolution , may also drive the extinction of ancient genes encoding once-essential functions ( Drinnenberg et al . , 2014 ) .
To knockdown HP1E expression , we acquired a fly line that encodes a UAS promoter-driven hairpin homologous to the D . melanogaster HP1E transcript ( line 24196 , Vienna Drosophila RNAi Center ) . We crossed this line to both an actin5C Gal4 driver stock ( ubiquitous expression , Bloomington #3954 ) and a vasa-driven Gal4 driver ( male germline expression , gift of L Jones ) . Similar data were obtained with both drivers; only actin5C driven-RNAi data are presented . We engineered a recoded version of HP1E in which all synonymous sites were changed ( Genscript Inc . ) , and PCR-stitched this coding sequence to the HP1E UTRs and 1080 bp and 550 bp of 5′ and 3′ flanking noncoding regions , respectively ( http://flybase . org ) . The recoded HP1E transgene was cloned into the pattB vector and engineered into cytolocations 68A4 or 25C6 via injection . The BestGene , Inc . ( Chino Hills , CA ) carried out this and all other embryo injections using standard procedures . Using genetic crosses , we generated a rescue genotype that encoded one copy of the recoded transgene ( cytolocation 68A4 ) in a background of the Gal4 driver and UAS-driven HP1E hairpin . We crossed UASp-driven HP1E transgene into the ‘MTD’ Gal4 driver background ( #31777; Bloomington ) to overexpress HP1E during oogenesis for deposition into the egg . We generated a 7-base pair lesion in the 5′ region of the HP1E coding sequence using a TAL-effector nuclease ( Genetic Services Inc . ) . We rescued fertility of HP1E homozygous mutant fathers by introducing the same transgene inserted at cytolocation 25C6 . To assess male fertility , we crossed five 0–5 day old virgin w1118 females to two 0–5 day old males containing Gal4 driver alone , HP1E-hairpin alone , both driver and hairpin , or driver , hairpin and recoded ‘rescue’ . Parents were discarded after 3 days and progeny counted on day 16 . We replicated each cross type four times . To determine if the HP1E-knockdown fathers produced motile sperm , we dissected 10 seminal vesicles in PBS , squashed the tissue between a cover slip and slide , and then examined them under a light microscope . To facilitate sperm imaging , we crossed flies encoding the ( donjuan ) dj-GFP construct ( Santel et al . , 1997 ) into an HP1E knockdown background and mounted the male seminal vesicle , the female seminal receptacle to which these males were crossed , and the 5 min-old embryos oviposited by these females . We used a similar scheme to visualize protamine:GFP ( Jayaramaiah Raja and Renkawitz-Pohl , 2005 ) in embryos fathered by HP1E-knockdown males . To assess the stage of embryonic arrest , we crossed males encoding both the actin5C driver and HP1E hairpin or the driver alone to wild-type sevelin females ( gift of B Wakimoto ) . After a 1-hr pre-lay , we collected embryos for 70 min , methanol-fixed each genotype separately ( Rothwell and Sullivan , 2007 ) , mounted in SlowFade Gold antifade with DAPI ( Molecular Probes , Life Technologies Inc . , Grand Island , NY ) , and counted nuclei number/embryo at 20× on a Leica DMI 6000 . A Mann–Whitney U test determined significance between the two frequency distributions . To assess embryonic mitotic rescue by sesame mothers , we set up the same cross but with virgin sesame females ( ssm185b , gift from Kami Ahmad ) . After a 1-hr pre-lay , sesame females oviposited for 45 min followed by 1 . 5 hr of aging . We then collected , fixed , and counted nuclei as above . For embryos fathered by HP1E knockdown vs wild-type males , we recorded the number of embryos that underwent early arrest ( cycle 3 or earlier ) and no arrest ( beyond cycle 12 ) . We tested for heterogeneity among the four categories using a Fisher's Exact Test . To characterize the progression of paternal chromatin and mitotic machinery leading up to and during the first embryonic mitosis , we conducted immunofluorescence and DAPI staining on wildtype- , HP1E knockdown- , or HP1E/HP1E fathered embryos that were 0–20 min old . We methanol-fixed embryos ( as above ) and then rehydrated in PBS plus a drop of PBS + 0 . 1% Triton . Next , we permeablized in PBS + 1% Triton for 30 min at room temperature . Embryos were blocked in the StartBlock reagent ( Thermo Scientific , Waltham , MA ) for 90 min at 4°C . We replaced the block with the primary antibody diluted in StartBlock and incubated overnight at 4°C . We then washed embryos in StartBlock for 1 hr following by a 2-hr room temperature incubation in a secondary antibody diluted in StartBlock . After washing embryos for 1 hr in PBS + 0 . 1% Triton , we mounted them as described above . Primary antibody dilutions were the following: anti-AcH4 ( Millipore , Billerica , MA; 1:1000 ) , anti-alpha tubulin ( Serotec , Kidlington , UK; 1:250 ) , anti-gamma-tubulin ( Sigma–Aldrich , St . Louis , MO ) , clone GTU-88 , 1:1000 , anti-Cenp-C ( 1:5000 , gift of C Lehner ) , anti-PCNA ( 1:300 , gift of P Fisher ) , and anti-HP1E ( 1:1000 ) . Alexa-Fluor goat secondary antibodies ( Life Technologies ) were diluted at 1:1000 . We acquired images from the Leica TCS SP5 II confocal microscope with LASAF software and present maximally projected . tif files . Finally , using ImageJ we quantified paternal—maternal metaphase asymmetry by tracing max projected AcH4-staining wild-type and PEL embryos . We measured ‘circularity’ , which is 4π × [Area]/[Perimeter]2 and calculated the paternal:maternal ratio . We raised an antibody against HP1E residues CKSLKRGQELNNQYETKAKRLKI and CRILDRRHYMGQLQYLVKWLDY . Covance Inc . ( Princeton , NJ ) immunized a single rabbit by injecting it with both peptides over two months . We confirmed HP1E antibody specificity by probing a western blot of nuclear extracts from Drosophila S2 cells transfected with a heat-shock inducible , N-terminally Flag-tagged and YFP-tagged HP1E fusion protein plasmids . We designed these constructs using Gateway technology ( emb . carnegiescience . edu/labs/murphy/Gateway%20vectors . html ) following standard procedures ( destination vectors pHFW and pHVW , respectively ) . We prepared cell lysates by transfecting ( FuGENE , Promega , Madison , WI ) S2 cells with 2 μg of plasmid DNA and incubating overnight . We transiently induced expression by heat shock ( following Ross et al . , 2013 ) . We washed the recovered cells in PBS and re-suspended in RIPA buffer , sonicated , pelleted , and re-suspended in SDS loading buffer . We probed the membranes with either anti-HP1E ( 1:500 ) or anti-Flag ( ‘M2’ , Sigma–Aldrich , St . Louis , MO; 1:2000 ) primary antibodies followed by goat anti-rabbit or goat anti-mouse IgG-HRP ( Santa Cruz Biotechnologies Inc . , Dallas , TX ) . Lysates for the western blot confirming deposition of ectopically-expressed HP1E into embryos were prepared from 0–40 min embryos laid by females encoding the MTD driver and either UASp-HP1E transgene ( described above ) or a balancer chromosome . We flash froze the embryos followed by grinding with a glass pipette in SDS loading buffer and boiling at 95°C for 5 min . We probed the membranes with anti-HP1E and anti-beta actin ( Abcam , Cambridge , UK , both at 1:5000 ) . To assess HP1E localization in testis , we used the HP1E antibody or generated transgenic flies encoding an N-terminal Flag- or YFP-tag fused to HP1E , flanked by the native promoter and 5′ and 3′ regions . To cytologically characterize HP1E localization without antibody staining , we fixed testis from 2–5 day old YFP-HP1E males in 4% paraformaldehyde ( PFA ) and mounted in SlowFade Gold antifade with DAPI . For immunofluorescence , we fixed Flag-tagged and untagged testis in 4% PFA in periodate-lysine-paraformaldehyde ( PLP ) for 1 hr followed by 30 min in PBS + 0 . 3% Triton , 0 . 3% sodium deoxycholate . After a 10 min wash in PBS + 0 . 1% Triton , we blocked testis for 30 min in PBS + 0 . 1% Triton + 3% BSA followed by standard IF procedures using the following dilutions: anti-Flag ( ‘M2’ , Sigma–Aldrich , St . Louis , MO; 1:2500 ) , anti-HP1E ( 1:1000 ) , and Alexa Fluor secondaries goat anti-mouse 488 ( 1:1000 ) and goat anti-rabbit ( 1:1000 ) . We acquired images from the Leica TCS SP5 II confocal microscope with LASAF software and present maximally projected tagged image files ( tifs ) . We prepared RNA from testis dissected from three biological replicates per genotype representing three independent crosses of males heterozygous for the UAS promoter-driven hairpin homologous to the D . melanogaster HP1E transcript and virgin females homozygous for vasaGAL4 inserted on chromosome II . The FHCRC Shared Resources Genomics Core prepared six libraries using Illumina TruSeq Sample Prep Kit v2 . We performed image analysis and base calling with Illumina's RTA v1 . 13 software and demulitplexed with Illumina's CASAVA v1 . 8 . 2 . We aligned reads to BDGP5r66 using TopHat v1 . 4 . 0 and converted files to sam format using samtools v0 . 1 . 18 . We used htseq-count v0 . 5 . 3 to generate counts/gene and removed genes that had 0 counts across all samples or less than 1 count/million in at least three samples . This culling resulted in 11 , 051 genes . We identified differentially expressed genes using edgeR v2 . 6 and tested for significant enrichment of up-regulated heterochromatin-embedded genes using binomial probability . We annotated heterochromatin-embedded genes using the D . melanogaster Release 5 . To determine if participation in chromatin bridging was chromosome-specific , we designed Cy3 or Cy5 conjugated probes ( IDT ) for in situ hybridization following ( Dernburg , 2011 ) : X ( 359 bp satellite ) , Y ( AATAC ) n , 2L + 3L ( AATAACATAG ) n , 4 ( AATAT ) n and IGS ( GTATGTGTTCATATGATTTTGGCAATTATA , ATATTCCCATATTCTCTAAGTATTATAGAG , designed by P Ferree ) . We co-hybridized two probes using the following conjugated tags and annealing temperatures: 32°C for Y-Cy5 + 2L/3L-Cy3 and Y + X-Cy3 , 23°C for Y + 4-Cy3 and 30°C for Y-Cy5 + IGS-Cy3 . For the FISH experiments , we modified the above embryo fixation protocol by replacing methanol:heptane with 3 . 7% paraformaldehyde:heptane . | The genetic information of cells is packaged into structures called chromosomes , which are made up of long strands of DNA that are wrapped around proteins to form a structure called chromatin . The cells of most animals contain two copies of every chromosome , but the egg and sperm cells contain only one copy . This means that when an egg fuses with a sperm cell during fertilization , the resulting ‘zygote’ will contain two copies of each chromosome—one inherited from the mother , and one from the father . These chromosomes duplicate and divide many times within the developing embryo in a process known as mitosis . The first division of the zygote is particularly complicated , as the egg and sperm chromosomes must go through extensive—and yet different—chromatin reorganization processes . For instance , paternal DNA is inherited via sperm , where specialized sperm proteins package the DNA more tightly than in the maternal DNA , which is packaged by histone proteins used throughout development . For paternal DNA to participate in mitosis in the embryo , it must first undergo a transition to a histone-packaged state . Despite these differences , both maternal and paternal chromosomes must undergo mitosis at the same time if the zygote is to successfully divide . Although it is known that the egg cell contributes essential proteins that are incorporated into the sperm chromatin to help it reorganize , the importance of paternal proteins in coordinating this process remains poorly understood . Many members of a family of proteins called Heterochromatin Protein 1 ( or HP1 for short ) have previously been shown to control chromatin organization in plants and animals . In 2012 , researchers found that several HP1 proteins are found only in the testes of the fruit fly species Drosophila melanogaster . They predicted that these proteins might help control the reorganization of the paternal chromosomes following fertilization . Levine et al . —including researchers involved in the 2012 study—have now used genetic and cell-based techniques to show that one member of the HP1 family ( called HP1E ) ensures that the paternal chromosomes are ready for cell division at the same time as the maternal chromosomes . Male flies that are unable to produce this protein do not have any offspring because , while these flies' sperm can fertilize eggs , the resulting zygotes cannot divide as normal . Further experiments revealed that HP1E is not inherited through the chromatin of mature sperm , but instead influences the structure of the chromosomes during the final stages of the development of the sperm cells in the fly testes . This study shows that both maternal and paternal proteins are needed to control how the paternal chromosomes reorganize in fruit fly embryos . Although difficult to discover and decipher , this work re-emphasizes the importance of paternal epigenetic contributions—changes that alter how DNA is read , without changing the DNA sequence itself—for ensuring the viability of resulting offspring . Future work will reveal both the molecular mechanism of this epigenetic transfer of information , as well as why certain Drosophila species are able to naturally overcome the loss of the essential HP1E protein . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"cell",
"biology"
] | 2015 | Mitotic fidelity requires transgenerational action of a testis-restricted HP1 |
The application of current channelrhodopsin-based optogenetic tools is limited by the lack of strict ion selectivity and the inability to extend the spectra sensitivity into the near-infrared ( NIR ) tissue transmissible range . Here we present an NIR-stimulable optogenetic platform ( termed 'Opto-CRAC' ) that selectively and remotely controls Ca2+ oscillations and Ca2+-responsive gene expression to regulate the function of non-excitable cells , including T lymphocytes , macrophages and dendritic cells . When coupled to upconversion nanoparticles , the optogenetic operation window is shifted from the visible range to NIR wavelengths to enable wireless photoactivation of Ca2+-dependent signaling and optogenetic modulation of immunoinflammatory responses . In a mouse model of melanoma by using ovalbumin as surrogate tumor antigen , Opto-CRAC has been shown to act as a genetically-encoded 'photoactivatable adjuvant' to improve antigen-specific immune responses to specifically destruct tumor cells . Our study represents a solid step forward towards the goal of achieving remote and wireless control of Ca2+-modulated activities with tailored function .
Microbial opsin-based optogenetic technologies have been widely adopted to modulate neural activity ( Fenno et al . , 2011 ) , but similar tools tailored for utilization in non-excitable tissues ( e . g . , the immune and hematopoietic system ) are still limited . The application of channelrhodopsin ( ChR ) -based optogenetic tools is limited by the lack of ion selectivity and the inability to extend the spectral sensitivity into the near-infrared ( NIR ) range ( Fenno et al . , 2011 ) . Here we present a tissue penetrable near infrared-stimulable optogenetic platform ( termed 'Opto-CRAC' ) that can be used to reversibly photo-manipulate Ca2+ influx through one of the most Ca2+-selective ion channels , the Ca2+ release-activated Ca2+ ( CRAC ) channel , which is abundantly present in most non-excitable cells ( Hogan et al . , 2010; Prakriya and Lewis , 2015 ) . Our tool is based on the engineering of light sensitivity into the CRAC channel and its subsequent coupling to lanthanide-doped upconversion nanoparticles ( UCNP ) , the latter of which act as nanotransducers to convert tissue penetrable NIR light into visible light emission ( Shen et al . , 2013; Chen et al . , 2014 ) . We demonstrate that Opto-CRAC tools can be applied to remotely control Ca2+ influx and generate repetitive Ca2+ oscillations , photo-tune Ca2+-dependent gene expression , and modulate a myriad of Ca2+-dependent activities in cells of the immune system , including effector T cell activation , macrophage-mediated inflammasome activation , dendritic cells ( DC ) maturation and antigen presentation . Our study set the stage for achieving the goal of remote optogenetic immunomodulation and spatiotemporal control over cellular immunotherapy in a wireless manner .
Following antigen presentation , T cell receptor ( TCR ) engagement triggers a cascade of signaling events in T lymphocytes that elicit the influx of extracellular Ca2+ through the CRAC channel , a classic example of store operated Ca2+ entry ( SOCE ) ( Hogan et al . , 2010; Prakriya and Lewis , 2015 ) . The molecular choreography of SOCE is mainly coordinated by two proteins that are located in distinct cellular compartments: ( i ) ORAI1 , a four-pass transmembrane protein that constitutes the CRAC channel pore-forming subunit in the plasma membrane ( PM ) ; and ( ii ) the stromal interaction molecule 1 ( STIM1 ) , an ER-resident Ca2+ sensor protein that is responsible for sensing ER Ca2+ depletion and directly gating ORAI1 channels through its cytosolic domain ( STIM1-CT ) . Store depletion induced Ca2+ influx through CRAC channels further activates calcineurin , a downstream Ca2+-dependent phosphatase that dephosphorylates the master transcriptional regulator NFAT ( nuclear factor of activated T cells ) and subsequently causes NFAT nuclear translocation ( Müller and Rao , 2010 ) . In the presence of the co-stimulatory pathway , which activates the activator protein 1 ( AP-1 ) , NFAT cooperates with AP-1 to turn on genes ( e . g . , IL-2 and IFN-γ ) that are characteristic of a productive immune response ( Müller and Rao , 2010 ) . To enable light control over the Ca2+/NFAT pathway , we set out to install light sensitivity into STIM1 by fusing a handful of STIM1-CT fragments with the genetically-encoded photoswitch LOV2 ( light , oxygen , voltage ) domain ( residues 404–546 ) of Avena sativa phototropin 1 ( Christie et al . , 1999; Harper , 2003; Yao et al . , 2008; Wu et al . , 2009 ) ( Figure 1a and Figure 1—figure supplement 1 ) . When expressed alone , these STIM1-CT fragments are capable of eliciting varying degrees of constitutive activation of ORAI1 channels to mediate Ca2+ entry from the extracellular space to the cytosol ( Yuan et al . , 2009; Park et al . , 2009; Zhou et al . , 2010a; Soboloff et al . , 2012 ) . In the dark , the C-terminal Jα helix docks to the LOV2 domain ( Harper , 2003; Yao et al . , 2008; Wu et al . , 2009 ) and keeps the ORAI1-activating STIM1-CT fragments quiescent . Upon blue light illumination , photoexcitation generates a covalent adduct between LOV2 residue C450 and the cofactor FMN ( Figure 1—figure supplement 1d ) , thereby promoting the undocking and unwinding of the Jα helix to expose the STIM1-CT fragments . Unleashed STIM1-CT fragments further move toward the plasma membrane to directly engage and activate ORAI1 Ca2+ channels ( Figure 1a , b ) . 10 . 7554/eLife . 10024 . 003Figure 1 . LOVSoc-mediated photoactivatable Ca2+ entry and nuclear translocation of NFAT in mammalian cells . ( a ) , Schematic of light-operated Ca2+ entry though engineered Opto-CRAC channels . Fusion with the lightswitch LOV2 domain confers photosensitivity to the ORAI1-activating STIM1-CT fragments . In the dark , STIM1-CT fragments are kept inactive presumably by docking toward the LOV2 domain . Upon blue light illumination , the undocking and unfolding of the LOV2 C-terminal Jα helix lead to the exposure of the STIM1-CT fragments , enabling their interaction with ORAI1 Ca2+ channels to trigger Ca2+ influx across the plasma membrane . See Figure 1—figure supplement 1 for the detailed design and comparison among the designed Opto-CRAC constructs . ( b ) , Light-inducible translocation of mCherry-LOV2404-546-STIM1336-486 ( designated as mCh-LOVSoc ) from the cytosol to the plasma membrane in HEK293T-ORAI1 stable cells . Upper panel , the images represent the same cells in the dark ( black bar ) or exposed to blue light at 470 nm ( 40 μW/mm2; blue bar ) . Scale bar , 10 μm . Lower panel , Kymograph of mCh-LOVSoc corresponding to the circled area ( top ) and quantification of mCherry signals over three repeated light-dark cycles ( bottom ) . n = 12 cells from three independent experiments . Error bars denote s . e . m . ( c ) , Light-induced Ca2+ influx reported by the green genetically-encoded Ca2+ indicator ( GECI ) GCaMP6s . The global cytosolic Ca2+ change was monitored after cotransfection of mCh-LOVSoc and GCaMP6s in HeLa cells; whereas the local Ca2+ change near the PM was reported by the PM-tethered GCaMP6s-CAAX construct . Shown were representative confocal or TIRF images following blue light stimulation ( 30 s , 40 μW/mm2 ) . The photo-activated Ca2+ response reflected in the fluorescence change was plotted on the right . n = 15 cells from three independent experiments . Error bars denote s . e . m . Scale bar , 10 μm . ( d ) , A representative example of light-inducible Ca2+ oscillation pattern generated by LOVSoc-expressing HeLa cells when exposed to repeated light-dark cycles ( 30 s ON and 120 s OFF ) . The red Ca2+ sensor , R-GECO1 . 2 , enabled recording of the whole course of intracellular Ca2+ fluctuation . n = 8 cells from three independent experiments . Blue bar indicates light stimulation at 470 nm with a power density of 40 μW/mm2 . Error bars denote s . e . m . ( e ) , Photo-triggered current-voltage relationships of CRAC currents in HEK293-ORAI1 cells transfected with mCh-LOVSoc . mCherry positive cells were subjected to whole-cell patch-clamp by a ramp protocol ranging from -100 mV to 100 mV in the presence ( blue ) or absence ( gray ) of light illumination . For the red curve , extracellular Na+ was replaced with a non-permeant ion NMDG+ to assess ion selectivity by examining the contribution of Na+ . ( f ) , Light-tunable nuclear translocation of GFP-NFAT1 and NFAT-dependent luciferase ( NFAT-Luc ) gene expression in HeLa cells transfected with mCh-LOVSoc . The HeLa-GFP-NFAT1 stable cells were subjected to light pulse stimulation for 30 s whilst the interpulse intervals were varied from 0 . 5 to 4 min . Representative snapshots of cells during GFP-NFAT1 nuclear translocation were shown in the middle panel . The corresponding time courses and dependence of NFAT nuclear translocation or NFAT-Luc activity on the interpulse interval were plotted on the right . n = 15–20 cells from three independent experiments . Error bars denote s . e . m . Scale bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 00310 . 7554/eLife . 10024 . 004Figure 1—figure supplement 1 . Design and characterization of engineered Opto-CRAC constructs ( related to Figure 1a ) . ( a ) , Fragments derived the cytoplasmic domain of STIM1 ( STIM1-CT ) that are capable of constitutively activating ORAI1 channels ( with the inclusion of the SOAR/CAD region ) were fused to the photoswitch moiety LOV2 . Residues KL or KLAAA was added between LOV2 and STIM1-CT fragments as linker . Five mutations were individually introduced into LOV2 domain . The chimera LOV2404-546-KL-STIM1336-486 ( designated as LOVSoc ) exhibited the highest dynamic range without showing significant dark activity ( judging from constitutive NFAT translocation in the dark ) . C450A traps the LOV2 domain in the dark state to acts as a negative control whilst I539E is known to keep LOV2 in its lit state ( 10 ) . G528A , I532E and N538E have been reported to increase the dynamic range of LOV2 domain ( 58 ) . However , those mutations did not seem to improve the overall performance of Opto-CRAC constructs in our hands . The previously reported LOVS1K ( LOV2-STIM1233-450 ) ( 47 ) , which is approximately 8 kDa larger than LOVSoc in size because of the inclusion of CC1 region , had a much narrower dynamic range with noticeable dark activity . The dark activity is gauged by the percentage of nuclear/total GFP-NFAT in the dark and defined as follows: “−” , no discernible activation; “+” , less than 10% activation; “++” , ≥ 10% . The dynamic range , reported by the averaged fluorescence changes ( ΔF/F0 ) of the Ca2+ sensor GCaMP6s , is categorized as: “−” , <0 . 2; “+” , 0 . 2–1 . 0; “++” , 1 . 0–2 . 5; “+++” , >2 . 5 . The domain architecture of STIM1 was shown on the top: SP , signal peptide; EF-SAM , EF-hand motif and sterile-alpha motif; TM , transmembrane domain; CC1 , predicted coiled-coil region 1; SOAR/CAD , the minimal ORAI-activating region in STIM1 or the CRAC activating domain; PS , proline/serine-rich region; K , poly-basic C-tail . ( b-–d ) , Normalized fluorescence changes in HeLa cells co-transfected with genetically-encoded Ca2+ sensors ( GCaMP6s or R-GECO1 . 2 ) and indicated opto-CRAC constructs ( b , various STIM1-CT fragments; c , optimization of the linker; d , LOV2 mutations ) . The LOV2 structure ( PDB entry: 2V0W ) , along with its co-factor FMN ( yellow sticks ) and mutated positions ( highlighted in red colors ) , was shown to the right of panel ( d ) . Data were shown as mean ± s . e . m . from 10–20 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 00410 . 7554/eLife . 10024 . 005Figure 1—figure supplement 2 . Light-dependent interaction between LOVSoc and ORAI1 ( related to Figure 1b ) . ( a ) , Schematic of the MBP- or mCherry ( mCh ) -tagged LOVSoc constructs used in the pulldown or CoIP assays . MBP fusion protein was expressed in bacteria and purified to probe its interaction with GB1-tagged ORAI1 C-terminus ( GB1-ORAI1-CT , residues 259–301 ) in vitro; whereas mCherry-tagged protein was used in the CoIP experiment to demonstrate its light-inducible interaction with FLAG-tagged ORAI1 ( FLAG-ORAI1 ) . ( b ) , Size-exclusion chromatography elution profile of purified MBP-LOVSoC . Inset , SDS-PAGE image of purified recombinant protein . c , UV-Vis spectra absorbance changes of LOV2 domain upon photoexcitation . The recovery of LOV2 to dark state was monitored every 25 s . ( d ) , In vitro light-inducible binding of recombinant GB1-ORAI1-CT ( MWtheoretic = 13 kDa; indicated by arrowhead ) to recombinant MBP-LOVSoc ( MWtheoretic = 76 kDa ) immobilized on the amylose resin . MBP ( MWtheoretic = 43 kDa ) was used as negative control and did not exhibit light-dependent association with GB1-ORAI1-CT . For the light stimulation groups , samples were constantly exposed to an external blue LED ( 470 nm , 40 μW/mm2 ) ( e ) , FLAG-ORAI1 coimmunoprecipitated with mCh-LOVSoc in a light-dependent manner . Samples in the light-simulated groups were constantly exposed to an external blue LED ( 470 nm , 40 μW/mm2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 00510 . 7554/eLife . 10024 . 006Figure 1—figure supplement 3 . Characterization of photoactivatable Ca2+ entry into mammalian cells ( related to Figure 1c , d ) . Data were shown as mean ± s . e . m . from 10–20 cells . ( a ) , Light-tunable Ca2+ entry reported by GCaMP6s in HEK293T cells transfected with mCh-LOVSoc . The response curves ( n= 10 cells ) were plotted as fold-changes of GCaMP6s signals following light stimulation at time 0 with indicated power densities at 470 nm . The bar graph on the right showed the maximal fold-changes of GCaMP6 at varying light power densities . ( b ) , Maximal Ca2+ response induced by light stimulation ( 470 nm for 1 min at 40 μW/mm2 ) in various cell types . The cells were derived from a wide range of mammalian tissues , including kidney ( HEK293T and Cos-7 ) , cervix ( HeLa ) , mammary gland ( MCF7 ) , prostate ( LNCaP ) , sternum ( WM793 ) , brain ( U87 ) , peripheral blood ( Jurkat ) , lymph nodes ( T cells ) , bone marrow ( macrophage ) and embryo ( MEF ) . Cultured cells were transiently transfected by 100 ng pTriEx-mCh-LOVSoc or transduced by retroviruses encoding mCh-LOVSoc . The cytosolic Ca2+ concentrations were calculated by using calibration protocols as described in our earlier studies ( Wang , 2014; Zhou et al . , 2010a; 2010b; 2013 ) . ( c-f ) , Light-inducible Ca2+ flux reported by Ca2+ indicators . HeLa cells were transfected with mCh-LOVSoc ( c , n = 12 ) , mCh-LOVSoc + R-CaMP2 ( d , n = 15 ) or mCh-LOVSoc + R-GECO1 . 2 ( e , n= 20 ) . Two-to-three dark-light cycles were applied to demonstrate the reversibility of photo-activated Ca2+ influx . Note that the acquisition of Fura-2 signals might induce the drift of the baseline , possibly due to residual preactivation of LOVSoc when excited at 380 nm in our imaging system . The half-life times of light-triggered cytosolic Ca2+ rise ( t1/2 , on ) and Ca2+ decay ( t1/2 , off ) after switching off the light stimulation were listed in panel ( f ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 00610 . 7554/eLife . 10024 . 007Figure 1—figure supplement 4 . Global and local Ca2+ influx generated by photo-activation of LOVSoc at defined spatial resolution ( related to Figure 1c ) . ( a ) , Confocal images of HEK293T cells transfected with GCaMP6s-CAAX and mCh-LOVSoc . The 488-nm confocal laser was able to globally activate mCh-LOVSoc to cause GCaMP6s signal increase within the whole illuminated field . Images were acquired within 1–2 s . The time course curve was plotted on the right . Data were collected from 10 cells from three independent experiments . Error bars denote s . e . m . ( b ) , Spatially-defined photoexcitation led to local Ca2+ influx at desired areas in HeLa cells coexpressing GCaMP6s-CAAX and mCh-LOVSoc . Shown were representative confocal images of cells with the framed areas ( box or circle ) photostimulated by a 488-nm laser ( 10 s at a power density of 40 μW/mm2 ) prior to the acquisition of GCaMP6s signals in the whole field . The pre-photostimulated areas showed constant high levels of fluorescence intensities , indicating the preactivation of Ca2+ influx at those defined regions before image acquisition . By contrast , the non-prestimulated area showed a time course that was comparable to panel a . See Vedio 4 for the dynamic changes of GCaMP6s intensities . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 00710 . 7554/eLife . 10024 . 008Figure 1—figure supplement 5 . Schematic representation and light-induced response curves of Opto-CRAC variants reported by GCaMP6s . ( a-c ) , Photoactivated Ca2+ response curves for HeLa-GCaMP6s cells transfected with 100 ng of ( a ) coexpression vectors pMIG-mCh-LOVSoc-IRES-ORAI ( IRES ) or pTriEx-ORAI1-T2A-mCh-LOVSoc ( T2A ) ; ( b ) a PM-tethering construct pTriEx-Lyn11-mCh-LOVSoc; ( c ) pTriEx-mCh- ( LOVSoc ) 2 or pTriEx-ORAI1- ( LOVSoc ) 2 constructs that harbor two covalently-linked copies of LOVSoc . Transfected cells were subjected to light stimulation at 470 nm with a power density of 40 μW/mm2 . Schematics of the design strategies were shown on the left . ( d-e ) , Comparison of maximal fold-changes of GCaMP6s signals ( d ) and the time ( e ) required to reach half maximal GCaMP6s fluorescence during photoactivatable Ca2+ influx . Data were shown as mean ± s . e . m from 10–15 cells in two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 00810 . 7554/eLife . 10024 . 009Figure 1—figure supplement 6 . Examples of light-tunable Ca2+ oscillation patterns generated in HeLa cells ( related to Figure 1d ) . To generate and monitor calcium oscillation patterns , HeLa cells were transfected with mCh-LOVSoc and the red GECI R-GECO1 . 2 . Blue light pulses were applied for 0 . 5 or 1 min with indicated intervals ranging from 0 . 5 min to 4 min . Data were shown as mean ± s . e . m from 6–8 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 009 We first created a series of Opto-CRAC constructs by varying the length of STIM1-CT fragments , introducing mutations into the LOV2 domain and optimizing the linker between these two moieties ( Figure 1—figure supplement 1a ) . After an initial screen of approximately 100 constructs using NFAT nuclear translocation and Ca2+ influx as readouts , we decided to use the LOV2-STIM1336-486 chimera ( designated as 'LOVSoc' ) in our following experiments because it showed no discernible dark activity and exhibited the highest dynamic range in terms of evoking light-inducible Ca2+ influx ( Figure 1—figure supplement 1a , b ) . When expressed as an mCherry-tagged fusion protein in HEK293-ORAI1 stable cells , LOVSoc underwent rapid translocation between the cytosol and the PM in response to blue light illumination ( t1/2 , on = 6 . 8 ± 2 . 3 s; t1/2 , off = 28 . 7 ± 6 . 5 s; Figure 1b and Video 1 ) . This process could be readily reversed by switching the light off , and could be repeated multiple times without significant loss in the magnitude of response . The light-dependent association between LOVSoc and ORAI1 or ORAI1 C-terminus ( ORAI1-CT ) was further confirmed by a pulldown assay using purified recombinant proteins and by coimmunoprecipitation assays ( Figure 1—figure supplement 2 ) . In mammalian cells expressing LOVSoc , the degree of Ca2+ influx could be tuned by varying the light power densities ( Figure 1—figure supplement 3a ) . After photostimulation for 1 min with a power density of 40 μW/mm2 at 470 nm , LOVSoc triggered significant yet varied elevation of cytosolic Ca2+ concentrations to approximately 500–800 nM in a dozen of mammalian cell types derived from various non-excitable tissues ( Figure 1—figure supplement 3b ) , likely owing to the varied endogenous levels of ORAI proteins among the tested cells . A Light-triggered global Ca2+ influx and oscillations in HeLa or HEK293T cells expressing mCherry-LOVSoc could be monitored in real-time by either Fura-2 ( Figure 1—figure supplement 3c ) or genetically-encoded Ca2+ indicators ( GECIs ) , including GCaMP6 ( Figure 1c and Videos 2 , 3 ) ( Chen et al . , 2013 ) , R-CaMP2 ( Figure 1—figure supplement 3d ) ( Inoue et al . , 2015 ) , and R-GECO1 . 2 ( Figure 1d and Figure 1—figure supplement 3e ) ( Wu et al . , 2013 ) . Notably , localized light stimulation can be applied to achieve local activation of Ca2+ influx at a defined spatial resolution ( Figure 1—figure supplement 4 and Video 4 ) , thereby providing a new approach to dissect the effect of Ca2+ microdomains in various biological processes ( Parekh , 2008 ) . Depending on the kinetic properties of the Ca2+ indicators used , the half-life time of the cytosolic Ca2+ rise in response to light stimulation ranged from 23 s to 36 s . After switching off the light , the cytosolic Ca2+ signal decayed with a half-life time of approximately 25–35 s ( Figure 1—figure supplement 3f ) . These values are largely in agreement with the time scale of SOCE under physiological stimulation ( Hogan et al . , 2010; Prakriya and Lewis , 2015; Soboloff et al . , 2012 ) . We further measured the photo-activated currents by whole-cell recording in HEK293 cells stably expressing ORAI1 ( Figure 1e ) . Following light stimulation , HEK293 cell transfected with LOVSoc developed a typical inward rectifying current , which is characteristic of the CRAC channel and distinct from the greater outward currents of non-selective cation channels such as TRPC ( Prakriya and Lewis , 2015 ) . Substitution of the most abundant extracellular cation Na+ by a non-permeant ion NMDG+ did not alter the amplitude or overall shape of the CRAC current , implying that Na+ has negligible contribution to LOVSoc- mediated photoactivatable Ca2+-selective CRAC currents . 10 . 7554/eLife . 10024 . 010Video 1 . Light-triggered reversible cytosol-to-PM translocation of mCh-LOVSoc . Three dark-light cycles were applied to HEK293-ORAI1 stable cells transfected with the Opto-CRAC construct mCh-LOVSoc . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 01010 . 7554/eLife . 10024 . 011Video 2 . Time-lapse imaging of light-triggered Ca2+ influx reported by cytosolic ( left ) or PM-tethered GCaMP6s . HeLa cells expressing mCh-LOVSoc was exposed to a 488-nm confocal laser . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 01110 . 7554/eLife . 10024 . 012Video 3 . Light-inducible Ca2+ influx monitored with TIRF microscopy in HeLa cells coexpressing GCaMP6s-CAAX and mCh-LOVSoc . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 01210 . 7554/eLife . 10024 . 013Video 4 . Sequential and localized activation of Ca2+ influx with defined spatial resolution . Imaging was performed on HeLa cells cotranfected with mCh-LOVSoc and GCaMP6s-CAAX . The boxed area was subjected to a brief photostimulation with the 488-nm laser for 10 s , followed by photoexcitation of the whole field at 488 nm to acquire GCaMP6s-CAAX signals . The boxed area showed preactivation of Ca2+ influx as reflected by the strong fluorescence signal at time point 0 s; whilst the other areas exhibited a gradual increase in fluorescence intensity following 488-nm light illumination . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 013 To confer more flexibility to the Opto-CRAC system with varied optical sensitivity , we explored the use of co-expression , membrane tethering or fusion strategies to generate five more variants of Opto-CRAC ( Figure 1—figure supplement 5 ) . We used either an internal ribosome entry site ( IRES ) -based bicistronic vector or a self-cleaving 2A peptide strategy ( de Felipe et al . , 2006 ) to enable the coexpression of ORAI1 and LOVSoc in the same cell with a single vector . Compared to LOVSoc alone , both co-expression systems resulted in ~1 . 4-fold increase in Ca2+ response ( Figure 1—figure supplement 5a ) . Tethering LOVSoc to the plasma membrane ( PM ) with an N-terminal PM-targeting sequence derived from the Src kinase Lyn ( Inoue et al . , 2005 ) ( Lyn11-LOVSoc ) expedited the photoactivation process by 3 . 5-fold ( Figure 1—figure supplement 5b ) , presumably owing to its increased local concentration and much closer proximity to the ORAI1 channels . By contrast , a concatemeric form of LOVSoc with two copies covalently connected in a single polypeptide or its fusion to ORAI1 substantially slowed down photoactivatable Ca2+ influx ( Figure 1—figure supplement 5c ) . Collectively , we have created a set of Opto-CRAC constructs that meet the varying needs on sensitivity and photoactivation kinetics ( Figure 1—figure supplement 5d , e ) . We next asked if we could manipulate the light pulse to generate diverse temporal patterns of Ca2+ signals to tune the degree of NFAT activation , which would be reflected in the efficiency of NFAT nuclear translocation and NFAT-dependent luciferase expression . We applied a fixed light pulse of 30 s while varying the interpulse intervals from 0 . 5 to 4 min to generate Ca2+ oscillation patterns with defined temporal resolution ( Figure 1d and Figure 1—figure supplement 6 ) and compared the levels of NFAT activation in HeLa cells . As shown in Figure 1f prolonged interpulse interval was largely accompanied by a decrease in the nuclear accumulation of NFAT . This observation agrees well with previous reports showing that higher Ca2+ oscillation frequencies , or faster repetitive Ca2+ pulses , tend to increase the ability to activate NFAT ( Lewis et al . , 1998 ) . Thus , we have demonstrated that the engineered Opto-CRAC tools are able to achieve remote and photo-tunable activation of NFAT in mammalian cells ( Figure 1f and Video 5 ) . We further confirmed the NFAT-dependent gene expression in HeLa cells transfected with an NFAT-driven luciferase ( NFAT-Luc ) reporter construct . In the presence of the co-stimulatory pathway ( mimicked by the addition of the pleiotropic PKC activator PMA ) , light illumination led to a robust increase in luciferase gene expression ( Figure 2a ) . A decrease in the light pulse frequency also caused a reduction in the efficiency of Ca2+/NFAT-driven luciferase expression ( Figure 1f ) . To obviate the use of carcinogenic PMA to photo-trigger gene expression , we also introduced a synthetic 5’ transcription regulatory region upstream of gene Ins1 ( Stanley et al . , 2012 ) , which contains a furin cleavage site that allows insulin processing in non-beta cells such as HEK293 cells ( Shifrin et al . , 2001 ) . The 5’ region is composed of three Ca2+-responsive elements in cis , including 2–3 copies of serum response elements ( SRE ) , cAMP response elements ( CRE ) and NFAT response elements with a minimal promoter . Upon light stimulation , we observed a robust production of insulin in cells transfected with LOVSoc , but not in those without LOVSoc expression ( Figure 2a ) . 10 . 7554/eLife . 10024 . 014Figure 2 . Photo-manipulation of Ca2+-dependent gene expression and immune response . All data were shown as mean ± s . d . from three independent experiments . *P<0 . 05; **P<0 . 01; ***P<0 . 001 ( paired Student’s t-test ) . ( a ) , Light-triggered Ca2+-dependent gene expression . Cells were either kept in the dark or exposed to pulsed blue light ( 30 s on with 30 s interval; 40 μW/mm2 ) for 6 hr prior to cell lysis to quantify luciferase activity ( middle ) or insulin production ( right ) . Iono , ionomycin . PMA , phorbol 12-myristate 13-acetate . Left panel , Schematic of experimental design . Three upstream Ca2+-responsive elements in the 5’ transcription regulatory region enable efficient initiation of gene expression of the downstream Ins1 gene encoding insulin following LOVSoc-mediated photoactivatable Ca2+ entry and NFAT nuclear translocation . SRE , serum-response element; CRE , cyclic adenosine monophosphate response element; NFAT RE , nuclear factor of activated T cells response element . Middle panel , Ca2+/NFAT-dependent luciferase activity in HeLa cells transfected with LOVSoc and an NFAT-dependent firefly luciferase reporter vector . A third plasmid encoding the Renilla luciferase gene was cotransfected as a reference gene for normalization of gene expression . Right panel , Photo-inducible insulin production driven by Ca2+-responsive elements in HEK293T cells . ( b ) , Photo-inducible expression of IL-2 and IFN-γ genes in mouse CD4+ T cells expressing the LOVSoc construct . Mouse CD4+ T cells were enriched and purified using an immunomagnetic negative selection kit and transduced with a retrovirus encoding mCh-LOVSoc . On day 5 after transduction and expansion in the presence of IL-2 , cells were treated with or without PMA , shielded from light or illuminated with blue light for 8 hr , and then lysed for qPCR ( upper panels ) or ELISA analyses ( lower panels ) . The schematic of the experiment was shown on the left . Upper panel , Optogenetic stimulation of cytokine production in mouse CD4+ effector T cells transduced with a retrovirus encoding mCh-LOVSoc . Right panel , Cytokine production ( IL-2 and IFN-γ that are characteristic of activated CD4+ T cells ) was determined by ELISA . ( c ) , Photo-tunable amplification of inflammasome activation in macrophages . Human THP-1-derived macrophages were transduced with lentiviruses expressing mCh-LOVSoc , primed with LPS ( 100 ng/ml ) and incubated with inflammasome inducer nigericin ( 10 μM ) for 6 hr . Cells were either shielded from light or illuminated with pulsed blue light for 8 hr at power densities of 5 or 40 μW/mm2 . The cell lysates were collected for ELISA analysis ( left ) and WB ( right ) . The schematic of the experiment was shown on the left . Left panel , the amount of secreted IL-1β in the culture supernatant quantified by ELISA . Right panel , NLRP3 inflammasome activation assessed by Western blotting of lysates and supernatants harvested from cells treated with indicated conditions . Arrowhead , processed caspase 1 ( Casp-1 ) subunit p20 . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 01410 . 7554/eLife . 10024 . 015Figure 2—figure supplement 1 . Retroviral transduction of CD4+ T cells and control experiments ( related to Figure 2b ) . ( a ) , Schematic illustration of the experimental protocol and NFAT-responsive cytokine expression in retrovirally transduced mouse CD4+ T cells . RV , retroviruses packaged from Platinum-E cells transfected with pMIG constructs; PMA , phorbol myristate acetate used to activate protein kinase C ( PKC ) ; Iono , the Ca2+ ionophore ionomycin used to elicit Ca2+ influx; AP-1 , the activator protein complex 1 that cooperates with NFAT to drive gene expression . ( b ) , Western blot analysis of cell lysates from mouse CD4+ T cells transduced with retroviruses encoding the empty vector ( lane 1 ) or mCh-LOVSoc ( lane 2 ) . ( c-d ) , IL-2 and IFN-γ expression quantified by qPCR ( c ) and ELISA ( d ) in mouse CD4+ T cells transduced with the mock retrovirus under indicated stimulation conditions . Light illumination in the presence or absence of PMA failed to cause productive cytokine expression . Data were shown as mean ± s . e . m from 12 wells in three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 01510 . 7554/eLife . 10024 . 016Video 5 . Light-inducible nuclear translocation of NFAT in HeLa cells . The HeLa GFP-NFAT stable cell line was transiently transfected with mCh-LOVSoc and exposed to pulsed light stimulation at 470 nm ( 30 s for every 1 min ) . Shown were fluorescence signals from the green ( GFP-NFAT , left panel ) and red ( mCh-LOVSoc , right panel ) channels in the same field . Only cells expressing the Opto-CRAC construct ( mCherry-positive , lower right corner ) showed light-dependent NFAT nuclear translocation . Note that the cytosol-to-PM translocation of mCh-LOVSoc is not evident as in Video 1 due to the low expression level of endogenous ORAI1 in HeLa cells and much more abundant expression of mCh-LOVSoc . Nonetheless , the light-triggered activation of endogenous ORAI1 channel was sufficient to activate the downstream GFP-NFAT nuclear translocation . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 016 In order to confirm light-inducible gene expression in a more physiologically relevant system , we retrovirally transduced the mCherry-tagged LOVSoc construct into naïve CD4+ T cells isolated from mice ( Figure 2b and Figure 2—figure supplement 1 ) . We then compared the expression levels of two signature genes that are characteristic of activated CD4+ T cells ( IL-2 and IFN-γ ) , in the presence or absence of light illumination , using qRT-PCR and ELISA ( Figure 2b ) . Again , in the presence of PMA , light stimulation faithfully mimicked ionomycin-induced effects on the Ca2+/NFAT pathway and remarkably boosted the cytokine production by over 15–30 fold in CD4+ T cells transduced with mCh-LOVSoc . By contrast , control cells transduced with the mock retrovirus failed to exhibit light-dependent production of cytokines ( Figure 2—figure supplement 1 ) . In addition to its well-established role in driving effector T cell activation , intracellular Ca2+ immobilization in macrophage is critical for the activation of the NLRP3 ( nucleotide-binding domain , leucine-rich-repeat-containing family , pyrin domain-containing 3 ) inflammasome ( Murakami et al . , 2012; Lee et al . , 2012; Horng , 2014 ) , which is accompanied by the release of processed caspase-1 ( p20 subunit ) and the proinflammatory cytokine IL-1β into culture supernatants ( Figure 2c ) . Following photostimulation at 5 or 50 μW/mm2 , we observed a notable light intensity-dependent boost in the production of IL-1β and processed caspase -1 ( p20 subunit ) in lipopolysaccharide ( LPS ) -primed THP1-derived macrophages in the presence of a commonly used inflammasome inducer nigericin ( Figure 2c ) , thus confirming the feasibility of harnessing the power of light to amplify macrophage-mediated inflammatory responses ex vivo . In aggregate , light-induced activation of the Opto-CRAC channel can generate both global and local Ca2+ signals and subsequently cause hallmark physiological responses in both model cellular systems ( e . g . , HeLa or HEK293 cells ) and rodent or human cells of the immune system . One fundamental roadblock that hampers the application of optogenetic tools in vivo is their inability to stimulate deep within tissues without the use of invasive indwelling fiber optic probes . In order to seek the possibility of controlling the Ca2+/NFAT pathway using light in the deep tissue penetrating near-infrared range , we explored the use of lanthanide-doped upconversion nanoparticles ( UCNPs ) as the NIR light transducer ( Sun et al . , 2015; Gnach et al . , 2015; Wu et al . , 2015 ) . Our UCNPs proved to be highly photostable , and their unique upconversion ( NIR excitation and emission at visible light range ) properties make them an ideal for the remote photoactivation of Opto-CRAC channel activities ( Wu et al . , 2009; Ostrowski et al . , 2012 ) . In order to match the absorption window of LOV2 , we chose mono-dispersed 40-nm β-NaYF4: Yb , Tm@β-NaYF4 UCNPs ( Figure 3—figure supplement 1 ) that exhibit bright blue emission upon 980 nm CW laser irradiation . When excited at 980 nm , the synthesized UCNPs displayed a sharp emission peak centered around 470 nm ( Figure 3a ) . Like direct blue light illumination , UCNPs were able to cause photoactivation of recombinant LOV2 proteins , as reflected by the absorbance changes following NIR light stimulation and the subsequent recovery to the dark state ( Figure 3b ) . This finding clearly validates the feasibility of shifting the spectral sensitivity toward the NIR window to activate LOV2-based optogenetic tools . 10 . 7554/eLife . 10024 . 017Figure 3 . NIR light control of Opto-CRAC by lanthanide-doped upconversion nanoparticles . ( a ) , Physiochemical properties of the synthesized upconversion nanoparticles . Upper panel , schematic illustration of the core/shell structure and energy transfer ( ET ) among lanthanide ions in the NaYF4: Yb , Tm@NaYF4 upconversion nanoparticles ( UCNPs ) . Lower panel , the emission spectrum of NaYF4: Yb , Tm@NaYF4 ( solid red line ) upon 980 nm CW laser irradiation ( 15 mW/mm2 ) superimposed by the absorbance spectrum of recombinant LOV2 protein ( dashed blue line ) . Inset: the bright blue emission could efficiently lighten the background upon NIR light illumination at 980 nm . ( b ) , NIR light-induced changes in the absorption spectra of purified MBP-LOVSoc at different time interval after mixing with UCNPs-Stv and irradiation with a 980 nm laser ( 1 min at a power density of 30 mW/mm2 ) . After blue ( excited at 470 nm as control , blue circle ) or NIR ( red triangle ) light stimulation , the recovery time course of LOV2 absorbance at 450 nm was plotted in the lower panel . ( c ) , Specific targeting of streptavidin-conjugated UCNPs to engineered ORAI1 channels in the plasma membrane of HeLa cells . Left panel , Schematic showing the interaction between streptavidin-coated upconversion nanoparticles ( UCNPs-Stv ) and the engineered ORAI1 Ca2+ channel that harbors a streptavidin-binding tag ( StrepTag ) in the second extracellular loop . The mCh-ORAI1StrepTag protein was able to efficiently recruit and anchor UCNPs-Stv to the plasma membrane of transfected HeLa cells . Right panel , Florescence microscopy imaging showing the accumulation of UCNPs-Stv ( green , λex: 980 nm , λem: 450–500 nm ) on the plasma membrane of cells transfected with mCh-ORAI1-StrepTag . Scale bar , 10 μm . ( d ) , NIR light-triggered Ca2+ influx and NFAT nuclear translocation in HeLa cells coexpressing mCh-ORAI1StrepTag and LOVSoc . Ca2+ influx was monitored by GCaMP6s fluorescence whilst GFP-NFAT translocation was reported by GFP signals . Transfected cells were mixed with UCNPs-Stv ( 20 μg/μl ) and illuminated by a 980-nm CW laser to trigger the Ca2+ influx . The relatively slow onset of Ca2+ influx and NFAT nuclear translocation provided us a time window to quickly capture the green signals without noticeably activating LOVSoc during image acquisition at low excitation energy ( <1 μW/mm2 ) . Scale bar , 10 μm . ( e ) , NIR light-induced reversible Ca2+ influx reported by R-GECO1 . 2 . HeLa cells were transfected with an IRES bicistronic pMIG retroviral construct that enabled coexpression of ORAI1StrepTag and mCh-LOVSoc . Transfected cells were mixed with 5 mg UCNPs-Stv and illuminated by a 980-nm laser at 30 mW/mm2 to trigger the Ca2+ influx . Data were shown as mean ± s . e . m . from 12 cells in two independent experiments . ( f ) , Flow cytometry analysis of IFN-γ production in mouse CD4+ T lymphocytes transduced with retroviruses co-expressing mCh-LOVSoc and ORAI1StrepTag . Freshly isolated CD4+ T cells were subjected to in vitro differentiation as described in Figure 2b , incubated with 20 μg/μl UCNPs-Stv and 1 μM PMA , and exposed to overnight NIR light pulse ( ON/OFF interval of 30 s , 980 nm , 30 mW/mm2 ) prior to analysis . ( g ) , NFAT-dependent luciferase expression in vivo triggered by NIR light stimulation . Left , Schematic of experimental setup . HeLa cells were transfected with NFAT-Luc and constructs encoding LOVSoc/ORAI1StrepTag . 48 hr post-transfection , cells were treated with 1 μM PMA , incubated with 10 mg UCNPs-Stv ( blue sphere ) and implanted to the flanks of mice subcutaneously . The implanted areas were then subjected to NIR light irradiation ( red ) with a 980 nm CW laser ( 50 mW/mm2 , 30 sec ON , 30 sec OFF for a total of 25 min ) . Right , Shown were bioluminescence imaging of three representative BALB/c mice , one implanted with HeLa cells expressing NFAT-Luc only ( left ) and the other two with cells expressing LOVSoc and NFAT-Luc ( middle and right ) . Mice were subjected to NIR light irradiation ( left and right ) with a 980 nm CW laser . The images were acquired 20 min after receiving a single dose of luciferin ( 100 μL , 15 mg/ml , s . c . ) . Luciferase-catalyzed bioluminescence was visualized as false color with the same rainbow scale bar for all acquired images . Red circle , implanted area . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 01710 . 7554/eLife . 10024 . 018Figure 3—figure supplement 1 . Synthesis scheme and in vitro characterization of UCNPs . ( a ) , TEM images of as-synthesized β-NaYF4: Yb , Tm core/shell UCNPs ( UCNPs-OA ) , after PAA surface coating ( UCNPs-PAA ) and further modifications with streptavidin ( UCNPs -Stv ) . Scale bar , 100 nm . ( b ) , Schematic illustration of the surface modification procedures for water-soluble and streptavidin functionalized β-NaYF4: Yb , Tm/NaYF4 UCNPs . ( c ) , FT-IR spectra of UCNPs with different surface modifications . In the spectrum representative of UCNPs-PAA , the peaks at 2898 and 2982 cm-1 were attributed to the resonance of COO-H . This PAA-specific peak disappeared after its amidation with zwitterion and streptavidin conjugation , but new peaks at 1554 , 1182 and 1042 cm-1 emerged . These new peaks were attributed to the C ( O ) -N , R-SO3- and C-N vibrations , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 01810 . 7554/eLife . 10024 . 019Figure 3—figure supplement 2 . Green-emitting UCNPs did not activate Opto-CRAC channels upon NIR light stimulation ( related to Figure 3d ) . ( a ) , Overlaid spectra of NaYF4: Yb , 2%Er@NaYF4 with green emission under 980 nm CW laser ( green line ) , NaYF4: Yb , Tm@NaYF4 with blue emission upon 980 nm irradiation ( blue line ) and the absorbance of LOV2 domain . ( b ) , Time-lapse images of GCaMP6s-CAAX in the illuminated region . Hela cells coexpressing mCh-ORAI1-StrepTag and LOVSoc were mixed with either blue-emitting ( used in Figure 3d ) or green-emitting UCNPs-Stv and then irradiated by a 980 nm CW laser ( 30 mW/mm2 ) . Blue-emitting UCNPs generated a steady increase in fluorescence whilst green-emitting UCNPs caused no significant change in the fluorescence intensity . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 01910 . 7554/eLife . 10024 . 020Figure 3—figure supplement 3 . No noticeable heat generation during the in vivo experiment . ( a ) , Thermal imaging at two minute intervals . ( b ) , Temperature change over time plot ( right ) of a Balb/c mouse exposed to 50 mW/mm2 980nm laser irradiation for 30 min . ( c ) , Histological sections of implantation positions 14 days after ectopic injection and NIR treatment in mice . Fixed tissues isolated around the injection sites were subjected to hematoxylin/eosin staining . The sections represent HeLa cells without ( upper ) or with UCNPs ( lower ) loaded during the injection . Scale bar , 200 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 020 In order to effectively and specifically illuminate the LOV2-based optogenetic construct in a cellular context , we first developed streptavidin-conjugated UCNPs , then engineered a genetically-encoded streptavidin-binding tag ( StrepTag ) into the second extracellular loop of the ORAI1 Ca2+ channel ( mCh-ORAI1StrepTag , Figure 3c ) and assessed its capability to recruit streptavidin-conjugated UCNPs ( UCNPs-Stv , Figure 3—figure supplement 1 ) . In HeLa cells expressing mCh-ORAI1-StrepTag , we detected remarkable local accumulation of UCNPs-Stv on the plasma membrane ( Figure 3c ) , confirming the cell-specific targeting of functionalized nanoparticles . To examine whether UCNPs-transduced blue light is sufficient to trigger the opening of Opto-CRAC channels , we monitored cytosolic Ca2+ changes using GCaMP6s in HeLa cells co-expressing LOVSoc , mCh-ORAI1-StrepTag and GCaMP6s following NIR light stimulation ( 980 nm ) . Within 20 s , transfected HeLa cells exhibited a significant increase in GCaMP6s fluorescence , indicating a rapid rise in the intracellular Ca2+ concentration that was evoked by NIR light ( Figure 3d ) . This was further confirmed by using a red-emitting Ca2+ indicator R-GECO1 . 2 , which enabled recording of reversible Ca2+ fluctuation cycles and circumvented the complications associated with potential direct activation of LOVSoc by the green light source used to excite GCaMP6 signals ( Figure 3e ) . This increase was found to be caused by Ca2+ influx through NIR-to-blue activated Opto-CRAC channels because cells incubated with the control NIR-to-green UCNPs ( β-NaYF4: Yb , 2% Er @ β-NaYF4; emission maxima at 510 nm ) did not show discernible changes in the GCaMP6s signal upon stimulation with the same NIR light ( Figure 3—figure supplement 2 ) . We then employed NIR light to remotely activate the downstream effector NFAT at the cellular level , and observed NFAT nuclear translocation ( Figure 3d ) , as well as NFAT-dependent IFN-γ production in CD4+ T lymphocytes ( Figure 3f ) . Next , we sought to demonstrate the potential application of NIR-triggered activation of the Opto-CRAC system in vivo . We performed a proof-of-principle experiment by implanting NFAT-Luc/LOVSoc expressing HeLa cells pre-incubated with UCNPs-Stv subcutaneously in the flanks of mice . The implanted site was irradiated by a 980-nm CW laser outside the body ( Figure 3e ) without noticeable heat production ( Figure 3—figure supplement 3a , b ) or severe damage to local tissues ( Figure 3—figure supplement 3c ) . Luciferase-catalyzed bioluminescence was readily detected after NIR irradiation , whereas no discernible background activation was observed in the negative controls where LOVSoc expression and/or NIR light were absent ( Figure 3g ) . To explore the application of the NIR Opto-CRAC system in a more disease-relevant context , we set out to combine the use of our optogenetic system with DC-mediated immunotherapy in the B16-OVA mouse model of melanoma ( Briles and Kornfeld , 1978; Fidler , 1975 ) , in which ovalbumin ( OVA ) ( Falo et al . , 1995; Mayordomo et al . , 1995 ) is used as a surrogate tumor antigen ( Figure 4a ) . Dendritic cells , which provide the essential link between the innate and adaptive immune responses , are adept at capturing tumor antigens and cross-presenting these antigens to T cells in tumor draining lymph nodes ( dLNs ) , thereby sensitizing and generating tumor-specific cytotoxic lymphocytes ( CTLs ) to cause tumor regression or rejection ( Palucka and Banchereau , 2012 ) . One of the major challenges of DC vaccination-based immunotherapy is how to efficiently maintain the maturational status of DCs . Pharmacological agents ( e . g . , ionomycin ) or signaling pathways controlling intracellular Ca2+ mobilization have been reported to facilitate immature dendritic cell maturation through up-regulation of co-stimulatory molecules CD80 or CD86 , major histocompatibility complex ( MHC ) class I and class II , as well as the chemokine receptor CCR7 ( Félix et al . , 2013; Matzner et al . , 2008; Hsu et al . , 2001; Koski et al . , 1999; Czerniecki , 1997 ) . We hypothesize that photoactivatable Ca2+ influx in DCs will lead to similar phenotypic changes to expedite and sustain DC maturation and promote antigen presentation , thereby maximally sensitizing T lymphocytes toward tumor antigens to boost anti-tumor immune response . To quickly test this in vitro , we transduced bone marrow-derived DCs ( BMDCs ) with retroviruses encoding both LOVSoc and ORAI1StrepTag ( termed 'Opto-CRAC DCs' ) , pulsed cells with a mixture of OVAp ( 257SIINFEKL264 ) and UCNPs-Stv nanoparticles . NIR light stimulation resulted in approximately 2–8 fold increase in the surface expression of MHC-I/II , CD86 , and CCR7 ( Figure 4b ) , which are characteristic of matured DCs that are capable of homing to dLNs to interact with T cells to modulate adaptive immune response ( Palucka and Banchereau , 2012 ) . We next used ex vivo cross-presentation assay to examine how CD8 T cells from OT-1 Rag1-/- mice respond to the OVA antigen presented by DCs . The isolated OT-1 CD8 T cells , bearing transgenic T cell receptors that specifically recognize processed OVA peptides ( Clarke et al . , 2000; Hogquist et al . , 1994 ) , were co-cultured with Opto-CRAC DCs in the presence of OVAp and UCNPs-Stv . After NIR stimulation , co-cultured OT-1 CD8 T cells exhibited over 2-fold increase in both proliferation ( Figure 4c ) and IFN-γ release ( Figure 4d ) , clearly attesting to the feasibility of using the NIR-stimulable Opto-CRAC system to expand and photo-prime antigen-specific T cells . 10 . 7554/eLife . 10024 . 021Figure 4 . NIR light control of Opto-CRAC DC-mediated antigen cross-presentation to OT-I CD8 T cells and B16-OVA melanoma killing . Data were shown as mean ± s . e . m . from at least three independent experiments . *p<0 . 05; **p<0 . 01; ***p<0 . 001 ( paired Student’s t-test ) . ( a ) , Scheme showing the experimental design . NIR-stimulated Ca2+ influx in Opto-CRAC DCs prompts immature DC maturation and OVA antigen cross-presentation to activate and boost anti-tumor immune responses mediated by OT-1 CD8 T cells ( cytotoxic T lymphocytes , CTLs ) , thereby sensitizing tumor cells to OVA-specific , CTL-mediated killing in the B16-OVA melanoma model . OVA peptide ( OVAp , 257SIINFEKL264 ) is used here as a surrogate tumor antigen . Rag1-/- mice were subcutaneously ( s . c . ) implanted in the flank or intravenously ( i . v . ) injected with 2 x 106 B16-OVA tumor cells per mice to induce melanoma and lung metastasis . Bone marrow-derived dendritic cells ( BMDCs ) expressing the Opto-CRAC system ( Opto-CRAC DCs ) were pulsed with UCNPs-Stv and OVA257-264 peptides and injected into Rag1-/- mice 3 days after tumor cell injection . Sorted OT-I CD8 cells from OT-I Rag1-/- mice , which are labeled by CSFE for monitoring DC antigen cross-presentation and T cell proliferation in vivo , were transferred into B16 tumor-bearing mice one day after Opto-CRAC DC infusion . Mice were kept in dark or exposed to NIR for 1 week ( 8 hr per day , 1 min ON/OFF pulse , 30 mW/mm2 ) after DC injection to stimulate Opto-CRAC DC maturation in vivo and photo-boost tumor antigen cross-presentation . 5 days after adoptive transfer , tumor draining lymph nodes ( dLNs ) and spleen were harvest for FACS ( panel e ) analysis on CFSE-labeled CD8 T cells . Tumor growth was measured by caliper ( panel f ) and mice were sacrificed on day 18 for lung metastasis analysis ( panel g ) . ( b ) , Flow cytometric analysis on the expression levels of MHC , co-stimulatory and chemokine receptor molecules in BMDCs . Cells were double-immunolabeled with CD11c-FITC vs MHC class I-PE , CD86-PE , MHC class II-APC or CCR7-PE and analyzed three days after viral transduction of Opto-CRAC . UCNPs-Stv/ OVA loaded Opto-CRAC DCs were exposed or shielded from NIR illumination for 48 hr ( 30 mW/mm2 with 1 min pulse interval ) prior to analysis . ( c ) , Proliferation of sorted naïve OT-I CD8 T cells co-cultured with UCNPs-Stv/OVA loaded Opto-CRAC DCs , with or without NIR illumination , was measured by the [3H] thymidine incorporation assay . ( d ) , Flow cytometric analysis of IFN-γ production in sorted naïve OT-I CD8 T cells co-cultured with UCNPs-Stv/OVA loaded Opto-CRAC DCs with or without NIR stimulation . ( e ) , Flow cytometric analysis of in vivo proliferation of CFSE-APC labeled OT-I CD8 T cells in dLNs and spleen 6 days after injection of UCNPs-Stv/OVA loaded Opto-CRAC DCs with or without NIR pulse excitation ( 30 mW/mm2 ) as indicated in panel a . ( f ) , Tumor-inoculated sites ( left ) were isolated from tumor-bearing mice ( n = 5 ) shielded or exposed to NIR and the tumor sizes ( mm3 ) were measured at indicated time points shown in the growth curve ( right ) after tumor implantation . ( g ) , Representative lungs with melanoma metastases ( left ) were isolated from tumor-bearing mice shielded or exposed to NIR . The histogram represents counted numbers of visible pigmented tumor foci ( as exemplified by the arrows ) with pulmonary melanoma metastases on the surface of lungs ( right panel; n = 5 mice ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10024 . 021 To further validate the immunomodulatory function in vivo , we injected UCNPs-Stv/OVA loaded Opto-CRAC DCs to the B16-OVA murine model of melanoma ( Falo et al . , 1995; Mayordomo et al . , 1995 ) , in which the B16 tumor cells bearing the OVA antigen could be readily recognized by OT-1 CD8 T cells to elicit anti-tumor immune responses ( Matzner et al . , 2008; Hsu et al . , 2001 ) . We next adoptively transferred CFSE-abled , OVA-specific OT-I CD8 T cells into the B16-OVA mice and examined their in vivo activation and phenotypic profiles following photoactivatable DC maturation . Compared to the control group shielded from NIR , the proliferation of CD8 T cells was substantially up-regulated after light stimulation , by judging from decreased CFSE staining due to proliferative dilution and increased population of OT-1 CD8T cells in tumor draining LNs and spleens ( Figure 4e ) . To assess the functional consequence of immunosensitization of tumor cells toward Opto-CRAC DC-activated immune response , we monitored the tumor growth in mouse melanoma models generated by either subcutaneous or i . v . injection of B16-OVA melanoma cells ( Figure 4a ) . NIR light stimulation significantly suppressed the tumor growth with diminished tumor volume ( Figure 4f ) or reduced numbers of tumor foci in the lungs ( Figure 4g ) . Both our ex vivo and in vivo data converge to support the conclusion that NIR-stimulable Opto-CRAC DC can robustly enhance tumor cell susceptibility to CTL-mediated killing , thereby improving antigen-specific immune responses to selectively destruct tumor cells . By acting as a genetically-encoded 'photoactivatable adjuvant' , the Opto-CRAC system may hold high potential for its future use in cancer immunotherapy .
In the present study , we described an NIR-stimulable optogenetic platform based on engineered CRAC channels and lanthanide-doped upconversion nanoparticles . Depending on the pulse and intensity of light input , the photosensitive module , LOVSoc , can reversibly generate both sustained and oscillatory Ca2+ signals . The magnitude and kinetics of photo-activated Ca2+ influx largely mimic the physiological responses following engagement of immunoreceptors or ligand binding to its cognate membrane receptors that leads to store depletion ( Prakriya and Lewis , 2015 ) . Ectopic expression of a single component of LOVSoc at endogenous levels of ORAI is sufficient to elicit strong intracellular Ca2+ elevation in a dozen of cell types derived from a wide range of human or rodent tissues . Most critically , light-generated Ca2+ signals can further lead to hallmark physiological responses in cells of the immune system . The sensitivity and photoactivation kinetics of this system can be further tuned by tethering LOVSoc to PM or through co-expression and fusion with ORAI1 . When paired with deep tissue-penetrant and NIR-stimulable UCNPs , we have successfully demonstrated the potential application of our Opto-CRAC system to drive Ca2+-dependent gene expression and to photo-modulate immune response both in vivo and ex vivo . Compared to other existing optical tools , our Opto-CRAC system that has several distinctive features: First , complementary to the existing ChR-based tools that exhibit less stringent ion selectivity and tend to perturb intracellular pH due to high proton permeability , our Opto-CRAC system is engineered from a bona fide Ca2+ channel that is regarded as one of the most Ca2+-selective ion channels . Although the unitary conductance of CRAC channel is estimated to be low ( <10 fS in 2 mM extracellular Ca2+ in T cells; compared to 4–10 pS for voltage-gated Ca2+ channels ) ( Prakriya and Lewis , 2015; Zweifach and Lewis , 1993 ) , sustained Ca2+ influx ( up to minutes ) through native ORAI1 channels is sufficient to activate downstream effectors . The high Ca2+ selectivity ( PCa/PNa: >1000 ) and its small unitary conductance is speculated to reduce the energy requirement of pumping out Na+ during sustained Ca2+ entry , thereby enhancing the specificity of downstream effector function ( Prakriya and Lewis , 2015 ) . Second , the Opto-CRAC tool has a relatively small size ( <900 bp , compared to >2 . 2 kb of ChR ) and is thus compatible with almost all existing viral vectors used for in vivo gene delivery . Indeed , we have successfully used retroviral and lentiviral expression systems to deliver Opto-CRAC into primary T cells , macrophages and dendritic cells . Its potential delivery into excitable tissues ( e . g . , muscle , heart and brain ) using adeno-associated viruses remains to be tested in follow-on studies . Third , the tunable and relatively slow kinetics make it most suitable for interrogating Ca2+-modulated functions in non-excitable cell types , such as cells in the endocrine , immune and hematopoietic system . We find that our system may find broad use in adoptive cell transfer experiments or adoptive immunotherapies , which are widely used in both basic research and the clinic settings ( Palucka and Banchereau , 2012; Restifo et al . , 2012 ) . Fourth , in conjunction with upconversion nano-transducers , the light harvesting window can be shifted to the NIR region where deep tissue penetration and remote stimulation are feasible . Results from our in vivo studies clearly indicate that the Opto-CRAC channel and its downstream effectors can be remotely activated using NIR light , thereby paving the way for its future applications in more ( patho ) physiologically-relevant mouse models , or ultimately , in cancer immunotherapies with improved spatiotemporal control over engineered therapeutic T cells or DCs to reduce off-tumor cross-reaction and mitigate toxicity ( Morgan et al . , 2010 ) . Given the spatial and temporal accuracy of NIR light , it is also possible to use guided NIR light to confine localized blue light generation , thus avoiding the photoactivation of off-target regions . Lastly , but critically , the lanthanide-doped UCNPs can be applied to activate other optogenetic tools that are dependent on blue light-absorbing cofactors ( e . g . , ChR2 and CRY2 ) . We anticipate that the flexible adaptability of our novel approach will lead to new opportunities to fine-tune Ca2+-dependent immune responses and interrogate other light-controllable cellular processes while minimally interfering with the host’s physiology .
Fura-2 AM calcium indicator was purchased form Life Technologies ( Carlsbad , CA , USA ) . Phorbol 12-myristate 12-acetate ( PMA ) , ionomycin , thapsigargin ( TG ) and isopropyl-ginethiogalactopyranoside ( IPTG ) were purchased from Sigma Aldrich ( St Louis , MO , USA ) . Tri ( 2-carboxyethyl ) phosphine ( TCEP ) was obtained from Pierce ( Life Technologies ) . Amylose resin used for MBP pulldown was purchased from New England Biolabs ( Ipswich , MA , USA ) . Ni-NTA resin used for purification of GB1-ORAI1-CT was purchased from Qiagen ( Valencia , CA , USA ) . The mouse monoclonal anti-Flag M2-HRP ( A859 , Sigma-Aldrich , St . Louis , MO , USA ) antibody , the rabbit anti-mCherry polyclonal antibody ( NBP2-25157 , Novus Biologicals , Littleton , CO , USA ) , the rabbit anti-Caspase-1 antibody ( D7F10 , Cell signaling , Danvers , MA , USA ) and the rabbit anti-IL-1β antibody ( sc-7884 , Santa Cruz Biotechnology , Dallas , TX , USA ) were used at a 1:1000 dilution . For flow cytometry ( FACS ) analysis , anti-mouse MHC Class II ( I-A/I-E ) APC ( ) , anti-mouse IFN gamma PE ( ) , anti-mouse CD86 ( B7-2 ) PE ( 12–7311 ) , anti-mouse CD197 ( CCR7 ) PE ( 12–1971 ) , anti-mouse MHC Class I PE ( 12–9558 ) , anti-mouse CD11c FITC ( 11–0114 ) , anti-mouse CD4 PerCP-Cyanine5 . 5 ( 45–0042 ) , and anti-mouse CD8a APC ( 17–0081 ) were purchased from eBioscience . All other reagents were form Sigma-Aldrich unless otherwise indicated . HeLa , HEK293/HEK293T and other indicated immortalized cell lines from the American Type Culture Collection ( ATCC ) were cultured in Dulbecco's modified Eagle's medium ( DMEM , Sigma-Aldrich ) supplemented with 10 mM HEPES and 10% heat-inactivated fetal bovine serum . All the cells were grown at 37°C in a 5% CO2 atmosphere . Cultured cells were seeded on 35-mm glass bottom dishes and an inverted Nikon Eclipse Ti-E microscope customized with Nikon A1R+ confocal laser sources ( 405/488/561/640 nm ) was used for confocal imaging . The same microscope body connected to a Ti-TIRF E motorized illuminator unit ( 488 nm/20 mW and 561 nm/20 mW lasers ) with a 60× , NA 1 . 49 oil-immersion TIRF objective was used for TIRF imaging . 100-nm fluorescent beads ( TetraSpeck microspheres , Life Technologies ) were deposited onto a coverslip and imaged as markers for later alignment . To monitor mCh-LOVSoc translocation from the cytosol to PM , 50–100 ng pTriEx-mCherry-LOVSoc was transfected to HEK293-ORAI1 stable cells using Lipofectamine 3000 ( Life Technologies ) . Cells were imaged 24 hr after transfection . Photostimulation was provided by an external blue light ( 470 nm , tunable intensity of 0–50 μW/mm2 , ThorLabs Inc . , Newton , NJ , USA ) . Light power density was measured by using an optical power meter from ThorLabs . Light cycles were applied either manually or programmed by connecting to a DC2100 LED Driver with pulse modulation ( ThorLabs ) . Time-lapse imaging of mCherry signal was carried out in the dark by turning on only the 561-nm laser channel . For measurements of Ca2+ influx using the green color calcium indicator GCaMP6s , 50–100 ng mCh-LOVSoc and 100 ng cytosolic GCaMP6s or membrane-tethered GCaMP6s-CAAX were cotransfected into HeLa or HEK293T or other indicated cells using Lipofectamine 3000 . Twenty-four hours after transfection , a 488-nm laser was used to excite GFP , and a 561-nm laser to excite mCherry at intervals of 1–5 s . The mCherry-positive cells were selected for statistical analysis . Since the excitation wavelength used to acquire the GCaMP6s signals ( 488 nm ) partially overlaps with the photo-activating wavelength of LOVSoc , Ca2+ influx was elicited when the 488-nm laser source was turned on , and thus GCaMP6s could only be used to monitor the ON phase of Ca2+ flux . For localized photostimulation , we took advantage of the NIKON component designed for fluorescence recovery after photobleaching ( FRAP ) to stimulate selected areas ( designated as pre-activated areas as exemplified in Figure 1—figure supplement 4 and Video 3 ) but only used 1–5% input of the 488-nm laser for 5–10 s . Next , we recorded the GCaMP6s-CAAX signals in the whole field . For measurements of Ca2+ influx using the red-emitting Ca2+ sensor ( R-GECO1 . 2 or R-CaMP2 ) , a total of 300 ng DNA ( 100 ng mCh-LOVSoc and 200 ng Ca2+ sensor ) was transfected into HeLa or HEK293 T cells . The 561-nm laser source as used to excite red emission with blue light stimulation imposed as described above . Because the 561-nm laser cannot activate LOVSoc , both the ON and OFF phases of Ca2+ fluctuation can be monitored by applying multiple dark-light cycles with an external pulsed LED light ( 470 nm at power intensity of 40 μW/mm2 ) or using the 488-nm laser source from the Nikon A1R+ confocal microscope . To monitor light-inducible NFAT nuclear translocation , we used a HeLa cell line stably expressing NFAT11-460-GFP . mCh-LOVSoc was transfected into this stable HeLa cell line and cells were imaged 24 hr posttransfection . A fixed blue light pulse of 30 s ( 40 μW/mm2 ) was applied to the transfected cells with the interpulse interval varying from 0 . 5 , 1 , 4 , to 8 min . A total of 24 min time-lapse images were recorded and the GFP signal ratio ( nuclear vs total GFP ) was used to report the efficiency of NFAT activation . At least 15 cells were analyzed for each condition in three independent experiments . Intracellular Ca2+ measurements with Fura-2 were performed using our previous protocols ( Wang , 2014; Zhou et al . , 2010a; 2010b; 2013 ) . Briefly , one day before imaging , HEK293 cells transiently expressing mCh-LOVSoc were seeded and cultured on cover slips . To load Fura-2 AM , cells were kept in the imaging solution with 0 mM CaCl2 and 2 μM Fura-2 AM for one hour . The imaging solution consists of ( mM ) 107 NaCl , 7 . 2 KCl , 1 . 2 MgCl2 , 11 . 5 glucose , 20 HEPES-NaOH ( pH 7 . 2 ) , and 0 or 1 mM CaCl2 . Fura-2 signals were then obtained using a ZEISS oberserver-A1 microscope equipped with a Lambda DG4 light source ( Sutter Instruments ) , Brightline FURA2-C-000 filter set ( Semrock Inc . ) . Fura-2 fluorescence at 509 nm generated by 340 nm excitation light ( F340 ) and 380 nm light ( F380 ) was collected every two seconds , and intracellular Ca2+ levels are indicated by F340/F380 ratio . To excite LOVSoc during light-on period , cells were continuously exposed to a 482 ± 9 nm light throughout each two-second interval immediately following the collection of every single F380 and F340 . After 1 min photostimulation ( 470 nm , 40 μW/mm2 ) , Ca2+ concentrations in cells were determined by using a Fura 2 calcium imaging calibration kit ( ThermoFisher Scientific ) as we routinely did in earlier studies ( Wang , 2014; Zhou et al . , 2010a; 2010b; 2013 ) . The resulting data collected with MetaFluor software ( Molecular Devices ) were then exported as txt file , analyzed with Matlab , and plotted using the Prism 5 software . HeLa cells were seeded in 24-well plates and transfected after reaching 40–50% confluence . mCh-LOVSoc , the firefly luciferase reporter gene ( NFAT-Luc ) and Renilla luciferase gene ( pRL-TK ) were co-transfected using Lipofectamine 3000 . 24 hr posttransfection , cells were treated with PMA ( 1 μM ) and/ or blue light ( pulse of 30 s for every 1 min , 40 μW/mm2 ) . Three duplicates were used for each treatment . After 8 hr , cells were harvested and luciferase activity was assayed by using the Dual Luciferase Reporter Assay System ( Promega ) on a Synergy luminescence microplate reader ( BioTek , Winooski , VT , USA ) . Renilla luciferase is used as control reporter for counting transfected cells and normalizing the luminescence signals . The ratio of firefly to renilla luciferase activity was calculated and normalized to un-treated control group . HEK EPC10 USB double patch amplifier controlled by Patchmaster software ( HEKA Elektronik ) was used for data collection . Conventional whole cell recordings were used to measure current in HEK293-ORAI1 stable cells transiently expressing mCh-LOVSoc as previously described ( Ma , 2015 ) . After the establishment of the whole-cell configuration , a holding potential of 0 mV were applied . A 50 ms step to −100 mV followed by a 50 ms ramp from −100 to +100 mV was delivered every 2 s . The intracellular solution contained ( mM ) : 135 Cs aspartate , 6 MgCl2 , 10 EGTA , 3 . 3 CaCl2 , 2 Mg-ATP , and 10 HEPES ( pH 7 . 2 by CsOH ) . The free Ca2+ concentration in this pipette solution is estimated to be 100 nM based on calculations from http://www . stanford . edu/~cpatton/webmaxcS . htm . The extracellular solutions contained ( mM ) : 130 NaCl ( or N-methyl-D-glucamine , NMDG+ ) , 4 . 5 KCl , 20 CaCl2 , 10 TEA-Cl , 10 D-glucose , and 5 Na-HEPES ( pH 7 . 4 ) . A 10 mV junction potential compensation was applied to correct the liquid junction potential between the pipette solution relative to extracellular solution . Currents from at least 6 cells for each condition were collected . HEKA Fitmaster and Matlab 2014a software were used for data analysis . Platinum-E ( Plat-E ) retroviral packaging cell Line ( Cell Biolabs , Inc , San Diego , CA ) was maintained in Dulbecco modified Eagle medium ( DMEM ) supplemented with 10% fetal calf serum ( FCS ) , 1% penicillin/streptomycin , and 1% glutamine . Plat-E cells were transiently transfected using Lipofectamine 3000 ( Life Technologies ) and retroviral stocks were collected twice at 24-hr intervals beginning 48 hr after transfection . Retrovirus-containing medium was centrifuged at 20 000 rpm for 2 hr at 4°C in a Beckman SW28 swinging bucket rotor lined with an Open-Top polyclear centrifuge tube ( Seton , Petaluma , CA ) . The retroviral pellet was resuspended in DMEM and retrovirus was titered by transduction of mouse T cells with serial dilutions of retrovirus in the presence of 8 μg/ml polybrene ( EMD Millipore , Merck KGaA , Darmstadt , Germany ) . 48 hr posttransduction , percentage of infected cells was determined by flow cytometric analysis of EGFP expression . The titer was calculated by multiplication of the total number of EGFP-positive cells with the dilution factor of the retroviral supernatant . Naive CD4+ T cells were purified ( >95% purity ) by negative selection ( Invitrogen ) with Mouse Depletion Dynabeads ( Life Technologies , Grand Island , NY ) from RBC-lysed single-cell suspensions of pooled spleen and lymph nodes isolated from 6-week-old female C57BL/6 mice . For stimulation , purified CD4+ T cells were cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum , 2 mM L-glutamine , penicillin-streptomycin , non-essential amino acids , sodium pyruvate , vitamins , 10 mM HEPES , and 50 µM 2-mercaptoethanol . Cells were plated at ~106 cells per ml in 6-well plates coated with anti-CD3 ( clone 2C11 , BioLegend , San Diego , CA , USA ) and anti-CD28 ( clone 37 . 51 , BioLegend ) ( 1 µg/ml each ) by pre-coating with 100 µg/ml goat anti-hamster IgG ( MP Biomedicals , Santa Ana , CA , USA ) . After 48 hr , cells were removed from the TCR signal and re-cultured at a concentration of 5x105 cells/ in T cell media supplemented with 20 U/ml recombinant human IL-2 ( rhIL-2 ) . For retroviral transduction , CD4+ T cells were re-suspended in concentrated viral supernatant containing 8 µg/ml polybrene and rhIL-2 and centrifuged at 2 , 000 x g for 90 min at 32°C then put back to the incubator . On day 5–6 , GFP+ cells were either left untreated ( resting ) , or re-stimulated with PMA ( 15 nM ) and ionomycin ( 0 . 5 µM ) , or subjected to blue light pulse for 6–8 hr ( 30 s pulse for every 1 min , 10–40 μW/mm2 ) , or treated with both PMA and blue-light pulse for 6–8 hr . Expression of cytokine production was assessed by real-time PCR and ELISA as described below . Total RNA was isolated from transduced CD4+ T cells and first-strand cDNA synthesis was performed using total RNA , oligo-dT primers and reverse transcriptase II according to manufacturer’s instructions ( Invitrogen ) . Real-time PCR was performed using the SYBR Green ER qPCR Super Mix Universal ( Invitrogen ) kit with specific primers using the ABI Prism 7000 analyzer ( Applied Biosystems ) . The sequences of the primers are as follows , Supernatants of transduced CD4+ T cells were collected at indicated time after stimulation . Cytokine concentrations were measured by using the mouse IL-2 ( OptEIA #555148 , BD Biosciences Inc . , San Jose , CA , USA ) and IFN-γ ELISA kits ( #88–7314 , eBiosciences Inc . , San Diego , CA , USA ) . ELISA assays were performed according to the manufacturer's instructions . In brief , 96-well plate was pre-coated with the capture antibody ( 1:500 in coating buffer ) at 4°C overnight . On the next day , the plate was washed with PBS/0 . 1%Tween 20 and blocked with 1%BSA/PBS or ELISA/ELISPOT diluent buffer for 2 hr at room temperature ( RT ) . Diluted supernatants and cytokine standards were then applied to the plate and incubated for 2 hr at RT . The plate was then washed and incubated with biotin-conjugated detection antibody ( 1:1000 in 1%BSA/PBS or ELISA/ELISPOT diluent buffer ) for 1 hr at RT . Next , the plate was washed and incubated with poly-HRP streptavidin ( 1:5000 in diluent buffer , Thermo Scientific ) for 30 min . The plate was finally washed and incubated with the tetramethylbenzidine substrate solution ( Sigma-Aldrich ) and the reaction was stopped with 2 M H2SO4 . For insulin reporter assay , 3x105 transfected HEK293T cells were cultivated in poly-L-lysine coated 24-well pates and starved in serum-free culture medium for 24 hr to ensure minimal activation of Ca2+ dependent pathways . On the day of experiment , cells were washed with PBS and maintained in serum-/glucose-starved Krebs buffer ( 118 mM NaCl , 4 . 7 mM KCl , 1 . 2 mM KH2PO4 , 1 . 2 mM MgSO4 , 4 . 2 mM NaHCO3 , 2 mM CaCl2 , 10 mM HEPES and 0 . 1 mg/ml BSA , pH7 . 4 ) with or without light stimulation . Supernatants were collected for insulin ELISA detection using a human insulin ELISA kit ( KAQ1251 , Life Technologies ) according to the manufacture’s instructions . Absorbance of each well was recorded at 450 nm . The absorbance of the standard sample was used to construct the standard curve . THP-1 cells from ATCC were maintained in RPMI-1640 medium containing 10% FBS and 0 . 05 mM 2-mercaptoethanol . Differentiated THP-1 cells were transduced with lentiviruses encoding LeGO-mCh-LOVSoc . THP-1 cells ( 5 × 105 ) were seeded in 24-well plates and cultured overnight , followed by priming with 100 ng/mlLPS for 3 hr and stimulating with Nigericin ( 10 μM ) for 6 hr with or without blue light stimulation . Medium from each well was mixed with 500 μl methanol and 125 μl chloroform , vortexed , and centrifuged for 5 min at 16 , 000 × g . The supernatant of each sample was removed and 500 μl methanol was added . Samples were centrifuged again for 5 min at 16 , 000 × g . Next , supernatants were removed and pellets were dried for 5 min at 50°C . 80 μl loading buffer was added to each sample , followed by boiling for 10 min prior to SDS-PAGE and immunoblot analysis with antibodies for the detection of activated caspase-1 ( D7F10; Cell Signaling ) . The amounts of processed IL-1β were measured using a human IL-1β ELISA kits ( R&D Systems ) according to the manufacturer’s instructions . Adherent cells in each well were lysed with the RIPA lysis buffer ( 50 mM Tris-HCl , pH 8 . 0 , with 150 mM sodium chloride , 1 . 0% Igepal CA-630 ( NP-40 ) , 0 . 5% sodium deoxycholate , and 0 . 1% sodium dodecyl sulfate ) with a protease inhibitors cocktail tablet ( Roche ) , followed by immunoblot analysis to determine the cellular content of various proteins . BL21 ( DE3 ) E . coli cells ( EMD Millipore ) were transformed with plasmids encoding MBP-LOVSoc or GB1-ORAI1-CT , and grown at 37ºC in LB medium with 100 mg/L of ampicillin . Protein expression was induced by the addition of 500 μM IPTG when OD600 of the culture reached between 0 . 6 and 0 . 8 . After IPTG induction , MBP-LOVSoc was incubated at 16ºC for additional 6–8 hr , whilst GB1-ORAI1CT incubated at 37ºC for 3–4 hr . Harvested cells were resuspended in 1X Phosphate Buffered Saline ( PBS ) and sonicated . The cellular debris was clarified by centrifugation . For His6-tagged GB1-ORAI1-CT , the cell lysates were applied to Ni2+-nitrilotriacetic acid ( Ni-NTA ) -agarose resin ( Qiagen ) . Bound recombinant proteins were eluted in PBS containing 250 mM imidazole and 1 mM TCEP . MBP and MBP-LOVSoc were purified through affinity purification with amylose resin ( New England Biolabs ) and finally eluted by PBS buffer containing 25 mM maltose and 1 mM TCEP . The proteins were further purified by gel filtration on Superose 6 10/300 GL or Superdex 200 10/300 GL columns ( GE Healthcare ) . For MBP pulldown assay , 400 μl 1 mg/ml of MBP ( used as negative control ) or MBP-LOVSoc was immobilized on 400 μl slurry of the amylose resin ( New England Biolabs ) , and incubated with each 800 μg recombinant GB1-ORAI1-CT proteins in 1 ml PBS buffer containing 1 mM TCEP . The mixtures were divided into two groups: one group is constantly exposed to an external blue LED ( 470 nm , 40 μW/mm2 ) for 4 hr at 4°C , and then followed by ten-time washing with PBS to minimize nonspecific binding; whereas the other group was similarly treated except that all steps were performed in the dark . After extensive wash , the resin was finally mixed with 100 μl PBS and 4x SDS gel loading buffer , heated at 100°C for 10 min , and briefly centrifuged prior to gel electrophoresis . Samples were separated on 15% SDS-PAGE or 4–12% gradient NuPAGE . Bound proteins were visualized on SDS-PAGE after Coomassie Brilliant Blue R-250 staining . For immunoprecipitation , HEK293 cells co-transfected with FLAG-ORAI1 and mCh-LOVSoc were lysed with 1x RIPA buffer containing protease inhibitor cocktails . Extracts were incubated for 1 hr with anti-FLAG M2 affinity resin ( A2220 , Sigma ) and the mixture was thoroughly washed with 1x RIPA buffer , denatured and eluted with 1x SDS sample buffer ( 62 . 5 mM Tris-HCl , pH 6 . 8 , 2% SDS , 10% ( v/v ) glycerol , and 0 . 002% bromphenol blue ) . For light-stimulated groups , all steps were performed by exposing to an external blue LED ( 470 nm , 40 μW/mm2 ) . All starting materials were obtained from commercial supplies and used as received . Rare earth oxides Y2O3 ( 99 . 9% ) , Yb2O3 ( 99 . 9% ) , Tm2O3 ( 99 . 9% ) , trifluoroacetic acid ( 99% ) , 1-octadecene ( ODE ) ( >90% ) , oleic acid ( 90% ) , 1-ethyl-3- ( 3-dimethylaminopropyl ) carbodiimide hydrochloride ( EDC⋅HCl ) , N-hydroxysulfosuccinimide sodium salt ( sulfo-NHS ) and poly ( acrylic acid ) ( PAA , Mw 1 , 800 ) were purchased from Sigma-Aldrich . All other chemical reagents with analytical grade were used directly without further purification . The size and morphology of UCNPs were determined at 200 kV at a JEM-2010 low to high- resolution transmission electron mircroscope ( JEOL Inc . , Peabody , MA , USA ) . The UCNP samples were dispersed in hexane and dropped on the surface of a copper grid for TEM test . The upconversion luminescence emission spectra were recorded on a Fluoromax-3 spectrofluorometer ( Horiba Scientific , Irvine , CA , USA ) that was equipped with a power adjustable collimated CW 980 nm laser . All the photoluminescence studies were carried out at room temperature . The β-NaYF4:Yb , Tm core UCNPs were prepared using a modified two-step thermolysis method ( Mai et al . , 2006 ) . In the first step , the CF3COONa ( 2 mmol ) and required Ln ( CF3COO ) 3 ( 0 . 5 mmol in total ) precursors were mixed with oleic acid ( 5 mmol ) , oleyl amine ( 5 mmol ) and 1-octadecene ( 10 mmol ) in a two-neck reaction flask . The mol-percentage of Tm ( CF3COO ) 3 was fixed at 0 . 5% , Yb ( CF3COO ) 3 was employed in 80% , and Y ( CF3COO ) 3 was used of 19 . 5% . The slurry mixture was heated to 110°C in order to form a transparent solution . This was followed by 10 min of degassing to remove the oxygen and water . The flask was then heated to 300°C at a rate of 15°C per min under dry argon flow , and remained at 300°C for 30 min . The α-NaLnF4 intermediate UCNPs were acquired by cooling down the reaction solution to room temperature , followed by centrifugation with excessive ethanol . In the second step , the α-NaYF4:Yb , Tm UCNPs were re-dispersed in oleic acid ( 10 mmol ) and 1-octadecene ( 10 mmol ) along with CF3COONa in a two-neck flask . After degassing at 110°C for 10 min , the flask was heated to 325°C at a rate of 15°C per min under dry argon flow , and remained at 325°C for 30 min . The β-NaYF4:Yb , Tm UCNPs were then centrifugally separated from the cooled reaction media and suspended in 10 ml of hexane as the stock solution for further use . In the thermolysis reaction , as-synthesized β-NaYF4:Yb , Tm UCNPs served as crystallization seeds for the epitaxial growth of undoped β-NaYF4 shell . Typically , a stock solution of β-NaYF4:Yb , Tm UCNPs ( 5 ml , ca . 0 . 26 μmol/L core UCNPs ) was transferred into a two-neck flask and hexane was sequentially removed by heating . Then CF3COONa and Y ( CF3COO ) 3 ( 0 . 5 mmol ) were introduced as UCNP shell precursors with oleic acid ( 10 mmol ) and 1-octadecene ( 10 mmol ) . After 10 min of degassing at 110°C , the flask was heated to 325°C at a rate of 15°C/min under dry argon flow and was kept at 325°C for 30 min . The products were precipitated by adding ethanol to the cooled reaction flask . After centrifugal washing with hexane/ethanol , the core/shell UCNPs were re-dispersed in 10 ml of hexane for further use . The hydrophobic UCNPs were first treated by surface ligand exchange using a modified literature method ( Dong et al . , 2011 ) . Briefly , nitrosonium tetrafluoroborate ( NOBF4 , 0 . 20 g ) was dissolved in dimethylformamide ( DMF , 5 ml ) , and β-core/shell UCNPs in hexane stock solution ( 1 ml ) was added , followed by 4 ml hexane and 2 hr of stirring at room temperature . Then BF4- capped UCNPs were precipitated by adding isopropanol ( 5 ml ) , and purified by 2 cycles of centrifugal wash with DMF . Subsequently , all UCNPs precipitates were dispersed in poly ( acrylic acid ) /DMF ( PAA , Mw 1800 , 10 mg/ml , 5 ml ) solution to replace surface BF4- by PAA . After overnight incubation , the PAA coated β-NaYF4:Yb , Tm/NaYF4 UCNPs were purified by centrifugal wash with deionized ( DI ) water . The streptavidin and zwitterion ligands ( Muro et al . , 2010 ) were conjugated to UCNPs-PAA surface by EDC ( 1-Ethyl-3- ( 3-dimethylaminopropyl ) -carbodiimide ) coupling approach . Generally , 50 mg hydrophilic PAA-coated UCNPs in 5 ml DI water were activated by EDC ( 50 mg ) and NHS ( 10 mg ) to form succinimidyl ester . After stirring at room temperature for 2 hr , the nanoparticles were collected by centrifugation followed by washing with DI water . The generated nanoparticles were then re-dispersed into 5 ml DI water , followed by adding 150 µg streptavidin and the mixture was further stirred at room temperature for 4 hr . Next , 100 mg zwitterion ligand was introduced to the solution . After overnight stirring at room temperature , the UCNPs-Stv were purified by washing with DI water , centrifugation and dispersion in DI-water for further use . 5 ml LOV2 or MBP-LOVSoc proteins were concentrated to 0 . 5 ml at a concentration of 50–100 μM using centrifugal filter devices with a cutoff of 10 kDa . The UV-Vis spectra were recorded with a Shimadzu or Nanodrop 2000 spectrophotometer ( Thermo Scientific , Waltham , MA , USA ) . The absorbance was recorded before and after the introduction of UCNPs-Stv . 10 mg of UCNPs-Stv was added to make a final concentration of 20 μg/μl . The mixed solution was then transferred to a thin glass tube ( with a diameter comparable to the CW laser spot ) and subjected to 980 nm CW laser excitation ( 15 mW/mm2 ) for 1 min . The control sample was exposed to blue light ( 470 nm , 40 μW/mm2 ) for 1 min . After light stimulation , the absorbance was monitored every 30–300 s till the LOV2 domain fully returned to its dark state . 20 mg of UCNPs with different surface modifications were mixed with 100 mg KBr , and then grounded into fine powder in a mortar . A piece of pre-cut cardboard was placed on top of a stainless steel disk and the cutout hole was filled with the finely ground mixture . A second stainless steel disk was put on top and the sandwich disks were transferred onto the pistil in the hydraulic press to obtain a homogenous and transparent film . The samples were then inserted into the IR sample holder for analysis . Black background ( KBr film only ) was subtracted from the corresponding spectrum . The upconversion quantum efficiency ( QE ) is used to precisely measure the upconversion ability of the characterized materials , which is defined as the fraction of the absorbed photons that successfully employed to generate upconversion emission . The upconversion QE was calculated based on the following equation: QE=i*QY; where QY represents the quantum yield and i equals to 3 as Tm3+ excited state produces three-photon luminescence at 480 nm ( from 1D2 state to 3H6 state ) . The upconversion QY was first measured on a relative basis , using a known QY ( 3 . 2% ) sample of α-NaYF4:Yb , Er @CaF2 as a standard ( Punjabi et al . , 2014 ) . The following equation was used to calculate the QY:QYSample=QYref ( EsampleEref ) ( ArefAsamples ) where ( E ) is the integrated emission intensity at 480nm , ( A ) is the absorption at 980 nm . The upconversion QE of the 40-nm β-NaYF4:Yb , Tm@β-NaYF4 UCNPs in the blue region was determined to be 2 . 7% at the power density of 10 W/cm2 . HeLa cells were cotransfected with a total of 500 ng DNA ( 200 ng of pTriEX-mCh-LOVSoc , 200 ng of pcDNA3 . 1-mCh-ORAI1-StrepTag , 100 ng of pGP-CMV-GCaMP6S-CAAX or 100 ng NFAT11-460-GFP in opti-MEM ) as described above . 16 hr posttransfection and 2 hr prior to imaging , 20 mg of UCNPs-Stv PBS solution was introduced into the cell culture media . For imaging , Petri dish was mounted on a Leica TCS SP 2 confocal microscope equipped with a 63×oil objective . mCherry was excited at 590 nm and emission was detected from 610 to 670 nm . A 488-nm laser with minimum power was used to acquire GFP signals whilst a 590-nm laser was applied to acquire mCherry signals . All images were collected at a scanning rate of 400 Hz . 980 CW laser was introduced into the system with a power density of 15 mW/mm2 , and each irradiation takes 5–10 s . The relatively slow onset of Ca2+ influx and NFAT nuclear translocation provided us a time window to quickly capture the green signals without noticeably activating LOVSoc during image acquisition . This allows us to confidently apply NIR light to monitor Ca2+ influx and NFAT nuclear translocation . HeLa cells were transfected with NFAT-Luc with and without the Opto-CRAC construct LOVSoc , as indicated . 48 hr after transfection , 5 × 105 cells suspended in 200 µL DMEM with 1 μM PMA were mixed with 10 mg UCNPs-Stv , then implanted i . v . into BALB/c mice ( female; 4–8 weeks; injected position: upper thigh , as indicated in red circle; from Jackson Laboratory ) . The hairs on the back of the mice were shaved , whilst the hairs on the belly remained unshaved . The implanted regions were subject to 980 nm CW laser irradiation ( 50 mW/mm2 , 30 sec every 1 min for a total of 25 min ) , during anesthesia using ketamine/xylazine ( 100 mg/kg , 10 mg/kg , i . v . ) . Five hours later , the cells implanted area was injected with D-luciferin ( s . c . , 100 µL , 15 mg/ml in PBS ) and imaged 20 min later with an IVIS-100 in vivo imaging system ( 2-min exposure; binning = 8 ) . Luciferase luminescence was plotted as false color with rainbow-scale bar set as the same for all acquired images . For thermal imaging , BALB/c mice were immobilized and exposed to 50 mW/mm2 980 nm CW laser under the same condition as we carried out for the in vivo luciferase experiment . Images at two-minute intervals were taken by a thermal imaging camera ( FLIR Instruments ) . To obtain murine bone marrow-derived dendritic cells ( DCs ) , bone marrow cells were washed out of the femurs of adult mice in RPMI-1640 using a syringe and a 25-gauge needle and depleted of red blood cells . Bone marrow cells ( 5x105 cells/well ) in 6-well plate were cultured in RPMI-1640 containing 2 mM-L-glutamine , 100 IU/ml penicillin , 100 mg/ml streptomycin , 10% FCS , 50 μM β-ME，20 ng/ml GM-CSF and 200 IU/ml IL-4 for dendritic cell differentiation . Bone marrow cells were transduced with MSCV expressing viral vector pMIG-mCh-LOVSoc-IRES-ORAI1StrepTag on day 3 at MOI of 20 for 6 hr . Next , 75% of the media and non-adherent cells were removed and replaced with fresh culture medium . On day 5 , transduced DCs were gently dislodged and pulsed for 3 hr at 37°C with 2 μg/ml OVA257–264 peptide ( GenScript ) and 1 mg/ml UCNP-Stv nanoparticles . Cells were then washed to remove unattached peptide and nanoparticles . To generate OT-I CD8 T cells , spleens and lymph nodes ( LN ) of OT-1 Rag1-/- mice ( purchased from the Jackson Laboratory ) were pressed through a 70 μm cell strainer ( BD Falcon ) . Untouched CD8 T cells were sorted by using mouse CD8 T Cell Isolation Kit ( Miltenyi Biotec ) . 2x104 irradiated peptide loaded UCNPs-Stv/OVAp Opto-CRAC DCs were seeded in triplicates in 96-well U-bottom plates containing 5x104 purified OT-I CD8 T cells in a total volume of 200 μl and co-cultured for 5 days with or without NIR light stimulation for 16h ( 1 min pulse , 15 mW/mm2 ) . T cell proliferation was determined by labeling cultured cells with [3H] thymidine at a concentration of 1 μCi/μL for 16 hr and the radioactivity was measured using a liquid scintillation counter ( PerkinElmer ) . To detect DCs maturation and migration , NIR-stimulated or unstimulated UCNPs-Stv Opto-CRAC DCs were stained with FITC-CD11c , PE-MHC-I , APC-MHC-II , PE-CD86 and PE-CCR7 and then subjected to flow cytometry analysis 3 days post-transduction . For intracellular IFN-γ staining , OT-I CD8 T cells were incubated with UCNPs-Stv/OVAp Opto-CRAC DCs for 6 hr at 37°C in the presence of GolgiStop ( monensin ) ( BD Pharmingen ) . Cells were then stained with surface marker using APC-CD8a antibody for 15 min on ice and permeabilized using cytofix/cytoperm ( BD Biosciences ) for 30 min on ice . Permeabilized cells were resuspended in BD Perm/Wash buffer ( BD Biosciences ) and stained with PE-anti-IFN-γ antibody for 20 min . Samples were run on a BD LSRII Flow Cytometer and analyzed by BD FACSDiva software . B16-OVA is an OVA-transfected clone derived from the murine melanoma cell line B16 ( ref . 35 ) . B16-OVA cells were cultured and maintained in Dulbecco's modified Eagle medium ( HyClone ) supplemented with 10% heat-inactivated fetal bovine serum ( FBS ) , 100 IU/ml penicillin , 100 mg/ml streptomycin under 37°C in 5% CO2 . B16-OVA cells ( 1x106 ) were injected s . c . into the flank region or i . v . via tail vein of Rag1-/- mice ( purchased from the Jackson Laboratory ) ( Overwijk and Restifo , 2001 ) . 3 days later , mice were injected i . v . with 2×105 Opto-CRAC DCs treated with UCNPs-Stv and the surrogate tumor antigen OVA257-264 . 1 . 5×106 OT-I T cells labeled with CellTrace far red CFSE were i . v . injected into tumor-bearing mice . Briefly , cells were incubated at 1x106 cells/ml in CFSE at a final concentration of 1 μM for 20 min at room temperature . The labeling reaction was stopped by adding the same volume of FBS . Recipient Rag1-/- mice were subjected to the excitation of NIR laser ( 8 hr per day , 0 . 5–1 min ON/OFF pulse , 30 mW/mm2 ) or shielded from NIR ( control group ) for 6 days to stimulate Opto-CRAC DC maturation , with the initial two days concentrating more on areas nearby the draining lymph nodes of restricted mice . For in vivo T cell proliferation , spleen and draining popliteal and inguinal LNs were harvested and injected with collagenase D ( 1 mg/ml; Boehringer-Mannheim , Mannheim , Germany ) in RPMI and 10% FBS for 20 min at 37°C . Digested LN or spleen were filtered through a stainless-steel sieve , and the cell suspension was washed twice in PBS and 5% FBS . CFSE-labeled OT-I CD8 T cells were analyzed by flow cytometry as described above . Tumor growth was measured at indicated time points using calipers shown in growth curve using the equation of V = Lx W2/2 . Lungs were isolated and tumor foci of lung melanomas were counted from tumor-bearing mice shielded or exposed to NIR pulse from day 3–9 after B16-OVA tumor cell injection . Hela cells and UCNPs were subcutaneously implanted into upper thigh of BALB/c mice , followed by 980 nm CW laser irradiation ( 50 mW/mm2 , 30 sec every 1 min for a total of 25 min ) , during anesthesia using ketamine/xylazine ( 100 mg/kg , 10 mg/kg , i . v . ) . Two weeks after subcutaneous implantation , mice were sacrificed and tissue samples under skin at the injection position were collected . Routine Hematoxylin and Eosin staining ( H&E ) was performed by University of Massachusetts Medical School morphology core . The fluorescence images were analyzed with the NIS-Elements imaging software ( Nikon ) or the Image J package ( NIH ) with the intensities plotted using the GraphPad Prism 5 graphing and statistical software . The mean lifetime of fluorescence signal change was calculated with a single exponential decay equation F ( t ) =F ( 0 ) *e^ ( -t/τ ) . Quantitative data are expressed as the mean and standard deviation of the mean ( s . e . m . ) unless otherwise noted . Paired Student’s t-test was used throughout to determine statistical significance . *P<0 . 05; **P<0 . 01; ***P<0 . 001 , when compared to control or WT . | Optogenetics is a technique that has been used to study nerve cells for several years . It involves genetically engineering these cells to produce proteins from light-sensitive bacteria , and results in nerve cells that will either send , or stop sending , nerve impulses when they are exposed to a particular color of light . Neuroscientists have learned a lot about brain circuits using the technique , and now researchers in many other fields are giving it a try . There are , however , several challenges to using optogenetics in other types of cells . Nerve cells create a tiny electrical impulses when they are activated , which helps them quickly transmit messages . But other types of cells use more diverse means to communicate and transmit signals . This means that optogenetics techniques must be adapted . Additionally , many cells are located deep in the body and so getting the light to them can be difficult . He , Zhang et al . have now developed an optogenetic system ( termed “Opto-CRAC” ) that can control immune cells buried deep in tissue . The action of immune cells can be tuned by controlling the flow of calcium ions through gate-like proteins in their membranes . He , Zhang et al . genetically engineered immune cells so that a calcium gate-controlling protein became light sensitive . When the cells were exposed to a blue light the calcium ion gates opened . When the light was turned off , the gates closed . More intense light caused more calcium to enter into the cells . Further experiments then revealed that exposing these engineered immune cells to blue light in the laboratory could trigger an immune response . The next obstacle was getting light to immune cells in a live animal . So , He , Zhang et al . used specific nanoparticles that have been shown to help transmit light deep within tissue . In these experiments , mice were injected with the light-sensitive immune cells and the nanoparticles . Then , a near-infrared laser beam that can transmit into the tissues was pointed at the mice . This caused calcium channels to open in the engineered cells deep in the mice . Finally , further experiments were used to show that this light-based stimulation could boost an immune response to aid the killing of cancer cells . Other scientists will likely use the technique to help them study immune , heart , and other types of cells that use calcium to communicate . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology",
"tools",
"and",
"resources"
] | 2015 | Near-infrared photoactivatable control of Ca2+ signaling and optogenetic immunomodulation |
Gamma activity in the subthalamic nucleus ( STN ) is widely viewed as a pro-kinetic rhythm . Here we test the hypothesis that rather than being specifically linked to movement execution , gamma activity reflects dynamic processing in this nucleus . We investigated the role of gamma during fast stopping and recorded scalp electroencephalogram and local field potentials from deep brain stimulation electrodes in 9 Parkinson’s disease patients . Patients interrupted finger tapping ( paced by a metronome ) in response to a stop-signal sound , which was timed such that successful stopping would occur only in ~50% of all trials . STN gamma ( 60–90 Hz ) increased most strongly when the tap was successfully stopped , whereas phase-based connectivity between the contralateral STN and motor cortex decreased . Beta or theta power seemed less directly related to stopping . In summary , STN gamma activity may support flexible motor control as it did not only increase during movement execution but also during rapid action-stopping .
Previous studies have described a neuronal stopping network involving prefrontal and supplementary motor cortical regions , as well as the subthalamic nucleus ( STN ) ( Aron et al . , 2014; Jahanshahi et al . , 2015; Rae et al . , 2015 ) . The STN is well-positioned to cancel actions as it receives cortical input via the hyperdirect pathway and can inhibit the thalamus and brainstem via the basal ganglia output nuclei as well as the striatum via the globus pallidus externus ( GPe ) ( Mink , 1996; Wei and Wang , 2016 ) . In spite of recent advances in understanding functional and effective connectivity within the stopping network using fMRI ( Rae et al . , 2015 , 2016; Xu et al . , 2016 ) , the fast temporal dynamics of population activity accompanying the stopping process are not entirely clear . When rats attempted to cancel an action , increased STN firing activity was found irrespective of whether cancellation was successful or not ( Schmidt et al . , 2013 ) , but more recently , micro-electrode recordings in the human STN revealed two distinct subpopulations that selectively increased firing rate either during successful response inhibition or during motor execution ( Bastin et al . , 2014; Benis et al . , 2016 ) . Also in the GPe a subpopulation termed arkypallidal cells , which seem to receive input not only from the striatum but also from the STN ( Nevado-Holgado et al . , 2014 ) , has specifically been linked to action cancellation ( Mallet et al . , 2016 ) . It is unclear , though , how different populations within the basal ganglia are activated in a selective and flexible way . Oscillations , particularly in the gamma band ( >30 Hz ) , have been proposed to be a key mechanism for coordinating spatially separate but functionally related assemblies ( Bosman et al . , 2012; Fries , 2015; Nikolić et al . , 2013; Schoffelen et al . , 2005 ) . We hypothesized that gamma activity may thus also facilitate coordinated activation of task-relevant subpopulations for efficient movement cancelation . A local field potential study in Parkinson’s disease patients , however , has shown increased 55–75 Hz gamma activity when patients failed to stop ( Alegre et al . , 2013 ) , which is in line with the prevailing view that gamma activity is pro-kinetic ( Cassidy et al . , 2002; Fogelson et al . , 2005; Litvak et al . , 2012 ) or related to response vigour ( Jenkinson et al . , 2013 ) . Beta activity , instead , is widely viewed as a marker of broad motor suppression within the STN ( Wessel et al . , 2016a ) as well as cortex ( Swann et al . , 2012 ) . High STN beta activity for example was linked to elongated response times during incongruent trials in a Stroop task ( Brittain et al . , 2012 ) and to stronger suppression of cortico-spinal excitability during speech inhibition ( Wessel et al . , 2016a ) . However , as movements are known to coincide with decreasing beta and increasing gamma activity ( Joundi et al . , 2012; Lalo et al . , 2008 ) , comparisons between executed and withheld movements might reflect the lack of movement rather than the stopping process per se . Ideally , stopping would be recorded as a continuous variable that measures how fast an ongoing movement is terminated instead of whether an action has been started at all . Motor inhibition has traditionally been investigated with stop signal or Go/NoGo tasks , in which movements are triggered by cues ( Huster et al . , 2013; Swick et al . , 2011 ) . In the stop signal paradigm , subjects press a button in response to a go cue and in some trials a stop signal instructs them to withhold the movement . Go/NoGo tasks instead rely on a large fraction of go trials to catch participants out on rare trials , in which a NoGo cue signals them to withhold the pre-potent motor response . Both tasks require participants to decide whether to stay or move but not to interrupt an ongoing action . Successful stopping is achieved in these tasks by successfully delaying or canceling action initiation rather than terminating an action that is already ongoing . Our aim was to extend existing studies by investigating rhythmic movements that can be interrupted halfway and are not directly preceded by go-cues but are self-initiated . Patients were asked to tap rhythmically to a metronome . Under these circumstances , subjects anticipate the metronome instead of reacting only after each sound , and so movements can be considered self-initiated . They were instructed to stop upon hearing a different cue that was timed such that they were able to stop only in approximately half of their attempts ( Figure 1 ) . The neural response to the stop signal was not intermixed with a foregoing response to a go cue as the last metronome sound was delivered about 700 ms prior to the stop signal . 10 . 7554/eLife . 23947 . 003Figure 1 . Behavioural task and representative data . ( A ) Schematic of the task in the STOP condition ( top row ) and in the control condition ( 2nd row ) . In the STOP condition participants had to tap ( =ellipses ) to a metronome ( =rectangles ) and stop after 5–9 taps . The red ellipse denotes a tap that was unsuccessfully stopped . ( B ) Pressure sensor , FDI muscle activity and goniometer data from one representative patient . Black lines are trials where the tapping movement after the stop signal was successfully stopped , red lines are trials where stopping failed . The markers around 0 ms represent the temporal offset between the last regular sound and the tap ( o = successful stop trials , x = failed stop trials ) . The markers at 680 ms show the time of the stop signal , which was always triggered relative to the last regular tap that was registered by the pressure sensor at 0 ms . Note that the black and red trajectories overlap , which shows that stopping performance did not depend on the preceding movement trajectory . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 003 The delay of the stop signal was set by the experimenter after a training period at the start and then kept constant for the rest of the experiment . It was delivered relative to the tap instead of the metronome sound to keep movement variability to a minimum and to prevent the strategy of delaying the tap relative to the metronome sound . This , in combination with the instruction to synchronize accurately to the metronome , provided trains of self-initiated actions that were well-matched across trials . The task was also well-suited to investigate endogenous fluctuations in readiness to stop . We analyzed STN local field potentials ( LFP ) and scalp electroencephalography ( EEG ) activity recorded in this task from nine Parkinson’s disease patients , who underwent deep brain stimulation surgery . To differentiate volitional motor inhibition from salience detection , six of them were recorded in an additional control condition with identical auditory cues but different instructions . Their task in this condition was to finish the tapping sequence with two more taps upon hearing the stop signal instead of attempting to stop ( Figure 1 ) .
The mean stop signal delay time with a 55% ± ( SD ) 10% successful stopping rate was 707 ± 49 ms ( range = 620–760 ms ) . The mean interval between the preceding tap and the unsuccessfully inhibited tap in trials where stopping failed was 864 ± 36 ms , and was significantly shorter than the 900 ms interval dictated by the metronome ( Wilcoxon signed-rank test , p=0 . 004 ) . In these trials , patients would have still had on average 156 ms to stop . Movement trajectories preceding successfully or unsuccessfully inhibited stops were overlapping ( Figure 1 , trajectories measured by a pressure sensor and goniometer ) . Thus any electrophysiological differences in this window are unlikely related to movement differences per se . Stopping performance was quantified as movement extent , which was the extent of downward movement after the stop signal relative to the amplitude of the preceding upward movement . 0% movement extent thus refers to a full stop . 50% describes a movement that was interrupted halfway and 100% would correspond to a full tap , i . e . failed stopping . Correlations between movement extent and various properties of the last regular tap were computed for each patient and then subjected to t-tests to assess if the Fisher’s z-transformed correlation coefficients significantly differed from zero on the group-level . In 7 of 9 subjects , movement extent correlated with the tap-to-sound offset , which indicates that stopping performance was worse when the foregoing tap was relatively late in a trial corresponding to previous results ( Fischer et al . , 2016 ) . However , none of the tested variables were associated with successful stopping after FDR-correction of the resulting p-values ( see Table 1 ) . 10 . 7554/eLife . 23947 . 004Table 1 . Correlations between movement parameters of the last regular tap and the movement extent after the stop signal ( mean ± SD ) . In 7 of 9 subjects , movement extent correlated with the soundOffset ( =tap-to-sound offset; negative values represent taps that occurred before the sound ) . But none of the p-values resulting from one-sample t-tests of the Fisher’s z-transformed intra-individual correlation coefficients of the nine subjects survived FDR-correction . downTime = duration of finger contact with the pressure sensor , maxPrs = peak pressure during the tap , tapNr = number of taps preceding delivery of the stop signal , peakVelDown=peak velocity of the downward movement of the previous tap , upMvmt = amount of up-movement , peakVelUp=peak velocity of the upward movement . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 004VariableRho±SDp-valueFDR-corrected p-valuesoundOffset0 . 29 ± 0 . 180 . 0200 . 137downTime−0 . 10 ± 0 . 190 . 1740 . 407maxPres−0 . 04 ± 0 . 230 . 4600 . 644tapNr−0 . 13 ± 0 . 200 . 0610 . 215peakVelDown−0 . 05 ± 0 . 230 . 3770 . 644upMvmt0 . 00 ± 0 . 230 . 9520 . 952peakVelUp0 . 04 ± 0 . 290 . 8100 . 945 Previous research suggests that a surprising sound alone already elicits motor slowing of verbal reports ( Wessel and Aron , 2013 ) . We thus checked if the tap performed after the salient ‘stop signal’ ( which served as ‘continue signal’ in the control condition ) was delayed or slowed down in the control condition when stopping was not even required . The median intertap interval directly preceding the stop signal ( median ITI = 893 ms ) did not differ significantly from the one directly after the stop signal ( median ITI = 889 ms , Wilcoxon signed-rank test p=1 . 0 ) . We tested for rapid LFP and EEG power changes between the stop cue and the average timing of the tap when inhibition failed , which was on average 156 ms after the cue and puts a limit on the window within which successful movement inhibition had to occur . The STN contralateral to the tapping hand responded to the stop signal with a 60–90 Hz gamma power increase when compared to activity from the tap before ( Figure 2 . 1 shows the reference data from the tap before aligned to where the stop signal would have occurred if it would have been presented one tap earlier; Figure 2 . 2 shows the response to the stop signal; Figure 3A shows the contrast between the two ) . Importantly , this gamma increase was significantly and consistently higher during successful movement inhibition ( Figure 2 . 3 + 2 . 4 , and 3B ) . The effect size of this difference was very large ( 60–90 Hz power difference between successful-failed stops: Cohen’s d meanwinOfInt = 1 . 2 , maxwinOfInt = 2 . 6 ) . Note that during regular tapping we observed the typical pattern of movement-related gamma power increase and beta power decrease ( Figure 2 . 1 and Figure 3—figure supplement 1 ) . Gamma power thus increased during both movement execution and movement inhibition . The movement-related peak was broader and weaker than the stop-related increase that peaked around 70 Hz ( Figure 3—figure supplement 2 ) . 10 . 7554/eLife . 23947 . 005Figure 2 . Contralateral STN power changes around the stop signal . T-scores calculated over all patients ( n = 9 , normalized by the average power during regular tapping ) for ( 1 ) the last regular tap aligned to the timepoint when the stop signal would have occurred if it would have been delivered one tap earlier ( vertical dashed line ) . The black line shows the tapping movement measured by the goniometer . The downward movement was accompanied by a beta decrease and gamma increase as expected . The following three columns show changes in response to the stop signal ( vertical dashed line ) ( 2 ) irrespective of whether stopping was successful or not , ( 3 ) during successful stops only , and ( 4 ) during failed stops only . Note that when a stop signal was present and especially when stopping was successful ( column 3 ) , gamma increased strongly . Differences between 2–1 and 3–4 are contrasted in Figure 3 . The tapping trajectory of failed stops does not reach the bottom line even though the finger touched the table because trajectories were normalized to the minimum of all four trajectories , which occured with the last regular tap , where the spring was extended more vigorously than during attempted inhibition . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 00510 . 7554/eLife . 23947 . 006Figure 3 . Contrasts between power changes following the stop signal . ( A ) T-scores calculated over all patients of the contrast between power aligned to the stop signal ( vertical dashed line ) averaged across all trials irrespective of stopping performance ( Figure 2 . 2 ) and the regular tap made before ( Figure 2 . 1 , aligned to where the stop signal would have occurred if it would have been presented one tap earlier ) . Red clusters denote that power significantly increased in response to the stop signal . ( B ) T-scores of power differences between successful and failed stops . Red clusters denote that power was significantly higher if participants successfully inhibited the upcoming tap ( Figure 2 . 3–2 . 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 00610 . 7554/eLife . 23947 . 007Figure 3—source data 1 . MATLAB data file containing source data related to Figure 3 . Data matrices ( subject * frequency * time ) for individual channels are stored at the respective fields in the structure data . Figure 3A . below40Hz for frequencies below 40 Hz and in data . Figure 3A . above60Hz for frequencies in the gamma range . The frequencies for each column are denoted in the field ‘freqs’ and the time in seconds in the field ‘time_in_sec’ . The field ‘stoppingWin’ provides the critical time window of interest between the stop signal and timing of the unsuccessfully stopped tap . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 00710 . 7554/eLife . 23947 . 008Figure 3—figure supplement 1 . Power time-course during regular tapping averaged across all patients . Spectra were tested for significant power modulation locked to the tap in a 0:156 ms window ( matched in size to the test-window for the main Figure 3 ) after tap onset ( = dashed line ) . As power was normalized by the average power of one full tap cycle including movement , the effects were relatively small and would not survive multiple-comparison correction over the full time-window . However , movement-related beta decrease and gamma increase relative to a pre-movement baseline has been repeatedly reported before ( Tan et al . , 2013; Androulidakis et al . , 2007 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 00810 . 7554/eLife . 23947 . 009Figure 3—figure supplement 2 . Peak frequencies of movement- and stop-related power changes . Power for individual frequencies between 3–120 Hz was averaged over time between the stop signal and the mean time of failed taps ( 156 ms later ) for stop-related changes ( right and middle plot ) . Bold thick lines show the average across subjects and coloured lines show individual subjects . For movement-related power changes ( left plot ) , the window was aligned to the time of each tap onset as shown in Figure 3—figure supplement 1 . The movement-related gamma increase was broader and weaker than the stop-related increase . Dotted lines show the peak frequency in the gamma range ( >40 Hz ) . The middle plot shows the difference between power in response to the stop signal irrespective of stopping outcome relative to the last regular tap ( corresponding to Figure 3A ) . The right plot shows the difference between successful vs . failed stops ( corresponding to Figure 3B ) . Power differences in frequencies lower than 50 Hz were highly variable , whereas power was consistently increased around 70 Hz in response to the stop signal , especially during successful stopping . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 00910 . 7554/eLife . 23947 . 010Figure 3—figure supplement 3 . Power changes following the stop signal when only successful stop trials are considered ( averaged across all patients ) . T-scores of the contrast between power of all successful stop trials aligned to the stop signal ( vertical dashed line ) and the regular tap done before ( aligned to where the stop signal would have occurred if it would have been presented one tap earlier ) . Significant clusters are similar as in the main Figure 3A . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 010 Cortical EEGs recorded a low-frequency increase in response to the stop signal in all channels ( Figure 3A ) , which was – in contrast to STN gamma activity – not significantly higher during successful stopping ( Figure 3B ) . Only 8–30 Hz power over contralateral C3/C4 , Cz and Fz was significantly higher when stopping was successful . However , there was no overall power increase following the stop signal in the 8–30 Hz band in these channels when compared to the tap before ( Figure 3A ) , not even when only successful stop trials were considered ( Figure 3—figure supplement 3 ) . In previous studies , such increase was observed when an action had to be withheld before being initiated ( Kühn et al . , 2004; Swann et al . , 2009 ) . To exclude that the STN gamma increase merely reflects processing of the salient stop cue , six patients additionally performed a control condition before and after the main stopping task ( Figure 1 ) . The stimulus sequence of the control condition was identical to the main condition and the instruction differed only in that patients had to finish the tapping sequence with two more taps upon hearing the stop signal instead of inhibiting the tap immediately . Importantly , no gamma increase was observed in this control condition , even though the difference between successful and unsuccessful stops was still significant despite the reduced sample size of 6 patients ( Figure 4C ) . 10 . 7554/eLife . 23947 . 011Figure 4 . Power time course in the STN averaged across patients relative to the stop signal . ( A ) 60–90 Hz gamma power was significantly higher when stopping was successful ( left , blue line ) . The first grey dashed line denotes the average time of the last regular tap . The grey dashed line after the stop signal ( red dashed line ) denotes the average time of all failed taps . This difference was consistent across patients ( middle panel; bold black line denotes the average difference between successful and failed trials with the individual differencess in grey; n = 9 ) . Filled blue areas show cluster-based corrected significant differences . ( B ) This difference was not present in the right STN ( n = 9; ipsilateral in 6 ) . ( C ) Gamma in contralateral STN did not increase when stopping was not attempted ( black line = control condition , the plot in the middle column shows individual power time courses in the control condition; n = 6 ) . Filled blue areas show cluster-based corrected significant differences between successful and unsuccessful stopping . The yellow filled area indicated by the purple arrow in the leftmost plot shows where power from successful stopping significantly differed from the control condition if uncorrected for multiple comparisons . ( D ) The 3–5 Hz increase in Cz ( n = 6 ) was similar irrespective of whether stopping was successful ( blue ) , unsuccessful ( red ) or whether it was not even attempted ( grey ) . Shaded areas denote standard errors of the mean . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 01110 . 7554/eLife . 23947 . 012Figure 4—source data 1 . MATLAB data file containing source data related to Figure 4 . The average power of successful and failed stopping attempts ( subject * time ) are stored in the fields ‘successStop’ and ‘failedStop’ , respectively . The field ‘time_in_sec’ provides the time in seconds for each data point . The field ‘stoppingWin’ provides the critical time window of interest between the stop signal and timing of the unsuccessfully stopped tap . For Figure 4C+D the field ‘controlCond’ contains the average power of the control condition , where stopping was not attempted . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 01210 . 7554/eLife . 23947 . 013Figure 4—figure supplement 1 . Scatter plot of correlations between movement extent ( x-axis ) and 60–90 Hz gamma relative to baseline ( y-axis ) . Subplots show individual participants . For each subject , gamma power yielding the maximum correlation ( detected anywhere between 60–90 Hz and 0:156 ms after the stop signal , considering that optimal frequencies and time points may differ across subjects ) is shown . The number of points can differ within each patient if an electrode was more prone to artefacts and thus more trials were excluded . Plot titles denote Spearman’s rho followed by its 95% bootstrapped confidence interval . In all but one subject correlations were significant in the contralateral STN . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 01310 . 7554/eLife . 23947 . 014Figure 4—figure supplement 2 . Power time course relative to the stop signal in patients who stopped fully in at least five trials . 60–90 Hz gamma power was highest during full stops ( defined as < 10% movement extent ) , it increased halfway when the tap was interrupted halfway and it did not increase when stopping failed . Thin blue lines denote the time course of full stops from the four individual subject . The FDI EMG activity to the left shows that the muscle activity pattern was reversed , i . e . EMG activity was absent when gamma increased quickly , which demonstrates that gamma did not only increase when the tap was interrupted halfway . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 01410 . 7554/eLife . 23947 . 015Figure 4—figure supplement 3 . 3–5 Hz power increase in contralateral and ipsilateral M1 , Fz and Pz . In contralateral and ispilateral M1 , as well as Fz , the stop signal-related theta increase seemed to be smaller in the control condition when stopping was not attempted ( grey line , n = 6 ) but this difference was not significant . Note also that the blue and red shaded areas ( succesful and failed stops ) , denoting standard errors of the mean , were highly overlapping . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 015 It has been suggested that specifically the right STN may mediate stopping ( Aron and Poldrack , 2006 ) . To evaluate the role of the right STN alone , individual gamma differences between successful and unsuccessful stops of all right STNs are displayed in Figure 4B , showing no significant increase . Three right-handed patients performed the task with the left hand and thus in those the right STN was the contralateral one . However , in the remaining six , the right STN was the ipsilateral STN , and thus the lack of significant right STN gamma increase indicates that the gamma increase was specific to the contralateral STN . To further corroborate the functional significance of our finding we also tested whether the average gamma increase peaked earlier during successful stops than during failed stops . Indeed , the average gamma peak of successful stops at 106 ± ( SD ) 59 ms occurred earlier than the average unsuccessful tap ( at 156 ± 50 ms ) , whereas the average gamma peak of failed stops occurred later ( at 179 ± 84 ms ) . These gamma peak timings significantly differed from each other ( t ( 8 ) =-2 . 9 , p=0 . 019 , CIdiff=[−131 , –16 ms] , Cohen’s d = 1 . 0 ) . We also examined within-subject correlations between movement extent ( i . e . inhibition failure ) and gamma within the stopping window ( 0–156 ms ) after the stop signal ( see Figure 4—figure supplement 1 ) . This was significant in 8 of 9 patients ( uncorrected tests; P3’s confidence intervals were borderline significant , Spearman’s rho p=0 . 049 ) when gamma was taken from the contralateral STN , meaning that in all but one patient we found that when gamma was higher , movement extent was less and stopping was more successful . In contralateral C3/C4 and in ipsilateral STN such a relationship was present only in three patients , and in ipsilateral motor cortex only in one patient , further indicating specificity to the correlation with contralateral STN activity . Note though that correlations might be harder to detect with EEG data due to the reduced signal-to-noise ratio in comparison with LFPs . To see if the gamma increase was highest specifically during full stops , we classified the movement after the stop signal into full stops ( <10% downward movement ) , intermediate stops ( >10% but pressure sensor was not touched ) and failed stops ( all trials where the pressure sensor was touched ) . Only four patients made five or more full stops ( mean number of full stops = 9 . 5 ) , so formal statistics were not applied . Still , in full stop trials , gamma increased most strongly . It increased moderately for intermediate stops and remained flat for failed stops ( Figure 4—figure supplement 2 ) . As expected , activity recorded from the first dorsal interosseous muscle of the tapping hand ( presented to the right in Figure 4—figure supplement 2 ) suggests an inverse relationship to the gamma increase . Finally , we examined if the cortical 3–5 Hz power increase , which was clearly present in the stop condition ( Figure 3A ) , was also present in the control condition when movement inhibition was not even attempted . The grey power trajectory representing the control condition shows a very similar peak in Cz ( Figure 4D , n = 6 ) . Significance testing within the crucial reaction time window ( ranging from the stop signal to the average time of the failed tap , 156 ms later ) resulted in no significant differences between the control condition and either the power increase during failed or successful stopping . The direct comparison between failed or successful stops was not significant either . Also a peak-extraction analysis failed to detect a difference between low-frequency peaks ( Cz successful stops vs . control: t ( 5 ) =0 . 2 , p=0 . 848 , CIdiff=[-−205 . 1 , 240 . 2%]; failed stops vs . control: t ( 5 ) =0 . 5 , p=0 . 641 , CIdiff=[−149 . 4 , 220 . 8%] ) . The 3–5 Hz increase only seemed to be reduced in the control condition in Fz and both M1 ( Figure 4—figure supplement 3 ) , however this was also not significant . In a next step , we computed intersite phase clustering ( ISPC ) values between filtered oscillations in the EEG recordings and the LFP signal from the STN contralateral to the tapping hand . To get an estimate of the temporal development , we subdivided a −350:160 ms time window around the stop signal into equal bins in which ISPC was computed for each trial and then averaged over trials ( see Materials and methods ) . ISPC describes whether phase differences between two sites are randomly distributed ( small ISPC → low connectivity ) or clustered ( high ISPC → high connectivity ) and was obtained by taking the length of the mean vector of all phase differences from all time points within one bin . ISPC of 60–75 Hz gamma between the contralateral motor cortex and the contralateral STN decreased strongly and significantly in response to the stop signal relative to the average of the −350:0 ms window preceding the stop signal ( Figure 5 ) . We also observed an increase of 6–12 Hz ISPC to all cortical channels . 10 . 7554/eLife . 23947 . 016Figure 5 . Connectivity changes following the stop signal . Intersite phase clustering ( ISPC ) values are normalized by a −350:0 ms baseline preceding the stop signal . The dashed line denotes the time of the stop signal . Gamma ISPC between contralateral STN and contralateral C3/C4 decreased significantly between 60–80 Hz ( encircled in red ) , whereas ISPC in low frequencies between STN and cortical electrodes increased . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 01610 . 7554/eLife . 23947 . 017Figure 5—source data 1 . MATLAB data file containing source data related to Figure 5 . Data matrices ( subject * frequency * time ) for individual channels are stored at the respective fields in the structure data . Figure 5 . below40Hz for frequencies below 40 Hz and in data . Figure 5 . above60Hz for frequencies in the gamma range . The frequencies for each column are denoted in the field ‘freqs’ and the time in seconds in the field ‘time_in_sec’ . The field ‘stoppingWin’ provides the critical time window of interest between the stop signal and timing of the unsuccessfully stopped tap . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 017 Finally we assessed if gamma power was already tonically elevated prior to the stop signal , before participants knew they had to stop . We tested for significant differences within a 350 ms window before the stop signal . If the upcoming tap was inhibited more successfully , STN gamma power was already higher prior to the stop signal ( Figure 6 ) . 10 . 7554/eLife . 23947 . 018Figure 6 . Power differences preceding the stop signal averaged across all patients . Around 150 ms before the stop signal ( at 0 ms ) gamma activity was significantly higher in the STN if stopping was successful . Beta power in ipsilateral C3/C4 was also increased prior to successful stops . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 01810 . 7554/eLife . 23947 . 019Figure 6—source data 1 . MATLAB data file containing source data related to Figure 6 . Data matrices ( subject * frequency * time ) for individual channels are stored at the respective fields in the structure data . Figure 6 . below40Hz for frequencies below 40 Hz and in data . Figure 6 . above60Hz for frequencies in the gamma range . The frequencies for each column are denoted in the field ‘freqs’ and the time in seconds in the field ‘time_in_sec’ . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 01910 . 7554/eLife . 23947 . 020Figure 6—figure supplement 1 . Power differences preceding the stop signal with the data aligned to the last regular tap before stop signal delivery ( averaged across all patients ) . This figure differs slighlty from the main Figure 6 as the delay between the tap and stop signal differed across patients in spite of being the same across trials in each subject . With the alignment to the last regular tap the beta difference is also significant in contralateral C3/C4 . No such difference was present in the STN . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 020 20–30 Hz beta power over C3/C4 ipsilateral to the tapping hand also was significantly higher preceding successful stops . If the data were re-aligned to the last regular tap instead of the stop signal , a second significant cluster at 20–30 Hz over C3/C4 contralateral to the tapping hand was found , in line with previous reports ( Fischer et al . , 2016 ) ( see Figure 6—figure supplement 1 ) .
We found that when finger tapping had to be stopped abruptly , the stop signal elicited a fast increase in 60–90 Hz gamma activity in the contralateral STN and a pronounced theta increase in cortex . However , only the former was significantly higher when stopping was successful . The gamma increase occurred within 156 ms , which was the brief time window between the stop signal and the average failed tap . In a control condition , in which participants were presented with the same stop signal while tapping , but stopping was not attempted , only cortical theta but not STN gamma power increased . This shows that STN gamma activity does not only reflect pro-kinetic activity as previously suggested ( Litvak et al . , 2012 ) nor does it merely reflect processing of the salient stop signal . The alternative hypothesis that stopping of the tapping movement itself involved an active movement seems unlikely on two grounds . First , the gamma increase was less if the tap was terminated mid-flight rather than before the downward finger movement was started . Second , gamma connectivity between the STN and C3/C4 also sharply decreased directly after the stop cue , which differs from the movement-related increase usually observed ( Litvak et al . , 2012 ) and may indicate disengagement from the obsolete motor plan . Two previous studies have reported a different relationship between gamma and stopping success to the one that we have found ( Alegre et al . , 2013; Ray et al . , 2012 ) . Ray and colleagues ( 2012 ) reported a gamma increase in response to the stop signal as we do ( see Ray et al . , 2012: Figure 4b ) but did not detect significantly higher gamma during successful stops . This discrepancy may result from extensive temporal smoothing ( their sliding window was 333 ms long ) , and from the pre-selection of a window of interest between 200–400 ms after the stop signal , which also would have failed to detect the gamma difference in our data occurring right after the cue . If we apply the same temporal smoothing to our data , the power trajectories of successful and failed stops would look very similar ( data not shown ) as extensive smoothing flattens the brief gamma increase , such that part of it appears before the stop signal . The late gamma increase during failed stopping , which would be too late to affect the stopping outcome , would remain as a prominent difference about 300 ms after the stop signal . In Alegre et al . ’s study , which differed methodologically in using a visual stop signal , one conclusion was that successful inhibition was associated with a bilateral gamma power decrease . The precise time-frequency decomposition parameters used in that study are unclear and so we have not re-analysed our data in the same way . However , as their window of interest was relatively long ( 0–0 . 4 s ) , they might have predominantly captured the pro-kinetic gamma component that is relatively reduced when a motor response is withheld . Even so , similar as in our study , gamma appeared to increase briefly when aligned to the stop signal in patients on medication during both successful and failed stopping attempts ( Alegre: Figure 5 ) and a drop in STN-M1 coherence during successful inhibition also was found ( Alegre: Figure 7 ) . In addition , when comparing results between studies it is important to acknowledge that differences in disease phenotype , medication history , electrode models and in the precision of the targeting achieved are factors that may also contribute to variability in study findings . The normalisation of LFP measures will have only partially mitigated this variability . Periods of high gamma activity in the STN have been reported to coincide with an overall increase in firing rate and phase-locking of spikes to the gamma cycle peak ( Pogosyan et al . , 2006; Trottenberg et al . , 2006 ) . From the LFP we cannot infer changes in firing rate , but it suggests that the number of neurons or inputs to these neurons synchronizing at 60–80 Hz was coupled with stopping outcome , and that increased synchronisation occurred early enough to influence such outcome . After observing that the strength of gamma synchronization in the STN or its coherence with C3/C4 did not depend on the exact movement performed , Litvak and colleagues suggested that STN gamma activity modulates rather than explicitly encodes motor commands ( Litvak et al . , 2012 ) . Our results take this hypothesis further by extending the concept of modulation to include a possible role for movement cancelation . This notion is also compatible with observations that have linked STN gamma activity to effort ( Jenkinson et al . , 2013; Oswal et al . , 2013; Tan et al . , 2013 ) and arousal ( Brücke et al . , 2013; Jenkinson et al . , 2013; Kempf et al . , 2009 ) . The fact that stopping was more likely successful after STN gamma was relatively high already 200 ms before the stop cue ( i . e . before patients knew they had to stop ) may reflect such arousal-related function and the need for proactive inhibition . The present study is correlative in nature , so we cannot infer that gamma oscillations are causally involved in stopping . However , we would like to speculate that a strong surge in STN gamma activity may shift the excitable period of the otherwise observed pro-kinetic gamma increase such that presynaptic spikes arrive at a period of relative inhibition and motor output thus may be interrupted . Inter-individual variability of the peak frequency and the strength of the STN gamma increase may have been related to differences in disease progression , individual stopping speed or electrode placement and type . We did not find a significant gamma increase in cortical electrodes , which may be due to the reduced signal-to-noise ratio of the EEG . However , a broad gamma increase was observed during stopping in electrocorticography recordings from the pre-supplementary motor area and right inferior frontal gyrus ( Swann et al . , 2012 ) , raising the possibility of cortical involvement in generating the gamma increase via the hyperdirect pathway . In comparison to tetrode recordings in rats ( Schmidt et al . , 2013 ) , our study is limited in that the recording contacts may not have been directly in the STN . The SNr is located in close proximity , ventrally adjacent to the STN , and thus we cannot exclude that we picked up activity from neighbouring structures . However , gamma has been reported to be specifically localized in the dorsal part of the STN ( Trottenberg et al . , 2006 ) , so that contacts selected according to the strongest gamma modulation are likely located closer to the dorsal border of this nucleus . However , this remains speculative . Another limitation of this study is that we recorded from patients that may exhibit pathological STN hyperactivity ( Hamani et al . , 2004 ) expressed in abnormal firing rates and patterns ( Magnin et al . , 2000; Remple et al . , 2011 ) . Even though these pathological changes are attenuated by dopaminergic medication ( Brown et al . , 2001; Heimer et al . , 2006; Levy et al . , 2002 ) , which was taken as usual , and patients were able to perform the task well , neuronal dynamics may still have differed from those of healthy subjects with intact basal ganglia circuits . It may also be argued that stopping may have involved muscle contractions , which were not picked up by the FDI EMG . But the short latency of the gamma increase and the absence of a similar increase in motor cortex in combination with the decrease in connectivity between STN and C3/C4 , which would be expected to increase during movements ( Litvak et al . , 2012 ) , renders this possibility unlikely . Additionally , we observed that gamma increased most strongly in trials where participants were able to stop fully instead of interrupting the downward movement halfway , showing that gamma increased not only during braking in the middle of a movement but that it increased even more in the absence of any movement . As reported previously , we confirmed a link between higher post-movement C3/C4 beta activity and subsequently improved stopping performance , which we suggest was related to fluctuations in cognitive load ( Fischer et al . , 2016 ) . Beta has also been implicated in time estimation ( Kononowicz and van Rijn , 2015 ) , thus it may also reflect an intention to delay the next tap’s timing , which would allow for more time to stop . Stopping success indeed was correlated with the tap-to-sound offset of the last regular tap in seven patients , such that relatively early taps ( early with respect to the sound , which should be compensated for by delaying the next tap ) were followed by higher stopping success . Note that this beta differencewas not present in the STN . We also observed significantly higher beta power over contralateral motor and frontal cortex when stopping was successful in comparison to when it failed . As beta oscillations are less likely to occur during movement execution ( Feingold et al . , 2015; Kilavik et al . , 2013 ) , this difference was expected . In the past , a number of studies have suggested that beta plays an active role in motor inhibition ( Bastin et al . , 2014; Brittain et al . , 2012; Wessel et al . , 2016a ) . Importantly , in the present study no beta increase was observed after the stop signal in comparison to the previous regular tap – not even when only successful stop trials were considered . Thus , it seems unlikely that bursts of beta oscillations per se implemented active braking in our task . Increased beta in other studies may have reinforced the resting position as current motor state that had to be maintained ( Gilbertson et al . , 2005 ) . Such resting posture was not present in our task given that the stop signal was delivered during ongoing tapping . How can we reconcile the above with the results reported by Benis et al . ( 2014 ) , who observed a weaker STN beta decrease during ‘proactively inhibited’ go-trials ( ‘proactively inhibited’ as participants were aware that a stop signal may come after the cue , although it did not appear in these trials ) in comparison to go-trials with a cue , which was never followed by a stop signal and thus resulted in faster reaction times ? The stronger beta decrease may have been related to a more vigorous response in fast go-trials ( Tan et al . , 2013 , 2015 ) or reduced response uncertainty ( Tzagarakis et al . , 2010 ) and thus does not necessarily need to reflect an inhibitory process . A stronger difference in beta decrease between the two trial types was also linked to shorter stop signal reaction times across patients . But this correlation may be mediated by symptom severity , as more severe symptoms could result in less beta reactivity ( Little et al . , 2012 ) , reduced modulation of response vigour and longer stop signal reaction times . Finally , our results may also be reconciled with those reported by Wessel et al . ( 2016a ) if elevated beta activity reflects better connectivity across task-relevant areas ( Gross et al . , 2004 ) or reduced cognitive load ( particularly for <20 Hz beta ) ( Fischer et al . , 2016; Rouhinen et al . , 2013 ) , it could support motor suppression without actually implementing movement inhibition . Recently , an influential hypothesis suggesting that motor suppression is implemented by fronto-central low-frequency activity has received further support ( Wessel et al . , 2016b ) . Even though the authors also observed an STN gamma increase concurrent with response slowing , this increase was associated with the cognitive demands of the verbal working memory task rather than motor inhibition . Similar to classical stop signal tasks , our auditory stop cue also elicited a slow-wave power increase . However , this increase occurred also when stopping was not even attempted . If the slow-wave power increase over Cz would have induced slowing or braking , then the intertap interval in the control condition between the tap before and the tap immediately after the stop signal should have been increased , and this was not the case . Our data suggest an alternative account , namely that the stop signal-evoked slow-wave response does not directly correspond to movement inhibition but instead registers salient sensory stimuli and alerts stopping-relevant areas , which in turn may trigger the STN gamma increase . The increase in cortico-subthalamic low-frequency connectivity might underpin this sequence , enabling the STN to trigger the stopping process . The event-related low-frequency response would thus be necessary for , but not equivalent to motor suppression per se . In Fz and M1 the average low-frequency response seemed to be diminished in the control condition . We would expect that registration of a salient stop signal and efficiency of the transmission process ( in terms of speed or extent of neuronal recruitment ) depends on endogenous fluctuations in arousal , attention and cognitive load , which would reconcile the hypothesis of low-frequency power-mediated salient stimuli processing with previous results regarding motor inhibition ( Wessel and Aron , 2014 ) . Taken together , our results showed that gamma oscillations in the contralateral STN were linked to successful stopping . This indicates that gamma oscillations in the STN are not simply pro-kinetic , but that they can also increase during movement termination . Though we can only infer an association and not causation from observational recordings , our data suggest that the observed gamma rhythm may underpin a fast stopping mechanism involving the STN . Gamma oscillations therefore seem to support fast changes in processing demands not only in cortical but also in cortico-basal ganglia networks in line with theories of gamma synchrony establishing effective , precise and selective neuronal communication ( Fries , 2015 ) .
Ten Parkinson’s disease patients ( mean disease duration = 8 ± 4 years , mean age = 59 ± 8 years; one left-handed/ambidextrous; two female ) were recorded after obtaining informed written consent to take part in this study , which was approved by the local ethics committee ( Oxfordshire REC A , 08/H0604/58 ) . One patient had to be excluded from the analysis as they intermittently fell asleep during the testing . All patients underwent bilateral implantation of deep brain stimulation electrodes into the STN two to six days before the recording with the aim to alleviate symptoms through chronic high-frequency deep brain stimulation . Surgeries and recordings were performed either at the University College Hospital in London or the John Radcliffe Hospital in Oxford , UK . For each patient one of the following three macroelectrode models were used: Medtronic 3389 ( quadripolar , for P1-4 and 8 ) , Boston Scientific , Vercise , DB-2201 ( octopolar , for P6 ) and Boston Scientific , Vercise directional , DB-2202 ( octopolar , directional , for P5 , 7 and 9 ) . Clinical details of the patients are given in Table 2 . Patients were tested on medication to ensure task performance and motor function were as normal as possible , although acknowledging that functional impairments , albeit lessened , still persist in this state . 10 . 7554/eLife . 23947 . 021Table 2 . Clinical details . Age and disease duration are given in years . UPDRS-III: Unified Parkinson’s disease rating scale part III . Levodopa equivalent dose was calculated according to Tomlinson et al . ( 2010 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23947 . 021IDAge/Sex/dom . HandUPDRS-III OFF/ON levodopaDisease durationMain symptomLevodopa equivalent dose ( mg / day ) DBS leadSurgical centre165/f/r33/115Tremor/Dyskinesia807 mgMedtronic 3389TMOxford255/m/r49/2510Leg dragging + tremor ( left side ) 2022 mgMedtronic 3389TMOxford366/f/r25/1417Freezing of gait , balance1089 mgMedtronic 3389TMLondon450/m/r37/175Tremor , Dyskinesia , especially in right foot958 mgMedtronic 3389TMLondon548/m/left-ambi46/186Frequent OFFs800 mgBoston Scientific DB-2202TMOxford654/m/r61/328Motor fluctuations455 mgBoston Scientific DB-2201TMOxford760/m/r37/66Rigidity left side , bradykinesia , dyskinesia2084 mgBoston Scientific DB-2202TMOxford867/m/r31/133 . 5Bradykinesia , Rigidity2173 mgMedtronic 3389TMLondon968/m/r33/1510Motor fluctuations1765 mgBoston Scientific DB-2202TMOxford Participants were asked to tap to an isochronous metronome ( 900 ms inter-trial interval , ITI , 700 Hz pitch , 40 ms duration ) and to interrupt tapping in response to a high pitched auditory stop cue ( 2000 Hz pitch , 40 ms duration ) after a random number of 5–9 taps . The metronome served as regular cue . However , the taps were not triggered in reaction to the metronome sounds but had to be initiated already before the sound to achieve synchronization that depends on how well the timing of the sound is anticipated . If the movement would be reactive , the tap would always lag behind the sound , which was not the case as evidenced by a negative tap-to-sound offset . Thus , this task is special as the metronome cues are not equivalent to go cues The sound was generated in Spike2 with a 1401 data acquisition unit ( Cambridge Electronic Design , Cambridge , UK ) , played by Creative Inspire T10 speakers and recorded by the EEG amplifier ( TMSi Porti amplifier , TMS International , Netherlands ) . The timing of the stop signal was adjusted in a training period at the beginning such that patients would be able to stop only in 50–60% of all trials . Importantly , the stop signal was triggered relative to the tap registered by the pressure sensor and not to the sound to prevent patients from delaying their taps relative to the metronome , which would improve stopping performance if the latter were the case . A ~50% success rate was desirable to capture fluctuations in alertness or stopping readiness and to distinguish related brain processes . The actual average stopping probability was 55% ± ( SD ) 10% . Six patients were additionally recorded in a control condition to assess if stopping-related activity was linked to active motor inhibition or whether it merely reflected registration of the more salient stop tone . In this control condition , patients were asked to end the tapping sequence with two more taps after hearing the high pitched sound instead of stopping immediately . The control condition thus posed much less of a challenge than the main stopping task . A nearly identical task has previously been studied in young healthy subjects ( Fischer et al . , 2016 ) . It differed from the present patient study only in the metronome interval duration , which was shorter ( 700 ms instead of 900 ms in the patients ) and the number of taps ( 6–10 taps until the stop signal may appear instead of 5–9 taps in the patients ) . Intervals were chosen to be longer because stopping proved to be more feasible for patients with longer intervals , and the number of taps was reduced to increase the number of trials obtained in the time-limited recording sessions . We planned to record 100 trials in the stopping condition and 20 trials before and after the main block in the control condition . Due to fatigue and time constraints in some cases less trials were recorded . As three patients ( P4 , P7 , P9 ) had severe motor symptoms on the right side , they performed the task with their left index finger . The remaining six patients used the right index finger . As we would expect the contralateral hemisphere to be more involved in the tapping , we analysed the data not separated between left and right motor cortex and STN , but between contra- and ipsilateral C3/C4 and STN . Behavioural outliers ( such as spurious goniometer deflexions ) prior to the stop signal were removed following visual inspection . After further exclusion of arrhythmic taps as defined by taps that deviated more than 300 ms from the metronome sound , an average number of 65 ± ( SD ) 24 trials remained for further analyses . Goniometer traces and the distribution of tap onsets were strongly overlapping prior to successfully vs . unsuccessfully stopped taps ( Figure 1 ) . To get a graded measure of stopping performance for correlations , the amount of downward movement measured by the goniometer was quantified as movement extent . It was defined as the extent of the downward movement normalized by the amplitude of the upward movement done before . The time between the stop signal and subsequently failed stops was quantified as median across trials for each patient and then averaged over subjects . Bilateral STN local field potentials and EEG was recorded at a sampling frequency of 2048 Hz . EEG electrodes were placed over ( or close to if sutures had to be avoided ) Fz , Cz , Pz , Oz , C3 and C4 according to the international 10–20 system . Electrooculogram ( EOG ) was recorded to remove eye blink artefacts in a subsequent procedure . For one patient , EEG channels could not be recorded because large DC drifts caused amplifier saturation . Tap onsets were registered by a force-sensitive resistor measuring the pressure of the finger on its surface . Finger flexion , i . e . the tapping trajectory , was recorded with a goniometer ( TMSi Goniometer F35 ) attached to the index finger over the metacarpophalangeal joint . To capture muscle activity , electromyogram ( EMG ) was recorded from the first dorsal interosseous muscle ( FDI ) . Events for tap and sound onsets were created in Spike 2 ( RRID:SCR_000903 , Cambridge Electronic Design ) . After DC component removal ( 2 s time constant ) , data were processed further with custom routines in MATLAB ( RRID:SCR_001622 , v . 2014b , The MathWorks Inc . , Natick , Massachusetts ) . EEG channels were re-referenced to linked earlobes if the latter were recorded ( n = 5 ) or to the average of all EEG channels if not ( n = 3 ) . LFP bipolars were computed by subtracting two channels of the same recording electrode ( bipolar combinations varied depending on the number of available contacts ) . Data were down-sampled to 1000 Hz and eye artefacts were removed from the EEG signals by subtracting the filtered EOG ( 40 Hz low-pass Butterworth filter with a filter order of 6 , passed forwards and backwards ) after amplitude matching via least-squares optimization ( MATLAB function fminocn ) . Power between 3–40 Hz was obtained by filtering the data into 3 Hz wide frequency bands shifted by 1 Hz ( Butterworth , filter order = 6 , two-pass , using fieldtrip functions_ft_preproc_lowpassfilter and ft_preproc_highpassfilter [RRID:SCR_004849 , Oostenveld et al . , 2011] ) and calculating the power of the Hilbert transform . Power between 50 and 120 Hz was calculated within 10 Hz wide frequency bands in 2 Hz steps . To reduce noise , power subsequently was temporally smoothed with a 100 ms sliding window . Before exporting the data , it was further down-sampled to 200 Hz . MATLAB analyses scripts for this procedure and subsequent steps to reproduce the figures are provided as source code files ( Source code 1 ) . The data can be downloaded from the Oxford University Research Archive: https://ora . ox . ac . uk/objects/uuid:54c00c3d-1809-4a52-bba8-b491b6075f35 . As we recorded from three different electrode models , with multiple contacts of which some may not have been located in the STN , we decided to pre-select the bipolar configuration that recorded the strongest gamma reactivity during regular tapping . We chose to select the contacts based on gamma activity because gamma has been found to be highly focal to the STN ( Trottenberg et al . , 2006 ) . For the quadripolar ( Medtronic 3389 ) and the unsegmented octopolar model ( Boston Scientific DB-2201 ) , bipolars were computed between neighbouring contacts or if channels saturated and thus could not be recorded , the surrounding contacts were instead used for the bipolar subtraction . For the directional contacts ( Boston Scientific DB-2202 ) , bipolar combinations were computed between the small segmented ones ( C2–C7 ) , plus C1 and C8 if more than two of these channels were saturated to increase the likelihood of including activity from the presumably focal gamma source . As power was converted into relative power changes with respect to a baseline , normalized power estimates were relatively comparable despite differently sized contact surfaces or distances between contacts , as was the case for the directional electrode model . For the selection process , we first computed the 60–90 Hz median power over all taps for each of the multiple bipolar pairs on each electrode in a time window spanning twice the tapping interval around each tap . Then the range between the maximum and minimum of the resulting power time course was divided by the average power within this window , providing the amount of movement-related gamma modulation captured by each bipolar configuration . For each recording electrode only the bipolar configuration with the highest modulation was analysed further . Note that these contacts also recorded significant movement-related beta modulation as shown in Figure 3—figure supplement 1 . Phase-based connectivity between the contralateral STN and the five EEG channels of interest ( Fz , C3 , C4 , Cz , Pz ) was computed based on the phase of the Hilbert-transformed filtered signal ( band-width and frequency shifts as described in Data pre-processing ) . Intersite phase clustering ( ISPC ) can be defined over trials or over time . As we did not expect high-frequency oscillations to be phase-locked across trials , we calculated ISPC for each trial over multiple fixed-width windows to get an estimate of changes in ISPC over time . The fixed width was 200 ms for 50–120 Hz and 250 ms for 6–40 Hz . The frequency cut-off was 6 Hz as 250 ms would have included only one and a quarter cycle of a 5 Hz oscillation or even less for lower frequencies . The window width was chosen to be longer for lower frequencies such that more cycles contributed to the estimate . 250 ms would for example encompass four cycles of a 16 Hz oscillation . ISPC was computed within each of these windows , which were shifted by 10 ms such that the overlapping bins resulted in a smooth image . ISPC was obtained by calculating the length of the average vector of phase ( ϕ ) differences represented as vectors with length one on a unit circle ( Lachaux et al . , 2000 ) based on the following equation ( n=number of samples , MATLAB code provided ) :|∑t=1nei∗ ( STNφt−EEGφt ) n| The amplitude of the signal thus did not contribute to the ISPC estimate . To assess whether ISPC changed in response to the stop signal , we compared whether it differed significantly from zero after normalizing it by the pre-stop signal period ranging from −350 to 0 ms before the stop cue . It should be noted that we analysed LFPs from electrode contact pairs of different surface areas ( according to electrode type ) and EEGs that had different references between subjects . Accordingly , we only considered normalised changes in power to mitigate this variability . All statistical analyses were performed in MATLAB . Correlations between stopping performance ( quantified as movement extent after the stop signal ) and movement parameters or features in the EEG/LFP were calculated as Spearman’s rank correlation coefficients with bootstrapped confidence intervals ( using the Spearman function from the Robust correlation toolbox [Pernet et al . , 2012] ) . To test if correlations with movement parameters differed significantly from zero on a group-level , correlation coefficients were Fisher’s z transformed for each patient and then subjected to a one-sample t-test ( n = 9 ) . The maximum correlation with EEG/LFP gamma power ( Figure 4—figure supplement 1 ) was determined for each patient by finding the maximum correlation within 60–90 Hz and 0:156 ms after the stop signal . Each time-frequency matrix was normalized for each subject and frequency by the average power across all regular taps ( excluding tap one and those directly followed by a stop signal ) to obtain a relative power percentage change before testing for differences . Multiple-comparison correction for power or ISPC comparisons in time-frequency or time windows of interest was performed by using a cluster-based permutation procedure ( Maris and Oostenveld , 2007 , MATLAB code provided ) : The original paired samples were randomly permuted 2000 times such that each pair was maintained but its order of subtraction may have changed to create a null-hypothesis distribution . For each permutation , the sum of the z-scores within suprathreshold-clusters ( pre-cluster threshold: p<0 . 05 ) was computed to obtain a distribution of the 2000 largest suprathreshold-cluster values . If the sum of the z-scores within a suprathreshold-cluster of the original difference exceeded the 95th percentile of the permutation distribution , it was considered statistically significant . Pairwise comparisons for behavioural data or peak timings were performed using t-tests or Wilcoxon signed-rank tests if the normality assumption ( assessed by Lilliefors tests ) was violated and if multiple comparisons were made , p-values were subjected to false discovery rate ( FDR ) -correction . | Being able to stop walking to allow a car to pass is one example of how terminating a movement midway through can be essential for surviving in an ever-changing world . However , people with Parkinson’s disease sometimes struggle to stop performing a repetitive movement . Also , they may find themselves stopping despite having intended to keep moving . This inability to control stopping and starting can play havoc with everyday activities such as walking . Some people with Parkinson’s disease find that their symptoms improve after a treatment called deep brain stimulation . Surgeons lower electrodes into specific regions of the brain and use them to block the abnormal electrical activity that causes problems with movement . One of the main brain regions targeted is an area called the subthalamic nucleus . Whenever people initiate a movement , nerve cells in the subthalamic nucleus start to become activated at the same time . This synchronization generates rhythmic waves of activity in the subthalamic nucleus , which are called gamma waves . To find out whether gamma waves are also involved in stopping a movement , Fischer et al . measured activity in the subthalamic nucleus of nine patients with Parkinson’s disease as they performed a finger tapping exercise . The patients had to tap their finger in time with a metronome , but refrain from tapping whenever they heard a high pitched noise . As expected , a burst of gamma waves accompanied the start of each finger tap . However , Fischer et al . showed that an increase in gamma waves also occurred whenever patients successfully stopped a finger tap midway . Gamma waves may thus help people to interact flexibly with the world around them . Techniques like deep brain stimulation have the potential to manipulate gamma waves . In order to treat symptoms without causing side effects , we need to work out how to target brain waves that are altered in patients , without disrupting other processes . A key step towards achieving this is to understand how brain waves change during essential behaviours such as stopping an on-going movement . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Subthalamic nucleus gamma activity increases not only during movement but also during movement inhibition |
CHARGE syndrome is caused by heterozygous mutations in the chromatin remodeler , CHD7 , and is characterized by a set of malformations that , on clinical grounds , were historically postulated to arise from defects in neural crest formation during embryogenesis . To better delineate neural crest defects in CHARGE syndrome , we generated induced pluripotent stem cells ( iPSCs ) from two patients with typical syndrome manifestations , and characterized neural crest cells differentiated in vitro from these iPSCs ( iPSC-NCCs ) . We found that expression of genes associated with cell migration was altered in CHARGE iPSC-NCCs compared to control iPSC-NCCs . Consistently , CHARGE iPSC-NCCs showed defective delamination , migration and motility in vitro , and their transplantation in ovo revealed overall defective migratory activity in the chick embryo . These results support the historical inference that CHARGE syndrome patients exhibit defects in neural crest migration , and provide the first successful application of patient-derived iPSCs in modeling craniofacial disorders .
CHARGE syndrome is an autosomal dominant genetic disorder characterized by coloboma of iridis , heart defect , atresia choanae , retarded growth , genital hypoplasia , and ear anomalies , a constellation of non-randomly associated malformations ( Blake and Prasad , 2006 ) . This syndrome is relatively common , occurring approximately one in 10 , 000 births . Since the first report that de novo mutations in CHD7 ( chromodomain helicase DNA binding protein 7 ) might be the cause of CHARGE syndrome ( Vissers et al . , 2004 ) , several groups have sought to identify genotype-phenotype correlations and to determine how various phenotypic features of CHARGE are contributed to by CHD7 mutations ( Aramaki et al . , 2006a; Sanlaville et al . , 2006; Zentner et al . , 2010 ) . CHD7 is expressed in various cell types , including pluripotent stem cells and cells of the neural tube and placodal regions ( Aramaki et al . , 2007 ) . CHD7 modulates chromatin formation by binding to genomic DNA and regulating the expression of downstream genes ( Martin , 2010 ) . For instance , several transcriptional factors , such as SOX2 , SOX9 , and SOX10 , have been reported to cooperate with CHD7 in regulating early development in various cell types ( Bajpai et al . , 2010 ) ( He et al . , 2016 ) ( Jones et al . , 2015 ) ( Micucci et al . , 2014 ) ( Schnetz et al . , 2010 ) . The hypothesis that clinical features observed in CHARGE syndrome patients are caused by abnormalities in neural crest development has been proposed for more than 30 years ( Siebert et al . , 1985 ) . The cells of the neural crest contribute to many different tissue lineages , including those of the craniofacial skeleton , cranial nerves ( VII , VIII , IX and X ) , ears , eyes , and heart . Since many of the defects observed in CHARGE syndrome appear to be related to abnormalities of cranial neural crest cells , this syndrome is considered as a ‘neurocristopathy’ ( Aramaki et al . , 2007; Sanlaville et al . , 2006; Siebert et al . , 1985 ) . A recent study supports this view by showing that the knockdown of CHD7 in human embryonic stem cells ( hESCs ) results in migratory neural crest formation defects ( Bajpai et al . , 2010 ) . Moreover , the knockdown of Chd7 in Xenopus laevis or zebrafish embryos led to abnormalities in neural crest specification and migration ( Asad et al . , 2016 ) ( Bajpai et al . , 2010 ) . The hypothesis that the neural crest pathophysiology observed in CHARGE syndrome is attributable to NCC defects has not been examined using patient-derived cells due to technical challenges . In addition , the phenotypic aspects of CHARGE patient-derived NCCs with respect to different migratory behaviors have not been examined in detail . Thus , there is great potential value in the establishment of in vitro models of this syndrome using patient-derived cells for use in the study of CHARGE pathophysiology . In the present study , we generated NCCs from CHARGE syndrome patient-derived iPSCs , established in vitro models of CHARGE syndrome , and observed defective migration in CHARGE NCCs via in vitro and in vivo experiments .
To gain mechanistic insights into the pathogenesis of CHARGE syndrome , we enrolled two CHARGE patients , designated patient 1 ( CH1 ) and patient 2 ( CH2 ) , in the present study in an effort to generate patient-derived iPSCs . Both patients have a heterozygous nonsense mutation in CHD7 , and exhibited the typical phenotype of CHARGE syndrome ( Figure 1—source data 1 ) . We collected fibroblasts from these patients and generated 25 and 23 iPSC clones from CH1 and CH2 , respectively , following the four-factor protocol first reported by Takahashi and Yamanaka ( Takahashi et al . , 2007 ) . We selected four lines from CH1 ( CH1#7 , CH1#11 , CH1#20 , CH1#25 ) and three lines from CH2 ( CH2#1 , CH2#16 , CH2#19 ) for further analyses . As shown in Figure 1A–C , these iPSC clones showed characteristics of pluripotent stem cells , including a morphology similar to that of human embryonic stem cells ( ESCs ) ( Figure 1A ) , the expression of pluripotent stem cell markers ( TRA1-60 and TRA1-81 ) ( Figure 1B ) , and the capacity for teratoma formation ( Figure 1C ) . We confirmed that the CH1-iPSCs and CH2-iPSCs retained the CHD7 mutations observed in the human dermal fibroblasts ( HDFs ) of origin , whereas none of the control iPSCs harbored mutations in CHD7 ( Figure 1D and Figure 1—figure supplement 1 ) . These patient-derived iPSCs enabled us to conduct further in vitro characterization of phenotypes relevant to CHARGE syndrome . Interestingly , the expression levels of CHD7 mRNA in the iPSCs derived from both patients were significantly lower than those in control iPSCs ( Figure 1E ) . Neural crest cells are thought to be the primary cells affected in CHARGE syndrome ( Blake and Prasad , 2006 ) ( Sanlaville et al . , 2006 ) . We therefore differentiated the patient-derived iPSCs into NCCs ( iPSC-NCCs ) using two protocols adapted from previous studies ( Bajpai et al . , 2010 ) ( Lee et al . , 2009 ) , which we refer to as Methods A and B , respectively ( Figure 2A and D ) . As shown in Figure 2B and E , control and CHARGE iPSC-NCCs obtained using each method displayed similar morphological features and were indistinguishable at the colony level . We initially examined the expression of neural crest markers , including CD271 ( NGFR ) and CD57 ( B3GAT1 ) ( Tucker et al . , 1984 ) , by flow cytometric analysis . As shown in Figure 2C and F , more than 90% of the cells obtained using these two methods from control and CHARGE iPSCs expressed both CD271 and CD57 . The ratio of CD271 ( + ) CD57 ( + ) cells per total induced cells from CHARGE iPSCs was same as that of control without regard to the NCC differentiation method . We additionally performed an immuno-cytochemical analysis using additional NCCs markers , including SOX10 and AP2a , and we found that the iPSC-NCCs expressed these neural crest markers similarly ( Figure 2G ) . These results indicate that iPSCs can be differentiated into NCCs , and that the NCC differentiation efficacy of CHARGE iPSCs was similar to that of control iPSCs . The phenotypes of CHARGE syndrome show clear associations with defects in cranial NCCs ( Siebert et al . , 1985 ) ( Blake et al . , 2008 ) , which prompted us to investigate the expression of OTX2 , a cranial marker ( Millet et al . , 1996 ) , by immunocytochemistry . We determined that OTX2 was expressed in both control and CHARGE iPSC-NCCs ( Figure 2H ) . Moreover , using ad hoc protocols , we were able to differentiate CHARGE iPSC-NCCs into adipocytes , chondrocytes , osteocytes , myofibroblasts , and peripheral neurons ( Figure 2I , J and K ) , indicating that these cells exhibit a potential for differentiation similar to control iPSC-NCCs ( Bajpai et al . , 2010 ) ( Lee and Studer , 2010 ) . This suggests that mutations in CHD7 do not affect NCC differentiation per se . These results also indicate that iPSCs from CHARGE patients can be efficiently differentiated into NCCs that express precise positional markers of cranial region and retain the ability to differentiate into cranial neural crest cells , potentially enabling the generation of disease-relevant cellular models of neurocristopathies , such as CHARGE syndrome . To identify the cellular functions dysregulated in CHARGE iPSC-NCCs , we performed a global gene expression analysis of NCCs derived from control and CHARGE iPSCs using a microarray . As shown in Figure 3A , we found that control and CHARGE iPSC-NCCs expressed essentially similar profiles of marker gene sets for early ( NGFR , B3GAT1 , ITGA4 ) , premigratory ( PAX3 , ZIC1 ) , and migratory ( TWIST1 ) NCCs , suggesting the acquisition of the fundamental NCC gene expression profile in CHARGE iPSC-NCCs . The detection of TWIST1 expression is notable , as one previous study reported that this gene , a marker of migratory NCCs , was downregulated in CHD7 shRNA-infected human ESC-NCCs compared with control shRNA-infected cells ( Bajpai et al . , 2010 ) . We next extracted 338 differentially expressed genes ( DEGs ) ( 238 upregulated and 100 downregulated in CHARGE iPSC-NCCs ) between control and CHARGE iPSC-NCCs ( fold change [FC]>1 . 25 ) in an effort to identify features common to genes with altered expression in CHARGE iPSC-NCCs ( Figure 3B ) . A Gene Ontology ( GO ) analysis of this set showed enrichment for genes involved in vasculature development ( p=2 . 08E-11 ) , blood vessel development ( p=1 . 24E-11 ) , and blood vessel morphogenesis ( p=1 . 23E-09 ) . Interestingly , GO terms associated with ‘cell migration’ and ‘cell motion’ were also significantly enriched in these genes ( Figure 3C ) . Given that NCC dysfunction is thought to be linked to the pathogenesis of CHARGE syndrome , we sought to examine the expression of genes associated with NCC behavior , specifically those related to cell migration and cell adhesion . We selected a set of genes listed under each GO term and compared their expression levels in control and CHARGE iPSC-NCCs . As shown in Figures 3D , E 56 genes under the GO terms ‘migration’ or ‘adhesion’ were differentially expressed . Quantitative real-time PCR ( qRT-PCR ) analyses performed for four selected genes , POU3F2 , OLFM3 , CTGF and EDN1 , confirmed that the changes observed in the microarray data set were indeed significant , thereby providing a validation of the analysis ( Figure 3—figure supplement 1A ) . These genes are also listed as CHD7 targets in the dataset . We also performed chromatin immunoprecipitation ( ChIP ) -qPCR for CHD7 using promoter primers for the genes to reveal direct CHD7 binding to these genes , and we found the direct binding of CHD7 to the distal promoter region in EDN1 ( Figure 3—figure supplement 1B ) . The results of these transcriptome analyses support the notion that NCCs exhibit migratory and/or cell adhesion defects during embryonic development in CHARGE patients . In the early stages of cranial neural crest cell migration , epithelial-to-mesenchymal transition ( EMT ) is thought to occur immediately prior to the delamination of NCCs from the neural crest . Migratory NCCs then begin their directed migration along the dorsolateral pathway , reaching their target and initiating their differentiation toward mature cell types ( Kulesa et al . , 2010 ) . Since CHARGE iPSC-NCCs showed only minimal defects in the initiation of NCC differentiation and subsequent differentiation into NCC derivatives , we hypothesized that sequential developmental processes , including delamination and migration , might be disrupted in CHARGE syndrome , as suggested by the results of our transcriptome analysis . We therefore next focused on the dysregulation of CHARGE iPSC-NCCs in cellular adhesion , migration , and cellular motion . The first step of the developmental journey of NCCs consists in their delamination from the region between the dorsal neural tube and the overlying ectoderm ( Kulesa et al . , 2010 ) . We first examined how control iPSC-NCCs migrated outward from spheres using iPSC-NCCs differentiated by Method A to model this particular event . As shown in Figure 4—video 1 , the cells began to spread out as a continuous monolayer ( Phase 1 ) once the sphere became attached to the culture dish . The control iPSC-NCCs residing at the outermost periphery then began to scatter apart ( Phase 2 ) ( Figure 4A–a ) . In contrast , CHARGE iPSC-NCCs exhibited a distinct behavior in Phase 2 , remaining closely associated with their neighbors ( Figure 4A–b ) . To clarify this difference , we performed a time-lapse analysis of the initial phase of cell dispersion from the sphere . We used a method to calculate cell dispersion by a Delaunay triangulation algorithm ( see Materials and methods ) ( Figure 4B ) . The distribution of the formed triangular area by the algorithm at 8 hr after the sphere attached to the plate was significantly increased relative to that at t = 0 for both control and CHARGE iPSC-NCCs ( Figure 4C ) . Next , we analyzed differences in the increased cell dispersion ( from t = 0 to t = 8 hr ) between control and CHARGE cells by calculating the median size of the triangular area . We revealed a delayed dispersion of cells from CHARGE spheres ( Figure 4D ) . Moreover , to determine whether CHARGE iPSC-NCCs have defects in premigratory-to-migratory transition , we analyzed the intercellular contacts of both control and CHARGE iPSC-NCCs at Phase 2 . We visualized the cell associations by F-actin and nuclear staining ( Figure 4E ) . Quantitative analysis of the number of intercellular contacts among the outermost migrating cells revealed significantly persistent intercellular contacts in CHARGE iPSC-NCCs in vitro ( Figure 4F ) . CHARGE NCCs were reluctant to disperse as single cells , in contrast with control NCCs . This suggests that NCC delamination from the neural tube may be affected in CHARGE syndrome . Following delamination from the neural tube , NCCs travel throughout the developing embryo and contribute to major NCC-derived organs ( Cordero et al . , 2011 ) ( Steventon et al . , 2014 ) ( Blake et al . , 2008 ) . To assess the migratory ability of CHARGE iPSC-NCCs , we assessed the trans-well migration of dissociated iPSC-NCCs using the xCELLigence system ( Roche ) . Using this system , cells migrating from the upper to lower well through fibronectin-coated microelectrode sensors were monitored automatically ( Figure 5A ) . As shown in Figure 5B , the migration index of CHARGE iPSC-NCCs became lower than that of control cells after approximately 8 hr of monitoring . At 20 hr , we observed a decrease of approximately 50% in the migration index of the CHARGE iPSC-NCCs compared with that of control cells ( Figure 5C ) . To exclude the possibility that the reduction in the number of migrating CHARGE iPSC-NCCs was due to reduced proliferation , we treated control iPSC-NCCs with an antimitotic , aphidicolin . Aphidicolin treatment did not have a significant effect on the migration index of iPSC-NCCs ( Figure 5—figure supplement 1A ) . Moreover , a BrdU incorporation assay indicated that the proliferative capacity of CHARGE iPSC-NCCs was not different from that of control iPSC-NCCs ( Figure 5—figure supplement 1B ) . Additionally , to exclude the possibility that lower CHARGE iPSC-NCC adherence to fibronectin caused the reduction in the number of migrating CHARGE iPSC-NCCs in this assay , we performed a cell adhesion assay to fibronectin; we found no differential adherence to fibronectin between control and CHARGE iPSC-NCCs ( Figure 5—figure supplement 1C ) . Taken together , these results suggest that CHARGE iPSC-NCCs exhibit aberrant migration , in contrast with their preserved capacity for proliferation and adherence to fibronectin . The defective scattering and trans-well migration of CHARGE iPSC-NCCs suggest that the collective migration of NCCs is affected . We wondered whether this might be attributable , at least in part , to an intrinsic motility defect of individual CHARGE iPSC-NCCs . Therefore , we performed a time-lapse analysis to examine the single-cell spontaneous motility of control ( 201B7 ) and CHARGE ( CH1#25 ) iPSC-NCCs . To exclude cell-density effects , we analyzed the motility of mixed iPSC-NCCs , i . e . , control + CHARGE , within the same well ( Figure 6A ) . The tracking of individual iPSC-NCCs revealed that the average velocities progressively increased over the course of the recording period ( Figure 6B ) . Notably , at any time interval , the average velocity of the CHARGE iPSC-NCCs was lower than that of control iPSC-NCCs ( Sidak's multiple comparisons test after two-way repeated measures ANOVA: pTime <0 . 001 , pCellType <0 . 001; 201B7 , N = 80 cells tracked; CH1#25 , N = 97 cells tracked ) . Sidak’s multiple comparisons tests confirmed the significantly reduced velocities of CHARGE iPSC-NCCs at multiple time intervals ( Figure 6B ) . In contrast , the directionality of the iPSC-NCCs was constant over time and was similar for both control and CHARGE iPSC-NCCs ( Figure 6B ) . A comparison of two different cell lines , WD39 and CH2#16 ( Figure 6—figure supplement 1A–B ) yielded similar results; the average velocity of CHARGE iPSC-NCCs migrating as single cells was significantly reduced compared with that of control iPSC-NCCs ( Sidak's multiple comparisons test after two-way repeated measures ANOVA: pTime <0 . 001 , pCellType <0 . 001; WD39 , N = 170 cells tracked; CH2#16 , N = 133 cells tracked ) . Sidak’s multiple comparisons tests confirmed the significantly reduced velocities of CHARGE iPSC-NCCs at multiple time intervals . In contrast , the directionality of CHARGE iPSC-NCCs was not different to that of control iPSC-NCCs , indicating that the abnormal migration of CHARGE iPSC-NCCs is due , at least in part , to a defective intrinsic motility . To examine whether CHARGE iPSC-NCCs also show defective migration in vivo , we grafted iPSC-NCCs into the dorsal edge of the hindbrain of chick embryos ( HH stage 8–10 ) . To compare the migration ability of CHARGE iPSC-NCCs and control cells under identical conditions , we transplanted a mixture of control and CHARGE iPSC-NCCs into the same embryo . To distinguish these cells , the iPSC-NCCs were stained with different lipophilic dyes ( Vybrant DiI or DiO ) before transplantation ( Figure 7A ) . First , to examine the serial migration of the transplanted cells , we transferred the transplanted embryo to a glass-bottomed plate ( IWAKI ) 6 hr after transplantation and then acquired time-lapse images every 20 min ( Figure 7–video 1 ) . We tracked 4–14 cells migrating well for both control and CHARGE cells in an embryo ( the average # of counted cells per experiment: control , 8 . 9; CHARGE , 9 . 0 ) , and we calculated their velocity at each time interval using ImageJ ( Figure 7—figure supplement 1A , B ) . In 6 of 9 transplanted embryos , the velocity of CHARGE iPSC-NCCs was significantly less than that of control cells . In the other 3 embryos , there were no differences between control and CHARGE cells . ( Figure 7—figure supplement 1—source data 1 -tab1 ) Collectively , CHARGE iPSC-NCCs exhibited a lower velocity compared with that of the co-transplanted control iPSC-NCCs ( p=0 . 03; Wilcoxon signed-rank test ) ( Figure 7—figure supplement 1C ) . Second , we examined the iPSC-NCCs that had migrated throughout the embryo in ovo thirty-six hours after transplantation . Interestingly , the iPSC-NCCs migrated from the site of transplantation ( dorsal area ) to the ventral area ( Figure 7B , lower panels ) . Both control and CHARGE iPSC-NCCs migrated in the expected direction , consistent with the normal developmental routes of NCCs . Notably , CHARGE iPSC-NCCs did not follow abnormal routes to ectopic sites in this model . To compare the migration of control and CHARGE iPSC- NCCs in vivo , we scored the maximum distance that the transplanted cells migrated in 17 surviving chick embryos ( Figure 7—source data 1 ) . We recorded the locations of the iPSC-NCCs-derived cells that had migrated the greatest distance from the transplant site and assigned a score from 1 ( dorsal area ) to 4 ( ventral area ) to each grafted embryo ( Figure 7B ) . As shown in Figure 7C , CHARGE iPSC-NCCs exhibited similar or lower migration scores compared with those of the co-transplanted control iPSC-NCCs . Adversely , the migration exhibited a large degree of variability among the embryos . These data suggest that the reduced migratory capability of CHARGE iPSC-NCCs observed in vitro reflects their reduced migration in vivo after transplantation in chick embryo . Our results show that NCCs differentiated from CHARGE iPSCs exhibit migration defects in vitro and in vivo that are consistent with the pathological features of CHARGE syndrome and thus may serve as a useful model for investigating the molecular causes of this condition .
We successfully generated iPSCs from CHARGE syndrome patient-derived fibroblasts and differentiated them into NCCs . We identified multiple functional abnormalities in CHARGE iPSC-NCCs , which may reflect a direct link between the NCC population affected in CHARGE syndrome and the multiple anomalies observed in CHARGE syndrome patients ( Figure 8 ) . It was previously shown by CHD7 knockdown in human ESCs that CHD7 controls EMT in multipotent NCCs ( Bajpai et al . , 2010 ) . Our results using CHARGE syndrome patient-derived iPSCs indicate that CHARGE iPSC-NCCs have migratory defects and that a series of migration-related behaviors following EMT , namely , delamination , migration , and motility , are affected . First , our scattering assay using migratory iPSC-NCCs , which is an in vitro model of the premigratory-to-migratory transition , akin to delamination in vivo , indicated defective delamination in CHARGE NCCs ( Figure 4 ) . In this assay , iPSCs were induced into neuroectodermal spheres ( Lee et al . , 2010 ) , and the cells migrated out from the spheres . This migration of iPSC-NCCs out from neuroectodermal spheres resembles the premigratory-to-migratory transition , after which the migrating cells scattered as single cells in a manner similar to delamination in vivo . Our finding of defective CHARGE iPSC-NCCs scattering is compatible with a previous report that CHD7 controls the transcriptional reprogramming of EMT . As shown in Figure 3E , FOXD1 expression was upregulated in CHARGE iPSC-NCCs . Since FOXD1 is known to be expressed in premigratory NCCs and extinguished once migration occurs ( Gómez-Skarmeta et al . , 1999 ) , altered FOXD1 expression may lead to a defective premigratory-to-migratory transition occurring in CHARGE iPSC-NCCs . While delamination in vivo is not so simple as this in vitro model , as it is subject to complex orchestration by various signals , this delamination model may be a very valuable tool , since it is impractical for ethical and technical reasons to observe human NCC delamination in early embryos directly . Second , our transwell migration assay using dissociated iPSC-NCCs , which occurred after delamination in vivo , showed defective CHARGE cell migration ( Figure 5 ) . As shown in Figure 3D–E , many genes referred to under the GO terms ‘migration’ and ‘adhesion’ were differentially expressed in CHARGE iPSC-NCCs , and the defective migratory phenotype of CHARGE iPSC-NCCs in the transwell assay is compatible with the results of this transcriptional analysis . This assay models cell migration toward chemoattractants . All cranial NCCs are suggested to have similar migratory potential , unlike trunk NCCs , which are known to be a heterozygous population consisting of cells such as leader cells and follower cells ( Richardson et al . , 2016 ) . Therefore , this transwell migration assay is adequate for assessing the migration of cranial NCCs such as our iPSC-NCCs that robustly express OTX2 ( Figure 2H ) . Of course , during the long journey from the dorsal neural tube to the ventral area , many signals influence NCC migration in a complex manner , and NCCs change their character during their migration . In our model , the in vivo migration provided additional evidence supporting the defective migration of CHARGE iPSC-NCCs . Third , a spontaneous motility assay allowed us to assess whether defective motility is a partial cause of the defective migration of CHARGE iPSC-NCCs ( Figure 6 ) . In this assay with a mixed population of control and CHARGE iPSC-NCCs ( co-culture system ) , autocrine or paracrine factors would likely diffuse within the wells and affect neighboring cells . The observed spontaneous defective motility of the CHARGE cells suggests that such soluble factors are not involved in the defective migration of CHARGE iPSC-NCCs . Our transcriptome analysis revealed that genes associated with ‘migration’ and ‘adhesion’ were altered in CHARGE iPSC-NCCs . CHD7 is an important chromatin remodeler and may thus play roles in various gene regulatory mechanisms ( Bajpai et al . , 2010 ) ( He et al . , 2016 ) ( Jones et al . , 2015 ) ( Micucci et al . , 2014; Schnetz et al . , 2010 ) . In particular , we focus on the PAX6 downstream and Hippo/YAP pathways . Importantly , CHD7 is considered to function cooperatively with SOX2 as a molecular partner ( Engelen et al . , 2011 ) , and PAX6 has also been reported to be a functional partner of SOX2 ( Thakurela et al . , 2016 ) . As shown in Figure 3—figure supplement 1A , the expression levels of POU3F2 ( BRN2 ) and OLFM3 ( Optimedin ) were significantly downregulated in CHARGE compared with control iPSC-NCCs , and these two genes have been reported to be targets of PAX6 ( Grinchuk et al . , 2005 ) ( Ninkovic et al . , 2013 ) ( Raviv et al . , 2014 ) . POU3F2 is involved in controlling the migration of melanocytes , which are neural crest derivatives ( Berlin et al . , 2012 ) . OLFM3 is considered to be involved in cell-cell adhesion and cell attachment to the extracellular matrix ( Grinchuk et al . , 2005 ) . Several downstream targets of Pax6 have been identified as cell adhesion molecules and structural proteins ( Cvekl and Callaerts , 2017 ) . Altered CHD7 expression resulted in the upregulation of PAX6 and the downregulation of PAX6 downstream genes ( Figure 3E ) . Therefore , it is conceivable that CHD7 regulates multipotent NCC migration by cooperating with PAX6 . Next , CTGF and EDN1 , known to be downstream factors in the Hippo-YAP signaling pathway , are highly expressed in CHARGE iPSC-NCCs compared with control cells ( Figure 3—figure supplement 1A ) . The Hippo-YAP signaling pathway is known to be regulated via cell density ( Zhao et al . , 2007 ) , and this pathway has recently been reported to inhibit migration and suggested to play important roles on the early stage of NCC specification and migration ( Lamar et al . , 2012; Wang et al . , 2016 ) . In particular , CTGF and EDN1 play important roles in craniofacial development , and the timing and regulation of their expression are crucial for their function ( Maj et al . , 2016; Mercurio et al . , 2004 ) . Altered CHD7 expression in iPSC-NCCs resulted in the upregulation of CTGF and EDN1 . Therefore , it is conceivable that CHD7 regulates the craniofacial phenotype of CHARGE syndrome through the Hippo-YAP pathway . To clarify this mechanistic insight into how NCCs are dysregulated in CHARGE syndrome patients , it is noteworthy that 202 of the 338 differentially expressed genes between the CHARGE and control iPSC-NCCs were listed as target genes of CHD7 in the ChIP-seq datasets from the ENCODE Transcriptional Factor Target dataset ( Rouillard et al . , 2016 ) . Although these target sites vary depending on cell type , we found the target site of CHD7 in the EDN1 distal promoter region by ChIP-qPCR for CHD7 using our cells . This result suggests that CHD7 regulates not only the expression of some specific key genes but also the robust gene expression in early NCCs . The current study represents the first model of a developmental morphogenetic disorder using patient-derived iPSCs . To date , the neural crest pathophysiology observed in CHARGE syndrome has not been examined directly using patient-derived cells due to technical challenges and ethical concerns surrounding the collection of NCCs from human embryos . Moreover , since the developmental regulation of NCCs is known to be unique to individual species ( Acloque et al . , 2009; Barriga et al . , 2015 ) , NCCs derived from CHARGE patient-derived iPSCs are an appropriate source for modeling the cellular features of this disease in vitro . We suggest that such cells may be used as a powerful assay system for evaluating NCC dysfunction in other morphogenetic disorders that could be considered neurocristopathies , such as craniofacial syndrome ( Minoux and Rijli , 2010 ) and infants of vitamin A exposure ( Kraft et al . , 1989; Rosa , 1983 ) . NCCs play important roles in the formation of sensory organs , such as ears , eyes , and olfactory organs , and some congenital neurocristopathies are caused by reproductive toxicity . These deformities of experimental animals have been used for the toxicity testing of newly developed drugs . The iPSC-NCC system presented herein could be used as an animal-free NCC system for reproductive toxicity testing .
As shown in Figure 1—source data 1 , patient1 ( CH1 ) , a Japanese male , was born at 39 weeks of gestation with a birth weight of 3 . 3 kg and a length of 50 . 5 cm . As major diagnostic criteria , he was noted to have external asymmetrical ear defects and bilateral sensorineural hearing loss ( >70 dB ) . A computed tomography ( CT ) scan of the temporal bones revealed that semicircular canals were bilaterally hypoplastic . He also showed velopharyngeal incoordination and gastroesophageal reflux . Development was severely delayed , with a developmental quotient of 15 at 3 years old . He had micropenis , cryptorchidism , and delayed incomplete pubertal development . His height was 126 . 0 cm ( – 2 . 1 s . D . ) , and his weight was 25 . 2 kg ( – 1 . 3 s . D . ) at 10 years and 6 months old . He was noted to have a distinctive CHARGE physiognomy ( Blake and Prasad , 2006 ) . By the direct sequencing of his genomic DNA , a heterozygous nonsense mutation in CHD7 , i . e . , c . 4171delC p . Gln1391fs*13 , was identified . Patient2 ( CH2 ) , a Japanese female , was born at 38 weeks of gestation with a birth weight of 3 . 03 kg and a length of 48 . 6 cm . She was noted to have external asymmetrical ear defects and bilateral sensorineural hearing loss ( >95 dB ) . A CT scan of the temporal bones revealed that semicircular canals were bilaterally hypoplastic , and the numbers of turns to the cochlea were decreased ( Mondini defects ) . She also showed velopharyngeal incoordination and bilateral retinal coloboma with visual impairment . As minor diagnostic criteria ( Blake and Prasad , 2006 ) , development was severely delayed with a developmental quotient of 50 at 5 years old . She had delayed incomplete pubertal development . Her height was 119 . 5 cm ( – 2 . 7 s . D . ) , and her weight was 21 . 5 kg ( – 1 . 6 s . D . ) at 10 years old . Her physiognomy showed features typical of CHARGE syndrome . By the direct sequencing of her genomic DNA , a heterozygous nonsense mutation in CHD7 , i . e . , c . 4480C > T p . Arg1493Ter , was identified . As a control , WD39-iPSCs were derived from the HDFs of a healthy 16-year-old Japanese female ( Imaizumi et al . , 2012b ) . 201B7-iPSCs and WA29-iPSCs were derived from the HDFs of a 36-year-old Caucasian female ( Cell Applications Inc . , San Diego , CA ) . 1210B2-iPSCs and 1201C1-iPSCs were derived from human peripheral blood mononuclear cells of a healthy 29-year-old African/American female ( Cellular Technology Limited ) . 201B7-iPSCs , 1210B2-iPSCs , and 1201C1-iPSCs were kindly provided by Shinya Yamanaka . ( Okita et al . , 2013; Takahashi et al . , 2007 ) KhES1-ESCs were kindly provided by Norio Nakatsuji ( Suemori et al . , 2006 ) . CH1-iPSCs and CH2-iPSCs were derived from the HDFs of a 10-year-old Japanese male patient and the HDFs of a 10-year-old Japanese female patient , respectively . The clinical diagnoses of these two CHARGE syndrome patients were made based on the Blake criteria ( Blake and Prasad , 2006 ) . WD39-iPSCs , 201B7-iPSCs , WA29-iPSCs were established through the retroviral transduction of four transcription factors ( KLF4 , OCT4 , SOX2 , and c-MYC ) into HDFs ( Takahashi et al . , 2007 ) , and 1210B2-iPSCs and 1201C1-iPSCs were established using the combination of KLF4 , OCT4 , SOX2 , L-MYC , LIN28 , EBNA and shRNA for TP53 , as previously described ( Okita et al . , 2013 ) . The maintenance of HDFs , stem cell culture , characterization and teratoma formation were performed as described previously ( Imaizumi et al . , 2012a2012; Ohta et al . , 2011; Takahashi et al . , 2007 ) . We performed mycoplasma contamination test using MycoAlert Mycoplasma Detection Kits ( Lonza Walkersville , Inc . , Walkersville , MD ) and confirmed all lines were not contaminated by mycoplasma . All human cell and tissue donors were provided explanatory materials and a verbal explanation of the procedure , detailing both the procedure and the purposes of the experiment , as well as their rights , prior to collection and use . All experimental procedures were reviewed and approved by the Keio University School of Medicine Ethics committee ( Approval Number: 20080016 ) . RRIDs ( Research Resource Identifiers ) were provided as below; KhES1 ( CVCL_B231 ) , WD39 ( CVCL_Y528 ) , 201B7 ( CVCL_A324 ) , WA29 ( CVCL_LJ40 ) , 1210B2 ( CVCL_LJ38 ) , 1201C1 ( CVCL_LJ37 ) , CH1#7 ( CVCL_LJ32 ) , CH1#11 ( CVCL_LJ31 ) , CH1#20 ( CVCL_Y955 ) , CK1#25 ( CVCL_Y956 ) , CH2#1 ( CVCL_LJ#33 ) , CH2#16 ( CVCL_LJ#34 ) and CH2#19 ( CVCL_LJ35 ) . To assess the pluripotency of generated iPSCs , we transplanted these iPSCs into the testis of 8-week-old NOD/SCID mice ( OYG International ) as previously described ( Ohta et al . , 2011 ) . Eight weeks after transplantation , teratomas were dissected and fixed with 4% PFA in PBS . Paraffin-embedded tissue was sectioned and stained with hematoxylin and eosin . Images were obtained with a BZ-9000 ( Keyence ) microscope . All experimental procedures were reviewed and approved by the Keio University Institutional Animal Care and Use Committee ( Approval Number: 09169 ) . The molecular tests for the CHD7 gene mutations were conducted as previously reported ( Aramaki et al . , 2006a ) . We confirmed that the fibroblasts and iPSCs from both CHARGE syndrome patients showed mutations in CHD7 , whereas the control fibroblasts and iPSCs did not , by sequencing of the PCR amplicons with the primers below using an automated sequencer ABI3100 ( Thermo Fisher Scientific , Waltham , MA ) as previously described . ( Aramaki et al . , 2006a ) Primer sets: CHD7 exon17 F: CTATGCGTCAGGCCTCCTT CHD7 exon17 R: TGGGTCTGACTGGTACTCTCTG CHD7 exon19 F: TGCAGCATTTGTTTAGTCTGC CHD7 exon19 R: TTCCCAATGCATCTTGTAAGC Total RNA was isolated and extracted as previously described . cDNA synthesis from RNA was performed using Superscript III reverse transcriptase ( Thermo Fisher Scientific ) , followed by digestion with RNase H ( Thermo Fisher Scientific ) . qRT-PCR was performed using a 7900HT Real-Time PCR system ( Thermo Fisher Scientific ) or a Viia7 Real-Time PCR system ( Thermo Fisher Scientific ) with SYBR green ( TaKaRa , Kusatsu , Japan ) . For every set of qRT-PCR analyses , we had three technical replicates and at least three biological replicates . Data were analyzed by Dunn’s multiple comparisons test after Kruskal-Wallis test using GraphPad Prism software version 7 . 0a ( GraphPad Software ) . The following primers were used: Primer sets: CTGF F: CAAGGGCCTCTTCTGTGACT CTGF R: ACGTGCACTGGTACTTGCAG EDN1 F: GACATCATTTGGGTCAACACTC EDN1 R: GGCATCTATTTTCACGGTCTGT OLFM3 F: CAGGAGGAAATTGGTGCCTA OLFM3 R: AGGGTCTGTCATCCAAGCAC POU3F2 F: CGGCGGATCAAACTGGGATTT POU3F2 R: TTGCGCTGCGATCTTGTCTAT TaqMan Gene Expression Assays , Inventoried CHD7 primer: Assay ID: Hs00214990_m1 GAPDH primer: Assay ID: Hs99999905_m1 Cells were crosslinked with 1% formaldehyde for 10 min , incubated with 200 mM glycine for 5 min and then stored at −80°C until use . The ChIP assay was performed as previously described ( Kimura et al . , 2008 ) . Co-immunoprecipitated DNA was used as a template for PCR of the genomic region . The genomic regions were determined by the NCC-specific enhancer regions identified , as previously described ( Rada-Iglesias et al . , 2012 ) . Data were analyzed by paired t test using GraphPad Prism software version 7 . 0a ( GraphPad Software ) . The following primers were used: Primer sets: hPOU3F2 distal enhancer F: CAGTAAGCTGCTTGGCCATT hPOU3F2 distal enhancer R: CAGCCCTCCCTCCTCTTAAC hOLFM2 distal enhancer F: CAATCCCATCTGACCCAACT hOLFM2 distal enhancer R: CTGGCTGGTTTCCAGGTTTA hEDN1 distal enhancer F: TTCCCTCAGCTTTTGCTTGT hEDN1 distal enhancer R: ATTTGGGGGCTTTTTGAGAA hCTGF distal enhancer F: GATTTCAGCTGCTGGCTACC hCTGF distal enhancer R: ATGGCTATCACTTGCCTGCT Day 10 iPSC-NCCs obtained by Method B were detached with Accutase ( Innovative Cell Technologies , San Diego , CA ) and collected with ice-cold MACS buffer ( Miltenyi Biotec , Bergisch Gladbach , Germany ) consisting of phosphate-buffered saline ( PBS ) , 0 . 5 M EDTA , and 5% bovine serum albumin . After washing , the cells were suspended in ice-cold MACS buffer at 2 × 105 cells/ml and stained for 30 min at 4°C using PE-conjugated anti-human CD271 ( NGFR ) mouse IgG1 antibody ( BioLegend , San Diego , CA ) and FITC-conjugated anti-human CD57 ( B3GAT ) mouse IgM antibody ( Beckman Coulter , Brea , CA ) . Propidium iodide staining allowed for the exclusion of dying/dead cells from the analysis . Isotype controls were used as negative controls . Flow cytometric analyses were performed using a FACS Calibur flow cytometer ( Becton Dickinson , Franklin Lakes , NJ ) . Cells were fixed with PBS containing 4% paraformaldehyde ( PFA ) for 15 min at room temperature . The cells were analyzed by immunofluorescence staining using the following antibodies: AP2α ( monoclonal , 1:100; Cell Signaling Technologies , Danvers , MA ) , β-III tubulin ( monoclonal , 1:1000; Sigma-Aldrich ) , CD90 ( monoclonal , 1:100; BD Pharmingen , San Diego , CA ) , FOXG1 ( polyclonal , 1:250; Abcam , Cambridge , UK ) , GFAP ( monoclonal , 1:200; Thermo Fisher Scientific ) , Mash1 ( monoclonal , 1:500; BD Pharmingen ) , OTX2 ( polyclonal , 1:100; R and D Systems , Minneapolis , MN ) , P75NTR ( polyclonal , 1:500; Abcam ) , SMA ( monoclonal , 1:500; Sigma-Aldrich ) , SOX10 ( polyclonal , 1:200; Abcam ) , Peripherin ( polyclonal , 1:500; Merck Millipore , Billerica , MA ) , TRA-1–60 ( monoclonal , 1:200; Millipore ) , and TRA-1–81 ( monoclonal , 1:200; Merck Millipore ) . Immunoreactivity was visualized with secondary antibodies conjugated with Alexa 488 , Alexa 568 , or Alexa 647 ( 1:1000 , Thermo Fisher Scientific ) . Nuclei were counterstained using Hoechst 33258 ( 10 μg/ml , Sigma-Aldrich ) . Images were obtained using an Apotome ( Carl Zeiss , Oberkochen , Germany ) or LSM-710 confocal ( Carl Zeiss ) microscope . The NCC differentiation of iPSCs was performed as previously described with some modifications ( Lee et al . , 2009; Lee and Studer , 2010 ) . Briefly , dissociated iPSCs were plated onto an AggreWell 400 plate ( Stem Cell Technologies , Vancouver , Canada ) at a density of 600 , 000 cells/ well in human ES medium consisting of DMEM/Ham’s F12 ( Sigma-Aldrich ) , 20% Knockout Serum Replacement ( Thermo Fisher Scientific ) , 2 mM L-glutamine ( Thermo Fisher Scientific ) , 1 × 10−4 M non essential amino acids ( Sigma-Aldrich ) , 1 × 10−4 M 2-mercptoethanol ( Sigma-Aldrich ) , and 0 . 5% penicillin and streptomycin ( Thermo Fisher Scientific ) , the mediun also contained 10 μM Y-27632 ( Wako Pure Chemical Industries , Osaka , Japan ) in order to make homogenous embryoid bodies ( EBs ) consisting of 400 cells . After 40 hr , the EBs were transferred to a bacteria dish and cultured in suspension for a week in human EB medium consisting of DMEM/Ham’s F12 ( Sigma-Aldrich ) , 5% Knockout Serum Replacement ( Thermo Fisher Scientific ) , 2 mM L-glutamine ( Thermo Fisher Scientific ) , 1 × 10−4 M non-essential amino acids ( Sigma-Aldrich ) , 1 × 10−4 M 2-mercaptoethanol ( Sigma-Aldrich ) , 0 . 5% penicillin and streptomycin , and containing 10 μM SB431542 ( R and D Systems ) , and 250 μg/ml Noggin-Fc ( R and D systems ) . At day 8 , the human EB medium was replaced with N2 medium consisting of DMEM/Ham’s F12 , GlutaMax-I ( Thermo Fisher Scientific ) , 0 . 5% GlutaMax ( Thermo Fisher Scientific ) , 1% N2 supplement ( Thermo Fisher Scientific ) , 0 . 5% insulin ( Thermo Fisher Scientific ) , 0 . 5% penicillin and streptomycin , and containing 10 μM SB431542 ( Sigma-Aldrich ) , and 250 μg/ml Noggin-Fc ( R and D Systems ) . At day 15 , the EBs were replaced in a 6-well plate coated with 10 ng/ml fibronectin ( Sigma-Aldrich ) and cultured in N2 medium supplemented with 20 ng/ml of human recombinant EGF ( PeproTech , Rocky Hill , NJ ) and 20 ng/ml of human recombinant FGF2 ( PeproTech ) . After 5–7 days of adhesion culture , the cells had migrated out from the colonies were collected and subjected to the analysis . The medium was changed every three days in this protocol . iPSCs were differentiated into NCCs , as previously described . ( Bajpai et al . , 2010 ) . Briefly , iPSCs were incubated with 2 mg/ml collagenase IV ( Thermo Fisher Scientific ) . Once the iPSCs were detached , the clusters were broken into pieces consisting of 100–200 cells and plated onto a 100 mm petri dish ( Becton Dickinson ) in hNCC medium ( NC medium ) . The medium consisted of 1:1 neurobasal medium ( Thermo Fisher Scientific ) and DMEM/F-12 medium containing 1x GlutaMax ( Thermo Fisher Scientific ) , 5 mg/ml insulin ( Sigma-Aldrich ) , 0 . 5% penicillin and streptomycin , 0 . 5x GEM 21 NeuroPlex serum-free supplement ( Gemini Bio Products , West Sacramento , CA ) , 0 . 5x N2 supplement and supplemented with 20 μg/ml human recombinant EGF and 20 μg/ml human recombinant FGF2 . The medium was changed every other day . After seven days of differentiation , migratory NCCs appeared from the attached spheres . At 3–4 days after their appearance , the cells were used for subsequent analysis . We induced in vitro differentiation into adipocytes , chondrocytes , and osteocytes as previously reported ( Lee et al . , 2010 ) . Differentiated cells were stained by Toluidine blue ( Wako Pure Chemical Industries ) , Safranin-O ( Wako Pure Chemical Industries ) , and Alizarin red ( Wako Pure Chemical Industries ) . We also differentiated them into myofibroblast ( SMA+ ) and peripheral neurons ( peripherin+ ) and performed an immunocytochemical analysis . In vivo differentiation of iPSC-NCCs into chondrocytes: We dissociated Method B iPSC-NCCs at day 10 with Accutase and purified the TRA-1–60-negative fraction using a MACS system ( Miltenyi Biotec ) . We next injected 1 . 0 × 106 TRA-1–60-negative cells into the testes of 8-week-old NOD-SCID mice , as previously described ( Ohta et al . , 2011 ) . Eight weeks after transplantation , the testes were dissected and fixed with 4% PFA in PBS . The paraffin-embedded tissue was sectioned and stained with toluidine blue ( performed by Dept . of Pathology , Keio University School of Medicine ) . Images were obtained using a BZ-9000 ( Keyence , Osaka , Japan ) microscope . Total RNA was isolated from day-10 iPSC-NCCs using TRIzol ( Thermo Fisher Scientific ) according to the manufacturer’s protocol and further purified with an RNeasy mini kit ( Qiagen , Hilgen , Germany ) . Two replicates were run per line from two independent inductions . For the microarray analysis , RNA quality was assessed using a 2100 Bioanalyzer ( Agilent Technologies Inc . , Santa Clara , CA , USA ) . Total RNA ( 100 ng ) was reverse-transcribed , biotin-labeled , and hybridized to a Human Genome U133 Plus 2 . 0 Array ( Affymetrix , Santa Clara , CA ) , which was subsequently washed and stained in a Fluidics Station 450 according to the manufacturer’s instructions ( Lockhart et al . , 1996 ) ( Heishi et al . , 2006 ) . The microarrays were scanned using a GeneChip Scanner 3000 7G ( Affymetrix ) , and the RMA algorithm was implemented for the background correction , normalization across arrays , and log2 transformation of the raw image files ( Bolstad et al . , 2003 ) . Normalized data were filtered based on gene expression level and analyzed using GeneSpring GX software 14 . 5 ( Agilent Technologies ) for producing scatter plots and using R package ( gplots ) for producing heatmaps ( Warnes et al . , 2015 ) . The GeneChip data were deposited in the NCBI Gene Expression Omnibus ( GEO; http://www . ncbi . nlm . nih . gov/geo/ ) and are accessible through the GEO series accession number GSE86212 . In this assay , we used iPSC-NCCs obtained by Method A . At day 15 , EBs replated onto a fibronectin ( 10 ng/ml ) -coated 8well-plastic-bottomed chamber ( ASAHI GLASS , Tokyo , Japan ) were imaged for 8 hr . To analyze how the cells dispersed from each sphere , the Delaunay triangulation algorithm was used ( Carmona-Fontaine et al . , 2011 ) . ALl cells around the spheres were connected to their closest neighboring cells , and the network shaped triangles by this algorithm . This algorithm is available as an ImageJ plugin . Data were analyzed by Mann-Whitney U test using GraphPad Prism software version 7 . 0a ( GraphPad Software ) . In this assay , we used iPSC-NCCs obtained by Method A . On day 16 , we attached a floating sphere onto the well of a 24-well plate coated with fibronectin ( 10 μg/ml ) and cultured it in N2 medium supplemented with 20 ng/ml of human recombinant EGF and 20 ng/ml of human recombinant FGF2 . After five days of adhesion culture , cells were fixed with PBS containing 4% PFA for 15 min at room temperature . F-actin and nuclei were stained using Alexa Fluor-488 phalloidin ( Thermo Fisher Scientific ) and Hoechst 33258 ( Sigma-Aldrich ) , respectively . Images were obtained using a BZ-9000 ( Keyence ) microscope . To quantify intercellular contacts of iPSC-NCCs , we analyzed the outermost nine cells in each of the eight 45 degree-sector of a sphere by counting the number of their contacting-neighboring cells ( Figure 4F ) , and we classified them into three groups , 0 , 1 , and >1 . Each cell line was analyzed in at least three independent experiments . Data were analyzed Dunnett’s multiple comparisons tests after one-way ANOVA or Dunn’s multiple comparison test after Kruskal-Walli test using GraphPad Prism software version 7 . 0a ( GraphPad Software ) . We used Method B NCCs for this assay . We dissociated day-10 iPSC-NCCs into single cells with Accutase ( Innovative Cell Technologies Inc . ) . We used the xCELLigence-DP system ( Roche ) with CIM-Plate 16 to measure the migration index of each type of iPSC-NCCs . The upper plate of CIM-Plate 16 was coated with fibronectin ( 10 μg/ml in PBS ) , and 100 , 000 cells were added to each upper well . NC medium without human recombinant EGF and human recombinant FGF2 was added into each upper well , and NC medium without human recombinant EGF and human recombinant FGF2 containing 10% fetal bovine serum was added to each lower well . Cells that migrated from the upper to the lower well were automatically measured by the xCELLigence system . Eventually , aphidicolin ( Sigma , Saint Louis , MO ) was used at the concentration of 10 μg/ml . Data were analyzed by Tukey's multiple comparisons test after one-way ANOVA and Sidak’s multiple comparisons test after two-way repeated measure ANOVA using GraphPad Prism software version 7 . 0a ( GraphPad Software ) . Passaged day10 iPSC-NCCs were seeded into wells of an 8-well glass-bottomed plate coated with poly-L-ornithine ( 0 . 1 mg/ml ) and fibronectin ( 10 μg/ml ) at a low density in NC medium supplemented with 10 μM BrdU ( Sigma-Aldrich ) . After 24 hr , the cells were fixed with PBS containing 4% PFA for 15 min at room temperature and immunostained with sheep polyclonal anti-BrdU antibody ( 1:500; Fitzgerald Industries International , Acton , MA ) . Images were randomly captured with an Apotome microscope , and the cells were manually counted . Data were analyzed by unpaired t test using GraphPad Prism software version 7 . 0a ( GraphPad Software ) . Control iPSC-NCCs and CHARGE iPSC-NCCs obtained by Method B were used for this assay . Each iPSC-NCCs type was cultured to semi-confluence in NC medium , detached by 5 min of treatment with Accutase , and were washed with NC medium twice . We resuspended the cells at a density of 1 × 105 cells per ml in NC medium , and added 100 μl of cell suspension to each well of a fibronectin-coated 96-well plate . After 60 min incubation at 37°C , we changed the medium and added 10 μl of WTS-1/ECS ( MerckMillipore , Billerica , USA ) per well except for 12 wells per 96-well plate . After 90 min of incubation at 37°C , the plate was shaken thoroughly for 1 min on a shaker , and then the absorbance at 450 nm of the treated and untreated samples was measured using a microplate reader . The average of absorbance values of the 12 wells without WTS-1/ECS was considered a baseline , and the data were normalized to that of 201B7 iPSC-NCCs in each experiment . Data were analyzed by Dunnett's multiple comparisons test after one-way repeated measures ANOVA using GraphPad Prism software version 7 . 0a ( GraphPad Software ) . ( Chen et al . , 2009 ) ( Mobley and Shimizu , 2001 ) . Control iPSC-NCCs and CHARGE iPSC-NCCs obtained by Method B were used for this assay . At day 10–12 , after the beginning of differentiation by Method B , the remaining spheres were removed by direct aspiration with a fine Pasteur pipette . The adherent NCCs were washed 2–3 times gently with PBS , and then stained for 3 hr at 37°C with either Vybrant DiI ( Thermo Fisher Scientific ) or Vybrant DiO ( Thermo Fisher Scientific ) diluted 1/300 in NC medium . Notably , permutations of the staining dyes confirmed that the nature of the dye had no effect on the migratory behavior of the cells . After 4–5 washes with PBS , the stained NCCs were then dissociated using Accutase , counted using Trypan Blue ( Wako Pure Chemical Industries ) , and then co-seeded in equal amounts at a density of 5 × 103 cells/well ( total of 10 × 103 cells per well ) onto 8-well , plastic-bottomed chambers that were previously coated with fibronectin at 10 μg/ml . Three hours after seeding , once the cells had attached , the chambers were transferred to an LSM 5 , PASCAL Exciter confocal microscope ( Carl Zeiss ) that was equipped with a heat- ( 37°C ) and gas-controlled incubation chamber ( 5% CO2 ) ( Tokai Hit , Shizuoka , Japan ) that was coupled to a heated motorized stage . The objective lens ( EC-Plan Neofluar , 10 X , Numerical Aperture 0 . 3 ) was maintained at 37°C and was used to acquire a Z-stack time-lapse series ( 7 Z-stacks spanning 30 μm , every 15 min ) of multiple locations . Z-projections were produced using Image Browser Zeiss software at the end of the analysis . The time-lapse recordings began at 4 hr after seeding and continued for at least 16 hr . Individual cells were manually tracked using the Manual Tracking plugin of the Fiji software ( 1 . 48 ) . Cells exhibiting abnormal morphologies ( e . g . , neurite-like , or with signs of apoptosis ) were excluded from the analysis . Calculations of individual velocities and directionalities were performed using the chemotaxis and migration tool from Ibidi ( Martinsried , Germany ) . Data were analyzed Sidak’s multiple comparisons tests after two-way repeated measures ANOVA using GraphPad Prism software version 7 . 0a ( GraphPad Software ) . Control and CHARGE floating spheres at day seven were stained with Vybrant DiI and Vybrant DiO respectively for 6 hr each and were seeded into the same 100 mm Petri dish to make mixture of dual colored NCCs sheet . On day 10 dual-colored NCCs were dissected with a needle as a cluster and transplanted into the dorsal side ( top ) of the developing neural tube at the hindbrain level of HH stage 8–10 chick embryos . The embryos were incubated for 36 hr , and imaged with an SVZ16 ( Olympus , Tokyo , Japan ) stereo microscope . We transplanted control and CHARGE iPSC-NCCs into chick embryos , as described above . At 6 hr after transplantation , we started to perform time-lapse imaging as previously described ( Tabata and Nakajima , 2003 ) . Briefly , transplanted chick embryos were placed on a Millicell-CM membrane ( pore size , 0 . 4 μm; Millipore ) and cultured in saline , which is described below . The dishes were then mounted onto a confocal microscope ( FV1000 , Olympus Optical ) . Approximately , 20 optical Z-section images were acquired at an interval of 5 μm every 15 min , and all focal planes ( 100 μm ) were merged . Individual cells were manually tracked using the Manual Tracking plugin of the Fiji software ( 1 . 48 ) . Data were analyzed by two-way repeated measures ANOVA using GraphPad Prism software version 7 . 0a ( GraphPad Software ) . The saline used consisted of the following: solution A ( for 1 l ) : 121 . 0 g of NaCl , 15 . 5 g of KCl , 10 . 4 g of CaCl2 . 2H2O , and 12 . 7 g of MgCl2 . 6H2O; solution B ( for 1 l ) : 2 . 4 g of Na2HPO4 . 2H2O and 0 . 2 g of NaH2PO4 . 2H2O . After autoclaving but prior to using the solutions , mix 120 ml of solution A with 2700 ml of H2O; then , add 180 ml of solution B , as previously described ( Psychoyos and Finnell , 2008 ) . | CHARGE syndrome is a disease in which organs including the heart , eyes and ears may not develop properly . The cells that form the tissues affected by CHARGE syndrome develop in embryos from precursor cells called neural crest cells . Individuals with CHARGE syndrome also have mutations in a gene called CHD7 . However , it is difficult to examine how CHD7 mutations affect neural crest cells in embryos . In recent years , cell reprogramming techniques have made it possible to create induced pluripotent stem cells ( iPSCs ) from the specialized somatic cells found in the human body . These iPSCs can be developed into many different cell types , including neural crest cells . Okuno et al . created iPSCs from the skin cells of people with CHARGE syndrome , developed these cells into neural crest cells , and compared them with neural crest cells that were developed from the skin cells of people without CHARGE syndrome . The neural crest cells developed from people with CHARGE syndrome showed multiple abnormalities . For example , they were not able to move around correctly . This is an important observation because neural crest cells must move through tissues to form the various organs affected by CHARGE syndrome . Okuno et al . also observed changes in the activity of many genes other than CHD7 in the neural crest cells developed from CHARGE patients . Further research is now needed to find out which genes are the most important for restoring the normal activity of neural crest cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"medicine"
] | 2017 | CHARGE syndrome modeling using patient-iPSCs reveals defective migration of neural crest cells harboring CHD7 mutations |
End-stage kidney disease ( ESKD ) patients are at high risk of severe COVID-19 . We measured 436 circulating proteins in serial blood samples from hospitalised and non-hospitalised ESKD patients with COVID-19 ( n = 256 samples from 55 patients ) . Comparison to 51 non-infected patients revealed 221 differentially expressed proteins , with consistent results in a separate subcohort of 46 COVID-19 patients . Two hundred and three proteins were associated with clinical severity , including IL6 , markers of monocyte recruitment ( e . g . CCL2 , CCL7 ) , neutrophil activation ( e . g . proteinase-3 ) , and epithelial injury ( e . g . KRT19 ) . Machine-learning identified predictors of severity including IL18BP , CTSD , GDF15 , and KRT19 . Survival analysis with joint models revealed 69 predictors of death . Longitudinal modelling with linear mixed models uncovered 32 proteins displaying different temporal profiles in severe versus non-severe disease , including integrins and adhesion molecules . These data implicate epithelial damage , innate immune activation , and leucocyte–endothelial interactions in the pathology of severe COVID-19 and provide a resource for identifying drug targets .
Coronavirus disease ( COVID-19 ) , caused by the SARS-CoV-2 virus , displays wide clinical heterogeneity from asymptomatic to fatal disease . Patients with severe disease exhibit marked inflammatory responses and immunopathology . The mechanisms underlying this remain incompletely characterised , and the key molecular mediators are yet to be determined . The first treatment shown to reduce mortality from COVID-19 in randomised trials was dexamethasone ( Horby et al . , 2020 ) , a corticosteroid that has broad non-specific effects on the immune system . Even with corticosteroid treatment , mortality in severe COVID-19 remains significant . There is a wide armamentarium of existing drugs that target inflammation more selectively , providing potential repurposing opportunities for the treatment of COVID-19 . Recently , the REMAP-CAP trial has demonstrated efficacy of anti-IL6 receptor blockade in patients admitted to intensive care units with severe disease ( Gordon et al . , 2021 ) . In order to select the most promising agents for future trials , we urgently need to better understand the molecular drivers of severe disease . Proteins are the effector molecules of biology and the targets of most drugs . Therefore , proteomic profiling to identify the key mediators of severe disease provides a valuable tool for identifying and prioritising potential drug targets ( Suhre et al . , 2021 ) . Risk factors for severe or fatal COVID-19 include age , male sex , non-European ancestry , obesity , diabetes mellitus , cardiovascular disease , and immunosuppression ( Williamson et al . , 2020 ) . End-stage kidney disease ( ESKD ) is one of the strongest risk factors for severe COVID-19 ( estimated hazard ratio for death 3 . 69 ) ( Williamson et al . , 2020 ) , and ESKD patients hospitalised with COVID-19 have a mortality of approximately 30% ( Docherty et al . , 2020; Corbett et al . , 2020; Ng et al . , 2020; Valeri et al . , 2020 ) . ESKD patients have a high prevalence of vascular and cardiometabolic disease ( e . g . hypertension , ischaemic heart disease , diabetes ) , either as a result of the underlying cause of their renal disease or as a consequence of renal failure . In addition , ESKD results in both relative immunosuppression and chronic low-grade inflammation , which may impact viral defence and the host inflammatory response . Here we performed proteomic profiling of serial blood samples of ESKD patients with COVID-19 , leveraging the unique opportunity for longitudinal sampling in both the outpatient and inpatient settings afforded by a large multi-ethnic haemodialysis cohort ( Figure 1a ) . These data revealed 221 proteins that are dysregulated in COVID-19 versus matched non-infected ESKD patients . Using linear mixed models , joint models , and machine learning , we identified proteins that are markers of COVID-19 severity and risk of death . Finally , we characterised the temporal dynamics of the blood proteomic response during COVID-19 infection in ESKD patients , uncovering 32 proteins that display altered trajectories in patients with severe versus non-severe disease .
Principal component analysis ( PCA ) of proteomic data from subcohort A demonstrated differences between samples from COVID-19-positive cases and controls , although the two groups did not separate into discrete clusters ( Figure 2a , b ) . To examine the effects of COVID-19 on the plasma proteome , we performed a differential expression analysis in subcohort A between COVID-19 cases ( n = 256 samples passing quality control [QC] from 55 patients ) and non-infected ESKD controls ( n = 51 ) using linear mixed models , which account for serial samples from the same individual ( see Materials and methods ) . This revealed 221 proteins associated with COVID-19 ( 5% false discovery rate , FDR ) ; the vast majority were upregulated , with only 40 downregulated ( Figure 3a , Supplementary file 1c ) . In order to provide a succinct and standardised nomenclature , we report proteins by the symbols of the genes encoding them ( see Supplementary file 1a for a mapping of symbols to full protein names ) . The most strongly upregulated proteins ( in terms of fold change ) were DDX58 , CCL7 , IL6 , CXCL11 , KRT19 , and CXCL10 , and the most strongly downregulated were SERPINA5 , CCL16 , FABP2 , PON3 , ITGA11 , and MMP12 ( Figure 3—figure supplement 1 ) . Notably , many of the upregulated proteins were chemotaxins . We observed that a high proportion of the measured proteins were associated with COVID-19 . Given the highly targeted nature of the Olink panels that we used ( enriched for immune and inflammation-related proteins ) , this was not surprising . Nevertheless , to ensure that the Benjamini–Hochberg adjustment of p-values was controlling the FDR at the 5% level , we performed two additional analyses ( see Materials and methods ) . First , we estimated the FDR using an alternative method ( the plug-in procedure ; Hastie et al . , 2001 ) ; this confirmed appropriate FDR control . Second , we used permutation to estimate the distribution of the number of proteins expected to be declared significant under the null hypothesis of no association between any proteins and COVID-19 . This showed that the probability of observing the number of differentially abundant proteins we identified was highly unlikely under the null ( empirical p<1×10−5; Figure 3—figure supplement 2 ) . Although our COVID-19-negative controls were well matched in terms of age , sex , and ethnicity ( Figure 1—figure supplement 1a–c ) , perfect matching of comorbidities was not feasible in the context of the healthcare emergency at the time of patient recruitment . There was a higher prevalence of diabetes in the COVID-19 cases compared to the controls ( 61 . 8% versus 47 . 1% , respectively; Table 1 ) . To evaluate whether differing rates of diabetes had impacted the proteins identified as differentially abundant between cases and controls , we performed a sensitivity analysis adding diabetes as an additional covariate in the linear mixed model . This did not materially affect our findings; estimated effect sizes and –log10 p-values from models with and without the inclusion of diabetes were highly correlated ( Pearson r > 0 . 99 , and r = 0 . 95 , respectively; Figure 3—figure supplement 3a , b ) . Full results from both models are shown in Supplementary file 1c . Similarly , there were also differences in the underlying cause of ESKD in cases compared to controls ( Table 1 ) . We therefore performed a further sensitivity analysis adjusting for underlying cause of renal failure . This did not make any meaningful difference to our results ( Figure 3—figure supplement 3c , d , Supplementary file 1c ) . We also considered the possibility that timing of haemodialysis might affect the plasma proteome . To minimise the impact of this , all samples were taken prior to haemodialysis . For the large majority ( 86 . 6% ) of samples , the most recent haemodialysis was between 48 and 72 hr prior to blood draw . This consistency in timing of blood sampling reduces the potential for impact of this issue . Nevertheless , to evaluate whether timing of haemodialysis might have impacted our results , we performed a sensitivity analysis including time from last haemodialysis as a covariate . Our results were not materially affected by this , with −log10 p-values and estimated effect sizes very highly correlated with those obtained without inclusion of this covariate ( Pearson r > 0 . 99 for effect size estimates and for −log10 p-values; Figure 3—figure supplement 4a , b , Supplementary file 1c ) . We used the smaller subcohort B ( n = 52 serum samples from 46 patients with COVID-19; see Materials and methods ) for validation . We first projected the data from subcohort B into the PCA space of subcohort A to examine how well the separation of cases and controls in the PCA space replicated ( see Materials and methods ) . This revealed clearer separation of infected and non-infected patients than in subcohort A ( Figure 2c , d ) , perhaps reflecting the higher proportion of hospitalised patients ( 41 of 46 patients ) in subcohort B ( Table 2 ) . We next performed differential abundance analysis in subcohort B and found 201 proteins that were dysregulated in cases versus controls ( 5% FDR ) ( Supplementary file 1c ) . Of the 221 differentially abundant proteins from subcohort A , 150 ( 69 . 7% ) were also identified in subcohort B at 5% FDR ( Figure 4a ) . Effect sizes in each dataset showed a strong correlation ( r = 0 . 80 , Figure 4b ) . This demonstrates that our findings are highly reproducible despite differences in sample sizes and blood materials ( plasma versus serum in subcohorts A and B , respectively ) . Examination of the principal components plot labelling samples by clinical severity at the time of sampling ( defined by WHO severity scores , graded as mild , moderate , severe , or critical ) demonstrated a gradient of COVID-19 severity , best captured by principal components 1 and 3 ( Figure 2—figure supplement 1a ) . To determine the proteomic effects of COVID-19 severity , we tested for associations between proteins and WHO severity score at the time of blood sampling , using linear mixed models with severity encoded as an ordinal predictor ( see Materials and methods ) . This analysis revealed 203 proteins associated with severity ( Figure 3b , Supplementary file 1d ) . The majority of these were upregulated in more severe disease , with only 42 downregulated . A sensitivity analysis adjusting for time since last haemodialysis made no significant impact on our results ( Figure 3—figure supplement 4c , d , Supplementary file 1d ) . Consistent with previous reports , we found that severe COVID-19 was characterised by elevated IL6 . In addition , we observed a signature of upregulated monocyte chemokines ( e . g . CCL2 , CCL7 , CXCL10 ) , neutrophil activation and degranulation ( e . g . PRTN3 , MPO ) , and epithelial injury ( e . g . KRT19 , AREG , PSIP1 , GRN ) . ( Figures 3b , c and 5 ) . SERPINA5 and leptin showed the greatest downregulation as COVID-19 severity increased ( Figure 3b , c ) . We next asked how does the COVID-19 severity protein signature relate to the proteins that are differentially abundant between cases and controls ? The majority ( 140/203; 69% ) of severity-associated proteins were also identified as differentially abundant in the COVID-19-positive versus -negative analysis ( Figure 6a ) . Log fold changes for proteins in COVID-19 versus non-infected patients were correlated with effect sizes in the severity analysis , such that the proteins most upregulated in cases versus controls also tended to show the greatest upregulation in severe disease ( Figure 6b ) . However , there were some notable exceptions ( e . g . CCL20 , IL17C , OSM ) that were strongly associated with severity , but not differentially expressed in infected versus non-infected patients ( Figure 6c ) . PCA revealed that some samples from patients who had mild or moderate disease at the time of sampling clustered with samples from patients with severe disease ( Figure 2—figure supplement 1a ) . Examination of the same PCA plot labelling samples according to the patient’s overall clinical course ( measured by peak WHO severity score over the duration of the illness ) ( Figure 2—figure supplement 1b ) revealed that these samples came from individuals who subsequently developed severe or critical disease . This suggested that molecular changes may predate clinical deterioration . To evaluate this further , we used supervised learning approaches to test whether the proteomic signature of the first blood sample for each patient in our dataset could identify whether the patient either had clinically severe COVID-19 at the time of sampling or would develop severe disease in the future . Whereas differential expression analyses consider each protein marker separately , machine-learning techniques allow examination of all proteins concurrently , thus capturing non-linear relationships in the dataset . Using Random Forests , we trained a classifier on the first sample for each COVID-19 patient to predict the overall clinical course , defined by peak WHO severity . For the purposes of this analysis , we binarised clinical course into either WHO mild/moderate or severe/critical . The Random Forests method achieved 71% accuracy in predicting peak severity . By contrast , using only clinically available predictors ( demographics , comorbidities , and clinical laboratory results ) , the Random Forests method achieved 66% accuracy in predicting peak severity . Combining clinical parameters plus proteins did not improve accuracy ( 71% ) compared to using proteomic predictors alone , suggesting that the information contained in the clinical predictors is captured at the proteomic level . While we do not believe that proteomic profiling is likely to enter clinical practice for risk stratification during this pandemic , the features selected by the classifier can highlight proteins of biological importance . We therefore interrogated the model to identify key proteins by calculating feature importance metrics ( see Materials and methods , Supplementary file 1e ) . The most important proteins for indicating the presence of current or future severe disease were IL18BP , CTSD ( Cathepsin D ) , GDF15 , KRT19 , TNFSF11 , and IL1RL1 ( ST2 ) ( Figure 7a ) . It is notable that through this distinct analytical approach , KRT19 again emerged as a key biomarker of severe disease . Nine of 55 patients in subcohort A died . We therefore sought to identify proteins associated with risk of death . To leverage the dynamic nature of repeated protein measurements for prediction of death , we utilised joint models , which combine linear mixed models and Cox proportional hazards models ( Ibrahim et al . , 2010; Rizopoulos , 2010 ) ( see Materials and methods ) . This analysis identified 44 proteins for which increased concentration was associated with increased risk of death ( Figure 7b , Supplementary file 1f ) , including CST3 , IL22RA1 , AZU1 , CCL28 , and SPON1 , and 25 proteins for which increased concentration was associated with reduced risk of death , including CD84 , TNFSF12 , TANK , PRKCQ , and ADM . A number of routine clinical laboratory tests have well-characterised associations with COVID-19 ( e . g . elevated inflammatory markers , d-dimer , and reduced lymphocyte count ) ( Guan et al . , 2020 ) . We therefore compared our proteomic data from COVID-19 patients at each timepoint to contemporaneous clinical laboratory measurements using linear mixed models ( see Materials and methods ) . We found associations between plasma proteins and all clinical laboratory measurements except troponin ( Figure 8 , Supplementary file 1g ) . Many of these proteins were also markers of severity ( e . g . IL6 , KRT19 , IFN-gamma , and CXCL10 were strongly associated with raised CRP and ferritin and reduced lymphocyte counts ) . Of note , CCL7 , a monocyte chemokine that was also identified as an important marker of severity by the Random Forests classifier , was associated with lower monocyte count and raised inflammatory markers . Elevated neutrophil count was associated with Oncostatin-M , which regulates IL6 , GCSF , and GMCSF production , and with the proteases MMP9 and defensin . The immune response to infection is dynamic , and therefore snapshot measurements provide only partial insights . Leveraging the dense serial sampling in our dataset ( Figure 1 ) , we modelled the temporal trajectory of each protein and asked whether or not any protein trajectories differed in patients with a severe/critical versus mild/moderate overall clinical course . This was achieved using linear mixed models that included a term for time from first symptoms and a time × severity interaction term ( see Materials and methods ) . One hundred and seventy-eight proteins displayed a significant association with time from first symptoms ( 5% FDR ) , demonstrating the temporal variability in plasma proteins across the disease course ( Supplementary file 1h ) . Moreover , we identified 32 proteins for which there was significant interaction between time and severity , that is , proteins displaying differential temporal trajectories between mild/moderate and severe/critical infections ( Supplementary file 1h , Figure 9 ) . Among the proteins with the strongest temporal differences according to clinical course were the integrins ITGA11 and ITGB6 , the adhesion molecule ICAM1 , TNFRSF10B ( a receptor for TRAIL ) , and PLAUR ( the receptor for urokinase plasminogen activator ) . Most of these proteins exhibited rising profiles in the more severe patients but flat profiles in milder cases . ACE2 , the receptor for SARS-CoV-2 , also displayed this pattern ( Figure 9 ) . In contrast , abundance of ITGA11 , which was also identified as reduced in the analysis of infected versus non-infected patients , fell over time in the severe group . In the UK , individuals from ethnic minorities are at higher risk of severe disease and death from COVID-19 ( Williamson et al . , 2020 ) . We therefore examined whether any of the proteins we measured exhibited differences across ethnicities , analysing COVID-19-positive cases and controls separately ( see Materials and Methods ) . In COVID-19-negative ESKD patients , no proteins were significantly associated with ethnicity in a multivariable model adjusting for age and sex . In COVID-19-positive ESKD patients , there is the potential for protein associations with ethnicity to be confounded by disease severity . To account for this , we included severity as well as age and sex as covariates . A single protein , LY75 , was associated with ethnicity in this multivariable model ( nominal p-value 0 . 0001 , Benjamini–Hochberg adjusted p-value 0 . 04 , with higher levels in white patients ) . Using the same within-case analysis strategy in subcohort B , we found no proteins were significantly associated with ethnicity after multiple testing correction , although the nominal p-value for LY75 was 0 . 025 . While these analyses failed to identify substantial ethnicity-related variation in the proteins we measured , an important caveat is that there were relatively modest numbers of individuals from each ethnic group , and so statistical power was limited . Larger multi-ethnic studies are needed to adequately address this question . Other studies have used a variety of proteomic platforms to investigate COVID-19 . We compared our findings to those of three published studies ( Shen et al . , 2020; Lucas et al . , 2020; Arunachalam et al . , 2020 ) and a preprint by Filbin et al . , 2020 . Of the 221 proteins that were differentially abundant in our analysis of COVID-19-positive versus -negative ESKD patients , 116 associations had been previously reported ( Supplementary file 1i ) . Of the 203 proteins associated with severity , 165 had previously been reported ( Supplementary file 1j ) . We focussed in more detail on the study by Filbin et al . , 2020 because of the large sample size and the breadth of proteomic assay used . This study comprised 384 patients with acute respiratory distress ( 306 COVID-19 positive and 79 COVID-19 negative ) and measured 1472 proteins using the Olink Explore platform . Four hundred and seventeen of these were also measured in our study . Of the 221 proteins differentially abundant in our case/control analysis , 210 were measured in their study . Of these , 100 ( 47 . 6% ) were significant in their analysis of COVID-19-positive versus COVID-19-negative respiratory distress . In addition , we observed strong correlation ( r = 0 . 69 ) between the estimated log fold changes in our and their studies ( Figure 4—figure supplement 1 ) . Of the 203 proteins associated with severity in our study , 192 were measured in their study . One hundred and fifty-seven of these were significantly associated with severity , giving a concordance of 81 . 8% . Thus , despite the differences in study design and clinical populations , we observed notable similarities in our results and those reported by Filbin et al . , 2020 .
In this study , we performed plasma proteomic profiling of haemodialysis patients with COVID-19 . A strength of our study was that we were able to perform serial blood sampling in both the outpatient and inpatient settings , including longitudinal samples from the same individual before and after hospitalisation . This was possible because haemodialysis patients are unable to fully isolate as they must continue to attend for regular dialysis sessions . Moreover , haemodialysis patients represent an important group since ESKD is one of the strongest risk factors for death from COVID-19 ( Williamson et al . , 2020; Corbett et al . , 2020; Ng et al . , 2020; Valeri et al . , 2020 ) . Data from the UK Renal Registry shows that 7 and 14 day mortality for COVID-19-infected in-centre haemodialysis patients was 11% and 19% , respectively ( COVID-19 Data , 2020 ) . Data from the Scottish Renal Registry estimates 30 day mortality following a positive COVID-19 test as 22% , and as of 31 May 2020 , 28 . 2% of renal replacement therapy patients who had a positive COVID-19 test had died ( Scottish Renal Registry COVID-19 report , 2020 ) . In our local population of 1352 in-centre haemodialysis patients , 315 patients had tested positive for COVID-19 by the end of our study period ( 31 May 2020 ) , of whom 53% required hospitalisation and 85 ( 27% ) died . The OpenSAFELY study ( Williamson et al . , 2020 ) examined ~17 million UK primary care records and linked these to the UK COVID-19 mortality register . Patients with estimated glomerular filtration rate ( eGFR ) < 30 ml/min/1 . 73 m2 had a hazard ratio ( HR ) for death of 3 . 56 after adjustment for age and sex . In part , the high mortality from COVID-19 in ESKD patients likely reflects the fact that these patients are enriched for cardiometabolic traits that predispose to severe COVID-19 . However , in multivariable analyses adjusting for these factors , impaired renal function remains an independent risk factor for severe COVID-19 ( Williamson et al . , 2020 ) . Moreover , there is an inverse relationship between renal function and risk of death from COVID-19 across the spectrum of chronic kidney disease . These observations support the notion that the state of ESKD per se is an important determinant of outcome in COVID-19 . ESKD is well-recognised as an immunosuppressed state ( Eiselt et al . , 2016; Girndt et al . , 1999; Sarnak and Jaber , 2000 ) , with defects in both innate and adaptive immunity ( Alexiewicz et al . , 1991; Massry and Smogorzewski , 2001; Girndt et al . , 2001; Meier et al . , 2002 ) . Accordingly , ESKD confers increased vulnerability to viral infections including influenza and respiratory syncytial virus ( Betjes , 2013; Boattini et al . , 2020; Prasad et al . , 2020; Usvyat et al . , 2012 ) . In addition , ESKD results in a chronic low-grade inflammatory state ( Carrero and Stenvinkel , 2010 ) . This tendency to a pro-inflammatory state , combined with reduced ability to respond to viruses , may contribute to the abnormal host response to SARS-CoV-2 infection , producing the immunopathology that leads to severe COVID-19 . Our comparison of COVID-19-positive and -negative haemodialysis patient plasma samples revealed 221 proteins that were differentially abundant in COVID-19 . The majority of these were upregulated , with strong representation of viral response proteins ( e . g . DDX58 , IFNG ) , cytokines/chemokines ( e . g . IL6 , CCL7 , CXCL10 , and CXCL11 ) , and epithelial proteins ( e . g . KRT19 , PSIP1 ) ( Figure 3a ) . The COVID-19-negative controls in this analysis were carefully matched to cases in terms of age , sex , and ethnicity . However , complete matching of clinical characteristics was not feasible; there were differences in the prevalence of diabetes and the underlying causes of ESKD between COVID-19-positive cases and controls ( Table 1 ) . Sensitivity analyses adjusting for these covariates gave highly consistent results , indicating that our findings are robust . In addition , we validated our findings when we analysed serum samples from a separate subcohort of COVID-19-positive ESKD patients . ESKD is itself likely to significantly impact the plasma proteome . Previous cross-sectional studies have shown that the levels of many circulating proteins have an inverse relationship with eGFR ( Naseeb et al . , 2015; Christensson et al . , 2018 ) . A longitudinal study using an Olink proteomics panel ( although not one used in our study ) found that for 74% of the 84 proteins measured , protein levels rose as eGFR fell ( Lind et al . , 2019 ) . For many proteins , it is unclear whether this inverse relationship with renal function reflects cause or effect . Some proteins may be increased in chronic kidney disease due to reduced renal clearance , some may be elevated secondary to tissue injury or chronic inflammation , and others may be drivers of renal injury . Regardless , this observation of widespread changes in the blood proteome of kidney disease patients emphasises the importance of using COVID-19 ESKD patients rather than healthy individuals as our control group . Analysis within COVID-19 cases revealed 203 proteins associated with disease severity , the strongest of which was IL6 ( Figure 3b ) . Association of IL6 with severe disease is well-established and has already received considerable attention ( Wu et al . , 2020; Li et al . , 2020 ) . Despite promising initial case reports of IL6R receptor blockade in COVID-19 , convincing efficacy was not demonstrated in early randomised trials ( Furlow , 2020 ) . More recently , the REMAP-CAP trial has shown the benefit of anti-IL6R therapy when given to critically ill patients on admission to intensive care units ( Gordon et al . , 2021 ) , indicating that IL6 does contribute to critical illness from COVID-19 . Our finding that IL6 was most strongly upregulated in severe disease demonstrates the value of plasma proteomic profiling in identifying putative drug targets . Members of the CCL and CXCL chemokine families ( e . g . CCL2 , CCL7 , CCL20 , and CXCL10 ) were strongly associated with severity . Likewise , higher levels of CCL2 , CCL7 , CCL20 , and CXCL10 were associated with lower blood lymphocyte count and higher inflammatory markers ( CRP and ferritin ) ( Supplementary file 1g ) , which are clinical markers of severe disease and poorer outcome in COVID-19 ( Gupta et al . , 2021 ) . Of note , CCL20 is a chemoattractant for lymphocytes ( Schutyser et al . , 2003 ) , and its negative association with lymphocyte count may reflect a direct effect on migration of lymphocytes from the blood into the tissues rather than simply marking severe disease . CCL2 ( also known as MCP-1 ) and CCL7 ( MCP-3 ) are both chemokines for monocytes , and CXCL10 has pleiotropic immunological effects including chemotaxis . These chemokines were also negatively correlated with blood monocyte count , suggesting recruitment of these innate immune cells into damaged tissues . The neutrophil proteases PRTN3 ( proteinase-3 ) and MPO ( myeloperoxidase ) ( Figure 5 ) and the neutrophil-derived protein AZU1 were associated with severe disease ( Supplementary file 1d ) , indicating that neutrophil activation and degranulation are features of severe COVID-19 . Degranulation of neutrophils releasing PRTN3 and MPO could potentially contribute to oxidative damage in the lungs and thus more severe disease . A striking finding of our study was the association of disease severity with upregulation of epithelial proteins ( e . g . KRT19 ) and epithelial tissue repair pathways ( e . g . PSIP1 , AREG , GRN [progranulin] ) , most likely reflecting lung and vascular damage . KRT19 was notably prominent in our analyses , as well as the study by Filbin et al . , 2020; Supplementary file 1j . KRT19 is an intermediate filament protein , important for the structural integrity of epithelial cells ( Saha et al . , 2017 ) . These data suggest that severe COVID-19 is characterised by destruction of the lung epithelium and vascular endothelium . Vascular injury might thus explain the high level of vascular thrombosis seen in patients in severe disease . In summary , our data reveal that severe COVID-19 is characterised proteomically by a signature of innate immune activation and epithelial injury . Sixty-nine percent of proteins associated with severity were also differentially abundant in the case versus control analysis ( Figure 6a ) , and for the large majority of proteins the within-case severity analysis , effect size was proportional to the fold change between cases and controls ( Figure 6b ) . This suggests that , in general , the distinction in the plasma proteome between severe and mild COVID-19 is a quantitative difference in the COVID-19 signature , rather than there being an orthogonal signature involving a different set of proteins . Consistent with this concept , examination of PCA plots coloured by severity revealed that while there was a gradient of COVID-19 severity , the samples from severe or critical patients did not form a discrete cluster distinct from those from patients with milder disease ( Figure 2—figure supplement 1 ) . However , there were a few exceptions where proteins that were associated with severity were not upregulated in the case–control analysis . These included OSM , IL17C , and CCL20 ( Figure 6c ) . These proteins therefore reflect biological processes specifically of severe disease and may represent therapeutic targets . Survival analysis identified 44 proteins associated with increased risk of death ( Figure 7b ) . As expected , many of these were also associated with disease severity , high CRP , and lower lymphocyte count ( Figure 7—figure supplement 1 ) . In contrast , 25 proteins were associated with reduced risk of death ( Figure 7b ) . One such protein is the multi-functional cytokine TNFSF12 ( TWEAK ) . Although TWEAK can exert pro-inflammatory effects , it also can inhibit the innate immune response ( Maecker et al . , 2005 ) and promote tissue repair and endothelial cell proliferation and survival ( Burkly et al . , 2007 ) , which may be beneficial responses in COVID-19 . This illustrates that although proteins associated with inflammation are often thought to be destructive , the inflammatory response also induces programmes for limiting injury and initiating tissue repair . Insufficient activation of such homeostatic mechanisms may contribute to why some individuals get severe COVID-19 . The host immune response to COVID-19 is a dynamic process , and clinical deterioration typically occurs 7–10 days after first symptoms . Temporal information may therefore be important in determining optimum timing of therapeutic intervention ( e . g . blockade of a particular cytokine ) . By taking serial samples and examining their patterns within individuals over time , we were able to model protein trajectories and found that many proteins display temporal variability during COVID-19 . Longitudinal measurements also allow molecular comparison of severe versus mild disease trajectories . By modelling the interaction term between time from first symptoms and overall disease course , we found 32 proteins that displayed distinct temporal profiles in severe versus mild disease . These results point to enhanced leucocyte–endothelial cell interactions indicated by upregulation of cell adhesion molecules ( e . g . ITGB6 , ICAM1 ) in severe disease . This endothelial activation may contribute to COVID-19-associated thrombosis discussed above . Management of thrombosis in COVID-19 currently consists of anticoagulation . Our results suggest that disrupting leucocyte–endothelial interactions may be a complementary therapeutic strategy . Several proteins associated with either risk of death or clinical severity lie in pathways targeted by existing drugs . PARP1 was identified as an important marker of current or future severe COVID-19 and also was associated with risk of death . PARP1 is associated with inflammatory and vascular disease ( Henning et al . , 2018 ) . PARP1 inhibitors are in use for cancer ( Rouleau et al . , 2010 ) , and our data suggest that re-purposing of PARP1 inhibition in COVID-19 should be explored further . IL33 was associated with both risk of death and clinical severity , and its receptor IL1RL1 ( ST2 ) was associated with clinical severity and identified as an important predictor of severe clinical course . Monoclonal antibodies against IL33 and its receptor are in late-stage development for asthma ( Corren , 2019 ) and could also be explored in COVID-19 . As discussed above , MPO was associated with clinical severity . MPO inhibitors ( Galijasevic , 2019 ) might have a role in reducing neutrophil-mediated tissue injury in COVID-19 . Finally , inhibitors of monocyte chemokines ( e . g . CCL2 ) and their receptors have been developed ( Vergunst et al . , 2008; Haringman et al . , 2006 ) , although drugging these pathways is made more challenging by molecular cross-talk . An important caveat is that we cannot determine whether the associations we observed are drivers of pathology in COVID-19 or simply reflect the downstream consequences of inflammation and tissue injury . Future studies using Mendelian randomisation analysis will provide a useful tool for assessing causality and prioritising drug targets . Other groups have studied the plasma or serum proteome in COVID-19 ( Shen et al . , 2020; Lucas et al . , 2020; Arunachalam et al . , 2020; Filbin et al . , 2020; Rodriguez et al . , 2020 ) , using either mass spectrometry or immunoassays including the Olink platform . Mass spectrometry is less sensitive than immunoassays and so it is likely to be unable to detect many of the cytokines measured here . Conversely , it can provide complementary information by measuring many proteins that our immunoassays did not target . A limitation of our study was that we used Olink panels that measured specific proteins selected on their relevance to inflammation , immunity , cardiovascular , and metabolic disease . This bias precluded formal pathway enrichment analysis of differentially abundant proteins . In general , our results had greater similarities to studies that used immunoassays over mass spectrometry ( Supplementary file 1i , j ) . 47 . 6% of proteins differentially expressed in COVID-19-positive versus -negative ESKD patients in our study were differentially expressed in COVID-19-positive versus -negative acute respiratory distress syndrome patients in the study of Filbin et al . , 2020 , who used a different Olink proteomics platform . Moreover , we observed consistent effect sizes ( Figure 4—figure supplement 1 ) . These similarities are striking given the difference in clinical populations and control groups; in Filbin et al . ’s report , the controls included patients with non-COVID-19 respiratory infections , whereas our control group did not have active infection . The concordance in proteins associated with COVID-19 severity within cases was even higher ( 81 . 8% ) . The similarities suggest a similar plasma proteomic signature of COVID-19 across different clinical populations , particularly the signature associated with severity . In summary , this study reveals proteins associated with COVID-19 infection and severity and demonstrates altered dynamic profiles between patients with severe disease and those with a more indolent course . Our results emphasise the importance of studying and targeting mechanisms that reduce the lung epithelial and endothelial damage to both alleviate the severity of the infection and reduce the chance of long-lasting complications . These data provide a valuable resource for therapeutic target prioritisation .
Severity scoring was performed based on WHO classifications ( WHO clinical management of COVID-19: Interim guidance 27 May 2020 ) adapted for clinical data available from electronic medical records . ‘Mild’ was defined as COVID-19 symptoms but no evidence of pneumonia and no hypoxia . ‘Moderate’ was defined as symptoms of pneumonia or hypoxia with oxygen saturation ( SaO2 ) greater than 92% on air , or an oxygen requirement no greater than 4 L/min . ‘Severe’ was defined as SaO2 less than 92% on air , or respiratory rate more than 30 per minute , or oxygen requirement more than 4 L/min . ‘Critical’ was defined as organ dysfunction or shock or need for high dependency or intensive care support ( i . e . the need for non-invasive ventilation or intubation ) . Severity scores were charted throughout a patient’s illness . We defined the overall severity/clinical course for each patient as the peak severity score that occurred during the patient’s illness . Plasma and serum proteomic measurements were performed using Olink proximity extension immunoassays ( https://www . olink . com/products/ ) . Five 92-protein multiplex Olink panels were run ( ‘inflammation’ , ‘immune response’ , ‘cardiometabolic’ , ‘cardiovascular 2’ , and ‘cardiovascular 3’ ) , resulting in 460 measurements per sample . Since a small number of proteins were measured on more than one panel , we measured a total of 436 unique proteins . The Olink assays were run using 88 samples/plate . All plates were run in a single batch . Plate layouts was carefully designed to avoid confounding of potential plate effects with biological or clinical variables of interest . To achieve this , we used an experimental design that combined ensuring case/control balance across plates with random selection of samples from each category and random ordering of allocation to wells . This is outlined in more detail as follows . We ensured that each plate contained a mixture of control and case samples . Specifically , a fixed proportion of each plate was designated for control samples . The allocation of specific control samples to each plate was performed using randomisation . For the case samples , we again used randomisation for plate assignment , with the constraint that once one sample from a given patient was allocated to a plate , all other longitudinal samples from that patient were assigned to same plate . Finally , once all the samples had been allocated to plates , the layout of samples within each plate was determined through a further randomisation step for well allocation . We used the Human Protein Atlas version 20 . 0 ( Uhlén et al . , 2015 ) for protein annotation ( Figure 1—figure supplement 1 ) . We performed enrichment analysis of the 436 proteins that we measured using string-db ( Szklarczyk et al . , 2019 ) . The data was normalised using standard Olink workflows to produce relative protein abundance on a log2 scale ( ‘NPX’ ) . Quality assessment was performed by ( 1 ) examination of Olink internal controls and ( 2 ) inspection of boxplots , relative log expression plots ( Gandolfo and Speed , 2018 ) , and PCA . Following these steps , three poor-quality samples were removed . In addition , five samples failed QC on a single proteomic panel only , with the remaining panels passing QC . For these samples , proteins on the panel that failed QC were set to missing , and the data for the remaining proteins was retained . PCA revealed no substantial impact of plate effects ( Figure 2—figure supplement 2 ) . Thirteen proteins were assayed more than once due to their inclusion in multiple Olink panels . For plasma , the median correlation between the assays was 0 . 986 with an inter-quartile range ( IQR ) of 0 . 974–0 . 993 and a range of 0 . 925–0 . 998 . For serum , the median correlation between the assays was 0 . 991 with an IQR of 0 . 952–0 . 995 and a range of 0 . 737–0 . 999 . We removed duplicate assays at random prior to subsequent analyses . For 11 ESKD controls , we had contemporaneous plasma and serum samples . To assess the comparability of these two matrices , we calculated the Pearson’s correlation coefficient between the assays for each protein ( Supplementary file 1k ) . Three hundred and forty-four of 436 ( 78 . 9% ) proteins had a Pearson’s r > 0 . 5 . We also report the variance of each protein in plasma and serum since low correlation may reflect low variance . The proteins with the lowest estimated Pearson correlation coefficient were AZU1 , STK4 , and TANK . We highlight that this comparison had small sample size ( only 11 samples ) and that the samples were from control patients without infection . Caution should be made in extrapolating these findings to the context of active infection where protein dynamic ranges may be different . Following QC , 0 . 22% data points were missing for the plasma dataset and 0 . 35% for the serum dataset . For analyses that required no missing values ( PCA and supervised learning ) , we imputed missing values as follows . The dataset was first scaled and centred , and missing values imputed using caret’s k-nearest neighbours method ( Kuhn , 2008 ) . The five closest samples ( by Euclidean distance ) were used to estimate each missing value . Singular value decomposition was used to perform PCA on the proteomic data from subcohort A ( plasma samples ) . We then used the loadings from subcohort A together with the proteomic data from subcohort B to calculate principal component scores . This enabled projection of subcohort B data into the PCA space of subcohort A . Differential protein abundance analyses between COVID-19 positive and negative samples were performed using linear mixed models , to account for the use of serial samples from the same individuals ( R lme4 package Bates et al . , 2015 ) . This analysis compared 256 samples from 55 COVID-19 patients with 51 non-infected patients ( one sample per non-infected patient ) . Age , sex , and ethnicity were included as covariates . We used a random intercept term to estimate the variability between individuals in the study and account for repeated measures . The regression model in R notation was:NPX covid−status+sex+age+ethnicity+ ( 1|individual ) where NPX represents the protein abundance and covid_status was a categorical variable ( infected/non-infected ) . Sex and ethnicity were also categorical variables . Age was a quantitative variable . We calculated P-values using a type 3 F-test in conjunction with Satterthwaite’s method for estimating the degrees of freedom for fixed effects ( Kuznetsova et al . , 2017 ) . The regression model was fitted for each of the 436 proteins individually . Multiple testing correction was performed using the Benjamini–Hochberg method and a 5% FDR used for the significance threshold . The same approach was used for subcohort B . This analysis comprised 52 serum samples from 46 COVID-19-positive patients versus 11 samples from non-infected patient samples ( one sample per non-infected patient ) . As sensitivity analyses , we repeated the differential abundance analyses between case and controls for the subcohort A adjusting for additional covariates and comparing this to the basic model ( i . e . using age , sex , and ethnicity alone ) . This was performed for each of the following parameters: diabetes status , cause of ESKD , and time to last haemodialysis . For testing the association of plasma proteins with the four-level WHO severity rating ( mild , moderate , severe , and critical ) within COVID-19-positive cases from subcohort A ( n = 256 samples from 55 patients ) , we used a similar linear mixed modelling approach to the COVID-19-positive versus -negative differential abundance analysis; for this analysis , the covid_status term was replaced by a severity variable encoded using orthogonal polynomial contrasts to account for ordinal nature of severity levels . As before , age , sex , and ethnicity were included as covariates . As a sensitivity analysis , we repeated the analysis with time to last haemodialysis as an additional covariate . The linear mixed modelling strategy was also employed for testing association of temporal clinical laboratory variables and protein levels , with the value of the clinical variable ( as a quantitative trait ) used in place of covid_status . Only COVID-19-positive patients were included in this analysis . Contemporaneous lab measurements were not available for all samples . This varied according to the clinical lab parameter . Some ( e . g . troponin , d-dimer ) were measured less frequently than full blood count and CRP . Details of the proportion of missing values for each lab parameter are included in Supplementary file 1g . We also calculated correlations between clinical laboratory variables and protein levels using the R package rmcorr , which determines the overall within-individual relationship among paired measures that have been taken on two or more occasion ( Bakdash and Marusich , 2017 ) . We performed testing of protein levels and ethnicity separately in COVID-19-negative ESKD patients and COVID-19-positive ESKD patients . These analyses were limited to individuals who were White , South Asian ( Indian , Pakistani , or Bangladeshi ancestry ) , or Black as there were too few individuals from other ethnic groups for meaningful interpretation . For COVID-19-negative patients ( one sample per patient ) , we performed linear regression for each protein with ethnicity as the predictor variable and age and sex as covariates . For COVID-19-positive patients , we used a linear mixed model to account for serial samples from the same individual , again with age , and sex as covariates . We used the Benjamini–Hochberg method to control the FDR at 5% for all statistical analyses . To provide additional support that the Benjamini–Hochberg procedure was providing adequate control of the FDR , we also used the plug-in procedure ( Hastie et al . , 2001 ) as an alternative method to estimate the FDR , as described below . We implemented a similar approach for the testing the association of proteins with severity scores within cases . Using this method , the estimated FDR for the case versus control analysis was 0 . 062 and for the severity analysis 0 . 057 , indicating that we had appropriately controlled the FDR . As a complementary analysis , based on the approach of Filbin et al . , 2020 , we estimated the empirical p-value for the likelihood of observing as many significant proteins as we identified in the real data if the null hypothesis of no differentially abundant proteins in cases versus controls were true . We again used 100 , 000 permutations of the case–control labels to estimate the null distribution . We performed Benjamini–Hochberg adjustment on the nominal p-values of each permutation and counted the number of proteins that were significant ( adjusted p-value<0 . 05 ) in each permutation . The distribution of the number of proteins declared significant is shown in Figure 3—figure supplement 2a; on no occasion in 100 , 000 permutations did we observe more proteins declared significant than in the real data . We can thus state that the empirical p-value ( the fraction of permutation runs where we observed ≥ the number of associations in the real data ) is less than 1/100 , 000 = 1×10−5 . We also applied this method to the association testing of proteins with severity scores within cases ( Figure 3—figure supplement 2b ) . Again , on no occasion in 100 , 000 permutations did we observe more proteins declared significant than in the real data ( empirical p-value<1×10−5 ) . Random forest models were fitted using R’s randomForest and caret packages ( Kuhn , 2008; Leo , 2001 ) . Data was centred , scaled , and imputed as in Missing values with the caveat that , during cross-validation , the pre-processing procedure was first applied on the resampled ( training ) data before the same method was applied without re-calculation to the holdout ( test ) set . To estimate model accuracy , we used fourfold cross-validation . The cross-validation procedure was repeated 100 times . The model’s parameters were kept constant at 500 trees and an mtry value ( number of proteins randomly sampled as candidates at each node ) calculated as the square root of the number of features . After parameter estimation , we fitted a final model trained using the entirety of the dataset . This model was used for subsequent feature extraction . Random forest feature extraction was carried out using the R randomForestExplainer package . We made use of the following importance measures: accuracy decrease ( the average decrease in prediction accuracy upon swapping out a feature ) , number of trees ( the number of trees with a node corresponding to a feature ) , and mean minimal depth ( the average depth at which a node corresponding to a feature occurs ) . Three models were generated with different input features: ( 1 ) proteomic data alone; ( 2 ) clinical parameters alone; and ( 3 ) proteomic data and clinical parameters . Clinical parameters included sex , age , ethnicity , cause of ESKD , comorbidities , smoking status , radiological evidence of pulmonary infiltrates , and clinical laboratory tests . Following scaling and centring , we fitted linear mixed models for each protein to capture the temporal trajectories of each individual . A polynomial spline of degree two was used to model protein concentration with respect to time ( from symptom onset , measured in days ) ; the spline was fitted for samples that were taken between 1 and 28 days from first symptoms , inclusive . Proteomic data after that point was censored . We estimated both random intercepts and random slopes for each individual , as per the following R formula notation:NPX∼time+ ( time|individual ) These were joined to a Cox regression model using the jointModel package ( Rizopoulos , 2010 ) in order to estimate the association of each protein with risk for death . P-values were calculated using a Wald test for the association between the linear mixed model and Cox regression . Benjamini–Hochberg adjustment was applied , with an adjusted p-value of 0 . 05 used as the significance threshold . We also used linear mixed models to estimate the temporal profile of each protein . For this longitudinal analysis , we explicitly modelled the time from first symptoms . We set up the model to test for each protein ( 1 ) whether the protein significantly change over time and ( 2 ) whether the protein changes over time differently in individuals with a mild versus severe disease course . The latter was performed statistically by testing for an interaction effect between time and clinical course . For the purposes of this analysis , we binarised patients into severe or non-severe clinical course according to the peak WHO severity disease of their illness . Patients with a peak WHO score of mild or moderate were considered non-severe and those with a peak score of severe or critical were considered severe . We then used R’s bs function to fit a polynomial spline of degree two to model protein concentration with respect time ( from symptom onset , measured in days ) ( Perperoglou et al . , 2019 ) . The spline was fitted for samples that were taken between 1 and 21 days from symptom onset , inclusive . We estimated random slopes with respect to time , in addition to random intercepts , to account for each individual’s unique disease course . For each protein , we fitted the following model ( R notation ) :NPX∼time∗severity+sex+age+ethnicity+ ( time|individual ) To identify proteins that changed significantly over time , we examined the p-values for the main effect of time . To identify proteins with distinct temporal profiles between severe and non-severe cases , we examined the p-values for the time × severity interaction term . For each of these two research questions , p-values were adjusted for the multiple proteins tested using the Benjamini–Hochberg method and 5% FDR used as the significance threshold . Code is available in the following GitHub repository: https://github . com/jackgisby/longitudinal_olink_proteomics; Gisby , 2021; copy archived at swh:1:rev:32f08137859d44707ec4f086eed9af9b9ee91a87 . | COVID-19 varies from a mild illness in some people to fatal disease in others . Patients with severe disease tend to be older and have underlying medical problems . People with kidney failure have a particularly high risk of developing severe or fatal COVID-19 . Patients with severe COVID-19 have high levels of inflammation , causing damage to tissues around the body . Many drugs that target inflammation have already been developed for other diseases . Therefore , to repurpose existing drugs or design new treatments , it is important to determine which proteins drive inflammation in COVID-19 . Here , Gisby , Clarke , Medjeral-Thomas et al . measured 436 proteins in the blood of patients with kidney failure and compared the levels between patients who had COVID-19 to those who did not . This revealed that patients with COVID-19 had increased levels of hundreds of proteins involved in inflammation and tissue injury . Using a combination of statistical and machine learning analyses , Gisby et al . probed the data for proteins that might predict a more severe disease progression . In total , over 200 proteins were linked to disease severity , and 69 with increased risk of death . Tracking how levels of blood proteins changed over time revealed further differences between mild and severe disease . Comparing this data with a similar study of COVID-19 in people without kidney failure showed many similarities . This suggests that the findings may apply to COVID-19 patients more generally . Identifying the proteins that are a cause of severe COVID-19 – rather than just correlated with it – is an important next step that could help to select new drugs for severe COVID-19 . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"medicine",
"immunology",
"and",
"inflammation"
] | 2021 | Longitudinal proteomic profiling of dialysis patients with COVID-19 reveals markers of severity and predictors of death |
The actin cortex plays a pivotal role in cell division , in generating and maintaining cell polarity and in motility . In all these contexts , the cortical network has to break symmetry to generate polar cytoskeletal dynamics . Despite extensive research , the mechanisms responsible for regulating cortical dynamics in vivo and inducing symmetry breaking are still unclear . Here we introduce a reconstituted system that self-organizes into dynamic actin cortices at the inner interface of water-in-oil emulsions . This artificial system undergoes spontaneous symmetry breaking , driven by myosin-induced cortical actin flows , which appears remarkably similar to the initial polarization of the embryo in many species . Our in vitro model system recapitulates the rich dynamics of actin cortices in vivo , revealing the basic biophysical and biochemical requirements for cortex formation and symmetry breaking . Moreover , this synthetic system paves the way for further exploration of artificial cells towards the realization of minimal model systems that can move and divide .
The actin cytoskeleton plays a central role in many cellular processes including polarization , cell shape determination , intracellular transport , cell division and movement ( Pollard and Cooper , 2009 ) . The structure and function of the cytoskeleton arise from the self-organized dynamics of numerous molecular building blocks . This self-organization spans several orders of magnitude in space and time and involves a complex interplay between biochemical and biophysical processes; A myriad of proteins interact with the actin cytoskeleton and influence its behavior , in a manner that is dependent on the global mechanical properties of the network but at the same time determines it ( Lecuit and Lenne , 2007; Pollard and Cooper , 2009; Mullins and Hansen , 2013 ) . Despite the significant progress in uncovering the molecular details underlying cytoskeletal dynamics , the principles governing large-scale coordination and polarization of the cytoskeleton are still not well-understood . The realization of biomimetic systems that reconstitute cellular processes in vitro , detached from the complexity of the cell , is a powerful approach for dissecting complex cellular phenomena . In particular , in vitro experiments have significantly advanced our understanding of the molecular requirements and the biophysical principles underlying actin-based motility and cytoskeletal organization in bulk ( Welch et al . , 1998; Cameron et al . , 1999; Loisel et al . , 1999; Gardel et al . , 2004; Van Der Gucht et al . , 2005; Bendix et al . , 2008; Field et al . , 2011; Kohler et al . , 2012 ) , and more recently in cell-sized compartments ( Pontani et al . , 2009; Stachowiak et al . , 2009; Pinot et al . , 2012; Sanchez et al . , 2012; Carvalho , 2013b ) . However , we are still far from understanding the complexity of cytoskeletal dynamics in vivo , and recapitulating even basic cellular phenomena such as polarization , division and directed movement in synthetic systems remains an outstanding challenge . The actin cytoskeleton undergoes continuous turnover and remodeling which are essential for its ability to perform its cellular tasks ( Pollard and Cooper , 2009 ) . In particular , the thin cortical actin shell underneath the cell membrane undergoes continuous assembly and disassembly processes , catalyzed by nucleation-promoting factors localized at the membrane and disassembly factors ( Fritzsche et al . , 2013 ) . Among the nucleation-promoting factors , Arp2/3 which nucleates branched networks localizes to cortical actin networks ( Machesky et al . , 1994 ) and is essential for cortex formation ( Bovellan , 2012 ) . Formins , which nucleate linear filaments , were also found to localize to cortical actin networks , yet their role is still not entirely clear ( Bovellan , 2012; Fritzsche et al . , 2013 ) . A host of actin binding proteins , including myosin motors , tethering proteins and various crosslinkers , further contribute to the spatio-temporal organization of actin cortices in cells ( Munro et al . , 2004 ) . This dynamic remodeling is responsible for the global rearrangements of the actin cortex which are essential for its function during polarization and movement ( Salbreux et al . , 2012 ) . Polarization in cells typically occurs in response to internal or external cues . Yet , the onset of polarity often reflects an inherent instability mechanism which can lead to symmetry breaking in the absence of any directional cues ( Van Oudenaarden and Theriot , 1999; Verkhovsky et al . , 1999; Wedlich-Soldner et al . , 2003; Boukellal et al . , 2004; Van Der Gucht et al . , 2005; Carvalho et al . , 2013a , 2013b ) . While biochemical signaling pathways appear to be important for transducing directional cues , the instability , at least in some cases , can be primarily mechanical ( Mullins , 2010; Van Der Gucht and Sykes , 2009 ) . For example , reconstitution of actin-based motility of bacterial pathogens such as Listeria monocytogenes revealed that directional movement can arise from spontaneous symmetry breaking within the actin network which ruptures and forms a polar comet-tail ( Cameron et al . , 1999; Van Oudenaarden and Theriot , 1999; Boukellal et al . , 2004; Van Der Gucht et al . , 2005 ) . In the cell cortex , the uniform actin shell can break symmetry to form a polar network which is essential for generating and maintaining cell polarity ( Munro et al . , 2004; Cowan and Hyman , 2007; Munro and Bowerman , 2009; Goehring et al . , 2011; Salbreux et al . , 2012 ) and inducing directional cell movement ( Hawkins et al . , 2011; Poincloux et al . , 2011 ) . Despite their importance , the factors controlling symmetry breaking in the cell cortex remain poorly understood . An attractive hypothesis is that the instability leading to cortical symmetry breaking is also mechanical as in the case of comet tail motility ( Van Oudenaarden and Theriot , 1999; Van Der Gucht et al . , 2005; Dayel et al . , 2009 ) , yet the mechanisms involved appear different . Comet tail formation is a myosin–independent process driven by actin polymerization ( Loisel et al . , 1999 ) , whereas cortical symmetry breaking appears to rely on myosin as the main force generating element; myosin contraction which generates cortical actin flows has been shown to be essential for inducing polarization and defining the anterior–posterior axis during the early stages of embryogenesis in Caenorhabditis elegans and in other species ( Munro et al . , 2004; Mullins , 2010; Munro and Bowerman , 2009; Mayer et al . , 2010 ) . Reconstitution of actomyosin networks is a valuable tool for dissecting the mechanisms underlying cortical symmetry breaking . Recent in vitro experiments revealed that small changes in the relative amounts of myosin and crosslinkers , or in their activity , can lead to a sharp transition in the overall contractile behavior of reconstituted actin networks ( Bendix et al . , 2008; Kohler et al . , 2012 ) . However , to emulate cortical dynamics it is essential to incorporate actin turnover dynamics , couple the actin network to a soft interface ( Murrell and Gardel , 2012 ) , and reproduce a cell-like geometry ( Pontani et al . , 2009; Pinot et al . , 2012; Carvalho et al . , 2013b ) . Recent reports ( Carvalho et al . , 2013a , 2013b ) suggest that myosin is responsible for building tension in actin networks tethered to soft interfaces , and can induce cortical rupture and generate asymmetry . Here we report the realization of artificial actin cortices that exhibit spontaneous symmetry breaking and display cell-like behavior . This novel synthetic system provides a basis for studying cortical actin dynamics and polarization in a simplified environment detached from the complexity of the living cell . Moreover , it presents an interesting example of a self-organized , far-from-equilibrium , active system that utilizes chemical energy to polarize and generate directional forces .
The formation of artificial actin cortices was induced by localizing the ActA protein from the pathogenic bacteria L . monocytogenes to the inner interface of water-in-oil emulsions ( Figure 1 ) . ActA is a nucleation-promoting factor , known for its role in activating the Arp2/3 complex and inducing nucleation of branched actin networks ( Welch et al . , 1998 ) . Such branched actin networks are found in the cortices of many cell types ( Machesky et al . , 1994; Medalia et al . , 2002; Bovellan , 2012; Fritzsche et al . , 2013 ) . A soluble ActA construct was purified from L . monocytogenes and conjugated to a fluorescent hydrophobic linker made with Bodipy-FL , to generate an amphiphilic complex ( ‘Materials and methods’ ) . This amphiphilic ActA complex was mixed with diluted Xenopus laevis egg cytoplasmic extract supplemented with labeled actin , and encapsulated within droplets surrounded by mineral oil ( ‘Materials and methods’ ) . The amphiphilic ActA spontaneously localized to the water–oil interface following droplet formation . Upon localization , ActA induced the assembly of a cortical actin network at the inner surface of the aqueous droplets ( Figure 1A , D ) . Cortex assembly initiated in patches scattered throughout the interface , consistent with the autocatalytic nature of Arp2/3 mediated actin filament nucleation ( Mullins and Hansen , 2013; Figure 1F ) . The system reached a steady state after ∼10 min , with the formation of thin ( ∼1 . 5 µm; Figure 1—figure supplement 1 ) , homogenous actin cortices . The assembled cortices are dynamic , as illustrated by the rapid cortical recovery following photobleaching ( Figure 1E , Figure 1—figure supplement 2; Video 1 ) . Thus , the cortical steady state reflects a balance between actin assembly and disassembly at the interface , as in cells ( Fritzsche et al . , 2013 ) . Control experiments with soluble ActA , or in the absence of ActA , resulted in an essentially uniform actin distribution throughout the emulsion droplets , with no apparent cortical localization ( Figure 1B , C ) , indicating that cortex formation depends on the localization of a nucleation-promoting factor at the interface . 10 . 7554/eLife . 01433 . 003Figure 1 . Reconstitution of actin cortices within water-in-oil emulsions . ( A ) Spinning disk confocal images of bodipy-conjugated ActA ( left ) and rhodamine-labeled actin ( right ) in a water-in-oil emulsion . The bodipy-conjugated ActA localizes to the water–oil interface , and induces the formation of an actin cortex there . The actin signal reflects the distribution of actin monomers and filaments . ( B ) Images of AlexaFluor488-conjugated ActA ( left panel ) and rhodamine-labeled actin ( right ) in a water-in-oil emulsion . The hydrophilic ActA and the actin remain dispersed throughout the emulsion . ( C ) Image of rhodamine-labeled actin in a water-in-oil emulsion in the absence of ActA . The actin is distributed within the emulsion . ( D ) Schematic illustration of the actin cortex formed at the inner interface of a water-in-oil emulsion . A bright-field image ( top left ) and a scheme ( top right ) of an actin cortex at the inner interface of an aqueous droplet surrounded by oil . The zoomed scheme ( bottom ) illustrates the localization of the amphiphilic bodipy-conjugated ActA to the water–oil interface , which leads to local activation of Arp2/3 and nucleation of a cortical actin network . ( E ) Scanning confocal images showing cortical recovery in a photobleaching experiment ( Video 1 ) . The time after photobleaching is indicated . ( F ) Spinning disk confocal images from a time-lapse video showing the formation of a homogenous actin cortex . The time after droplet formation is indicated . Scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 00310 . 7554/eLife . 01433 . 004Figure 1—figure supplement 1 . Actin and ActA distributions in water-in-oil emulsions . The fluorescence intensity cross-sections of rhodamine-labeled actin ( top ) and bodipy-conjugated ActA ( bottom ) extracted from spinning disk confocal images ( inset ) are plotted as a function of distance from the emulsion boundaries . Data for individual emulsions incubated at 30°C ( grey lines ) is shown together with the population average ( thick lines ) . The cross-section intensity for each emulsion was normalized to the mean intensity along the emulsion boundary ( Figure 2A , B; bottom ) . The actin signal reflects the distribution of actin monomers and filaments . A homogeneous cortical actin shell with a thickness of ∼1 . 5 µm is apparent . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 00410 . 7554/eLife . 01433 . 005Figure 1—figure supplement 2 . Analysis of photobleaching experiments . FRAP experiments were performed on reconstituted cortices at 30°C using a scanning confocal microscope ( ‘Materials and methods’ ) . A region of the cortex at the bottom of an emulsion was bleached and its recovery was followed over time . Analysis of the recovery following photobleaching was done using the ZEN software by fitting to a single exponent , and using an unbleached region of the cortex as a reference . Left: histograms of the half time for recovery are shown . The average half time at 30°C was τ½ = 81 ± 18 s ( mean ± STD; N = 18 ) . Right: graphs showing the normalized intensity as a function of time following photobleaching . The mean normalized intensity ( mean ± STD ) averaged over different emulsions at 30°C is shown . The solid line depicts a fit to a single exponent . The cortices typically exhibited only partial recovery after bleaching , usually returning to ∼50% of their initial intensity . This partial recovery appeared to be a result of photodamage during the bleaching step . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 00510 . 7554/eLife . 01433 . 006Figure 1—figure supplement 3 . Actin cortices formed with different types of labeled actin . Spinning disk confocal images of actin cortices prepared with different types of labeled actin , as indicated . Similar behavior was observed in all cases . Scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 00610 . 7554/eLife . 01433 . 007Video 1 . Cortical recovery after photobleaching . This video shows scanning confocal images ( at a single z-plane ) of rhodamine-labeled actin within a water-in-oil emulsion incubated at 30°C . A small region of the cortex was photobleached using a high intensity laser pulse . The video follows the dynamics of cortex recovery over several minutes . The lower panel depicts the whole droplet . The field of view is 132 µm wide . The upper panel shows a zoomed area in the proximity of the bleached region . The time relative to the bleaching event is indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 007 Our reconstituted actin cortices can exhibit spontaneous symmetry breaking ( Figure 2 ) . The actin network assembled within emulsion droplets incubated at 20°C broke symmetry and formed polar actin caps ( Figure 2A ) , in contrast to the homogenous spherical cortices formed at higher temperature ( 30°C; Figure 2B ) . We followed the dynamics of actin cap formation as a function of time ( Figure 2C , D; Video 2 ) . Initially , actin assembly appeared nearly uniform throughout the interface , but within minutes we observed the onset of a directional actin network flow along the interface . The cap formation process developed over ∼10–20 min , with cortical flows initiating at ∼1–4 µm/min ( Figure 2E ) , and slowing down over time , as the cap contracted ( Figure 2D ) . The local turnover due to actin assembly and disassembly continued throughout the cap formation process , as illustrated by photobleaching experiments in which a cortical region within the asymmetric actin cap was bleached ( Video 3 ) . Cortical recovery in the cap was observed within minutes ( Figure 2—figure supplement 2 ) , similar to the recovery seen in photobleaching experiments of homogenous cortices ( Figure 1—figure supplement 2 ) . Despite this rapid turnover at the molecular level , the position of the actin caps remained stable over much longer time scales , typically exhibiting persistent polarization for hours ( Figure 2—figure supplement 3 ) . Notably , the magnitude of the cortical actin flows and the characteristic length and time scales for polarization observed in our artificial cortices are similar to those seen in developing embryos ( Munro et al . , 2004; Cowan and Hyman , 2007; Munro and Bowerman , 2009; Mayer et al . , 2010 ) . 10 . 7554/eLife . 01433 . 008Figure 2 . Temperature-dependent symmetry breaking in reconstituted actin cortices . ( A and B ) Spinning disk confocal images of bodipy-conjugated ActA ( top left ) and rhodamine-labeled actin ( top right ) in water-in-oil emulsions incubated at 20°C ( A ) or at 30°C ( B ) . A polar actin cortex with a single cap is observed at 20°C , in contrast to the uniform spherical cortex observed at 30°C . The fluorescence intensity profiles as a function of the angle along the contour are shown for different water-in-oil emulsions at 20°C ( A; bottom panels ) or at 30°C ( B ) . Data for individual emulsions normalized to the mean intensity in each contour ( grey lines ) is shown together with the population average ( thick line ) . The actin distributions ( bottom right ) are peaked at the center of the cap for all the emulsions at 20°C , in contrast to the uniform distributions at 30°C . The ActA distributions ( bottom left ) are essentially flat at both temperatures . The residual ActA observed outside the polar caps at 30°C suggests that ActA is in excess in our system ( Figure 2—figure supplement 5 ) . ( C ) Spinning disk confocal images from a time-lapse video ( Video 2 ) showing the development of a polar actin cap in an emulsion incubated at 20°C . The time after droplet formation is indicated . Cortex assembly starts uniformly along the interface , and becomes polar within minutes . ( D ) A kymograph of the video in ( C ) , showing the fluorescence intensity along the contour of the emulsion ( vertical axis ) as a function of time ( horizontal axis ) . The cortical actin flow which slows down as the cap contracts is evident in the kymograph . ( E ) A histogram of the initial cortical flow speeds measured for a population ( N = 32 ) of emulsions followed by time-lapse microscopy during the symmetry breaking process as in ( C ) . Inset- a schematic illustration of the cortical actin flow during the symmetry breaking process . ( F ) Bulk assay for actin network contractility . The relative width of the actin containing strip in comparison to the width of the incubation chamber after 30 min is plotted as a function of temperature . At temperatures <25°C the network contracts into a thin strip ( left inset ) , while at higher temperatures ( >25°C ) no bulk contractility is observed ( right inset ) . ( G ) Samples were incubated for 30 min at 30°C to generate homogenous cortices , and then moved to 20°C . Cortex polarity developed over time , with symmetry breaking often initiating in more than one position ( Figure 2—figure supplement 4A ) . The relative proportion of symmetric ( dark ) and polar cortices ( light ) with one ( bare ) or multiple ( stripes ) actin caps , at different time windows following the temperature shift are shown , in comparison to control samples incubated at 30°C or at 20°C . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 00810 . 7554/eLife . 01433 . 009Figure 2—figure supplement 1 . Actin and ActA distributions in water-in-oil emulsions at 20°C . The fluorescence intensity cross-sections of rhodamine-labeled actin ( top ) and bodipy-conjugated ActA ( bottom ) extracted from spinning disk confocal images ( inset ) are plotted as a function of distance from the emulsion boundaries . Data for individual emulsions incubated at 20°C ( grey lines ) is shown together with the population average ( thick lines ) . The cross-section intensity for each emulsion was normalized to the mean intensity along the emulsion boundary . The cortical actin distribution is peaked on one side at 20°C , in contrast to the homogenous cortices formed at 30°C . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 00910 . 7554/eLife . 01433 . 010Figure 2—figure supplement 2 . Analysis of photobleaching experiments at 20°C . FRAP experiments were performed on reconstituted cortices at 20°C using a scanning confocal microscope ( ‘Materials and methods’ ) . A region of the cortex at the bottom of an emulsion was bleached and its recovery was followed over time . Analysis of the recovery following photobleaching was done using the ZEN software by fitting to a single exponent , and using an unbleached region of the cortex as a reference . Left: histograms of the half time for recovery are shown for emulsions . The average half time at 20°C was τ½ = 73 ± 24 s ( mean ± STD; N = 9 ) . The average half times at 30°C and at 20°C were not significantly different ( p>0 . 1 ) . Right: graphs showing the normalized intensity as a function of time following photobleaching . The mean normalized intensity ( mean ± STD ) averaged over different emulsions at 20°C is shown . The solid line depicts a fit to a single exponent . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 01010 . 7554/eLife . 01433 . 011Figure 2—figure supplement 3 . The actin cap remains stable following symmetry breaking . ( A ) Spinning disk confocal images from a time lapse video following the dynamics of cap evolution over 2 hr . Once a stable cap is formed , its position remains relatively stable over time . ( B ) A histogram of the net cap rotation speed measured for a population ( N = 30 ) of emulsions followed by time-lapse microscopy after the symmetry breaking process . The cap position was determined along the perimeter as the center of mass of the intensity distribution along the boundary . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 01110 . 7554/eLife . 01433 . 012Figure 2—figure supplement 4 . A temperature shift from 30°C to 20°C leads to symmetry breaking . Samples were incubated for 30 min at 30°C to generate homogenous cortices , and then moved to 20°C . ( A ) Spinning disk confocal images of the actin distribution in representative emulsions at different time points following the temperature shift . ( B and C ) The actin ( B ) and ActA ( C ) density profiles within different time windows following the shift in temperature are shown as a function of the angle along the contour . The time window after the temperature shift for each graph is indicated . Data for individual emulsions normalized to the mean intensity in each contour ( grey lines ) is shown together with the population average within the time window ( thick lines ) . Cortex polarity develops over time , with symmetry breaking often initiating in more than one position ( e . g . , image at 22′ ) . Both the ActA and the actin distributions become polar , with the actin distribution exhibiting a more pronounced peak . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 01210 . 7554/eLife . 01433 . 013Figure 2—figure supplement 5 . The temperature–dependent behavior of artificial cortices as a function of ActA concentration . Artificial actin cortices were prepared with different ActA concentrations which were higher ( 3 µM ) or lower ( 0 . 5 µM ) than the typical ActA concentration used ( 1 . 5 µM ) . ( A and B ) Bar plots showing the average fluorescence ( ±SEM ) intensity of bodipy-conjugated ActA ( A ) and rhodamine-labeled actin ( B ) in artificial cortices prepared at different ActA concentrations and incubated at 20°C or 30°C as indicated . The fluorescence intensity along the interface was extracted from spinning disk confocal images , and the mean cortex intensity was averaged over a population of emulsions for each condition . The intensity values were normalized to the average intensity with 1 . 5 µM ActA at 30°C . The ActA density at the interface was higher at 20°C and increased with the bulk concentration of ActA . The cortical actin density was also higher at 20°C , but was nearly independent of the ActA concentration . ( C ) Bar plot showing the relative proportion of symmetric ( dark ) and polar cortices ( light ) with one ( bare ) or multiple ( stripes ) actin caps for samples of emulsions made with different amounts of ActA and incubated at 20°C or 30°C . The proportion of symmetric and polar cortices does not depend on the ActA concentration within the range of concentrations examined . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 01310 . 7554/eLife . 01433 . 014Figure 2—figure supplement 6 . ActA dynamics at the interface . Scanning confocal images showing the ActA distribution in an emulsion incubated at 20°C . A small patch of the ActA at the interface was photobleached and the recovery of the fluorescence signal was followed over time . The ActA appeared to recover homogenously within the bleached region . The time from the bleaching event is indicated . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 01410 . 7554/eLife . 01433 . 015Video 2 . Actin cortical flow during symmetry breaking . This video shows spinning disc confocal images ( at a single z-plane ) of rhodamine-labeled actin within a water-in-oil emulsion incubated at 20°C . The actin signal reflects the distribution of actin monomers and filaments . The actin cortex initiates in patches throughout the droplet’s interface . Cortical actin flows appear at the interface within minutes , leading to the formation of a polar actin cap . The field of view is 55 µm wide , and the time from droplet formation is indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 01510 . 7554/eLife . 01433 . 016Video 3 . Cortical recovery after photobleaching in an asymmetric actin cap . This video shows scanning confocal images ( at a single z-plane ) of rhodamine-labeled actin within a water-in-oil emulsion incubated at 20°C . A small region at the tip of the actin cap was photobleached using a high intensity laser pulse . The video follows the dynamics of cortex recovery over several minutes . The lower panel depicts the whole droplet . The field of view is 132 µm wide . The upper panel shows a zoomed area in the proximity of the bleached region . The time relative to the bleaching event is indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 016 To examine whether the temperature dependent contractile behavior we observe in the emulsions is related to the confined geometry as suggested in Pinot et al . ( 2012 ) , or reflects changes in bulk contractility ( Bendix et al . , 2008; Kohler et al . , 2012 ) , we designed a bulk assay to assess the behavior of our system . We introduced the same actin machinery into chambers ( ∼100 µm × 100 µm × 5 mm ) , replacing the amphiphilic ActA with fixed L . monocytogenes which express ActA on their surface ( ‘Materials and methods’ ) . At high temperatures , the fixed bacteria remained distributed throughout the sample , exhibiting comet tail motility , and no overall contraction was observed ( Figure 2F , right ) . At lower temperatures , we observed rapid contraction of the entire sample , which concentrated actin filaments in the extract and bacterial comet tails into a narrow strip of actin gel ( Figure 2F , left ) . The transition between non-contractile and contractile behavior occurred abruptly at ∼25°C . We found that temperature controls the onset of symmetry breaking in our reconstituted actin cortices ( Figure 2A , B ) . To investigate whether a temperature shift could lead to symmetry breaking in pre-assembled actin cortices , we first incubated emulsions for 30 min at 30°C to prepare homogenous cortices , and then shifted the temperature to 20°C . Cortical polarity developed as a function of time following the temperature shift ( Figure 2G , Figure 2—figure supplement 4 ) . Initially , the actin network density along the droplet interface was uniform . Within minutes , the homogenous cortical network began to rupture , often in more than one place ( Figure 2—figure supplement 4A; 22′ ) , leading to the formation of gaps in the actin cortex . After ∼30 min , a polar distribution of actin was observed in all emulsions , characterized in most cases by a single polar cap , similar to the distributions measured in emulsions prepared at 20°C ( Figure 2G ) . At longer times , we often observed accumulation of actin in the interior of the droplet ( near the cap ) driven by the actin flows initiated at the interface ( Figure 2—figure supplement 4A; 75′ ) . The cortical flows also generated a bias in the distribution of the nucleation-promoting factor ActA towards the cap ( Figure 2—figure supplement 4C ) . Similar recruitment of nucleation-promoting factors by polymerizing actin networks was previously observed in comet tail motility on soft interfaces ( Boukellal et al . , 2004 ) . We further characterized the temperature–dependent behavior of the artificial actin cortices as a function of ActA concentration ( Figure 2—figure supplement 5 ) . As expected , the ActA density at the water–oil interface increased as a function of the initial bulk ActA concentration used ( Figure 2—figure supplement 5A ) . The ActA densities at the interface were also higher at 20°C , which can be understood from simple thermodynamic considerations; the dynamic localization of ActA at the interface ( Figure 2—figure supplement 6 ) depends on the interplay between enthalpy and entropy which favors localization at lower temperatures . However , within the range of ActA concentrations examined , the intensity of the actin cortices and their symmetry breaking properties were essentially unaffected by the ActA concentration ( Figure 2—figure supplement 5B , C ) , suggesting that ActA is in excess . These results suggest that the temperature-dependent symmetry breaking observed reflects the changing properties of the cortical actin networks , rather than differences in ActA localization . The intensity of the actin cortices was found to be higher at lower temperatures , most likely due to the temperature-induced changes in the balance between actin assembly and disassembly dynamics in our system , also exemplified by the increased lengths of actin comet tails at 20°C compared to 30°C in the same motility mix ( data not shown ) . Our reconstituted system allowed us to investigate the biochemical requirements for symmetry breaking in the artificial cortices ( Figure 3 ) . Myosin motors are involved in contractile behavior and symmetry breaking in the cortex of cells ( Munro et al . , 2004; Cowan and Hyman , 2007; Munro and Bowerman , 2009; Van Der Gucht and Sykes , 2009 ) . To examine the role of myosin in our reconstituted system , we depleted myosin II from the Xenopus egg extracts by immunodepletion ( Bendix et al . , 2008 ) ( ‘Materials and methods’ ) . Emulsions made from myosin-depleted extracts did not exhibit cortical actin flows and maintained a symmetric actin shell when incubated at 20°C ( Figure 3A , C ) , in contrast to the behavior observed at the same temperature in the presence of myosin . Symmetry breaking was not affected in control experiments with mock immunodepletion ( Figure 3B ) . To further establish the essential role of myosin in the symmetry breaking process , we added different amounts of purified myosin to myosin–depleted extracts and showed that the symmetry breaking is restored ( Figure 3B , C ) . The cap morphology was dependent on the amount of added myosin; as the concentration of myosin increased , we observed more cortices with multiple caps , rather than a single cap ( Figure 3C ) . Moreover , at even higher myosin concentrations ( 0 . 66 µM ) most of the cortices appear fractured in many places with multiple small interconnected cortical patches which exhibited jittery motion . The decrease in the typical size of cortical patches with increasing amounts of myosin is probably due to the decrease in the contractile unit size as a function of myosin concentration ( Thoresen et al . , 2013 ) . Overall our results suggest that the symmetry breaking observed in our artificial cortices is driven by the same myosin-induced mechanism found in cells ( Munro et al . , 2004; Mayer et al . , 2010 ) . 10 . 7554/eLife . 01433 . 017Figure 3 . The effect of myosin and crosslinkers on symmetry breaking . ( A ) Spinning disk confocal images of actin cortices in water-in-oil emulsions made with myosin–depleted ( upper panel ) or mock-depleted ( lower panel ) extracts and incubated at 20°C . Myosin depletion eliminates symmetry breaking , whereas a mock depletion does not . ( B ) Spinning disk confocal images of actin cortices made with myosin-depleted extracts supplemented with different amounts of purified myosin . At intermediate myosin concentrations ( 0 . 24 µM; left images ) the actin cortices typically have one or few actin caps . At higher myosin concentrations ( 0 . 66 µM; right image ) the actin cortices appear fractured with many discrete puncta along the interface . ( C ) Bar plot showing the relative proportion of symmetric ( dark ) , polar cortices ( light ) with one ( bare ) or multiple ( stripes ) actin caps , or fractured cortices ( dense stripes ) . Data is shown for a control sample of emulsions at 20°C , in comparison to samples prepared at the same temperature with myosin-depleted , mock-depleted extracts or myosin-depleted extracts supplemented with different amounts of purified myosin . ( D ) Spinning disk confocal images of actin cortices formed with extracts supplemented with different amounts of α-actinin or filamin crosslinkers and incubated at 30°C . ( E ) Bar plot showing the relative proportion of symmetric and polar cortices ( as in ( C ) ) for samples of emulsions made with different amounts of crosslinkers ( 2–8 µM α-actinin; 4 µM filamin ) relative to a control sample at 30°C . The addition of crosslinkers induces symmetry breaking at 30°C . Note the large fractions of emulsions at higher α-actinin concentrations that exhibit multiple actin caps . Scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 017 The contractile behavior of actin networks is also dependent on the degree of crosslinking within the network; Sufficient network connectivity is required to enable myosin motors to generate forces rather than merely induce sliding of filaments , whereas excessive crosslinking stiffens the network and prevents contraction ( Bendix et al . , 2008; Kohler et al . , 2012 ) . We used two different kinds of actin crosslinkers which are known to localize to cortical actin networks in vivo ( Charras et al . , 2006 ) , α-actinin which preferentially generates anti-parallel actin bundles and filamin which is a more flexible crosslinker that can bind to disordered actin networks . We showed that symmetry breaking can be induced at 30°C by adding crosslinkers ( Figure 3D , E ) . Increasing α-actinin concentration led to the formation of polar cortices in a concentration dependent manner . At low concentrations ( <4 µM ) the cortices remained largely symmetric , at intermediate concentrations ( 4–6 µM ) we mostly observed the formation of a single actin cap , while at higher concentrations ( 8 µM ) we often observed multiple actin domains . Addition of filamin ( 4 µM ) also led to the formation of a polar actin cap . Despite the different properties of these crosslinkers , they were both successful at inducing symmetry breaking , highlighting the importance of the overall connectivity of the network rather than the detailed characteristics of individual crosslinkers . Interestingly , our artificial actin cortices exhibit additional cell-like behaviors ( Figure 4 ) . The cortex , which is under contractile stress , could spontaneously detach from the interface locally ( Figure 4A; Video 4 ) , as observed in vivo during cellular blebbing ( Charras et al . , 2006 ) . Detachment of the cortex did not lead to outward bulging of the interface due to the high surface tension of the water–oil interface in our system ( ∼700 pN/µm , ‘Materials and methods’ ) . As in cells ( Charras et al . , 2006 ) , the actin cortex healed over time and a new cortex assembled at the interface within minutes . In addition , the asymmetric actin cortices generated polar forces which could induce reorganization in the interior of the droplet and global deformation of the droplet’s shape ( Figure 4B , C; Video 4 ) . Such deformations were observed in ∼10% of the emulsions , and their incidence increased after longer incubation times compared to the initial polarization event . Similar reorganization and changes in global morphology have been observed in cells and embryos ( Lecuit and Lenne , 2007; Munro and Bowerman , 2009 ) . The shape deformation of the droplet can be used to measure the forces generated by the cortical actin network ( Boukellal et al . , 2004; Figure 4C ) . The high curvature at the tip of the actin cap indicates that the cytoskeleton exerts a net protrusive force there , which we estimate to be ∼20 pN/µm2 ( Figure 4C ) . Away from the tip , on the sides of the actin cap , the local curvature is lower implying that the cytoskeleton is exerting pulling forces on the interface . These pulling forces are likely responsible for the accumulation of actin and ActA near the cap ( Figure 4B , Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 01433 . 018Figure 4 . Cell-like behavior of reconstituted actin cortices . ( A ) Scanning confocal images from a time-lapse video ( Video 4 ) showing local detachment of an actin cortex from the interface followed by regrowth of a new cortex within minutes . The time after droplet formation is indicated . Images of the bodipy-conjugated ActA ( green ) and rhodamine-labeled actin ( magenta ) are depicted , together with an overlay of the detachment region ( top ) . ( B ) Spinning disk confocal images of two representative water-in-oil emulsions incubated at 20°C , exhibiting polar actin cortices and deformed morphology . Images of the bodipy-conjugated ActA ( green ) , rhodamine-labeled actin ( magenta ) and their overlay are shown . The asymmetric actin cortices generate polar forces which induce deformation of the droplets . Note the formation of a cytoplasmic concentration of actin and ActA near the deformed cap . The ActA distribution is polar in the example shown on the right , where the overall amount of ActA appears lower ( Figure 4—figure supplement 1 ) . ( C ) A schematic illustration of a deformed droplet and the forces at its interface . The cytoskeletal forces F and the pressure difference ΔP at the interface are balanced by the Laplace pressure which is proportional to the surface tension σ and the local mean curvature κ , so that F+ΔP=2σκ . The pressure difference can be estimated from the curvature outside the cap and the known surface tension , since the cytoskeletal forces are assumed to be zero outside the cap . The protrusive force within the cap was estimated from the difference in the local curvatures in the cap region and outside the cap: F=2σκcap−ΔP=2σ ( κcap−κno cap ) . Scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 01810 . 7554/eLife . 01433 . 019Figure 4—figure supplement 1 . Actin and ActA distributions in a deformed droplet . The fluorescence intensity of rhodamine-labeled actin ( top ) and bodipy-conjugated ActA ( bottom ) extracted from spinning disk confocal images of the deformed emulsion shown in Figure 4B ( insets ) are depicted . The intensity across the droplet ( left ) is plotted together with the profiles along the boundary ( right ) . The actin distribution is strongly peaked along the boundary in the cap region ( top right ) . The ActA is also distributed asymmetrically along the boundary , yet exhibits a less pronounced peak at the cap ( bottom right ) . Both the actin and ActA show extensive cytoplasmic localization near the cap in deformed emulsions ( left ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 01910 . 7554/eLife . 01433 . 020Video 4 . Cytoskeletal forces in the actin cortex lead to bleb-like cortex detachment and shape deformation . This video shows scanning confocal images ( at a single z-plane ) of bodipy-conjugated ActA ( left ) , rhodamine-labeled actin ( center ) and an overlay ( right ) , within a large water-in-oil emulsion incubated at 20°C . The contractile forces within the cortex led to detachment of the cortex from the interface . The cortex recovered within minutes by regrowth of a new cortex at the interface . The polar forces generated by the cortex also led to shape deformation which developed over time . The field of view is 109 µm wide , and the time from droplet formation is indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 01433 . 020
In this work we present a novel reconstituted system that self-organizes into dynamic actin cortices which exhibit spontaneous symmetry breaking in the absence of any directional cue . Our system integrates actin assembly at the interface , actin turnover , myosin motor activity and crosslinkers within a confined geometry , and displays interesting cell-like behavior . The water-in-oil emulsions provide an easy and reproducible approach to generate cell-like compartments ( Tawfik and Griffiths , 1998; Pinot et al . , 2012; Good et al . , 2013; Hazel et al . , 2013 ) . Giant vesicles ( GUVs ) have also been used as compartments for reconstituting cellular phenomena ( Noireaux and Libchaber , 2004; Pontani et al . , 2009; Carvalho et al . , 2013b ) . Both approaches have their advantages and limitations; encapsulation in GUVs is more challenging and the GUVs limited stability in cell extracts can be problematic , while the use of oil-based emulsions limits the applicability of various cell permeable biochemical drugs ( e . g . , blebbistatin ) which tend to be hydrophobic and segregate into the oil phase . In addition , the actin machinery within emulsions appears more sensitive to phototoxicity effects , probably due to elevated oxygen levels within the emulsions compared to bulk . Despite these limitations , our system offers new possibilities for studying cytoskeletal dynamics; we are able to vary the identity and concentrations of actin nucleators , myosin motors and crosslinkers within a controlled environment , and follow their dynamics over time . Using this novel reconstituted system we show that the assembly of dynamic actin cortices requires localization of nucleation factors at the interface , and identify the conditions for cortical symmetry breaking . Specifically , we show that symmetry breaking requires myosin motors and sufficient network connectivity , but does not depend upon pre-patterned localization of actin nucleators , the involvement of microtubules ( Munro et al . , 2004 ) , or any local changes in the properties of the interface . The dynamics of the artificial cortices are temperature–dependent; a shift by a few degrees leads to a large qualitative change in the behavior of the system , going from homogenous cortices at high temperature to asymmetric actin caps at low temperature ( Figure 2 ) . Obviously , temperature is a convenient control parameter . Experimentally , this allowed us to show that cortical polarization can be induced at high temperature by adding crosslinkers and prevented at low temperature by depleting myosin ( Figure 3 ) , implying that symmetry breaking in the artificial cortices is driven by contraction of the cortical actomyosin network . This contractile behavior observed at low temperatures is consistent with previous experiments in cell extracts which were all done at temperatures below 20°C ( Bendix et al . , 2008; Field et al . , 2011; Pinot et al . , 2012 ) . At 30°C the cortical network appears weakly crosslinked , and hence unable to transmit the contractile stresses generated by myosin motor activity . Visualization of the ultrastructure of the actin networks could provide additional insight into the mechanism of symmetry breaking . However , analysis of actin structures by electron microscopy within cell-like compartments is technically challenging due to difficulties in fixation and sample preparation . We hypothesize that the transition to contractile behavior at lower temperature is primarily due to an increase in the intrinsic crosslinking activity in the system . This increase could emerge from longer dwell times of individual myosin motor heads at low temperature , which would turn the multi-motor myosin filaments into more effective crosslinkers . Alternatively , decreasing temperature could increase the affinity of endogenous actin crosslinkers . Our observations are analogous to recent results , which revealed a sharp transition in the contractile behavior of active actin networks as a function of pH ( Kohler et al . , 2012 ) . Overall these findings indicate that actin cortices are close to an instability threshold; small changes in the composition of the system , or in the activity of its components , can lead to dramatic changes in the global characteristics of the system . We believe that this reflects a general organizational principle in cellular systems , whereby cells are often posed near an instability threshold to enable rapid responses to changing conditions . As individual modules within the cells are being unveiled at the molecular level , understanding their integration at the whole-cell level is becoming a central challenge in cell biology . Reconstituting different functional modules together in a meaningful way can be invaluable in that respect , but such reconstitutions have proven to be challenging experimentally . Here we report the successful integration of different cytoskeletal modules into a single reconstituted system which recapitulates several cellular phenomena including cortical symmetry breaking ( Figures 2 and 3 ) , polar force generation , and blebbing ( Figure 4 ) . Our reconstituted system provides novel insights into the complex behavior of the actin cortex in living cells , presenting new opportunities for investigating cortical dynamics , both experimentally and theoretically ( Joanny et al . , 2013 ) . More generally , our system opens the way for future exploration of cytoskeletal dynamics within cell-like compartments , and the integration of additional modules , such as protein expression systems ( Noireaux and Libchaber , 2004 ) and cell cycle ( Good et al . , 2013; Hazel et al . , 2013 ) , towards the realization of artificial cells that can move and divide . Advances in this direction will promote our understanding of basic cellular functions , as well as present ample opportunities for future application in therapeutics and bioengineering .
Actin was purified from chicken skeletal muscle and labeled in filamentous form on either amines or cysteine 374 with tetramethylrhodamine iodoacetamide ( #T6006; Molecular Probes , Grand Island , NY ) , Tetramethylrhodamine Succinimidyl Ester ( #T6105 ) or AleaFluor647 Succinimidyl Ester ( #A20006 ) at a ratio of 1:10 using standard protocols . The behavior of the actin cortices was similar with the different types of labeled actin used ( Figure 1—figure supplement 3 ) . The purified actin was stored in G-buffer ( 10 mM Tris–HCl pH 8 . 6 , 0 . 1 mM DTT , 0 . 2 mM ATP , 0 . 1 mM CaCl2 ) on ice for up to 3 weeks , since the assembly of the cortices was sensitive to the quality of the actin . Chicken skeletal myosin was purified according to standard protocols and stored lyophilized in high salt buffer ( 500 mM KCl , 50 mM Hepes pH 7 . 5 , 5% sucrose ) . Before use , dead myosin heads were removed by spinning the myosin hexamers in the presence of 1 mM ATP and preformed actin filaments , as described in Thoresen et al . ( 2011 ) but without adding phalloidin . ActA-His-Cys was purified from strain DP-L4363 of L . monocytogenes ( a gift from Julie Theriot , Stanford University ) expressing a truncated actA gene encoding amino acids 1–613 with a COOH-terminal six-histidine tag replacing the transmembrane domain and containing an additional cysteine amino acid ( Welch et al . , 1998 ) . ActA was conjugated through a heterobifunctional linker LC-SMCC ( #22362; Thermo Scientific/Pierce , Rockford , IL ) , to poly-D-lysine ( #P0296; Sigma–Aldrich , St . Louis , MO ) labeled with ∼6–8 molecules of Bodipy FL-X-SE ( #D6102; Molecular Probes ) per peptide . The poly-D-lysine was first incubated with the linker and Bodipy at room temperature for 1 hr . The reaction was quenched with 1M Tris pH 7 . 5 . Reduced ActA-His-Cys was added to the mixture and incubated for 2 hr at room temperature . The unbound reagents were removed by dialysis against XB buffer ( 100 mM KCl , 0 . 1 mM CaCl2 , 2 mM MgCl2 , 5 mM EGTA , 10 mM K-HEPES pH 7 . 7 ) . α-actinin was purchased from Cytoskeleton Inc . ( Denver , CO ) , and reconstituted to final concentration of 40 µM with water . Filamin was purchased from Prospec ( East Brunswick , NJ ) , dialyzed against Hepes pH 7 . 5 and reconstituted to final concentration of 20 µM in XB buffer with 50 mM Hepes . Concentrated M-phase extracts were prepared from freshly laid Xenopus laevis eggs as previously described ( Cameron et al . , 1999 ) . Immunodepletion of myosin II was performed as in Bendix et al . ( 2008 ) with minor modifications . In brief , anti-myosin antibody raised against a C-terminal peptide of Xenopus myosin II-A heavy chain ( a gift from Aaron Straight , Stanford University ) or random rabbit IgG ( #011-00-003; Jackson Laboratories , Bar Harbor , ME ) were covalently bound to protein-A Dynabeads ( #100-02; Dynal Biotech , Grand Island , NY ) and resuspended in XB . The extract was incubated with the conjugated Dynabeads for two sequential depletion rounds of 10 min each at RT , and used immediately . A motility mix was prepared by mixing the following: crude extract ( 4 µl ) , XB containing 30 mM MgCl2 ( 4 µl ) , 0 . 13–0 . 2 mM tetramethylrhodamine labeled actin in G-buffer ( 1 µl ) , 20 × ATP regenerating mix ( 150 mM creatine phosphate , 20 mM ATP , 20 mM MgCl2 and 20 mM EGTA ) and 30 µM bodipy-conjugated ActA ( 0 . 5 µl each ) . We estimate the total actin concentration in the motility mix to be ∼25 µM . Emulsions were made by adding 1% ( vol/vol ) motility mix to mineral oil ( Sigma ) containing 4% Cetyl PEG/PPG-10/1 Dimethiocone ( Abil EM90; Evnok Industries , Essen , Germany ) and stirring for 2 min at room temperature . The oil and surfactant mixture was degassed under vacuum overnight prior to use to reduce phototoxicity . Samples were made in chambers assembled from a glass slide and a coverslip which was passivated with trichloromethyl silane ( Sigma ) , and sealed with vaseline:lanolin:paraffin ( at 1:1:1 ) . Samples were either imaged immediately ( for time-lapse videos ) or incubated for 30 min at the indicated temperature and then imaged for population analysis . The surface tension between the aqueous and the oil phase was measured using the pendant-drop assay . An extract containing droplet was injected into an oil cuvette , and the droplet was imaged from the side to determine its shape . The surface tension was determined by analyzing the shape of the droplet using standard procedures . Emulsions were imaged on a 3I spinning disk confocal microscope running Slidebook software , or a laser scanning confocal microscope ( Zeiss LSM 700 ) running ZEN 2009 software , using a 63 × oil objective ( NA = 1 . 4 ) . Images were acquired using 488 nm and 561 nm laser illumination and appropriate emission filters in a temperature controlled incubator . Images on the spinning disk confocal were collected with an EM-CCD ( QuantEM; Photometrics , Tucson , AZ ) . Images of emulsions were typically taken at a single plane at ∼1/3 of the emulsion height , since imaging at the mid-plane or higher suffered from optical artifacts due to the differences in the indices of refraction between the water and oil phase . Bright field images were taken at the mid-plane . Quantitative image analysis was done using the Celltool package developed by Zach Pincus ( Pincus and Theriot , 2007 ) and custom code written in Matlab . Analysis was done on emulsions with a diameter of 30–80 µm . Actin contractility in bulk was imaged using either a 20 × air objective ( NA = 0 . 6 ) or a 40 × oil objective ( NA = 1 . 3 ) . FRAP experiments were done on a laser scanning confocal microscope ( Zeiss LSM 700 ) using a 10 mW 555 nm laser . Bleaching was done on a ∼1 . 5 × 1 . 5 µm region of a cortex either at the bottom of an emulsion or at its side . FRAP analysis was done using the ZEN software by fitting the recovery to a single exponent . An unbleached cortical region was used as a reference region in the analysis . | Cells are extremely complex because they have to perform a vast number of processes . However , this also makes it difficult for researchers to figure out how the individual parts of the cell work . There is interest , therefore , in developing simple artificial cells that can accurately mimic how specific parts of a cell behave . An important process for a cell is called polarization . This is where the contents of the cell arrange themselves in a way that is not symmetrical . Polarization is necessary for many cellular functions , and is particularly important during embryonic development where it helps to form the complex shape of the developing embryo . The cytoskeleton—a dynamic structure that supports the cell and enables it to move—is crucial for polarization . An important part of the cytoskeleton is the actin cortex . This is a thin active sheet made up of a network of tiny filaments of a protein called actin that assembles at the inner face of the cell membrane . Many aspects of the structure and behavior of the actin cortex are not understood . Abu Shah and Keren have now developed an artificial cell system using aqueous droplets surrounded by oil that can reproduce the behavior of actin cortices in real cells . An actin cortex forms upon the localization of specific nucleation factors at the inner surface of the droplets . The artificial cortices are capable of spontaneous symmetry breaking , similar to the initial polarization in embryonic cells during development . This symmetry breaking is driven by molecular motors called myosins and depends on the connectivity of the actin network in the cortex . Experiments on the artificial cells also rule out several other mechanisms that have been proposed to explain symmetry breaking . The work of Abu Shah and Keren represents a further step towards the goal of creating simple artificial cells that can move and divide . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | Symmetry breaking in reconstituted actin cortices |
Pif1 family helicases are conserved from bacteria to humans . Here , we report a novel DNA patrolling activity which may underlie Pif1’s diverse functions: a Pif1 monomer preferentially anchors itself to a 3′-tailed DNA junction and periodically reel in the 3′ tail with a step size of one nucleotide , extruding a loop . This periodic patrolling activity is used to unfold an intramolecular G-quadruplex ( G4 ) structure on every encounter , and is sufficient to unwind RNA-DNA heteroduplex but not duplex DNA . Instead of leaving after G4 unwinding , allowing it to refold , or going beyond to unwind duplex DNA , Pif1 repeatedly unwinds G4 DNA , keeping it unfolded . Pif1-induced unfolding of G4 occurs in three discrete steps , one strand at a time , and is powerful enough to overcome G4-stabilizing drugs . The periodic patrolling activity may keep Pif1 at its site of in vivo action in displacing telomerase , resolving R-loops , and keeping G4 unfolded during replication , recombination and repair .
Approximately , 1% of eukaryotic genes encode DNA or RNA helicases . These enzymes function in nearly all aspects of nucleic acid metabolism in living organisms ( Singleton et al . , 2007; Lohman et al . , 2008 ) . Although helicases were originally recognized as enzymes that catalyze the strand separation of double-stranded nucleic acids , it is now evident that helicases , as defined by a series of characteristic sequence motifs , may have additional functions aside from unwinding duplexes , including protein displacement from DNA and RNA , the unwinding of G-quadruplex ( G4 ) structures , remodelling of chromatin and ribonucleoprotein complexes , promotion of Holliday junction branch migration and the catalysis of a range of nucleic acid conformational changes ( Singleton et al . , 2007; Lohman et al . , 2008; Pyle , 2008; Paeschke et al . , 2013 ) . The Pif1 DNA helicase is an example of a multi-functional helicase . The Pif1 helicase family is a group of 5′→3′ directed , ATP-dependent , super-family ( SF ) 1B helicases that are evolutionarily conserved from bacteria to humans ( Bochman et al . , 2010; Paeschke et al . , 2013 ) . The Saccharomyces cerevisiae Pif1 helicase ( Pif1 ) , the prototypical member of the Pif1 helicase family , plays critical roles in inhibiting telomerase activity at telomeres and double-stranded DNA breaks ( DSBs ) , processing Okazaki fragments , promoting break-induced replication , maintaining mitochondrial DNA , and preventing replication pausing and DSBs at G-quadruplex ( G4 ) motifs ( Boule and Zakian , 2006; Bochman et al . , 2010; Lopes et al . , 2011; Paeschke et al . , 2011 , 2013; Wilson et al . , 2013 ) . However , the molecular mechanisms responsible for these diverse Pif1 functions remain elusive . For example , Pif1 is known to unwind RNA/DNA hybrids better than dsDNA but its mechanistic basis is unknown ( Boule and Zakian , 2007 ) . In addition , how Pif1 may selectively function on certain DNA structures such as stalled replication forks or G4 structures is unclear . Here , we show that a Pif1 monomer is preferentially recruited to 3′ ss-dsDNA junctions and induces repetitive DNA looping that is tightly coupled to its translocation activity powered by its ATPase . This periodic DNA patrolling activity of a Pif1 monomer can be used to unwind RNA–DNA hybrids in its path but not DNA–DNA duplexes because DNA unwinding requires the cooperation of multiple Pif1 monomers . Furthermore , we show that a Pif1 monomer has the capacity to unwind an intramolecular G4 structure in its patrolling path with near unity yield . This novel activity would keep the enzyme at its site of in vivo action in displacing telomerase from 3′ ssDNA ends ( Boule et al . , 2005 ) , resolving biologically relevant ‘R-loops’ ( Aguilera and Garcia-Muse , 2012 ) , and keeping G4 DNA sequences unfolded during DNA replication , recombination and repair ( Paeschke et al . , 2011; Wilson et al . , 2013 ) .
We employed single-molecule ( sm ) FRET assays with total internal reflection ( TIR ) fluorescence microscopy ( Roy et al . , 2008 ) to study the ATP-dependent 5′ to 3′ ssDNA translocation activity of Saccharomyces cerevisiae Pif1 ( Galletto and Tomko , 2013 ) . We first measured the binding constant ( KD ) of Pif1 to a partial duplex ( pd ) DNA with a 3′ 40-nt Poly ( dT ) overhang ( Figure 1A ) . The donor ( Cy3 ) and the acceptor ( Cy5 ) were attached to the overhang separated by 16 nt so that a relatively high FRET efficiency ( EFRET ) of ∼0 . 58 was observed ( Figure 1B ) due to the high flexibility of ssDNA ( Murphy et al . , 2004 ) . Upon addition of Pif1 , a new population with EFRET = 0 . 37 appeared because protein binding stretches ssDNA ( Yang et al . , 2013 ) . Half of the DNA was occupied by Pif1 at 7 ± 2 nM Pif1 concentration after 2-min incubation , and Pif1 dissociation was minimal 30 min after washing out unbound proteins ( Figure 1—figure supplement 1 ) . Unless specified otherwise , all experiments were performed in the absence of free proteins after 2-min pre-incubation with 10 nM Pif1 followed by wash ( referred to as ‘monomer condition’ ) . 10 . 7554/eLife . 02190 . 003Figure 1 . Periodic DNA patrolling by a Pif1 monomer . ( A ) A schematic representation of reaction steps for smFRET measurements under the monomer condition . ( B ) smFRET histograms obtained before and after adding 10 nM Pif1 in the absence of ATP . Peaks centered at EFRET = 0 . 58 and 0 . 37 were observed represent 0 and 1 Pif1 molecules binding to DNA , respectively . ( C and D ) , smFRET-time traces showing repetitive translocation of a Pif1 monomer on ( dT ) 40 . The labeling positions of Cy3 and Cy5 differ in ( C ) vs ( D ) . ( E ) smFRET-time trace for a DNA substrate containing a 5′ Poly ( dT ) ssDNA overhang ( referred to as ( dT ) 40-5′ ) showing no sawtooth pattern . 10 nM Pif1 and 20 μM ATP were added . ( F ) A schematic representation of the repetitive looping model . A Pif1 monomer remains bound at the 3′ ss-dsDNA junction and reels in the 3′ ssDNA overhang . ( G ) A smFRET-time trace showing repetitive translocation of a surface-immobilized Pif1 monomer on non-biotinylated ( dT ) 40 . ( H ) smFRET-time trace showing repetitive looping by a Pif1 monomer on a forked DNA substrate . ( I ) Histograms of the time interval of repetition ( Δt ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 00310 . 7554/eLife . 02190 . 004Figure 1—figure supplement 1 . Pif1 binding to ( dT ) 16+24 DNA . ( A ) A partial duplex DNA with a 3′ 40-nt Poly ( dT ) overhang was used to examine Pif1 binding to DNA . Cy3 and Cy5 were attached to the ss–dsDNA junction and the middle of the ssDNA overhang , respectively , separated by 16 nt ( this DNA substrate is referred to as ( dT ) 16+24 ) . smFRET histograms were obtained 2 min after introducing the indicated Pir1 concentration to the sample chamber . ATP is absent for all the conditions except for the last condition where the excess Pif1 proteins are flushed . Two distinct peaks centered at EFRET = 0 . 58 and 0 . 37 were observed at varying Pif1 concentrations , representing 0 and 1 Pif1 molecules binding to DNA , respectively . ( B ) Fraction of DNA with Pif1 bound vs Pif1 concentration . ( C ) Time evolution of the smFRET histogram in the absence of ATP after incubating with 50 nM Pif1 and flushing out the excess unbound proteins ( D ) Fraction of DNA with Pif1 bound vs time after flushing out the unbound proteins . It should be noted that 7 nM may be larger than the actual KD value , because the 2-min incubation time may not be enough for the system to achieve the equilibrium . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 00410 . 7554/eLife . 02190 . 005Figure 1—figure supplement 2 . Pif1 binding to ( dT ) 16+24 , ( dT ) 40 , and ( dT ) 21 . ( A ) Single molecule time traces for ( dT ) 16+24 DNA only or Pif1-bound ( dT ) 16+24 in the absence of ATP . Unbound proteins were flushed out after incubating 10 nM Pif1 with surface-tethered ( dT ) 16+24 . ( B ) Single molecule time traces for ( dT ) 40 DNA only or Pif1-bound ( dT ) 40 in the absence of ATP . Unbound proteins were flushed out after incubating 10 nM Pif1 with surface-tethered ( dT ) 40 . ( C ) Single molecule time traces of ( dT ) 21 obtained before and after flushing out 10 nMPif1 and at 20 μM ATP . Pif1 binding and dissociation events ( marked by black arrows ) were observed only if 10 nM Pif1 was maintained in the sample chamber . Sawtooth-shaped patterns appear only when Pif1 binds to ( dT ) 21 . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 00510 . 7554/eLife . 02190 . 006Figure 1—figure supplement 3 . Investigation of Pif1’s periodic patrolling activity using protein-induced fluorescence enhancement ( PIFE ) . It has been shown that Cy3 intensity would increase as a protein approaches ( Hwang et al . , 2011 ) . This PIFE effect can be used to infer distance change between Pif1 and the labeled position . ( A ) When Cy3 is placed at the 3′ end of ( dT ) 40 , periodic patterns were observed in the Cy3 intensity time traces under the same monomer condition as in Figure 1D . ( B ) When Cy3 is placed at the 3′ ss–dsDNA junction of ( dT ) 40 , no periodic pattern was observed in the Cy3 intensity time traces under the same monomer condition as in Figure 1D . ( C ) When Cy3 is placed at the 3′ ss-dsDNA junction of ( dT ) 40 , about 20% increase in Cy3 intensity was observed after Pif1 binding to DNA . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 006 When ATP was added , the smFRET histogram became broader ( Figure 1—figure supplement 1A ) , and smFRET-time traces exhibited periodic fluctuations between EFRET = 0 . 37 and 0 . 58 ( Figure 1C ) , likely due to Pif1 translocation on and off the 16-nt segment separating the fluorophores . When we repositioned the fluorophores near the two ends of the ( dT ) 40 region ( the construct is referred to as ( dT ) 40 ) , smFRET-time traces showed a periodic sawtooth-shaped pattern where FRET gradually increases until it abruptly drops to a low level and repeats the process many times ( Figure 1D ) . Observation of the sawtooth pattern required both Pif1 and ATP ( Figure 1—figure supplement 2 ) , and the sawtooth pattern was not observed with a 5′ overhang ( Figure 1E ) . We hypothesized that Pif1 remains anchored to the 3′ ss-dsDNA junction while translocating on and reeling in the 3′ overhang , extruding a ssDNA loop ( Figure 1F ) . To test this hypothesis , we used the fact that Cy3 gets brighter when a protein is in close proximity ( Fischer et al . , 2004; Hwang et al . , 2011 ) . A sawtooth pattern was observed when Cy3 was placed at the 3′ ssDNA end but not at the ss-dsDNA junction , supporting the model that Pif1 approaches the 3′ end while remaining anchored to the junction ( Figure 1—figure supplement 3 ) . To ascertain the monomeric state of Pif1 , Pif1 was pulled down to the surface ( Jain et al . , 2011 ) using an antibody against the Histidine6-tag ( Figure 1G ) . Pif1 pulled-down in this way should maintain its monomeric state in solution ( Lahaye et al . , 1993; Barranco-Medina and Galletto , 2010 ) . The same sawtooth patterns were observed when 5 nM ( dT ) 40 was added with ATP ( Figure 1G ) , indicating that a Pif1 monomer is responsible for the observed repetitive DNA looping that we term here as ‘periodic patrolling’ . The presence of a 5′ overhang in addition to the 3′ overhang did not inhibit periodic patrolling ( Figure 1H ) . The histograms of the patrolling period , Δt , were nearly identical among all three schemes ( >50 molecules each; Figure 1I ) , yielding an average patrolling period τ = 2 . 1 ± 0 . 1 s . Pif1 monomers are ATP-dependent translocases on ssDNA with 5′ to 3′ directionality ( Galletto and Tomko , 2013 ) . Periodic patrolling slowed down as we lowered [ATP] from 500 to 5 μM ( Figure 2A ) . Plotting the inverse of τ vs [ATP] and fitting to the Michaelis–Menten equation , we obtained KM = 110 ± 17 μM ( Figure 2B ) . We examined Pif1’s periodic patrolling on 3′ Poly ( dT ) overhangs of varying lengths ( N = 32 , 40 , 56 , and 72 nt; referred to as ( dT ) N ) at a fixed [ATP] ( 20 μM ) and found that τ increased linearly with the overhang length ( Figure 2C , D ) , showing that the patrolling is indeed coupled to ssDNA translocation with a speed of 13 ± 2 nt/s at 20 μM ATP , corresponding to 85 ± 13 nt/s at saturated [ATP] , consistent with previous ensemble studies ( Galletto and Tomko , 2013; Ramanagoudr-Bhojappa et al . , 2013 ) . 10 . 7554/eLife . 02190 . 007Figure 2 . ATP and ssDNA length dependence of the repetition period and step size determination . ( A ) smFRET time traces showing Pif1 repetitive looping on ( dT ) 40 at varying ATP concentrations . ( B ) Michaelis–Menten fit of repetition rate ( 1/τ ) vs ATP concentration . Error bars denote SD . Errors in the fit results are shown in SEM . ( C ) smFRET time traces of repetitive looping on ( dT ) N at 20 μM ATP . To obtain the average translocation speed , v , we examined RL with pdDNA containing 3′ Poly ( dT ) overhangs of varying lengths ( N = 32 , 40 , 56 , and 72 nt; referred to as ( dT ) N ) . ( D ) Average repetition period , τ , vs 3′ ssDNA overhang length . The error bar denotes SD . Errors in the fit result are SEM . ( E ) Δt histogram obtained from a single Pif1 monomer showing 190 repetitive looping events on ( dT ) 32 . The solid line is a fit to the Г-distribution . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 00710 . 7554/eLife . 02190 . 008Figure 2—figure supplement 1 . Histograms of the repetition period , Δt , for Pif1 periodic patrolling on ( dT ) N . ( A ) Δt histograms obtained from single Pif1 molecules showing the periodic patrolling behavior on ( dT ) 32 . Solid lines are fits with Г-distribution , ( Δt ) n−1exp ( −k·Δt ) . We found that the translocation speeds ( or repetition rates ) of single Pif1 translocases differ , and these differences persist during our observation time window of minutes . Similar molecular heterogeneity in the translocation speed was also previously observed for PcrA ( Park et al . , 2010 ) . ( B ) Δt histograms obtained from >50 Pif1 molecules showing the periodic patrolling behavior on ( dT ) N ( N = 32 , 40 , 56 , and 72 ) . Due to the molecular heterogeneity , the combined Δt histogram of ( dT ) 32 from an ensemble of single Pif1 molecules is broader than that from a single Pif1 molecule , leading to a smaller n value ( ∼10 ) from the Γ-distribution fit . Hence , we used the n values obtained in ( A ) for the step size determination . Overall , our data consistently place an upper limit on the kinetic step size of <2 nt , and assuming that there exists a defined integer value for the kinetic step size , the kinetic step size of ssDNA translocation is thus 1 nt . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 008 To determine the kinetic step size for Pif1 translocation , we assumed that Pif1 takes n hidden irreversible Poissonian steps with identical rate k in each cycle of duration Δt . A Δt histogram obtained from a single Pif1 that showed hundreds of cycles on ( dT ) 32 was fit with the Г-distribution , ( Δt ) n−1exp ( −k·Δt ) ( Figure 2E ) ( Park et al . , 2010 ) . The fit yielded n = 26 ± 3 , k = 25 ± 3 s−1 for one molecule of ( dT ) 32 , suggesting 26 or more steps required to translocate over 32 nt , which supports a 1-nt step size . There is variation of Pif1 translocation speed among different Pif1 molecules , but a 1-nt step size could be consistently determined ( Figure 2—figure supplement 1 ) . Related to our findings , a recent ensemble study proposed a 1-bp step size of dsDNA unwinding by Pif1 ( Ramanagoudr-Bhojappa et al . , 2013 ) . It has been proposed that Pif1 inhibits telomerase activity by unwinding the RNA–DNA hybrid formed between telomerase RNA and the 3′ end of telomeric DNA ( Boule et al . , 2005; Boule and Zakian , 2007 ) and that Pif1 removes R-loops , that is , RNA–DNA hybrids that occur naturally during replication and transcription ( Boule and Zakian , 2007; Aguilera and Garcia-Muse , 2012 ) . Whether a monomer or dimer of SF1/SF2 helicases is responsible for nucleic acid unwinding has been a subject of debate ( Singleton et al . , 2007; Lohman et al . , 2008 ) , so we asked if a patrolling Pif1 monomer can unwind dsDNA or RNA–DNA hybrids . We designed gapped substrates by placing an additional 31-bp DNA–DNA or RNA–DNA duplex next to the 3′ end of the ( dT ) 40 substrate ( referred to as DD31 and RD31 , respectively; Figure 3A ) . If Pif1 fully unwinds the duplex during its periodic patrolling , the Cy3-labeled strand should be released from the surface , resulting in a decrease of the number of Cy3 spots over time . DD31 unwinding was not detectable under the monomer condition but was observed at 100 nM Pif1 , likely due to multiple Pif1 proteins acting in concert ( Figure 3A ) . In contrast , efficient unwinding of RD31 was detected under both conditions ( Figure 3A ) . Under the monomer condition , unwinding curves for RD31 can be fit to a single exponential function , yielding the unwinding times tunwind = 6 . 3 ± 0 . 9 and 1 . 6 ± 0 . 2 min at 20 μM and 1 mM ATP , respectively ( Figure 3—figure supplement 1C ) . smFRET-time traces taken under the monomer condition showed that Pif1 performs periodic patrolling on both DD31 and RD31 ( Figure 3B , Figure 3—figure supplement 1A ) with an average period of τ = 2 . 4 ± 0 . 1 s at 20 μM ATP ( Figure 3C , Figure 3—figure supplement 1A ) . Our data hence suggest that a Pif1 monomer takes about 200 cycles on average ( = 6 . 3 min/2 . 4 s ) to unwind the 31 bp RNA–DNA hybrid , whereas dsDNA unwinding requires the cooperation of multiple Pif1 monomers or protein oligomerization on DNA . At 100 nM Pif1 , FRET fluctuations became irregular ( Figure 3B ) , likely due to the binding of multiple Pif1 monomers to the gapped substrate at this Pif1 concentration . The dsDNA unwinding speed previously reported was much higher ( ∼0 . 3 s for unwinding 16 bp dsDNA ) ( Ramanagoudr-Bhojappa et al . , 2013 ) , either because they pre-assembled a functional helicase unit on the DNA before adding ATP to start the reaction or in that study , or due to a limited processivity of dsDNA unwinding by a Pif1 oligomer . 10 . 7554/eLife . 02190 . 009Figure 3 . Periodic DNA patrolling of a Pif1 monomer can unwind RNA-DNA hybrids but not dsDNA . ( A ) Unwinding curves for DD31 ( top ) and RD31 ( bottom ) obtained under different conditions . ( B ) A smFRET time trace of RD31 obtained under the monomer condition shows Pif1 repetitive looping ( top ) and a smFRET time trace of RD31 obtained at 100 nM Pif1 shows irregular FRET fluctuations ( bottom ) . ( C ) Δt histograms for Pif1 periodic patrolling on ( dT ) 40 , DD31 , and RD31 . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 00910 . 7554/eLife . 02190 . 010Figure 3—figure supplement 1 . Translocation and unwinding activity of Pif1 on gapped substrates . ( A ) Single molecule time traces of Pif1 dynamics on DD31 at two different protein concentrations . After incubating 10 nM Pif1 and flushing out unbound Pif1 ( we refer to this condition as the Pif1 monomer condition ) , single molecule time traces were obtained at 20 μM ATP ( top panel ) . Periodic wave patterns were observed , indicating that a single Pif1 monomer periodically patrols the ( dT ) 40 ssDNA segment . In contrast , in single molecule time traces obtained at 100 nM Pif1 and 20 μM ATP , the regular wave patterns were disrupted ( bottom panel ) , likely due to the binding of multiple Pif1 monomers or protein oligomerization on the substrates at 100 nM Pif1 ( Barranco-Medina and Galletto , 2010; Galletto and Tomko , 2013 ) . Similar time traces were obtained under the same conditions for RD31 , a gapped substrate containing a 31-bp RNA–DNA hybrid ( Figure 3B ) . ( B ) Unwinding curves for DD31′ obtained under different conditions . DD31′ is a gapped DNA substrate slightly modified from DD31 by repositioning the Cy3 fluorophore , and is used to test whether the presence of a Cy3 fluorophoe at the ss–dsDNA junction would affect Pif1 unwinding . ( C ) Unwinding curves for RD31 with exponential fits . Solid lines are single exponential fits . The red dashed line is a bi-exponential fit . At 10 nM Pif1 ( the monomer condition ) , the unwinding curves can be fit well with a single exponential function . However , at 100 nM Pif1 , the unwinding curve cannot be fit well with a single exponential function ( black solid line ) , and a bi-exponential function characterized by two unwinding phases ( red dashed line ) can well describe the curve . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 010 In a gapped DNA substrate , both 3′- and 5′-ss-dsDNA junctions are present for Pif1 loading . To determine Pif1’s preferential binding site on a gapped substrate , we used two labeling schemes: ( 1 ) ( dT ) 32+18-S1: FRET probes were attached to the middle and 5′ end of the gap region; ( 2 ) ( dT ) 32+18-S2: the probes were attached to the middle and 3′ end of the gap region so that FRET reports on Pif1 binding to the respective ssDNA segments . At 10 nM Pif1 , a FRET decrease was observed only for ( dT ) 32+18-S1 , whereas at 100 nM Pif1 , FRET decreased for both schemes ( Figure 4 ) , indicating that Pif1 prefers to load at 3′-ss-dsDNA junctions and in agreement with our observation that at 100 nM Pif1 , binding of multiple Pif1 monomers to a gapped substrate abolishes the regular sawtooth pattern ( Figure 3B , Figure 3—figure supplement 1C ) . 10 . 7554/eLife . 02190 . 011Figure 4 . Pif1 monomers are preferentially recruited to 3′ ss-dsDNA junctions . ( A and B ) Schematic representations of reaction steps to identify the preferential Pif1 binding site to a gapped DNA substrate . The two gapped DNA substrates used ( ( dT ) 32+18-S1 and ( dT ) 32+18-S2 ) only differ in Cy5 labeling position . smFRET histograms were determined at different Pif1 concentrations . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 011 Pif1 is also a potent unwinder of G-quadruplexes ( G4s ) ( Paeschke et al . , 2011 ) , which are held together by noncanonical Hoogsteen G–G base pairs and stabilized by monovalent cations ( Huppert , 2008 ) . Intramolecular G4 motifs ( four runs of ‘GGG’ , with loops of 1–25 nt between them of any sequence ) are prevalent in eukaryotic genomes ( gene promoter regions , telomeres , etc ) ( Bochman et al . , 2012 ) , and G4 formation may slow replication and regulate transcription ( Paeschke et al . , 2011; Bochman et al . , 2012 ) . Growing evidence indicates that G4 structures indeed form in vivo ( Lipps and Rhodes , 2009; Paeschke et al . , 2011 , 2013; Biffi et al . , 2013 ) . To test whether a patrolling Pif1 monomer can unwind intramolecular G4 DNA , we placed a 31-nt standard G4 sequence ( termed TP ) from the mouse immunoglobulin locus ( Paeschke et al . , 2013 ) at the 3′ end of the ( dT ) 40 region ( referred to as TP-S1; Figure 5A ) . Cy3 and Cy5 were placed at the two ends of TP so that a folded G4 structure would yield high FRET . As a control , we replaced the 31-nt TP sequence with Poly ( dT ) of the same length ( Figure 5B ) . We confirmed G4 formation by performing K+ titration ( Figure 5—figure supplement 1 ) . In 5 mM Mg2+ and 60 mM K+ , TP-S1 gave a single population of EFRET = ∼0 . 75 for the folded G4 structure in contrast to the control that gave EFRET = ∼0 . 3 ( Figure 5A , B ) . Addition of 10 nM Pif1 alone did not alter the smFRET histogram of TP-S1 ( Figure 5A ) , suggesting Pif1 cannot unwind G4 in the absence of ATP , when K+ is present . 10 . 7554/eLife . 02190 . 012Figure 5 . Periodic DNA patrolling of a Pif1 monomer can unwind G4 structures . ( A and B ) smFRET histograms of TP-S1 ( A ) and ( dT ) 40+31 ( B ) obtained under different conditions . ( C ) A smFRET time trace showing Pif1 repetitive unwinding on TP-S1 at varying ATP concentrations . ( D ) Michaelis–Menten fit of the repetition rate ( 1/τ ) vs ATP concentration . Error bars denote SD . Errors in the fit results are SEM . ( E ) A smFRET time trace showing Pif1 periodic patrolling on ( dT ) 40+31 . EFRET changes periodically between ∼0 . 3 and 0 . 4 , likely due to Pif1 translocation on and off the ( dT ) 31 segment separating Cy3 and Cy5 . ( F ) smFRET histograms for a DNA substrate devoid of the ( dT ) 40 region ( referred to as TPΔ ) obtained under different conditions . The G4 structure is not unwound when the ssDNA region is absent , likely due to lack of Pif1 loading . ( G ) A smFRET time trace showing repetitive G4 unwinding of a surface-immobilized Pif1 monomer on non-biotinylated TP-S1 . ( H and I ) smFRET histrograms ( H ) and time traces ( I ) of TEL . ( J ) smFRET histograms of TELΔ obtained under different conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 01210 . 7554/eLife . 02190 . 013Figure 5—figure supplement 1 . K+ titrations to study G4 formation for the TP sequence in the presence and absence of Mg2+ . ( A ) smFRET histograms of TP-S1 at varying K+ concentrations in the absence and presence of Mg2+ . ( B ) smFRET histograms of ( dT ) 40+31 at varying K+ concentrations in the absence and presence of Mg2+ . TP-S1 and ( dT ) 40+31 only differ in the sequence of the 31-nt ssDNA region flanked by Cy3 and Cy5 . Therefore , G4 formation in TP-S1 should result in increased EFRET values compared to the EFRET values of ( dT ) 40+31 at the same [K+] . TP G4 formation is evident when [K+] ≥ 60 mM in the absence of Mg2+ , or when [K+] ≥ 30 mM in the presence of Mg2+ , indicating that Mg2+ may further stabilize G4 structures in K+ . Previous ensemble G4 studies have shown that Mg2+ may facilitate G4 formation ( Yan et al . , 2010 ) . For ( dT ) 40+31 , the FRET peak slightly moves to a higher FRET value , due to increased compaction of ssDNA at higher salt concentrations ( Murphy et al . , 2004; Chen et al . , 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 01310 . 7554/eLife . 02190 . 014Figure 5—figure supplement 2 . Monomeric Pif1 repetitively unwinds TP G4 in the presence and absence of BRACO19 . ( A ) smFRET time traces for TP-S1 DNA only and Pif1-bound TP-S1 in the absence of ATP . ( B ) Single molecule time traces showing periodic DNA patrolling by monomeric Pif1 . ( C ) Single molecule time traces of TP-S1 showing periodic G4 unwinding in the presence and absence of BRACO19 . The total fluorescence intensity of Cy3 and Cy5 fluctuates periodically between two levels , and the low intensity level ( yellow regions ) coincides well with the folded G4 state . The repetitive unwinding is unlikely caused by successive binding of proteins because the excess unbound proteins have been removed from the sample chamber . ( D ) The average total fluorescence intensity of Cy3 and Cy5 from >10 , 000 bare DNA molecules , before and after addition of 1 μM BRACO19 . The total fluorescence intensity became 40% of its original level after adding BRACO19 to bare DNA substrates , suggesting that BRACO19 binding to DNA quenches the total fluorescence intensity . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 01410 . 7554/eLife . 02190 . 015Figure 5—figure supplement 3 . Pif1 binding to TPΔ and TELΔ in the presence of Na+ or Li+ . TPΔ and TEL-DD21 are both partial duplex DNA substrates that contain a G4 sequence but are devoid of the 40-nt ssDNA region at the 5' end of G4 sequence . Figure 5F , J show that Pif1 cannot unwind the G4 structure for both TPΔ and TELΔ in the presence of K+ . ( A ) smFRET histograms of TPΔ obtained under different conditions to study G4 unwinding by Pif1 in Na+ . Upon addition of Pif1 , similar TP G4 unwinding activity was observed both in the absence and presence of ATP . ( B ) Single molecule time traces of TPΔ showing an event representing ScPif dissociation and G4 folding obtained after adding 10 nM Pif1 and 20 μM ATP . Sawtooth patterns were observed for TPΔ in the smFRET-time traces , consistent with Pif1's periodic patrolling on unfolded G4 sequence ( Figure 2C ) . The fact that the 0 . 7 FRET species is more stable in K+ than in Na+ is in line with known G4 stabilizing properties of these monovalent cations and is hence further evidence that the 0 . 7 FRET peak represents folded G4 molecules . ( C ) smFRET histograms of TELΔ obtained under different conditions to study G4 unwinding by Pif1 in Na+ or Li+ . As in K+ , no TEL G4 winding was observed in Na+ , but in Li+ , TEL G4 winding activity was detected even in the absence of ATP . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 01510 . 7554/eLife . 02190 . 016Figure 5—figure supplement 4 . K+ and Na+ titrations to study G4 formation for the TEL sequence in the presence and absence of Mg2+ . ( A ) smFRET histograms of TEL at varying K+ or Na+ concentrations in the absence and presence of Mg2+ . ( B ) smFRET histograms of TEL-DD21 at varying K+ or Na+ concentrations in the presence of Mg2+ . The TEL sequence used contains 21 nt , so G4 formation in TEL or TEL-DD21 should result in increased EFRET values compared to the EFRET values of ( dT ) 21DD21 at the same [K+] . TEL G4 formation is nearly complete when [K+] ≥ 30 mM in the absence of Mg2+ , and when [K+] ≥ 10 mM in the presence of Mg2+ , indicating that Mg2+ may further stabilize G4 in K+ . ( C ) smFRET histograms of ( dT ) 21DD21 at varying K+ concentrations in the presence of Mg2+ . The FRET peak of ( dT ) 21DD21 slightly moves to a higher FRET value as [K+] increases , due to increased compaction of ssDNA at higher salt concentrations ( Murphy et al . , 2004; Chen et al . , 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 01610 . 7554/eLife . 02190 . 017Figure 5—figure supplement 5 . Presence of a DNA duplex next to the 3′ end of the G4 sequence does not affect Pif1′s periodic G4 unwinding activity . We modified the TEL construct used in Figure 5H by adding a 21-bp DNA duplex next to the 3′ end of the TEL G4 sequence ( referred to as TEL-DD21 ) . ( A ) smFRET histograms of TEL-DD21 obtained under different conditions to study G4 unwinding by Pif1 ( left panel ) . smFRET time traces of TEL-DD21 obtained for DNA only or after adding 10 nM Pif1 in the absence of ATP ( right panel ) . ( B ) smFRET histograms of ( dT ) 21DD21 obtained under different conditions to study Pif1 binding to ( dT ) 21DD21 ( left panel ) . ( dT ) 21DD21 serves as a control DNA substrate here , because Cy3 and Cy5 were placed at the two ends of a 21-nt Poly ( dT ) region for ( dT ) 21DD21 and were placed at the two ends of a 21-nt TEL sequence for TEL-DD21 . smFRET time traces of ( dT ) 21DD21 were obtained under the monomer condition ( flushing out unbound proteins after incubation in 10 nM Pif1 ) or after adding 100 nM Pif1 and 20 µM ATP ( right panel ) . Under the Pif1 monomer condition ( top right ) , the wave patterns seen for DD31 ( Figure 3—figure supplement 1A ) were not clearly observed for ( dT ) 21DD21 , likely due to a shorter ssDNA region available for Pif's periodic patrolling . However , similar irregular dynamics in the FRET time traces were observed at 100 nM Pif1 and 20 µM ATP as seen for DD31 ( bottom right ) . ( C ) A representative smFRET time trace showing Pif1 repetitive unwinding on TEL-DD21 . Addition of a DNA duplex next to the 3′ end of TEL does not affect the periodic G4 unwinding , indicating that Pif1 monomers unwind G4 but do not go beyond to unwind dsDNA . This is consistent with the findings from the gapped substrates ( Figure 3 ) that the periodic patrolling of a Pif1 monomer cannot unwind dsDNA . ( D ) The average EFRET values and dwell times at each step ( or FRET plateau ) were determined by a previously described automated step-finding algorithm ( Kerssemakers et al . , 2006; Myong et al . , 2007; Lee et al . , 2012 ) . The distribution of the average EFRET values determined from many unwinding cycles collected from 100–200 molecules is shown in the right panel . The red line shows the Gaussian fit . Peak values are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 01710 . 7554/eLife . 02190 . 018Figure 5—figure supplement 6 . Dwell time histograms of each unwinding step and histogram of G4 refolding time built from TP-S1 . For the Δt1 Δt2 , Δt3 , and Δt4 histograms ( the dwell times for each unwinding step ) , solid lines are Г-distribution fits . For the Δt5 histogram ( the time for the TP G4 structure to refold ) , the solid line is a single exponential fit . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 01810 . 7554/eLife . 02190 . 019Figure 5—figure supplement 7 . Δt histograms of different DNA substrates . ( A ) Histograms of the time interval of repetition ( Δt ) for TP-S1 , TP-S2 , and ( dT ) 40+31 . ( B ) Histograms of the time interval of repetition ( Δt ) for TEL and TEL-DD21 . τ is the average repetition period . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 019 Upon addition of ATP , a lower FRET species appeared , indicating G4 unwinding . Remarkably , smFRET-time traces showed a periodic G4 unwinding pattern . In each unwinding cycle ( Δt ) , EFRET starts at ∼0 . 75 and slowly decreases to ∼0 . 3 , followed by an abrupt increase back to 0 . 75 ( Figure 5C ) . For a modified labeling scheme of placing Cy3 at 3′ end of TP sequence and Cy5 at 3′ ss-dsDNA junction ( TP-S2; Figure 5—figure supplement 2B ) , smFRET-time traces resembled the sawtooth patterns seen for ( dT ) N , indicating that before unwinding G4 , a Pif1 monomer patrols the ( dT ) 40 region . The observation indicates that Pif1 unwinds G4 with near unity yield on each encounter . The average period of the unwinding pattern , τ , depends on [ATP] with a KM of 118 ± 7 μM ( Figure 5C , D , Figure 5—figure supplement 2A ) and is close to the average patrolling period of Pif1 on a 3′ Poly ( dT ) overhang of the same overhang length ( 71 nt; Figure 2D ) . The same periodic unwinding pattern was observed with pulled down monomeric Pif1 molecules ( Figure 5G ) . The ( dT ) 40 region is necessary for G4 unwinding in K+ solution ( Figure 5F ) but not in Na+ solution ( Figure 5—figure supplement 3 ) , indicating that the G4 structure is easier to be unwound in Na+ than in K+ . In Na+ solution , addition of 10 nM Pif1 led to the unwinding of the G4 structure without an adjacent ssDNA region , even in the absence of ATP ( Figure 5—figure supplement 3A ) , suggesting that Pif1 binding alone may unfold TP G4 in Na+ . This is consistent with the known order of cations’ ability to stabilize G4: K+>Na+>Li+ ( Simonsson , 2001 ) , further confirming that the secondary structure formed by TP sequence is indeed G4 . We next tested another well-studied G4 sequence from human telomeres ( Ying et al . , 2003; Lee et al . , 2005 ) by replacing TP with A ( GGGTTA ) 3GGGTT ( referred to as TEL ) . K+ and Na+ titrations with and without Mg2+ indicated that Mg2+ facilitates TEL G4 formation both in K+ and Na+ ( Figure 5—figure supplement 4 ) . Similar smFRET histograms ( Figure 5H ) and periodic unwinding time traces ( Figure 5I ) were obtained , implying that the periodic G4 unwinding by a Pif1 monomer may be general and not specific to sub-species of G4 structures . Similar to the observations for TP , deletion of the ssDNA region at the 5′ end of TEL ( Figure 5J ) inhibited Pif1 loading and G4 unwinding in K+ . The inhibitory effect remained in Na+ but was largely reduced in Li+ ( Figure 5—figure supplement 3C ) , likely due to the inability of Li+ to stabilize TEL G4 . Addition of a 21-bp DNA duplex next to the 3′ end of TEL ( Figure 5—figure supplement 5 ) did not affect the periodic G4 unwinding , indicating that even though a Pif1 monomer repetitively unwinds G4 DNA with a very high yield , it does not go beyond it to unwind dsDNA . This intrinsic activity of a Pif1 monomer would be useful in keeping G4 DNA unfolded until other DNA metabolic enzymes can take over without causing the side effect of uncontrolled dsDNA unwinding . To test if the periodic patrolling activity alone is sufficient to unwind G4 , we analyzed the Bacillus stearothermophilus PcrA , which also exhibits periodic DNA patrolling but with reverse translocation directionality ( Park et al . , 2010 ) . We designed new versions of ( dT ) 40 and TEL with reverse polarity ( Figure 6A , B ) . Under monomer conditions ( 1 nM PcrA ) ( Park et al . , 2010 ) , PcrA showed robust periodic patrolling , but no G4 unwinding was detected ( Figure 6C , D ) . Even at 100 nM , PcrA efficiently unwound dsDNA ( Figure 6E ) , but efficient G4 unwinding was not detected . These data suggest that having the periodic patrolling activity per se is not sufficient for G4 unwinding , and robust G4 unwinding is a specific property of Pif1 . 10 . 7554/eLife . 02190 . 020Figure 6 . Periodic DNA patrolling of a PcrA monomer cannot unwind G4 structures . ( A ) smFRET histograms of ( dT ) 40-5′ obtained for DNA only or at 1 or 100 nM PcrA ( ATP was added ) . ( B ) smFRET histograms of TEL-5′ obtained for DNA only or at 1 or 100 nM PcrA ( ATP was added ) . ( C ) Single molecule time traces showing PcrA dynamics on ( dT ) 40-5′ obtained at 1 or 100 nM PcrA . When 1 nM PcrA was added , periodic sawtooth patterns were observed for periodic DNA patrolling by a PcrA monomer as previously demonstrated ( Park et al . , 2010 ) . However , when 100 nM PcrA was added , the sawtooth patterns were disrupted , and only irregular FRET dynamics were observed , indicating the binding of multiple PcrA monomers per DNA or protein oligomerization . ( D ) Single molecule time traces of TEL-5′ obtained under the same conditions as in ( C ) . No efficient G4 unwinding was observed . ( E ) The dsDNA unwinding activity at 1 and 100 nM PcrA . At 1 nM PcrA , no efficient unwinding was observed in 10 min at 1 nM PcrA , whereas efficient unwinding was observed at 100 nM PcrA . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 020 Upon close examination of the G4 unwinding phase , we noticed that the FRET decrease occurs in three discrete steps ( Figure 7A ) . We determined the average EFRET values and dwell times of each step ( Δt1 , Δt2 , Δt3 and Δt4 ) using an unbiased step-finding algorithm ( Kerssemakers et al . , 2006 ) ( Figure 7B ) . The distributions of average EFRET values of the steps thus identified ( 100–200 molecules each ) revealed four peaks , showing that G4 unwinding indeed occurs in three steps ( Figure 7C , Figure 5—figure supplement 5D ) . The highest FRET state was assigned to the folded G4 DNA and the lowest FRET state to the fully unfolded state . The two additional FRET states likely represent partially folded states ( G-hairpin and G-triplex ) , as previously proposed based on ensemble kinetic and simulation studies ( Mashimo et al . , 2010; Zhang and Balasubramanian , 2012 ) ( Figure 7D ) . The dwell time histogram of each FRET state exhibits a Г-distribution ( Figure 7E , Figure 5—figure supplement 6 ) , suggesting that Pif1 reels in DNA in multiple 1-nt steps before the next strand is unraveled and so on . The refolding occurred rapidly with a single rate-limiting step ( ∼0 . 2 s refolding time , see the Δt5 histogram ) even for the TP sequence with 7 nt of intervening ssDNA between G repeats ( Mashimo et al . , 2010; Zhang and Balasubramanian , 2012 ) , and the subsequent three-step unfolding events by Pif1 indicate that the G4 structure is indeed reestablished rapidly . Instead of resolving a G4 structure once and leaving the DNA , which would allow the intramolecular G4 structure to form again , Pif1 stays in proximity and continually resolves the G4 DNA using its periodic patrolling activity . 10 . 7554/eLife . 02190 . 021Figure 7 . Pif1 unwinds G4 in three discrete steps . ( A ) A representative unwinding cycle showing a stepwise pattern for G4 unwinding followed by fast G4 refolding . The dwell times of different parts of the cycle , Δt1 , Δt2 , Δt3 , Δt4 and Δt5 , are indicated . ( B ) The average EFRET values and dwell times at each step were determined by an automated step-finding algorithm ( red ) . ( C ) Distributions of the average EFRET values determined from many unwinding cycles collected from 100 to 200 molecules . ( D ) A proposed model for G4 unwinding involving two intermediates . ( E ) Dwell time histograms of each unwinding step and histogram of G4 refolding time Δt5 ( gray ) . A histogram of Δt1 ( F ) obtained in the presence of BRACO19 is shown in black . Solid lines are Г-distribution fits . ( F ) Single molecule time traces showing repetitive G4 winding by a Pif1 monomer in the presence of BRACO19 . The total fluorescence intensity of Cy3 and Cy5 fluctuates periodically between two levels , and the low intensity level ( yellow regions , Δt1 ) coincides well with the folded G4 state . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 021 Finally , we found that Pif1’s patrolling activity is strong enough to unwind G4 structures stabilized by a small compound , BRACO19 ( De Cian et al . , 2007 ) . smFRET-time traces of TEL ( Figure 7F ) and of TP-S1 ( Figure 5—figure supplement 2C , Figure 5—figure supplement 7 ) showed periodic unwinding patterns in the presence of 1 μM BRACO19 with a nearly identical τ value ( ∼4 . 7 s; Figure 5—figure supplement 7 ) to what was obtained without BRACO19 . Periodic fluctuations of the total fluorescence intensity caused by fluorescence quenching by BRACO19 confirmed the binding of the compound to G4 ( Figure 5—figure supplement 2D , Figure 7E ) .
Our study establishes a novel DNA patrolling activity by a Pif1 monomer . We discovered that monomeric Pif1 is preferentially recruited to 3′-ss-dsDNA junctions and reels in ssDNA while staying at the junction , repeatedly looping out a ssDNA segment . This periodic behavior is ATP dependent and tightly coupled to Pif1’s translocase activity . Previous single-molecule studies have shown that B . stearothermophilus PcrA and Escherichia coli UvrD , two bacterial SF 1A helicases , are both preferentially recruited to 5′-ss-dsDNA junctions and exhibit similar activities with reverse translocation directionality ( i . e . , 3′→5′ ) ( Park et al . , 2010; Tomko et al . , 2010 ) . In addition , we showed that Pif1′s patrolling activity persists in forked and gapped DNA substrates containing a 3′-ss-dsDNA junction , raising the possibility that it may happen when Pif1 binds to various types of DNA intermediates during DNA replication , repair , and recombination where a 3′-ss-dsDNA junction is present . Figure 8 illustrates several possible biological contexts in which Pif1′s periodic patrolling activity may play important roles ( Figure 8 ) . First , the patrolling activity of a Pif1 monomer can be used to unwind ‘R-loops’ ( RNA-DNA hybrids ) ( Figure 8A ) . Growing evidence indicates that R-loops occur much more frequently during replication and transcription than previously foreseen , and persistent R-loop formation in cells is a potential source of genome instability including mutations , recombination , chromosomal rearrangements , and chromosome loss ( Aguilera and Garcia-Muse , 2012 ) . We showed that a Pif1 monomer can utilize its patrolling activity to remove the RNA strand from 31-bp RNA–DNA hybrids through ∼200 cycles ( and fewer cycles may be needed for shorter heteroduplexes ) . In contrast , dsDNA unwinding required the cooperation of multiple Pif1 monomers . Therefore , it is possible that Pif1’s periodic patrolling activity contributes to the cell’s ability to remove deleterious R-loops . Second , periodic patrolling may be responsible for Pif1’s telomerase removal activity from 3′ ssDNA ends found at telomeres or DSBs ( Boule et al . , 2005 ) ( Figure 8B ) . Many previous studies show that the translocation activity of monomeric helicases can help displace proteins from DNA ( Lohman et al . , 2008 ) . In fact , PcrA utilizes periodic patrolling activity to dismantle potentially deleterious RecA filaments from 5′ ssDNA ends ( Park et al . , 2010 ) . As it has been proposed that Pif1 inhibits telomerase by unwinding the RNA–DNA hybrid formed between the telomerase RNA and the telomeric DNA end ( Boule and Zakian , 2007 ) , Pif1’s periodic patrolling activity , which is capable of unwinding RNA–DNA heteroduplexes , may be responsible for the Pif1-mediated telomerase inhibition at 3′ ssDNA ends . Third , periodic patrolling can be used to keep G4 DNA unfolded at transcription bubble ( Figure 8A ) , at DSBs or telomeres ( Figure 8C ) or on the lagging and/or leading strand during replication ( Figure 8D , E ) . We showed that periodic patrolling by a Pif1 monomer can unwind intramolecular G4 DNA structures at every patrolling cycle . Because , as shown here , an unwound G4 DNA refolds immediately ( ∼0 . 2 s ) , the periodic patrolling activity of Pif1 due to its anchoring at the junction is critical to ensure that G4 structures are kept resolved . In gapped DNA substrates , the patrolling activity can keep G4 sequences unfolded in the ssDNA gap while leaving the flanking dsDNA regions intact . This intrinsic property of a Pif1 monomer would suppress G4 formation until other DNA metabolic enzymes take over without causing the side effect of uncontrolled dsDNA unwinding . 10 . 7554/eLife . 02190 . 022Figure 8 . Possible cellular sites for the patrolling activity of a Pif1 monomer . The periodic patrolling activity may keep Pif1 at its site of in vivo action in resolving biologically relevant ‘R-loops’ ( A ) , displacing telomerase from 3′ ssDNA ends at telomeres or DSBs ( B ) , unwinding G4 structures on the 3′ ssDNA tails at telomeres or DSBs ( C ) , and unwinding G4 structures on the lagging and/or leading strands during DNA replication ( D and E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02190 . 022 Additionally , our results provide an example in which separation of helicase function can be achieved by the assembly state . Many SF1 and SF2 helicases , including PcrA and UvrD , display limited dsDNA unwinding activity as monomers , and their helicase activity can be greatly enhanced through self-assembly or interactions with accessory proteins ( Singleton et al . , 2007; Lohman et al . , 2008 ) . Our data suggest that the monomeric form of Pif1 is able to unwind RNA–DNA heteroduplexes and G4 structures , whereas dsDNA unwinding requires the binding of multiple Pif1 molecules per DNA substrate . As a result , DNA binding-induced enzyme oligomerization ( Barranco-Medina and Galletto , 2010 ) and/or interactions with accessory proteins might play a role in regulating the translocation vs helicase activity of Pif1 . Thus far , more than 20 helicases have been shown to bind and/or unwind G4 structures in vitro , among which Pif1 displays the highest specificity in G4 unwinding ( Paeschke et al . , 2013 ) . We have for the first time resolved the intermediates for G4 unwinding by a helicase . The three discrete steps observed are likely due to unraveling of G4 DNA one strand at a time ( Figure 7D ) , which indicates that the G4 unwinding by Pif1 is tightly coupled to its translocation activity . It has been shown that the stabilization energy of a Hoogsteen GG base pair is comparable to that of AT base pair ( Mashimo and Sugiyama , 2007 ) , and that the lifetime of G-triplex and G-hairpin could be as long as 0 . 1–1 s , consistent with our observations ( Li et al . , 2013 ) . In addition , we also showed that a G4 stabilizing compound , BRACO19 , cannot inhibit Pif1-mediated G4 unwinding . In conclusion , our study provides the mechanistic details on how Pif1 may achieve its distinct in vivo functions at the molecular level . The periodic patrolling activity by its monomeric form may keep the enzyme at its sites of in vivo action in displacing telomerase from 3′ ssDNA ends , resolving biologically relevant ‘R-loops’ , and keeping G4 DNA sequences unfolded during DNA replication , recombination and repair . The inability of a Pif1 monomer to display dsDNA unwinding activity might be important for preventing Pif1 monomers from unwinding dsDNA after the Pif1 monomer has displaced proteins , removed RNA strands or resolved G4 structures from within a ssDNA gap .
Partial duplex ( pd ) DNA substrates with a 3′ Poly ( dT ) ssDNA overhang5′-/Cy3/GCC TCG CTG CCG TCG CCA/biotin/-3′5′-TGG CGA CGG CAG CGA GGC ( T ) 15/iAmMC6T/ ( T ) 24-3′5′-/Cy5/GCC TCG CTG CCG TCG CCA/biotin/-3′5′-TGG CGA CGG CAG CGA GGC ( T ) N/Cy3/-3′ , where N = 21 , 32 , 40 , 56 , 725′-GCC TCG CTG CCG TCG CCA/biotin/-3′5′-TGG CGA CGG CAG CGA GGC ( T ) 39/iAmMC6T/ ( T ) 31/3AmMO/-3′5′-/Cy5/GCC TCG CTG CCG TCG CCA-3′ Forked DNA substrates5′- ( dT ) 19/iAmMC6T//GCC TCG CTG CCG TCG CCA/biotin/-3′5′-TGG CGA CGG CAG CGA GGC ( T ) 40/Cy3/-3′ Gapped substrates5′-/Cy5/GCC TCG CTG CCG TCG CCA/biotin/-3′5′-TGG CGA CGG CAG CGA GGC ( T ) 40 GCA CTC GGA TCA CCA TGG CGG ACT CTC TGC T-3′5′-AGC AGA GAG TCC GCC ATG GTG ATC CGA GTG C /Cy3/-3′5′-/Cy3/AGC AGA GAG TCC GCC ATG GTG ATC CGA GTG C-3′5′-AGC AGA GAG UCC GCC AUG GUG AUC CGA GUG C /Cy3/-3′ ( RNA ) 5′-TGG CGA CGG CAG CGA GGC ( T ) 31/iAmMC6T/ ( T ) 18 CAC CAT GGC GGA CTC TCT GCT-3′5′-AGC AGA GAG TCC GCC ATG GTG/3AmMO/-3′5′-GCC TCG CTG CCG TCG CCA/biotin/-3′5′-TGG CGA CGG CAG CGA GGC ( T ) 20 CAC CAT GGC GGA CTC TCT GCT-3′ Intramolecular G4 substrates ( for Pif1 ) 5′-GCC TCG CTG CCG TCG CCA/biotin/-3′5′-TGG CGA CGG CAG CGA GGC ( T ) 39/iAmMC6T/GGG GGA GCT GGG GTA GAT GGG AAT GTG AGG G/3AmMO/-3′5′-TGG CGA CGG CAG CGA GGC AAA GGG GGA GCT GGG GTA GAT GGG AAT GTG AGG G AAA/Cy3/-3′5′-TGG CGA CGG CAG CGA GGC ( T ) 39/iAmMC6T/A GGG TTA GGG TTA GGG TTA GGG TT/3AmMO/-3′5′-TGG CGA CGG CAG CGA GGC ( T ) 37/iAmMC6T/A GGG TTA GGG TTA GGG TTA GGG TCA CCA TGG CGG ACT CTC TGC T-3′5′-TGG CGA CGG CAG CGA GGC/iAmMC6T/A GGG TTA GGG TTA GGG TTA GGG TCA CCA TGG CGG ACT CTC TGC T-3′5′-AGC AGA GAG TCC GCC ATG GTG/3AmMO/-3′ pdDNA substrates with a 5′ Poly ( dT ) ssDNA overhang5′-/biotin/TGG CGA CGG CAG CGA GGC/Cy5/-3′5′-/Cy3/ ( dT ) 40 GCC TCG CTG CCG TCG CCA-3′ Gapped substrates ( for PcrA ) 5′-/biotin/TAC TCG CGC CGT CGC TCC G/Cy5/-3′5′-TGG CGA CGG CAG CGA GGC ( T ) 40 CGG AGC GAC GGC GCG AGT-3′5′-/Cy3/GCC TCG CTG CCG TCG CCA-3′ Intramolecular G4 substrates ( for PcrA ) 5′-/biotin/TGG CGA CGG CAG CGA GGC/Cy5/-3′5′-/5AmMC6/TA GGG TTA GGG TTA GGG TTA GGG/iAmMC6T/ ( T ) 39 GCC TCG CTG CCG TCG CCA-3′ DNA strands were purchased from Integrated DNA Technologics , Coralville , IA . The TP sequence is bold , and the TEL sequence is underlined . The amine-modified thymine ( /iAmMC6T/ ) , 3′ amino-modification ( /3AmMO/ ) , and 5′ amino-modification ( /5AmMC6/ ) shown in the sequence enable the oligonucleotides to be labeled with the monofunctional NHS ester form of Cy3 or Cy5 dye ( GE Healthcare , Piscataway , NJ ) . /Cy3/ and /Cy5/ represent Cy3 and Cy5 dyes , respectively , linked to the DNA backbone using phosphoramidite chemistry . The substrates were annealed by mixing ∼5 μM of each strand in 10 mM Tris:HCl ( pH 8 . 0 ) and 200 mM KCl , followed by slow cooling from 90°C to room temperature for ∼2 hr . The S . cerevisiae Pif1 was purified as previously reported ( Boule et al . , 2005 ) . The B . stearothermophilus PcrA was purified as previously described ( Niedziela-Majka et al . , 2007; Park et al . , 2010 ) . All smFRET experiments were performed with total internal reflection fluorescence ( TIRF ) microscopy ( Roy et al . , 2008 ) at 22 ± 1°C in imaging buffer composed of 20 mM Tris:HCl ( pH 7 . 5 ) , 5 mM MgCl2 , 60 mM KCl , 0 . 1 mg/ml BSA , 1 mM DTT , 2% ( vol/vol ) glycerol , and an oxygen scavenging system ( 0 . 5% wt/vol glucose , 3 mM Trolox , 165 U/ml glucose oxidase , and 2170 U/ml catalase ) unless specified otherwise . Cy3-Cy5 labeled pdDNA substrates ( 50–100 pM ) were immobilized on a quartz slide surface coated with polyethyleneglycol ( mPEG-SC , Laysan Bio , Arab , AL ) in order to eliminate nonspecific surface adsorption of proteins ( Ha et al . , 2002; Roy et al . , 2008 ) . Surface immobilization was mediated by biotin–Neutravidin interactions between biotinylated DNA , Neutravidin ( Thermo Scientific , Newington , NH ) , and biotinylated polymer ( Bio-PEG-SC , Laysan Bio , Arab , AL ) . To ensure no more than one Pif1 monomer is loaded per DNA ( i . e . , under monomer conditions ) , 10 nM Pif1 was incubated with the surface-immobilized DNA for 2 min , and excess unbound proteins were then flushed out from the sample chamber using five chamber volumes of imaging buffer . Next , imaging buffer devoid of proteins was added with 20 μM ATP ( unless specified otherwise ) into the chamber . Finally , the Cy3/Cy5 fluorescence intensities from single DNA molecules were recorded with 30-ms time resolution . For some experiments , ATP was added together with proteins into the chamber for data acquisition after immobilizing the DNA substrates . In the pull-down assays , Histidine6-tagged Pif1 was pulled down to the surface using an antibody against the Histidine6-tag following the procedure described previously ( Jain et al . , 2011 ) . To study the unwinding of dsDNA or RNA–DNA hybrids , gapped substrates were immobilized on the PEG surface , and the mean Cy3 spot count per image ( imaging area = ∼2500 μm2 ) was then determined from images taken from 5–10 different slide regions at different times after introducing ATP . Apparent FRET efficiency was calculated from the fluorescence intensities of the donor ( ID ) and acceptor ( IA ) using the formula EFRET = IA/ ( IA + ID ) . The background and the cross-talk between the donor and acceptor were considered as previously described ( Roy et al . , 2008 ) . Analysis of individual FRET-time traces was performed using Origin or programs written in Matlab . Single molecule FRET histograms were generated by averaging for the time period of 0 . 15 s from ∼10 , 000 DNA molecules . The time period Pif1 takes for one cycle of looping , Δt , was determined by visually picking the moments when EFRET reaches its maximum value ( i . e . , the peaks ) in the single molecule FRET-time traces showing periodic sawtooth or wave patterns . The time period Pif1 takes for one cycle of G4 unwinding , Δt , was also determined by visually picking the moments when EFRET recovers to the EFRET value that represents folded G4 structures ( Figure 5A ) . The Δt histograms were generated by collecting Δt from >50 molecules . To quantify the stepwise patterns observed in single molecule FRET-time traces for G4 unwinding , average EFRET values and the dwell time of each unwinding step were determined using the automated step-finding method used in previous single molecule FRET studies ( Kerssemakers et al . , 2006; Myong et al . , 2007; Lee et al . , 2012 ) . | Helicases are enzymes that are best known for their ability to separate the two strands of DNA that make up the famous double-helix structure . Many important processes within cells—including the expression of genes as proteins , and the replication of DNA before cell division—rely on DNA molecules being separated in this way . However , these enzymes can perform many other roles that help maintain the integrity of a cell’s DNA . The genetic code is written using four DNA bases—called A , C , G and T—and if a stretch of DNA contains lots of G bases , then one of the strands can loop back upon itself three times to form a structure known as a ‘G-quadruplex’ . These structures can prevent the expression of genes , and slow the replication of DNA . However , a helicase called Pif1 can unwind G-quadruplexes to allow these activities to continue . This helicase is found in many organisms , from bacteria to humans , and carries out multiple functions for a cell . However , the exact mechanisms underlying these activities are unknown . Now , Zhou et al . have used biophysical techniques to reveal that individual Pif1 proteins bind to single-stranded overhangs at one end of a DNA molecule . Pif1 also binds to forks in DNA where the double helix separates into two single strands . And once Pif1 has bound to the DNA , it works to ‘reel in’ the overhang or a single strand , one base at a time . This activity can unwind a G-quadruplex , and individual Pif1 proteins will patrol DNA to keep this structures unwound without unraveling the double helix itself . Separating the two strands of DNA actually needs multiple Pif1 proteins to join and work together . As it patrols , Pif1 also displaces other proteins from DNA and removes unusual , and potentially harmful , structures in DNA ( such as RNA molecules that have displaced one of the strands of DNA double helix ) . The next challenge will be to address important questions that remain unanswered including: how does Pif1 recognize DNA structures and change its activity; and how does it coordinate with other proteins that target the same structures ? | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | Periodic DNA patrolling underlies diverse functions of Pif1 on R-loops and G-rich DNA |
Oncogenic mutations in BRAF and NRAS occur in 70% of melanomas . In this study , we identify a microRNA , miR-146a , that is highly upregulated by oncogenic BRAF and NRAS . Expression of miR-146a increases the ability of human melanoma cells to proliferate in culture and form tumors in mice , whereas knockdown of miR-146a has the opposite effects . We show these oncogenic activities are due to miR-146a targeting the NUMB mRNA , a repressor of Notch signaling . Previous studies have shown that pre-miR-146a contains a single nucleotide polymorphism ( C>G rs2910164 ) . We find that the ability of pre-miR-146a/G to activate Notch signaling and promote oncogenesis is substantially higher than that of pre-miR-146a/C . Analysis of melanoma cell lines and matched patient samples indicates that during melanoma progression pre-miR-146a/G is enriched relative to pre-miR-146a/C , resulting from a C-to-G somatic mutation in pre-miR-146a/C . Collectively , our results reveal a central role for miR-146a in the initiation and progression of melanoma .
Melanoma is the deadliest form of skin cancer accounting for ∼80% of skin cancer-related deaths ( Miller and Mihm , 2006 ) . The most commonly observed oncogenic events in melanomas are activating mutations in the BRAF and NRAS proto-oncogenes , which occur in 70% of cases ( Miller and Mihm , 2006; Tsao et al . , 2012 ) . Activating mutations in BRAF and NRAS genes cause constitutive activation of downstream signaling pathways , resulting in pro-proliferative and anti-apoptotic effects that promote cellular transformation , tumor growth and metastasis ( Downward , 2003; Wellbrock et al . , 2004; Karnoub and Weinberg , 2008 ) . Oncogenic BRAF mutants ( typically BRAFV600E ) primarily activate the MAPK pathway ( Wellbrock et al . , 2004 ) . Genetic and pharmacological studies have shown that disruption of the BRAF-MEK-ERK pathway blocks the growth of melanoma cells harboring an oncogenic BRAF mutation and thus represents an attractive therapeutic target ( Wellbrock et al . , 2004; Miller and Mihm , 2006; Tsao et al . , 2012 ) . Oncogenic NRAS is capable of activating multiple downstream pathways , including BRAF-MEK-ERK , all of which are thought to play an important role in NRAS driven oncogenesis ( Downward , 2003; Karnoub and Weinberg , 2008 ) . The Notch pathway is an evolutionary conserved signaling cascade that has an essential role in embryonic development and cell renewal in the adult ( Guruharsha et al . , 2012 ) . Notch signaling has a key role in melanoblast and melanocyte homeostasis ( Haass and Herlyn , 2005; Aubin-Houzelstein et al . , 2008; Kumano et al . , 2008 ) . For example , conditional ablation of Notch signaling in the melanocyte lineage leads to drastic elimination of melanoblasts and melanocyte stem cells ( Moriyama et al . , 2006 ) . NOTCH1 expression is normally decreased in mature melanocytes , whereas melanomas regain expression and activity of NOTCH1 ( Balint et al . , 2005; Pinnix et al . , 2009 ) . NOTCH1 is required for melanoma formation , can transform primary human melanocytes and can confer metastatic properties to primary melanoma cells ( Liu et al . , 2006; Asnaghi et al . , 2012 ) . miRNAs are small non-coding RNAs that function by regulating the stability or translation of mRNAs ( Bartel , 2004; He and Hannon , 2004; Leung and Sharp , 2006 ) . miRNAs have been implicated in essentially all aspects of tumor biology including tumorigenesis , angiogenesis and metastasis ( Croce , 2009; Garzon et al . , 2009 ) indicating that , similar to protein-coding genes , miRNAs function as crucial regulators of tumor initiation and progression . Interestingly , miRNAs can act as either tumor suppressors or oncogenes depending on the functions of their targets ( Croce , 2009; Garzon et al . , 2009 ) . High-throughput profiling has revealed dysregulation of miRNAs in a variety of cancers ( Croce , 2009; Garzon et al . , 2009 ) . For example , more than half of miRNA genes in human cancers are located in chromosomal regions that frequently exhibit amplification , deletion , or translocation ( Croce , 2009; Garzon et al . , 2009 ) . Whether microRNAs ( miRNAs ) have a role in BRAF- and NRAS-driven melanoma initiation and progression remains to be determined . In this study , we perform small RNA profiling and identify miR-146a as an oncogenic BRAF- and NRAS-regulated miRNA that promotes the initiation and progression of melanoma . We show that miR-146a functions as an oncogene by activating Notch signaling , and that during melanoma progression pre-miR-146a can acquire a somatic mutation that enhances its oncogenic activity .
To identify possible BRAFV600E-regulated miRNAs , we generated miRNA libraries from primary lung fibroblast WI-38 cells transduced with retrovirus expressing BRAFV600E or an empty vector and performed deep sequencing analysis . To rule out miRNAs that are altered due to cell cycle arrest , we also sequenced a miRNA library generated using quiescent WI-38 cells ( Figure 1A ) . Our analysis identified miR-146a as the most upregulated miRNA by BRAFV600E ( Figure 1B ) . To confirm that miR-146a is a target of BRAFV600E , we transduced human fibroblasts WI-38 , IMR-90 and primary human melanocytes ( hereafter referred to as melanocytes ) with retroviruses expressing BRAFV600E . We found that BRAFV600E activates miR-146a expression in WI-38 as well as in IMR-90 and melanocytes ( Figure 1—figure supplement 1 , Figure 1C ) . 10 . 7554/eLife . 01460 . 003Figure 1 . BRAFV600E and NRASQ61K upregulate miR-146a expression . ( A ) Schematic summary of study design . ( B ) Volcano plot showing miR-146a ( red arrow ) as the most upregulated miRNA . ( C ) qRT-PCR analysis of miR-146a expression in primary melanocytes transduced with BRAFV600E ( gray ) or Vector control ( black ) retrovirus particles . Quiescent cells were used as a control ( light gray ) . ( D ) qRT-PCR analysis ( top ) measuring miR-146a expression in MEL-ST cells expressing BRAFV600E in the presence ( + ) or absence ( − ) of the MEK inhibitor U0216 ( 10 µM ) relative to Vector control . Immunoblots ( bottom ) of phosphorylated ( p- ) ERK ( upper ) and total ( t- ) ERK ( lower ) show that BRAFV600E activates MEK-dependent phosphorylation of ERK . ( E ) qRT-PCR analysis ( top ) measuring miR-146a expression in MEL-ST cells expressing NRASQ61K relative to Vector control . Immunoblots ( bottom ) of phosphorylated ( p- ) ERK ( upper ) and total ( t- ) ERK ( lower ) show that NRASQ61K activates phosphorylation of ERK . ( F ) qRT-PCR analysis ( top ) measuring miR-146a expression in MEL-ST cells expressing constitutively active MEK ( DD ) , relative to the Vector control . Immunoblots ( bottom ) of p-ERK ( upper ) and t-ERK ( lower ) show that MEK ( DD ) activates phosphorylation of ERK . ( G ) qRT-PCR analysis monitoring miR-146a expression in melanocyte and melanoma cell lines . BRAF ( blue ) and NRAS ( orange ) mutant cell lines are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 00310 . 7554/eLife . 01460 . 004Figure 1—figure supplement 1 . Ectopic expression of BRAFV600E in WI-38 and IMR-90 cells stimulates miR-146a expression . qRT-PCR analysis of miR-146a expression in WI-38 or IMR-90 cells transduced with BRAFV600E ( gray ) or Vector control ( black ) retrovirus particles . Quiescent cells were used as a control ( light gray ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 00410 . 7554/eLife . 01460 . 005Figure 1—figure supplement 2 . Activation of MAP Kinase pathway target genes upon ectopic expression of BRAFV600E in MEL-ST cells . qRT-PCR analysis of MEK-ERK transcriptional targets FOS , EGR1 and FOSL1 under indicated conditions in MEL-ST cells expressing either an empty vector or BRAFV600E . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 00510 . 7554/eLife . 01460 . 006Figure 1—figure supplement 3 . Ectopic expression of HRAS v12 or HRAS v12 S35 in MEL-ST cells stimulates miR-146 expression . qRT-PCR analysis of miR-146a ( left ) and immunoblot analysis ( right ) of phosphorylated ( p- ) ERK , total ( t- ) ERK in MEL-ST cells transduced with empty vector , HRAS v12 , or HRAS v12s35 . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 00610 . 7554/eLife . 01460 . 007Figure 1—figure supplement 4 . Inhibition of MAP kinase signaling by MEK inhibitor blocks HRAS v12-mediated miR-146a upregulation . qRT-PCR analysis ( left ) to monitor miR-146a expression and immunoblot analysis ( right ) of phosphorylated ( p- ) ERK and total ( t- ) ERK in MEL-ST cells transduced with HRAS v12 in the presence ( + ) or absence ( − ) of the MEK inhibitor U0216 . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 00710 . 7554/eLife . 01460 . 008Figure 1—figure supplement 5 . Ectopic expression of MEK DD stimulates the transcription of MAP kinase target genes . qRT-PCR analysis of MEK-ERK transcriptional targets FOS , EGR1 and FOSL1 in MEL-ST cells expressing either a empty vector or constitutively active MEK ( MEK DD ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 008 Next , we asked whether BRAF-MEK-ERK signaling is required for miR-146a upregulation . We stably expressed BRAFV600E or an empty vector in MEL-ST cells , which are immortalized melanocytes that have basal levels of BRAF-MEK-ERK signaling and can be transformed by single oncogenes ( Gupta et al . , 2005 ) . Figure 1D shows that , as in melanocytes , introduction of BRAFV600E into MEL-ST cells upregulated miR-146a expression . To confirm that miR-146a upregulation requires BRAF-MEK-ERK signaling , we treated MEL-ST/BRAFV600E cells with the MEK inhibitor U0216 . Figure 1D shows that treatment of MEL-ST/BRAFV600E cells with U0216 resulted in decreased miR-146a expression . As expected , transcriptional targets of the MAP kinase pathway were also downregulated following U0216 treatment ( Figure 1—figure supplement 2 ) . Similar to BRAFV600E , stable expression of NRASQ61K , HRAS v12 or a mutant that selectively activates BRAF-MEK-ERK signaling also resulted in increased miR-146a levels ( Figure 1E , Figure 1—figure supplement 3 ) , which was attenuated by addition of U0216 ( Figure 1—figure supplement 4 ) . To further confirm that increased BRAF-MEK-ERK signaling is sufficient to induce miR-146a expression , we stably expressed a constitutively active MEK derivative ( MEK DD ) in MEL-ST cells . Figure 1F shows that expression of MEK DD substantially increased BRAF-MEK-ERK signaling and stimulated miR-146a expression . As expected , transcriptional targets of the MAP kinase pathway were also upregulated in cells expressing MEK DD ( Figure 1—figure supplement 5 ) . Over 60% of melanomas harbor mutations in the BRAF and NRAS genes ( Davies et al . , 2002; Chin et al . , 2006 ) . We therefore asked whether miR-146a expression is upregulated in melanoma cell lines containing mutant BRAF or NRAS . We monitored miR-146a expression in a panel of BRAF/NRAS wild-type or BRAF or NRAS mutant melanoma cell lines , short-term patient-derived melanoma cultures and melanocytes . Figure 1G shows that miR-146a expression was significantly higher in a subset of BRAF and NRAS mutant melanoma cell lines compared to both BRAF/NRAS wild-type melanoma cell lines and melanocytes . Collectively , our results show that BRAF-MEK-ERK signaling is necessary and increased BRAF-MEK-ERK signaling is sufficient for upregulation of miR-146a . Several previous studies have shown that the BRAF-MEK-ERK pathway regulates transcription , protein stability and activity of MYC ( Sears et al . , 1999 , 2000; Pintus et al . , 2002; Marampon et al . , 2006; Tsai et al . , 2012 ) . Analysis of the miR-146a promoter sequence using the PROMO 3 . 0 bioinformatics program ( Messeguer et al . , 2002 ) identified a potential MYC binding site ( Figure 2A ) , suggesting that MYC promotes miR-146a upregulation . Stable expression of BRAFV600E in MEL-ST cells led to increased MYC expression , increased MYC phosphorylation ( Figure 2B ) and , as monitored in a chromatin immunoprecipitation ( ChIP ) assay , binding of MYC to the miR-146a promoter ( Figure 2C ) . Notably , binding of MYC to miR-146a promoter was inhibited by the treatment of cells with U0216 ( Figure 2C ) . Furthermore , shRNA-mediated knockdown of MYC in melanoma cell lines , SKMEL-28 and M14 , substantially decreased miR-146a levels and the expression of other MYC target genes ( Figure 2D , Figure 2—figure supplements 1–3 ) . As expected , shRNA-mediated knockdown of MYC also resulted in decreased binding of MYC to the miR-146a promoter ( Figure 2E ) . In addition to MYC , we also identified binding sites for transcription factors ETS1 , ELK1 , NF-κB , and c/EBPβ in the miR-146a promoter . However , unlike MYC , shRNA-mediated knockdown of these other transcription factors did not significantly affect miR-146a expression ( Figure 2—figure supplement 4 ) . Collectively , our results show that increased BRAF-MEK-ERK signaling results in activation and recruitment of MYC to the miR-146a promoter , which stimulates miR-146a transcription . 10 . 7554/eLife . 01460 . 009Figure 2 . BRAFV600E upregulates miR-146a through MYC oncogene . ( A ) Schematic representation of MYC promoter and the miR-146a binding site that has been identified by PROMO analysis software . ‘+1’ indicates the transcription start site . ( B ) Immunoblot analysis of p-MYC and total MYC in whole cell lysates of MEL-ST cells transduced with empty vector or BRAFV600E . Actin was used as a loading control . ( C ) Chromatin immunoprecipitation ( ChIP ) assay measuring MYC binding to the miR-146a promoter in MEL-ST cells stably expressing BRAFV600E in the presence ( + ) or absence ( − ) of the MEK inhibitor U0216 relative to cells transduced with the empty vector . A non-specific control region served as a negative control for MYC recruitment . ( D ) qRT-PCR analysis of MYC mRNA ( left ) and miR-146a ( right ) expression in SKMEL-28 cells infected with MYC shRNAs . ( E ) qPCR analysis of MYC ChIP of miR-146a promoter and negative control in SKMEL-28 cells infected with MYC shRNAs . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 00910 . 7554/eLife . 01460 . 010Figure 2—figure supplement 1 . Analysis of MYC and miR-146a expression in M14 cells expressing shRNAs against MYC . qRT-PCR analysis of MYC and miR-146a expression in M14 cells transduced with MYC shRNA expression vectors relative to cells transduced with a non-specific ( NS ) shRNA vector . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 01010 . 7554/eLife . 01460 . 011Figure 2—figure supplement 2 . shRNA-mediated downrgulation of MYC in M14 cells inhibits the expression of MYC transcriptional target genes . qRT-PCR analysis of MYC targets CCDN1 and CDC25C in M14 cells transduced with MYC shRNA expression vectors relative to cells transduced with a non-specific ( NS ) shRNA vector . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 01110 . 7554/eLife . 01460 . 012Figure 2—figure supplement 3 . shRNA-mediated downrgulation of MYC in SKMEL-28 cells inhibits the expression of MYC transcriptional target genes . qRT-PCR analysis of MYC targets CCDN1 and CDC25C in SKMEL-28 cells transduced with MYC shRNA expression vectors relative to cells transduced with a non-specific ( NS ) shRNA vector . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 01210 . 7554/eLife . 01460 . 013Figure 2—figure supplement 4 . Transcriptional regulation of miR-146a . qRT-PCR analysis of ETS1 , ELK1 , NF-κB , c/EBPβ and miR-146a in SKMEL-28 expressing indicated shRNA . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 013 Several recent studies have found that a pre-miR-146a SNP ( C>G rs2910164 ) alters the expression of mature miR-146a and correlates with an increased risk to several cancers ( Jazdzewski et al . , 2008; Hezova et al . , 2012; Hung et al . , 2012; Lung et al . , 2012; Wang et al . , 2012; Yamashita et al . , 2013 ) . This SNP has been shown to occur in the pre-miR-146a sequence and does not alter the sequence of mature miR-146a ( Jazdzewski et al . , 2008; Hezova et al . , 2012; Hung et al . , 2012; Lung et al . , 2012; Wang et al . , 2012; Yamashita et al . , 2013 ) . The mechanism by which this SNP promotes tumorigenesis and its potential role in melanomagenesis remain to be determined . To address this question , we expressed both pre-miR-146a/C and pre-miR-146a/G ( Figure 3A ) in highly-tumorigenic human melanoma cell lines that efficiently formed colonies in soft-agar and tumors in immunocompromised mice . Significantly , consistent with a previous report ( Jazdzewski et al . , 2008 ) , the amount of mature miR-146a produced from pre-miR-146a/G was higher than that from pre-miR-146a/C ( Figure 3—figure supplement 1 ) . Ectopic expression of pre-miR-146a/G promoted proliferation at a higher rate than pre-miR-146a/C , as evidenced by increased colony formation and increased proliferation in two of the three melanoma cell lines analyzed ( Figure 3B , Figure 3—figure supplement 2 ) . We also compared the ability of pre-miR-146a/C and pre-miR-146a/G to promote anchorage-independent growth in soft-agar . Again , pre-miR-146a/G-stimulated colony formation more efficiently than pre-miR-146a/C ( Figure 3C ) . Notably , although expression of pre-miR-146a/G in A375 cells did not increase proliferation in liquid culture ( Figure 3B ) , it did increase colony formation in soft-agar ( Figure 3C ) . Conversely , inhibition of miR-146a by miRZip-146a in SKMEL-28 and M14 cells reduced colony formation in liquid culture and soft-agar , and inhibited tumor formation in mice ( Figure 3D–G and Figure 3—figure supplement 3 ) . Similarly , expression of a miR-146a locked nucleic acid ( LNA ) -based antagomiR in SKMEL-28 and M14 cells reduced colony formation in liquid culture and soft-agar ( Figure 3—figure supplements 4 and 5 ) . By contrast , expression of a miR-146a antagomiR in YUSIV cells , which express low levels of miR-146a , did not significantly affect colony formation in either liquid culture or soft-agar ( Figure 3—figure supplement 6 ) . 10 . 7554/eLife . 01460 . 014Figure 3 . Oncogenic activity of pre-miR-146a/C and pre-miR-146a/G . ( A ) Schematic representation of pre-miR-146a/C and pre-miR-146a/G sequences . ( B and C ) Number of colonies formed in liquid ( B ) or soft-agar ( C ) by M14 , SKMEL-28 or A375 melanoma cells expressing pre-miR-146a harboring a C or G at position 40 , as compared to the Vector ( − ) control . Colonies were counted after 2 weeks ( B ) or 4 weeks ( C ) of growth . ( D ) qRT-PCR analysis of miR-146a expression in SKMEL-28 cells infected with miRZip-146a ( + ) or an empty vector ( − ) . ( E and F ) Number of colonies formed in liquid ( E ) or soft-agar ( F ) by SKMEL-28 expressing miRZip-146a ( + ) or the vector ( − ) control . Colonies were counted after 2 weeks ( E ) or 4 weeks ( F ) of growth . ( G ) Average tumor volumes at 1 month time points for mice injected with SKMEL-28 expressing a control miRZip or miRZip-146a . ( H ) Average volume of tumors formed by SKMEL-28 cells ( 2 . 5 × 106 ) expressing pre-miR-146a/C or pre-miR-146a/G , as compared to the Vector control , injected subcutaneously into the flanks of nude mice ( n = 5 ) . ( I ) Representative images ( left ) and colony number in soft-agar ( right ) for the NRASQ61K transformed MEL-ST cells that express an empty vector or miRZip-146a . ( J ) Average tumor volumes at the indicated time points for mice injected with NRASQ61K transformed MEL-ST cells expressing a control miRZip or miRZip-146a . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 01410 . 7554/eLife . 01460 . 015Figure 3—figure supplement 1 . Analysis of miR-146a expression in SKMEL-28 , A375 and M14 cells expressing either pre-miR-146a/G or pre-miR-146a/C construct . SKMEL-28 , M14 and A375 cells stably expressing pre-miR-146a/C ( C ) or pre-miR-146a/G ( G ) or an empty vector ( − ) were analyzed for miR-146a expression by qRT-PCR . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 01510 . 7554/eLife . 01460 . 016Figure 3—figure supplement 2 . pre-miR-146a/G promotes proliferation of melanoma cells more effective than pre-miR-146a/C . M14 , SKMEL-28 and A375 cells stably expressing pre-miR-146a/C ( blue ) or pre-miR-146a/G ( red ) or an empty vector ( black ) were analyzed for proliferation at indicated days . Relative proliferation is plotted . * , ** and *** represents p values <0 . 01 , <0 . 001 and <0 . 0001 respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 01610 . 7554/eLife . 01460 . 017Figure 3—figure supplement 3 . Inhibition of miR-146a expression blocks the proliferation and anchorage-independent growth of M14 cells . M14 cells either expressing a control miRZip vector or miRZip-146a were analyzed for miR-146a expression , colony formation , or growth in soft-agar or tumor formation in mice . Tumor volume at 1 month timepoint is plotted . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 01710 . 7554/eLife . 01460 . 018Figure 3—figure supplement 4 . Inhibition of miR-146a expression blocks the proliferation and anchorage-independent growth of SKMEL-28 cells . SKMEL-28 cells either expressing a scrambled LNA-antagomiR or LNA-based miR146a antagomiR were analyzed for miR-146a expression ( left ) , colony formation ( middle ) or growth in soft-agar ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 01810 . 7554/eLife . 01460 . 019Figure 3—figure supplement 5 . Inhibition of miR-146a expression blocks the proliferation and anchorage-independent growth of M14 cells . M14 cells either expressing a scrambled LNA-antagomiR or LNA-based miR146a antagomiR were analyzed for miR-146a expression ( left ) , colony formation ( middle ) or growth in soft-agar ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 01910 . 7554/eLife . 01460 . 020Figure 3—figure supplement 6 . Inhibition of miR-146a expression does not block the proliferation and anchorage-independent growth of YUSIV cells . ( Top panel ) YUSIV cells either expressing a control miR-ZIP vector or miR-Zip-miR146a were analyzed for miR-146a expression ( left ) , colony formation ( middle ) or growth in soft-agar ( right ) . ( Bottom Panel ) YUSIV cells either expressing a control LNA-based control antagomiR or LNA-based miR-146a antagomiR were analyzed for miR-146a expression ( left ) , colony formation ( middle ) or growth in soft-agar ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 02010 . 7554/eLife . 01460 . 021Figure 3—figure supplement 7 . Analysis of miR-146a expression in indicated melanoma cell lines transfected with increasing concentration of synthetic miR-146a . qRT-PCR analysis of miR-146a expression in indicated melanoma cell lines . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 02110 . 7554/eLife . 01460 . 022Figure 3—figure supplement 8 . miR-146a enhances the growth of melanoma cell lines in a concentration dependent manner . Relative proliferation of indicated melanoma cell lines 5 days after the transfection of synthetic mature miR-146a at indicated concentrations . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 02210 . 7554/eLife . 01460 . 023Figure 3—figure supplement 9 . NRASQ61K is sufficient to transform MEL-ST cells . Representative images of soft-agar colonies formed by MEL-ST cells expressing BRAFV600E . NRASQ61K was used as a positive control . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 023 Finally , we subcutaneously injected SKMEL-28 cells stably expressing pre-miR-146a/C or pre-miR-146a/G , or an empty vector into immunocompromised mice . Although both pre-miR-146a/C and pre-miR146a/G enhanced tumor growth , the effect was much larger with pre-miR-146a/G ( Figure 3H ) . To confirm that the stronger oncogenic effect of pre-miR-146a/G is due to increased abundance of mature miR-146a , we transfected SKMEL-28 cells with increasing amounts of synthetic miR-146a and monitored cell proliferation . Notably , increased levels of miR-146a enhanced proliferation in a dose-dependent manner ( Figure 3—figure supplements 7 and 8 ) . Collectively , our results show that the oncogenic activity of pre-miR-146a/G is greater than that of pre-miR-146a/C both in vitro and in vivo due to increased abundance of mature miR-146a . Next , we asked whether miR-146a plays a role in BRAFV600E-mediated cellular transformation . For these experiments , we used immortalized but not transformed MEL-ST cells that can be transformed by a single oncogene . In agreement with a previous report ( Chudnovsky et al . , 2005 ) , we found that BRAFV600E was not sufficient to transform the immortalized melanocytes ( Figure 3—figure supplement 9 ) . Activated alleles of the NRAS gene are the second most common oncogenic mutations in melanoma ( Tsao et al . , 2012 ) . Therefore , we chose to analyze the role of miR-146a in the context of NRASQ61K-induced melanomagenesis . Similar to BRAFV600E , NRASQ61K transcriptionally activates miR-146a expression ( Figure 1E ) . Notably , inhibition of miR-146a expression substantially reduced the ability of NRASQ61K-expressing cells to form colonies in soft-agar ( Figure 3I ) and tumors in mice ( Figure 3J ) . Collectively , our results show that miR-146a promotes the initiation and progression of melanoma . To gain insight into the mechanism of miR-146a-mediated melanomagenesis and melanoma growth , we ectopically expressed pre-miR-146/C or pre-miR-146a/G in SKMEL-28 cells and performed microarray analysis . Of the genes downregulated by both pre-miR-146a/C and pre-miR-146a/G , TargetScan analysis identified 20 mRNAs with potential miR-146a binding sites in the 3′-UTR ( Supplementary file 1A ) . We elected to focus on the NUMB mRNA because it encodes a repressor of NOTCH and several previous studies have shown that NOTCH functions as an oncogene in melanoma ( Liu et al . , 2006; Pinnix et al . , 2009 ) . We identified candidate miR-146a binding sites in both the 3′ UTR and coding region of NUMB mRNA ( Figure 4A ) . Consistent with its increased oncogenic activity , pre-miR-146a/G decreased NUMB mRNA and protein levels more efficiently than pre-miR-146a/C ( Figure 4B , Figure 4—figure supplement 1 ) . Using a luciferase reporter assay , we found that miR-146a does not target the site within the NUMB 3′-UTR ( Figure 4—figure supplement 2 ) . By contrast , ectopic expression of miR-146a downregulated the expression of wild-type NUMB ( NUMB-WT ) open reading frame ( ORF ) , whereas expression of a miR-146a-resistant NUMB ORF ( NUMB-MUT ) lacking the binding site in the coding region was unaffected ( Figure 4C ) . 10 . 7554/eLife . 01460 . 024Figure 4 . Downregulation of NUMB and activation of NOTCH signaling by pre-miR-146a/C and pre-miR-146a/G . ( A ) Schematic representation of the NUMB mRNA and potential miR-146a target sites in the coding region ( blue arrow ) and 3′-UTR . ( B ) qRT-PCR analysis of NUMB mRNA levels in SKMEL-28 cells expressing the indicated pre-miR-146a allele relative to the Vector control . ( C ) Western blot of NUMB expression in SKMEL-28 cells transfected with either wild-type NUMB ( NUMB-WT ) or miR-146a-resistant NUMB ( NUMB-MUT ) and increasing amount of synthetic miR-146a . ( D ) Dual luciferase assay using a CSL-Luciferase reporter to measure NOTCH activity in SKMEL-28 cells expressing the indicated pre-miR-146a alleles relative to the Vector control . ( E ) qRT-PCR analysis of the NOTCH targets HES1 , HEY2 and CCDN1 mRNA in indicated samples . Actin mRNA was used as an internal control . ( F ) Dual luciferase assay using a CSL-Luciferase reporter to measure NOTCH activity in SKMEL-28 cells expressing miRZip-146a relative to the Vector control . ( G ) qRT-PCR analysis of HES1 , HEY2 and CCDN1 mRNA in indicated samples . Actin mRNA was used as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 02410 . 7554/eLife . 01460 . 025Figure 4—figure supplement 1 . Ectopic expression of miR-146a downregulates NUMB protein expression . Immunoblot analysis of NUMB protein in SKMEL-28 cells stably transduced with empty vector , pre-miR-146a/C or pre-miR-146a/G . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 02510 . 7554/eLife . 01460 . 026Figure 4—figure supplement 2 . Relative luciferase activity of NUMB 3′UTR luciferase construct in SKMEL-28 cells expressing an empty vector ( − ) , pre-miR-146a/C ( C ) or pre-miR-146a/G ( G ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 026 To gain additional support for the idea that pre-miR-146a/G can activate Notch signaling more effectively than pre-miR-146a/C , we monitored NOTCH activity using a NOTCH responsive reporter plasmid CSL-pGL3-Luciferase ( CSL-Luc ) in SKMEL-28 cells expressing either pre-miR-146a/C or pre-miR-146a/G . Notably , both pre-miR-146a/C- and pre-miR-146a/G-stimulated CSL-Luc activity but again the effect was greater with pre-miR-146a/G ( Figure 4D ) . Additionally , we analyzed expression of transcriptional target genes of Notch signaling . In agreement with the reporter assays , we found that Notch signaling targets were upregulated to a greater extent by pre-miR-146a/G compared to pre-miR-146a/C ( Figure 4E ) . Conversely , stable knockdown of miR-146a decreased CSL-Luc activity and reduced Notch target gene expression ( Figure 4F , G ) . Collectively , these results confirm that NUMB is a direct target of miR-146a and that miR-146a downregulates NUMB to activate Notch signaling . To investigate the role of NUMB downregulation and activated Notch signaling in melanoma cell growth , we knocked down NUMB in SKMEL-28 cells ( Figure 5A ) and monitored cellular proliferation in liquid culture , anchorage-independent colony formation in soft-agar and tumor formation in mice . Similar to ectopic expression of miR-146a , depletion of NUMB increased cellular proliferation ( Figure 5—figure supplements 1 and 2 ) , soft-agar colony formation ( Figure 5B ) and tumor formation in mice ( Figure 5C ) . Finally , to test if increased NOTCH1 expression recapitulates NUMB loss , we ectopically expressed activated intracellular NOTCH1 ( ICN ) in SKMEL-28 cells ( Figure 5D ) . As predicted , ectopic expression of ICN-enhanced cellular proliferation in liquid culture ( Figure 5—figure supplements 3 and 4 ) , colony formation in soft-agar ( Figure 5E ) and tumor formation in mice ( Figure 5F ) . To confirm that NUMB downregulation is necessary for the ability of miR-146a to promote melanoma growth , we expressed either the miR-146a-resistant NUMB-MUT or NUMB-WT in SKMEL-28 cells expressing pre-miR-146a/G . We found that the expression of NUMB-MUT , but not NUMB-WT , prevented pre-miR-146a/G from promoting colony formation in soft-agar ( Figure 5G ) and activating NOTCH signaling ( Figure 5H ) . Moreover , NUMB-MUT , but not NUMB-WT blocked the ability of NRASQ61K transformed MEL-ST cells to form colonies in soft-agar ( Figure 5I ) . Notably , simultaneous antagonism of miR-146a and knockdown of NUMB restored proliferation of SKMEL-28 and M14 cells in both liquid culture and soft-agar ( Figure 5J–M ) . 10 . 7554/eLife . 01460 . 027Figure 5 . miR-146 oncogenic activity depends on the activation of the NOTCH signaling through downregulation of the tumor suppressor NUMB . ( A ) qRT-PCR analysis ( top ) and immunoblot ( bottom ) of NUMB expression levels in SKMEL-28 cells infected with two different shRNAs against NUMB relative to the control non-silencing shRNA ( NS ) . ( B ) Number of colonies in soft-agar of SKMEL-28 cells expressing NUMB shRNAs relative to the control non-silencing shRNA ( NS ) . ( C ) Average volume of tumors formed by SKMEL-28 cells ( 2 . 5 × 106 ) expressing NUMB shRNAs , relative to the non-silencing shRNA ( NS shRNA ) control , injected subcutaneously into the flanks of nude mice ( n = 5 ) . ( D and E ) Immunoblot of NOTCH ( D ) and colony formation assay ( E ) of SKMEL-28 cells stably transduced with the activated intracellular NOTCH domain ( ICN ) or empty vector . ( F ) Average volume of tumors formed by SKMEL-28 cells expressing the activated intracellular NOTCH domain ( ICN ) relative to vector control . 2 . 5 × 106 cells were injected subcutaneously into the flank of nude mice ( n = 5 ) . ( G ) Colony formation in soft-agar of SKMEL-28 cells expressing pre-miR-146a/G and transfected with either an empty vector , NUMB wild-type ( WT ) or an miR-146a-resistant NUMB ( MUT ) . ( H ) qRT-PCR analysis of HES1 , HEY2 and CCDN1 mRNA in indicated samples . Actin was used as an internal control . ( I ) Colony formation in soft-agar of MEL-ST/NRASQ61K transfected with either an empty vector , NUMB wild-type ( WT ) or miR-146a-resistant NUMB ( MUT ) . ( J and K ) Number of colonies formed in liquid ( J ) or soft-agar ( K ) by SKMEL-28 expressing miRZip-146a ( + ) or the control miRZip ( − ) that either express a non-silencing shRNA or an shRNA against NUMB . Colonies were counted after 2 weeks ( J ) or 4 weeks ( K ) of growth . ( L and M ) Number of colonies formed in liquid ( L ) or soft-agar ( M ) by M14 expressing miRZip-146a ( + ) or the control miRZip ( − ) that either express a non-silencing shRNA or an shRNA against NUMB . Colonies were counted after 2 weeks ( L ) or 4 weeks ( M ) of growth . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 02710 . 7554/eLife . 01460 . 028Figure 5—figure supplement 1 . Inhibition of NUMB expression promotes growth of SKMEL-28 cells . Colony formation assay of SKMEL-28 cells expressing two different NUMB shRNAs relative to a non-silencing ( NS ) shRNA . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 02810 . 7554/eLife . 01460 . 029Figure 5—figure supplement 2 . Inhibition of NUMB promotes proliferation of melanoma cells . Proliferation assay of SKMEL-28 cells expressing two different NUMB shRNAs relative to a non-silencing ( NS ) shRNA . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 02910 . 7554/eLife . 01460 . 030Figure 5—figure supplement 3 . Ectopic expression of activated Notch promotes the growth of melanoma cells . Colony formation assay of SKMEL-28 cells stably transduced with the activated intracellular NOTCH domain ( ICN ) or empty vector . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 03010 . 7554/eLife . 01460 . 031Figure 5—figure supplement 4 . Ectopic expression of intracellular Notch promotes proliferation of melanoma cells . Proliferation rate of SKMEL-28 cells stably expressing Intracellular NOTCH ( ICN ) relative to cells with empty vector . Proliferation rate was measured after 72 hr of growth using a colorimetric MTT assay . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 031 The results described above suggested that melanomas with elevated miR-146a expression might be sensitive to combined inhibition of BRAF-MEK-ERK and Notch signaling . To test this idea we treated BRAF mutant SKMEL-28 and SKMEL-19 cells , with the MEK inhibitor U0216 , the gamma secretase inhibitor DAPT , or both drugs . Notably , combined U0216 and DAPT treatment inhibited proliferation of SKMEL-28 and SKMEL-19 cells much strongly than either of the drugs alone ( Figure 6A ) . As expected , U0216 efficiently blocked MAPK signaling ( Figure 6B ) and DAPT inhibited the transcriptional targets of Notch signaling ( Figure 6C ) . These results further establish the role of miR-146a in activating Notch signaling . Collectively , these results indicate that the ability of miR-146a to block NUMB expression and activate Notch signaling is necessary for miR-146a to promote melanoma initiation and progression . 10 . 7554/eLife . 01460 . 032Figure 6 . Synergistic melanoma growth inhibition by simultaneous blockage of Notch and BRAF signaling . ( A ) MTT proliferation assay . ( B ) Immunoblot analysis of phosphorylated ( p- ) ERK and total ( t- ) ERK and ( C ) qRT-PCR analysis of NOTCH target HES1 in SKMEL-28 and SKMEL-19 cells left untreated or treated with U0216 , DAPT or combination of both drugs . ( D ) Model . DOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 032 To investigate the clinical relevance of our findings , we cloned and sequenced pre-miR-146a from the genomic DNA of three independent melanocyte cultures and 18 established melanoma cell lines or short-term melanoma cultures . We found that two out of three melanocytes were homozygous for pre-miR-146a/C and one was a pre-miR-146a/C:pre-miR-146a/G heterozygote . By contrast , all melanoma cell lines were either pre-miR-146a/C:pre-miR-146a/G heterozygotes or pre-miR-146a/G homozygotes ( Table 1 ) . 10 . 7554/eLife . 01460 . 033Table 1 . Sequence at position 40 of pre-miR-146a in primary human melanocytes , human melanoma cell lines and clinical samplesDOI: http://dx . doi . org/10 . 7554/eLife . 01460 . 033Human melanoma cell linesCell Line Melanocytes-1CC Melanocytes-2CC Yale SPORE melanocytesCG WM3918CG YUHEFCG YUVONCG YUSIVGG YUROBGG YUROLCG MeWoGG UCC257CG SKMEL28CG A375GG SKMEL5GG M14CG SKMEL19GG YULACGG YUGEN8GG YUSAC2GG YURIFCG SKMEL103CGMatched clinical melanoma samples ( Nevus/Primary ) SampleNevus ( Type ) *Primary Melanoma1CG ( IM ) CG2CC ( C ) CG3CG ( C ) GG4CG ( C ) GG5CG ( LJ ) CG6CG ( IM ) CG7CG ( IM ) CG8CG ( LJ ) CG9CG ( IM ) CG10CC ( C ) CGMatched Clinical Melanoma Samples ( Primary/Metastatic ) SamplePrimary MelanomaMetastatic Melanoma1CGGG2CGCG3CGGG4CGCG5CCGG6CGGG7CGGG8CCCC9CGCG10CGCG11CGCG12CGGG13CGGG14CGGG*IM , Intradermal melanocytic nevus . C , Compound nevus . LJ , Lentigenous juctional nevus . Samples highlighted in gray indicate C-to-G mutation during melanoma progression . Next , we sequenced 48 melanoma samples consisting of 10 matched pairs of melanocytic nevi and primary melanomas , as well as 15 matched pairs of primary and metastatic melanomas ( Table 1; Supplementary file 1B , C ) . As shown in Table 1 , in 4 of 10 cases the nevus contains a pre-miR-146a/C allele that is a pre-miR-146a/G allele in the matched primary melanoma . Remarkably , in 8 of 15 cases analyzed , a pre-miR-146a/C allele in the primary melanoma is a pre-miR-146a/G allele in the matched metastases . In one case , both pre-miR-146a/C alleles in the primary melanoma are pre-miR-146a/G alleles in the matched metastasis . Finally , we also analyzed a publically available melanoma SNP datasets ( Gast et al . , 2010 ) and also observed an enrichment for pre-miR-146a/G during melanoma progression ( Supplementary file 1D ) . Collectively , these results reveal a selection for pre-miR-146a/G during melanoma progression . Moreover , the results with matched patient samples indicate that enrichment of pre-miR-146a/G results from a C-to-G mutation in pre-miR-146a .
In this report , we demonstrate a critical role for miR-146a in the initiation and progression of BRAF/NRAS-positive melanomas , which is summarized in Figure 6D and discussed below . In addition , our results reveal a pharmacologically tractable pathway for the treatment of melanoma . We identified miR-146a as the microRNA whose expression was most upregulated by activated BRAF . Upregulation of miR-146a by activated BRAF , as well as activated NRAS , occurs through the MAPK signaling pathway . Accordingly , we find that BRAF and NRAS mutant melanoma cell lines and short-term melanoma cultures show higher levels of miR-146a compared to those that are wild type for these genes . A major function of the MAPK pathway is to activate transcription by regulating the stability and expression of multiple transcription factors primarily through direct phosphorylation ( Qi and Elion , 2005 ) . We show that the MAPK pathway regulates the phosphorylation of the transcription factor MYC , which in turn binds to the promoter of miR-146a and stimulates its transcription . Notably , MYC has been found to stimulate transcription of several other miRNAs ( Chang et al . , 2008 ) . For example , MYC has been shown to directly activate transcription of the oncogenic miR-17-92 cluster and thereby promote cell proliferation , survival , angiogenesis , and metabolic reprogramming in a number of tumor cell lines ( Mu et al . , 2009; Dews et al . , 2010 ) . miRNAs and components of miRNA biogenesis pathways such as Dicer have been implicated in several aspects of melanocyte biology as well as in melanoma initiation and progression ( Levy et al . , 2010; Bonazzi et al . , 2012 ) . Previous studies have shown that miR-146a can function either as an oncogene or as a tumor suppressor depending upon the cell type ( Boldin et al . , 2011; Chen et al . , 2013 ) . For example , miR-146a has been shown to function as an oncogene in a variety of human cancers including papillary thyroid carcinoma ( PTC ) , triple negative sporadic breast cancers and anaplastic thyroid carcinoma ( Jazdzewski et al . , 2008; Pacifico et al . , 2010; Garcia et al . , 2011 ) . miRNAs function primarily by targeting mRNAs and either promoting their degradation or blocking their translation ( He and Hannon , 2004 ) . Our analysis identified 20 potential targets of miR-146a , including NUMB , which is a well-characterized Notch signaling inhibitor . It is thought that NUMB negatively regulates NOTCH , potentially through a direct protein–protein interaction that requires the phosphotyrosine-binding ( PTB ) domain of NUMB and either the RAM23 region or the very C-terminal end of NOTCH ( Guo et al . , 1996 ) . Consistent , with our results , miR-146a was previously reported to target NUMB mRNA in mouse myocytes C2C12 cells ( Kuang et al . , 2009 ) . We confirmed NUMB as a direct target of miR-146a and showed that NUMB targeting and activation of Notch signaling was necessary for the ability of miR-146a to promote the initiation and progression of melanoma . Many cancers , including melanoma , have been found to have increased Notch signaling , which results from two distinct mechanisms ( Ranganathan et al . , 2011 ) . First , gain-of-function mutations in NOTCH1 can increase Notch signaling ( Ranganathan et al . , 2011 ) . Second , decreased expression of NUMB can result in increased Notch signaling ( Guo et al . , 1996 ) . For example , NUMB expression is decreased in breast and lung cancer and NUMB is considered to be a tumor suppressor in these malignancies ( Lindsay et al . , 2008 ) . In this study , we have shown that Notch signaling can also be increased by upregulation of a miRNAs that targets NUMB . Several previous studies have shown that pre-miR-146a contains a single nucleotide polymorphism ( SNP ) ( C>G rs2910164 ) , which has been associated with various aspects of tumor biology ( Jazdzewski et al . , 2008; Hezova et al . , 2012; Hung et al . , 2012; Lung et al . , 2012; Wang et al . , 2012; Yamashita et al . , 2013 ) . We found that the oncogenic activity of pre-miR-146a/G is substantially greater than that of pre-miR-146a/C , which results , at least in part , from more efficient processing of pre-miR-146a/G compared to pre-miR-146a/C . Our results are consistent with a previous report showing that pre-miR-146a/G undergoes a more efficient nuclear processing compared to pre-miR-146a/C and thus give rise to a higher amount of mature miR-146a ( Jazdzewski et al . , 2008 ) . Most importantly , our analysis of melanoma cell lines and matched patient samples of nevi , primary , and metastatic melanoma reveal enrichment of the more oncogenic variant , pre-miR-146a/G , during melanoma progression . Our results are consistent with previous studies that observed enrichment for pre-miR-146a/G ( Gast et al . , 2010; Supplementary file 1D ) and increased miR-146a expression ( Philippidou et al . , 2010 ) during melanoma progression . Finally , our findings have important therapeutic implications . Both the Notch and BRAF-MEK-ERK signaling pathways are amenable to pharmacological inhibition ( Rizzo et al . , 2008; Bollag et al . , 2010; Flaherty et al . , 2010 ) . Our results suggest that combining clinically approved γ–secretase inhibitors , which block Notch activation , with BRAF inhibitors , which block BRAF-MEK-ERK signaling , might be a more effective treatment of BRAF/NRAS mutant melanomas .
Small RNAs isolated from Vector-infected , BRAFV600E-infected and quiescent cells were cloned essentially as described ( Ambros et al . , 2003 ) . The Firecrest , Bustard and Gerald analysis module ( Illumina ) were used for image analysis , base calling and filtering the raw data ( 36 bp reads ) from each run to generate the sequence reads . Sequences were further filtered to those containing 18–30 bp followed by the first 6 bp of the 3′ linker ( CTGTAG ) . Sequences that passed filters were clustered and mapped to known human microRNAs using BLAST . The Fisher exact test was used to identify differentially-expressed miRNAs . The odds ratio and 95% confidence interval were computed for each treatment using the ‘fisher test’ function in R 2 . 7 . 0 based on conditional maximum likelihood estimation . Adjusted p values were obtained using the Benjamini–Hochberg method to correct for multiple comparisons , and miRNAs with adjusted p value<0 . 05 were considered significant . Deep sequencing results were submitted to Gene Expression Omnibus ( Accession No . GSE39983 ) . WI-38 and IMR-90 cells were purchased from ATCC and grown as recommended . Different primary human melanocyte cultures were purchased from Lonza , Invitrogen , and Yale SPORE in Skin Cancer . Melanoma cell lines were purchased from ATCC and grown as recommended . All short-term melanoma cultures were purchased from Yale Skin SPORE ( Yale University ) and grown as recommended . The plasmids pBabe puro-BRAFV600E ( 15269; Addgene , Cambridge , MA ) , pBabe puro-HRAS V12 ( 15269; Addgene ) , pBabe puro-HRAS V12 S35 ( 12274; Addgene ) , pBabe puro-MEK-DD ( 15268; Addgene ) , and EF . hICN1 . Ubc . GFP ( 17626; Addgene ) were purchased from Addgene . FG12-NRAS 61K was a kind gift of Maria Soengas and Mikhail A Nikiforov . pre-miR146a/C , pre-miR146a/G were cloned into the lentiviral expression vector FG12-CMV . A 1081 bp fragment of the NUMB 3′UTR , including miR146a seed sequence , was PCR amplified by using XhoI-NotI primers ( Supplementary file 1E ) and cloned into psiCHECK-2 luciferase vector . NUMB coding sequence ( CDS ) was PCR amplified by using HindIII-NotI primers ( Supplementary file 1E ) and cloned into pcDNA3 . 1-Hygro plasmid . The miR146a seed sequence was mutated by using specific primers listed in Supplementary file 1E 5 and the QuikChange XL Site-Directed Mutagenesis Kit ( Agilent Technologies , Santa Clara , CA ) by following manufacturer’s instruction . MYC , NUMB , ETS1 , ELK1 , c/EBPβ and RelA ( p65 subunit of NF-κB ) shRNAs in the pLKO . 1 lentiviral expression vector were obtained from Open Biosystems . The anti-miR-146a vector miRZip-146a was obtained from System Biosciences . For viral particles production viral expression vectors and viral packaging plasmids were co-transfected into 293T cells using Effectene ( Qiagen , Valencia , CA ) as per manufacturer’s recommendations . Purified virus particles were infected into primary or melanoma cell lines , and cell lines stably transduced with viral DNA were selected by growth on puromycin or by sorting GFP-positive cells using a flow cytometer . To monitor NOTCH activity SKMEL28 cells stably expressing pre-miR-146a/C , pre-miR-146a/G or empty vector were transfected with a NOTCH responsive reporter plasmid CSL-pGL3-Luciferase ( CSL-pGL3-Luc ) . After 48 hr cells were lysed and luciferase activity was measured by using Dual-Luciferase reporter assay system ( Promega , Madison , WI ) . For mRNA expression analyses , total RNA was extracted with TRIzol ( Invitrogen ) and purified using RNAeasy mini columns ( Qiagen ) , and cDNA was generated using M-MuLV first-strand cDNA synthesis kit ( New England Biolabs ) as per manufacturer’s instructions . For miR-146a expression analyses , total RNA was prepared using TRIzol ( Life Technologies , Grand Island , NY ) , small RNAs were enriched using the miRVana kit ( Ambion ) , and cDNA was prepared using the miScript Reverse Transcription Kit ( Qiagen ) as per manufacturer’s instructions . Quantitative RT-PCR was performed using Power SYBR-green kit ( Applied Biosystems , Foster City , CA ) for mRNA expression analysis or the miScript SYBR-green PCR assay kit ( Qiagen ) , as per manufacturer’s instructions . GAPDH was used as an internal control . ChIP experiments were performed as described previously ( Raha et al . , 2005 ) . MYC binding to the miR-146a promoter and a negative control region was determined using the primers listed in Supplementary file 1E . Whole cell protein extracts were prepared using IP lysis buffer ( Pierce ) containing Protease Inhibitor Cocktail ( Roche , Madison , WI ) and Phosphatase Inhibitor Cocktail ( Sigma–Aldrich , St . Louis , MO ) . Protein concentration was estimated using a Bradford Assay kit ( Bio-Rad ) . Proteins separated on 10% or 12% polyacrylamide gels were transferred to PVDF membranes using a wet transfer apparatus from Biorad . Membranes were blocked with 5% skim milk and probed with primary antibodies followed by the appropriate secondary HRP-conjugated antibody ( GE healthcare , UK ) . Blots were developed using the Supersignal Pico Reagent ( Pierce , Rockland , IL ) . Antibody information is provided in Supplementary file 1E . This study was approved by Boston University School of Medicine , Institutional Review Board ( IRB docket #H-29789 and H-29979 ) and University of Massachusetts Medical School , Instiutional Review Board ( IRB docket# H00001007 ) . Archived tissue with a diagnosis of primary cutaneous malignant melanoma and nevus from the same patient ( n = 10 ) and primary cutaneous malignant melanoma and metastases from the same patient ( n = 15 ) were retrieved from the pathology files of the Skin Pathology Laboratory , Boston University School of Medicine , Boston , MA , USA and University of Massachusetts Medical School , Worcester Histopathologic sections of all cases were reviewed by two board-certified dermatopathologists ( initial sign-out on all by a dermatopathologist; cases were then re-reviewed , and the diagnoses were confirmed by MM ) . All patient data were de-identified . The informed consent was not required because all the samples used in study were archival tissues . For soft-agar and colony formation assays , individual cell lines were seeded in triplicates at three different dilutions , ranging from 1 × 103 to 1 × 104 cells . For soft-agar assays , cells were seeded into a 0 . 35% soft-agar layer . Each experiment was repeated at least twice . Colonies were stained with 0 . 005% crystal-violet solution and counted after 4 weeks . Athymic nude ( NCr nu/nu ) mice ( 8 week old ) were injected subcutaneously with SKMEL-28 or M14 cell lines expressing pre-miR146a/C or /G ( 2 . 5 × 106 cells ) , miRZip-146a or transduced with the control construct . For melanomagenesis experiments , athymic nude ( NCr nu/nu ) mice ( 8 week old ) were injected with 1 . 0 × 106 NRASQ61K transformed MEL-ST cells expressing a control construct or miRZip-146a . Tumor volume was calculated using the formula: length × width2 × 0 . 5 . Genomic DNA was isolated either from the indicated cell lines or from formalin-fixed paraffin-embedded nevi , primary or metastatic melanoma patient samples . For melanoma cell lines , cell pellets were incubated at 50°C overnight in 100 mM NaCl , 10 mM Tris–HCl , pH8 , 25 mM EDTA , 0 . 5 SDS and 0 . 1 mg/ml Proteinase K and the genomic DNA was extracted with phenol-chloroform . For paraffin-fixed tissues , tissue sections scraped from five 10-μm samples were incubated at 60°C overnight ( in SSC buffer , 180 mM NaCl , 0 . 45% SDS , 2 mg/ml Proteinase K and 1 mM DTT ) and the DNA was extracted with phenol–chloroform . A 227 bp fragment including pre-miR146a was amplified by PCR using the primers listed in Supplementary file 1E 5 , cloned using pGEM-T kit ( Promega ) , and plasmid DNA isolated from 24 bacterial colonies were sequenced . Similarly , fragments of BRAF and NRAS were amplified for genotyping using primers listed in Supplementary file 1E from genomic DNA isolated from patient-derived samples , cloned using pGEM-T kit ( Promega ) , and plasmid DNA isolated from 24 bacterial colonies were sequenced using S6 primers . All the sequencing was performed using Sanger sequencing method that typically have the error rate that range from 0 . 001 to 1% ( Hoff , 2009 ) . For microarray experiments , total RNA isolated from SKMEL-28 cells transduced by pre-miR-146a/C or /G or the Vector control was used to generate labeled antisense RNA using the Ambion MessageAmp Kit and hybridized to Illumina HumanHT-12 V4 . 0 expression beadchip using Illumina’s protocol . The microarray data were processed using GenomeStudio ( Illumina ) , log2-transformed , and quantile-normalized using the ‘lumi’ package of Bioconductor . All samples passed quality-control ( QC ) assessment , which included checking various control plots as suggested by Illumina as well as other standard microarray-related analyses . Differential expression analyses were performed using the ‘limma’ package , and a moderated t-test with a Benjamini–Hochberg multiple testing correction procedure was used to determine statistical significance ( adjusted p<0 . 05 ) . Pathway analysis of differentially expressed genes for each comparison was performed using MetaCore ( version 6 . 8 build 29806; GeneGo ) . Microarray data were submitted to Gene Expression Omnibus ( Accession No . GSE39294 ) . For analyzing the SNP dataset the CEL files were downloaded from gene expression omnibus and genotype callings by the CRLMM method on the data of GSE17534 and GSE7822 were done by the R program ‘crlmm’ in the BioConductor ‘oligo’ package ( Scharpf et al . , 2011 ) . The cell line stages of GSE17534 dataset was extracted from Table 1 of the previously published study ( Gast et al . , 2010 ) . All the experiments were done at three different times using independent sample preparation in triplicate . Mean values for individual experiments are expressed as mean ± SEM . | Cancer is a leading cause of death worldwide , and although cancer can have many causes , mutations to a small number of genes are often responsible for a large number of cancer cases . For example , about 70% of cases of the deadliest form of skin cancer have mutations in two specific genes . Many cancer-causing genes are regulated by small RNA molecules that impede the function of other genes by blocking the machinery that turns a gene into a functional protein . However , until recently there was a limited understanding of the role of these microRNAs in the skin cancers caused by the most common mutations . Now , Forloni et al . have looked at all the microRNAs present in cells carrying a mutated form of the one of these genes , and compared these with the microRNAs found in healthy cells . A microRNA , called miR-146a , was discovered to be much more common in the cancer cells . Moreover , production of this microRNA was increased by the two cancer-causing mutations . Forloni et al . found that increasing production of this microRNA caused human skin cancer cells to grow faster , and also caused tumors to develop in mice . Reducing the level of miR-146a had the opposite effect . Forloni et al . also looked at cancer cells taken from individuals at different stages of skin cancer and found that , as the disease progresses , an unprocessed form of miR-146a tends to acquire a mutation , which leads to much higher levels of the active , processed form of miR-146a in the cancer cells . High levels of miR-146a also switched off a gene that encodes for a protein that , in turn , switches off a protein called NOTCH that is linked to skin cancer . As such , excess miR-146a actually increases the activation of the NOTCH protein . These results led Forloni et al . to test if delivering drugs that block the production of miR-146a by inhibiting skin cancer-causing genes , along with other drugs that inhibit the Notch signaling pathway , could be more effective than treatment with either drug on its own . The combined treatment was very effective against human skin cancer cells and could represent a promising development in the treatment of this disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2014 | miR-146a promotes the initiation and progression of melanoma by activating Notch signaling |
Bacterial factors favoring the unprecedented multidrug-resistant tuberculosis ( MDR-TB ) epidemic in the former Soviet Union remain unclear . We utilized whole genome sequencing and Bayesian statistics to analyze the evolutionary history , temporal emergence of resistance and transmission networks of MDR Mycobacterium tuberculosis complex isolates from Karakalpakstan , Uzbekistan ( 2001–2006 ) . One clade ( termed Central Asian outbreak , CAO ) dating back to 1974 ( 95% HPD 1969–1982 ) subsequently acquired resistance mediating mutations to eight anti-TB drugs . Introduction of standardized WHO-endorsed directly observed treatment , short-course in Karakalpakstan in 1998 likely selected for CAO-strains , comprising 75% of sampled MDR-TB isolates in 2005/2006 . CAO-isolates were also identified in a published cohort from Russia ( 2008–2010 ) . Similarly , the presence of mutations supposed to compensate bacterial fitness deficits was associated with transmission success and higher drug resistance rates . The genetic make-up of these MDR-strains threatens the success of both empirical and standardized MDR-TB therapies , including the newly WHO-endorsed short MDR-TB regimen in Uzbekistan .
Multidrug-resistant tuberculosis ( MDR-TB ) , caused by Mycobacterium tuberculosis complex ( MTBC ) strains that are resistant to the first-line drugs isoniazid and rifampicin , represent a threat to global TB control . Barely 20% of the estimated annual 480 , 000 new MDR-TB patients have access to adequate second-line treatment regimens . The majority of undiagnosed or ineffectively treated MDR-TB patients continue to transmit their infection and suffer high mortality ( WHO , 2016 ) . Based on early observations that the acquisition of drug resistance could lead to reduced bacterial fitness ( Middlebrook and Cohn , 1953 ) , it was hypothesized that drug-resistant MTBC-strains had a reduced capacity to transmit , and would not widely disseminate in the general population ( Borrell and Gagneux , 2009; Billington et al . , 1999; Burgos et al . , 2003; Dye and Espinal , 2001; Andersson and Levin , 1999 ) . This optimistic scenario has been invalidated by the now abundant evidence for transmission of MDR and extensively drug-resistant MTBC-strains ( XDR-TB; MDR-TB additionally resistant to at least one fluoroquinolone and one injectable aminoglycoside ) in healthcare and community settings ( Borrell and Gagneux , 2009; Gagneux et al . , 2006; Müller et al . , 2013; Pym et al . , 2002; Comas et al . , 2012 ) . In former Soviet Union countries , which experience the highest MDR-TB rates worldwide , the expansion of drug-resistant MTBC-clones is thought to be promoted by interrupted drug supplies , inadequate implementation of regimens , lack of infection control and erratic treatment in prison settings ( Balabanova et al . , 2004; Casali et al . , 2014a ) . Continued transmission is thought to be aided by the co-selection of mutations in the bacterial population that compensate for a fitness cost ( e . g . growth deficit ) associated particularly with the acquisition of rifampicin resistance mediating mutations ( Borrell and Gagneux , 2009; Andersson and Levin , 1999; Gagneux et al . , 2006; Müller et al . , 2013; Pym et al . , 2002; Comas et al . , 2012 ) . The compensatory mechanism for rifampicin-resistant MTBC-strains is proposed to be associated with structural changes in the RNA-polymerase subunits RpoA , RpoB , and RpoC that increase transcriptional activity and as a consequence enhance the growth rate ( Comas et al . , 2012 ) . However , the impact of these bacterial genetic factors on the epidemiological success of MDR-MTBC strains and implications for current and upcoming MDR-TB treatment strategies remain unexplored . We utilized whole-genome sequencing ( WGS ) to retrace the longitudinal transmission and evolution of MTBC-strains toward MDR/pre-XDR/XDR geno- and phenotypes in Karakalpakstan , Uzbekistan . In this high MDR-TB incidence setting , the proportion of MDR-TB among new TB-patients increased from 13% in 2001 to 23% in 2014 despite the local introduction of the World Health Organization ( WHO ) recommended DOTS strategy in 1998 and an initially limited MDR-TB treatment program in 2003 ( Cox et al . , 2007; Ulmasova et al . , 2013 ) . We expanded our analyses by including a WGS dataset of MDR-MTBC isolates from Samara , Russia ( 2008–2010 ) ( Casali et al . , 2014a ) to investigate clonal relatedness , resistance and compensatory evolution in both settings .
Despite differences in sampling for cohort 1 ( cross-sectional , 2001–2002 ) and cohort 2 ( consecutive enrollment of MDR-TB patients , 2003–2006 ) ( see Materials and methods ) , patients showed similar age , sex distributions , and proportion of residence in Nukus , the main city in Karakalpakstan ( Uzbekistan ) ( Appendix—table 1 ) . While the majority of strains from both cohorts were phenotypically resistant to additional first-line TB drugs ( i . e . beyond rifampicin and isoniazid ) , combined resistance to all five first-line drugs was significantly greater in cohort 2 ( 47% in cohort 2 compared to 14% in cohort 1 , p<0 . 0001 ) . The same was true for resistance to the second-line injectable drug capreomycin ( 23% in cohort 2 compared to 2% in cohort 1 , p=0 . 0001 ) ( Appendix—table 1 ) . This finding was surprising as the isolates from cohort two patients - who were treated with individualized second-line regimens predominately comprising ofloxacin as the fluoroquinolone and capreomycin as the second-line injectable - were all obtained before the initiation of their treatment . In addition , there was no formal MDR-TB treatment program in Karakalpakstan prior to 2003 . These elements imply that the higher rate of resistance to capreomycin was attributable to infection by already resistant strains ( i . e . to primary resistance ) . Utilizing WGS , we determined 6979 single-nucleotide polymorphisms ( SNPs ) plus 537 variants located in 28 genes and upstream regions associated with drug resistance and bacterial fitness ( Supplementary file 1 ) . The corresponding phylogeny revealed a dominant clade comprising 173/277 ( 62 . 5% ) closely related isolates within MTBC lineage 2 ( particularly Beijing-genotype ) ( Figure 1 ) . This group , termed Central Asian Outbreak ( CAO ) , showed a highly restricted genetic diversity ( median pairwise distance of 21 SNPs , IQR 13–25 ) and was differentiated from a set of more diverse isolates by 38 specific SNPs ( Appendix—figure 1 , Supplementary file 1 ) . The proportion of CAO-isolates was similar between 2001–2002 and 2003–2004 ( 49% and 52% , respectively ) , but increased to 76% in 2005–06 ( p<0 . 01 ) . Over the same time periods , the proportions of other groups remained stable or decreased ( Appendix—figure 2 ) . We then sized transmission networks ( measured by transmission indexes , see Materials and methods ) supposed to reflect human-to-human transmission over the last ~10 years based on a maximum of 10 differentiating SNPs between two isolates . Transmission rates varied , even among closely related outbreak isolates ( Figure 1 ) . Beijing-CAO-isolates formed particularly large transmission networks ( >50 patients; Figure 1 ) ; 96 . 0% ( 166/173 ) of all Beijing-CAO isolates were associated with recent transmission ( i . e . transmission index ≥1 ) , versus 48 . 4% ( 31/64 ) of non-CAO Beijing isolates ( p<0 . 0001 ) and 57 . 5% ( 23/40 ) of non-Beijing isolates ( p<0 . 0001 ) ( Supplementary file 1 ) . In addition , the large CAO transmission network exhibited higher levels of drug resistance relative to non-Beijing strains , as reflected by the larger number of drugs for which phenotypic ( p=0 . 0079 ) and genotypic drug resistance ( p=0 . 0048 ) was detected Appendix—figure 3 ) . In order to gain more detailed insights into the emergence of resistance mutations in the evolutionary history of the CAO clade , we sought to employ a Bayesian phylogenetic analysis for a temporal calibration of the CAO phylogeny and an estimation of the mutation rate . Using an extended collection of more diverse CAO isolates ( n = 220 ) from different settings ( see Materials and methods ) , we initially compensated for the restricted sampling time frame of the Karakalpakstan dataset ( 2001–2006 ) . A linear regression analysis showed correlation between sampling year and root-to-tip distance and even a moderate temporal signal ( p=0 . 00039 , R2 = 5 . 2% , Appendix—figure 4 ) , allowed for a further estimation of CAO mutation rates and evaluation of molecular clock models using Bayesian statistics as discussed previously ( Duchêne et al . , 2016 ) . Based on the marginal L estimates collected by path sampling , we found a strict molecular clock with tip dates to be a reasonable model for CAO isolates ( Appendix—table 2 ) . Mutation rate estimates ( under a relaxed clock model ) ranged on average from 0 . 88 to 0 . 96 × 10−7 substitutions per site per year ( s/s/y ) , depending on the demographic model , in favor for the Bayesian skyline model with a mutation rate of 0 . 94 × 10−7 ( s/s/y ) ( 95% HPD 0 . 72–1 . 15 × 10−7 ( s/s/y ) ) ( Appendix—table 2 ) . Comparing different demographic models for the CAO-Karakalpakstan dataset ( n = 173 ) an exponential growth model and a Bayesian skyline model were superior over the constant size demographic prior . Employing the Bayesian skyline model with a strict molecular clock set to 0 . 94 × 10−7 ( s/s/y ) specifically we determined that the most recent common ancestor ( MRCA ) of the CAO-clade emerged around 1974 ( 95% highest posterior density ( HPD ) 1969–1982 ) . The time to the MRCA was confirmed with the exponential growth demographic model ( 1977 , 95% HPD 1977–1982 , Appendix—table 2 ) . The MRCA already exhibited a streptomycin resistance mutation ( rpsL K43R ) ( Figure 2 ) , and acquired isoniazid resistance ( katG S315T ) in 1977 ( 95% HPD 1973 – 1983 ) . The CAO-population size then rose contemporaneously with multiple events of rifampicin , ethambutol , ethionamide , and para-aminosalicylic acid resistance acquisition in different branches ( Figure 2 ) . As an illustration , the most frequent CAO-clone ( upper clade in Figure 2 ) acquired ethambutol and ethionamide resistance mutations ( embB M306V , ethA T314I ) around 1984 ( 95% HPD 1982–1989 ) , and an MDR-genotype ( rpoB S450L ) around 1986 ( 95% HPD 1985–1992 ) . The effective population size reached a plateau before fixation of mutations in the ribD promoter region ( leading to para-aminosalicylic acid resistance ) and rpoC N698S , putatively enhancing its fitness around 1990 ( 95% HPD 1989–1994 ) ( Figure 2 ) . Independent fixation of pyrazinamide ( pncA Q10P and I133T ) and kanamycin ( eis −12 g/a ) resistance-associated mutations was detected in 1992 and 1991 ( both with 95% HPD rounded to 1991–1996 ) ( Figure 2 ) . To further account for uncertainties of substitution rates and thus fixation of drug resistance within the CAO-clade we ran the best models ( Bayesian skyline and exponential growth ) with the upper and lower HPD interval of the best clock estimate ( see above ) . Similarly , the most recent fixation of the putative compensatory mutation rpoC N698S was 1994 ( 95% HPD 1992–1996 ) , still years before implementation of the systematic DOTS-program in Karakalpakstan in 1998 . Interestingly , the DOTS implementation coincided with a second effective population size increase ( Figure 2 ) . At that time , distinct CAO-clades already exhibited pre-XDR ( in this context MDR plus kanamycin resistance ) resistance profiles , mediating resistance to as many as eight different anti-TB drugs . Of note , only a single isolate was identified as harboring a gyrA mutation ( A90V ) , associated with fluoroquinolone resistance ( Supplementary file 1 ) . At the end of the study period in 2006 , we observed a pre-XDR rate among CAO isolates of 52 . 0% ( 90/173 ) , compared to 35 . 9% ( 23/64 ) among other Beijing isolates ( p=0 . 03 ) and compared to 42 . 5% ( 17/40 ) among non-Beijing isolates ( p=0 . 30 ) ( Supplementary file 1 ) . Overall , 62 . 1% ( 172/277 ) of all MDR-MTBC isolates carried putative compensatory mutations ( Comas et al . , 2012; Casali et al . , 2014a ) in rpoA ( n = 7 ) , rpoC ( n = 126 ) and rpoB ( n = 43 ) ( Supplementary file 1 ) . These mutations were almost completely mutually exclusive , as only 4/172 isolates harbored variants in more than one RNA polymerase-encoding gene . While mutations in rpoA and rpoB were equally distributed between Beijing-CAO isolates and other non-outbreak Beijing isolates , CAO-isolates had more rpoC variants ( 56% vs 28% , p=0 . 003 ) ( Appendix—table 3 ) . The mutation rpoC N698N accounted for 79/124 ( 63 . 7% ) of CAO isolates with putative compensatory effects . The mean number of resistance mutations was higher among isolates carrying compensatory mutations ( Figure 3A ) , 4 . 77 vs 3 . 35 mutations ( two-sample t-test p=1 . 2×10−10 ) . Notably , isolates with compensatory mutations also showed larger transmission indexes than isolates presenting no compensatory mutation , 37 . 16 vs 9 . 22 ( Welch two-sample t-test p<2 . 2×10−16 ) ( Figure 3B ) . CAO-isolates with compensatory mutations also had more resistance-conferring mutations than CAO-isolates lacking such mutation ( ANOVA , Tukey multiple comparisons of means P adj = 0 . 0000012 ) . There was no difference observed for the means of resistance-conferring mutations among non-CAO isolates; compensatory mutation present vs . absent ( P adj = 0 . 1978623 ) ( Figure 3C ) . Regression-based analyses of transmission success scores in the Beijing-CAO clade confirmed that the presence of compensatory mutations was strongly associated with cluster sizes independent of the accumulation of resistance mutations ( Figure 4 ) . This pattern was mostly observed for clusters initiated in the late 1980s and the 1990s . To place our analyses in a broader phylogenetic and geographic context , we combined our Karakalpakstan genome set with previously published genomes of 428 MDR-MTBC isolates from Samara ( Casali et al . , 2014a ) , a Russian region located ~1700 km from Nukus , Karakalpakstan . This analysis showed that Beijing-CAO isolates accounted for the third largest clade in Samara ( Casali et al . , 2014a ) . Conversely , the second largest clade in Samara , termed Beijing clade B according to Casali et al ( Casali et al . , 2014a; Casali et al . , 2012 ) , or European/Russian W148 ( Merker et al . , 2015 ) , was represented in Karakalpakstan by a minor clade ( Figure 5 ) . Considering a third Beijing clade ( termed clade A ) restricted to Samara ( Casali et al . , 2014a ) , three major Beijing outbreak clades accounted for 69 . 6% ( 491/705 ) of the MDR-TB cases in both regions . The three Beijing clades ( A , B , and CAO ) in Samara and Karakalpakstan had more drug resistance conferring mutations ( in addition to isoniazid and rifampicin resistance ) with means of 5 . 0 ( SEM 0 . 07 ) , 4 . 2 ( SEM 0 . 18 ) , and 4 . 7 ( SEM 0 . 11 ) , respectively ( Appendix—figure 5 ) , than compared to only 3 . 6 ( SEM 0 . 20 ) additional genotypic drug resistances ( p<0 . 0001 , p=0 . 0143 , p<0 . 0001 ) for other Beijing isolates in both settings . Isolates belonging to other MTBC genotypes ( mainly lineage four clades ) were found with a mean of 2 . 6 ( SEM 0 . 20 ) additional drug resistance mediating mutations , lower than any Beijing-associated group ( p≤0 . 0009 ) ( Appendix—figure 5 ) . Similar to Karakalpakstan , MDR-MTBC isolates from Samara with compensatory mutations also accumulated more resistance-associated mutations ( 4 . 57 vs 2 . 30 mutations per genome; two-sample t-test p<2 . 2×10−16 ) and had higher transmission indexes ( 50 . 32 vs 0 . 46; Welch two-sample t-test p<2 . 2×10−16 ) compared to isolates lacking compensatory mutations ( Appendix—figure 6 ) . The impact of resistance conferring and compensatory mutations on the transmission success score in Beijing-A clade from Samara ( Appendix—figure 7 ) was strikingly similar to the one observed in CAO isolates from Karakalpakstan . The presence of compensatory mutations , but not the accumulation of resistance mutations , was significantly and independently associated with network size in clusters originating in the 1980s and 1990s , with a maximum influence found in clusters starting in the late 1990s . Critically , the high proportions of isolates detected in both settings with pre-XDR and XDR resistance profiles among the three major Beijing clades ( clade A , 96%; clade B , 62%; clade CAO , 50%; Appendix—table 4 , Figure 6 ) reveal the low proportion of patients that are or would be eligible to receive the newly WHO endorsed short MDR-TB regimen . As per definition of the WHO exclusion criteria , for example any confirmed or suspected resistance to one drug ( except isoniazid ) in the short regimen , only 0 . 5% ( 1/191 in Karakalpakstan ) and 2 . 7% ( 8/300 in Samara ) of the patients infected with either a Beijing clade A , B or CAO strain would benefit from a shortened MDR-TB therapy ( Supplementary file 1 ) .
Using WGS combined with Bayesian and phylogenetic analyses , we reveal the evolutionary history and recent clonal expansion of the dominatant MDR/pre-XDR MTBC-clade in Karakalpakstan , Uzbekistan , termed the Central Asian outbreak ( CAO ) . Strikingly , CAO-isolates were also found also in Samara , Russia , and vice versa isolates belonging to the second largest clade in Samara ( Beijing clade B , i . e . European/Russian W148 ( Casali et al . , 2014a; Merker et al . , 2015 ) were identified in Karakalpakstan , suggesting that the MDR-TB epidemic in this world region is driven by few outbreak clades . During the three last decades , these strains gradually accumulated resistance to multiple anti-TB drugs that largely escaped phenotypic and molecular diagnostics , and reduced treatment options to a restricted set of drugs that often cause severe side effects . In addition , our results suggest that compensatory mutations ( in RNA-polymerase subunit coding genes ) that are proposed to ameliorate growth deficits in rifampicin resistant strains in vitro are also crucial in a global epidemiological context allowing MDR and pre-XDR strains to form and maintain large transmission networks . The predominance of these strain networks , seen in two distant geographic regions of the former Soviet Union clearly limit the use of standardized MDR-TB therapies , for example the newly WHO endorsed short MDR-TB regimen , in these settings . Temporal reconstruction of the resistance mutation acquisition and of changes in bacterial population sizes over three decades demonstrates that MDR outbreak strains already became resistant to both first- and second-line drugs in the 1980s . Fully first-line resistant strains massively expanded in the 1990s , a period that shortly preceded or immediately followed the end of the Soviet Union , years before the implementation of DOTS and programmatic second-line MDR-TB treatment . This is in line with the known rise in TB incidence that accompanied the economic breakdown in Russia during the 1990s ( Institute of Medicine Forum on Drug Discovery , Development , and Translation and Russian Academy of Medical Science , 2011 ) . From a bacterial genetic point of view , our data show that particular MDR and pre-XDR clades are highly transmissible despite accumulation of multiple resistance mutations . The acquisition of compensatory mutations after introduction of low fitness cost resistance mutations ( e . g . katG S315T ( Pym et al . , 2002 ) , rpoB S450L ( Gagneux et al . , 2006 ) , rpsL K43R ( Böttger et al . , 1998 ) seems the critical stage allowing for higher transmission rates . Multiple regression analyses further strengthened this hypothesis by demonstrating that the presence of fitness compensating variants was positively associated with transmission success in different settings and outbreak clades , independently of the accumulation of resistance mutations . Compensatory evolution thus appears to play a central role in driving large MDR-TB epidemics such as that seen with the Beijing CAO-clade . A particular concern is the high prevalence of mutations conferring resistance to second-line drugs currently included in treatment regimens , among the dominant MDR-MTBC strains . Their detected emergence in a period preceding DOTS implementation , for example in Karakalpakstan , can be explained by past , largely empirical treatment decisions or self-medication . For instance , high frequencies of mutations in the ribD promoter region , and folC among Beijing-CAO isolates , associated with para-aminosalicylic acid resistance ( Zheng et al . , 2013; Zhao et al . , 2014 ) , are a likely consequence of the use of para-aminosalicylic acid in failing treatment regimens in the late 1970s to the early 1980s in the Soviet Union ( USSR Ministry of Health , 1976; USSR Ministry of Health , 1983; Mishin , 2008 ) . Likewise , the frequent independent emergence of mutations in the eis promoter and of rare variants in the upstream region of whiB7 , both linked to resistance to aminoglycosides ( mainly streptomycin and kanamycin ) ( Zaunbrecher et al . , 2009; Reeves et al . , 2013 ) , probably reflects self-administration of kanamycin that was available in local pharmacies . Of note , prominent mutations such as katG S315T or rpoB S450L might have occurred multiple times independently in a bacterial population and inferring the common ancestor could lead to an overestimate of the TMRCA . However , this is not the case for rare and more diverse mutations , for example conferring resistance to pyrazinamide , PAS or kanamycin , thus further strengthening the historic fixation mentioned above . The pre-existence of fully first-line resistant strain populations ( e . g . CAO-Beijing in Karakalpakstan ) likely contributed to the poor treatment outcomes observed among MDR-TB patients following the implementation of first-line DOTS treatment in 1998 ( Cox et al . , 2006 ) . This period coincides with a detected CAO population size increase , likely reflecting the absence of drug susceptibility testing and therefore appropriate second-line treatment during extended hospitalization at the time , resulting in prolonged infectiousness of TB-patients and further spread of these strains . The frequencies of fluoroquinolone resistance , mediated by gyrA and gyrB mutations , remained low among the Karakalpakstan MDR-MTBC isolates , which is consistent with the notion that such drugs were rarely used for treating TB in former Soviet Union countries ( see Discussion ( Casali et al . , 2014a; USSR Ministry of Health , 1976; USSR Ministry of Health , 1983; Mishin , 2008 ) . This observation explains the generally favorable MDR-TB treatment outcomes observed with the use of individualized second-line regimens , including a fluoroquinolone , in the latter MDR-TB treatment program in the Karakalpakstan patient population ( Cox et al . , 2007; Lalor et al . , 2011 ) . However , fluoroquinolone resistance , representing the last step towards XDR-TB , is already emerging as reported for strains in Beijing clade A and B ( Casali et al . , 2014a ) . In conclusion , the ( pre- ) existence and wide geographic dissemination of highly resistant and highly transmissible strain populations most likely contributes to increasing M/XDR-TB incidence rates despite scaling up of the MDR-TB programs in some Eastern European and Russian regions ( Ulmasova et al . , 2013; Institute of Medicine Forum on Drug Discovery , Development , and Translation and Russian Academy of Medical Science , 2011; Medecins Sans Frontiere , 2013 ) . Importantly , from the large spectrum of resistance detected among dominating strains in this study , it can be predicted that standardized therapies , including the newly WHO endorsed short MDR-TB regimen in Uzbekistan , are/will be largely ineffective for many patients in Samara and Karakalpakstan , and likely elsewhere in Eurasia . In order to successfully control the worldwide MDR-TB epidemics , universal access to rapid and comprehensive drug susceptibility testing , best supported by more advanced technologies , will be crucial for guiding individualized treatment with existing and new/repurposed TB drugs and to maximize chances of cure and prevention of further resistance acquisition .
A total of 277 MDR-MTBC isolates derived from two separate cohorts were sequenced . The first cohort comprised 86% ( 49/57 ) of MDR-MTBC isolates from a cross-sectional drug resistance survey conducted in four districts in Karakalpakstan , Uzbekistan between 2001–2002 ( Cox et al . , 2006 ) . An additional 228 isolates were obtained from TB-patients enrolled for second-line treatment in the MDR-TB treatment program from 2003 to 2006 . These isolates represented 76% ( 228/300 ) of all MDR-TB cases diagnosed over the period . While the MDR-TB treatment program covered two of the four districts included in the initial drug resistance survey , the majority of isolates from both cohorts , 69% and 64% respectively , were obtained from patients residing in the same main city of Nukus ( Appendix—table 1 ) . To set the MDR-MTBC isolates from Karakalpakstan into a broader geographical perspective , raw WGS data of 428 MDR-MTBC isolates from a published cross-sectional prospective study in Samara , Russia from 2008 to 2010 ( Casali et al . , 2014a ) were processed as described below and included into a composite MDR-MTBC dataset . Drug susceptibility testing ( DST ) was performed for five first-line drugs ( isoniazid , rifampicin , ethambutol , streptomycin , pyrazinamide ) , and three second-line drugs ( ofloxacin , capreomycin and prothionamide ) for cohort 1 , and six second-line drugs for cohort 2 ( capreomycin , amikacin , ofloxacin , ethionamide , para-aminosalicylic acid and cycloserine ) by the reference laboratory in Borstel , Germany as described previously ( Kent and Kubica , 1985 ) . WGS was performed with Illumina Technology ( MiSeq and HiSeq 2500 ) using Nextera XT library preparation kits as instructed by the manufacturer ( Illumina , San Diego , CA ) . Fastq files ( raw sequencing data ) were submitted to the European nucleotide archive ( see Supplementary file 1 for accession numbers ) . Obtained reads were mapped to the M . tuberculosis H37Rv reference genome ( GenBank ID: NC_000962 . 3 ) with BWA ( Li and Durbin , 2009 ) . Alignments were refined with GATK ( McKenna et al . , 2010 ) and Samtools ( Li et al . , 2009 ) toolkits with regard to base quality re-calibration and alignment corrections for possible PCR artefact . We considered variants that were covered by a minimum of four reads in both forward and reverse orientation , four reads calling the allele with at least a phred score of 20 , and 75% allele frequency . In the combined datasets , we allowed a maximum of 5% of all samples to fail the above-mentioned threshold criteria in individual genome positions to compensate for coverage fluctuations in certain genome regions; in these cases , the majority allele was considered . Regions annotated as ‘repetitive’ elements ( e . g . PPE and PE-PGRS gene families ) , insertions and deletions ( InDels ) , and consecutive variants in a 12 bp window ( putative artefacts flanking InDels ) were excluded . Additionally , 28 genes associated with drug resistance and bacterial fitness ( see Supplementary file 1 ) were excluded for a conservative and robust phylogenetic reconstructions . The remaining single-nucleotide polymorphisms ( SNPs ) were considered as valid and used for concatenated SNP alignments . Further detailed methods of the phylogenetic reconstruction , molecular resistance prediction , strain-to-strain genetic distance , and Bayesian models are given as Appendix 1 . Based on the distance matrix ( SNP distances ) , we further determined for every isolate the number of isolates that were in a range of 10 SNPs or less ( in the following referred to as ‘transmission index’ ) . This 10 SNP-threshold was used to infer the number of recently linked cases , as considered within a 10-year time period , based on previous convergent estimates of MTBC genome evolution rate of ≈ 0 . 5 SNPs/genome/year in inter-human transmission chains and in macaque infection models ( Ford et al . , 2011; Walker et al . , 2013; Roetzer et al . , 2013; Walker et al . , 2014 ) . This can include direct transmission events among the study population but also cases which are connected by a more distant contact which was not sampled . In the latter case , we assumed that two isolates with a maximum distance of 10 SNPs share a hypothetical common ancestor that is 5 SNPs apart from the two sampled isolates ( considering a bifurcating phylogeny ) and thus covers a timeframe of 5 SNPs over 0 . 5 SNPs/year equals 10 years between the two actual samples and a shared recent ancestor node/case ( see also Appendix 1 ) . Mutations ( small deletions and SNPs ) in 34 resistance-associated target regions ( comprising 28 genes ) were considered for a molecular resistance prediction to 13 first- and second-line drugs ( Supplementary file 1 ) . Mutations in genes coding for the RNA-Polymerase subunits rpoA , rpoB ( excluding resistance mediating mutations in the rifampicin resistance determining region ( RRDR ) , and in codons 170 , 400 , 491 ) , and rpoC were reported as putative fitness compensating ( e . g . in vitro growth enhancing ) variants for rifampicin-resistant strains as suggested previously ( Comas et al . , 2011; de Vos et al . , 2013; Casali et al . , 2014b; Cohen et al . , 2015 ) . A detailed overview of all mutations considered as genotypic resistance markers is given in Supplementary file 1 . Mutations that were not clearly linked to phenotypic drug resistance were reported as genotypic non wild type and were not considered as genotypic resistance markers . When no mutation ( or synonymous , silent mutations ) was detected in any of the defined drug relevant target regions the isolate was considered to be phenotypically susceptible . We used jModelTest v2 . 1 and Akaike and Bayesian Information Criterion ( AIC and BIC ) to find an appropriate substitution model for phylogenetic reconstructions based on the concatenated sequence alignments ( Appendix—table 5 ) . Maximum likelihood trees were calculated with FastTree 2 . 1 . 9 ( double precision for short branch lengths ) ( Price et al . , 2010 ) using a general time reversible ( GTR ) nucleotide substitution model ( best model according to AIC and second best model according to BIC ) , 1000 resamplings and Gamma20 likelihood optimization to account for evolutionary rate heterogeneity among sites . The consensus tree was rooted with the ‘midpoint root’ option in FigTree ( resulting in the expected tree topology of lineage 2–4 strains ) and nodes were arranged in increasing order . Variants considered as drug resistance markers ( see above ) and putative compensatory variants were analyzed individually and mapped on the phylogenetic tree to define resistance patterns of identified phylogenetic clades . In order to compute a time scaled phylogeny and employ the Bayesian skyline model ( see below ) for the identified Central Asian outbreak ( CAO ) clade , we sought to define an appropriate molecular clock model ( strict versus relaxed clock ) and a mutation rate estimate . Due to the restricted sampling timeframe of the Karakalpakstan dataset ( 2001–2006 ) , we extended the dataset for the model selection process with CAO isolates from Samara ( 2008–2010 ) and ‘historical’ CAO isolates from MDR-TB patients in Germany ( 1995–2000 ) thus allowing for a more confident mutation rate estimate . The strength of the temporal signal in the combined dataset , assessed by the correlation of sampling year and root-to-tip distance , was investigated with TempEst v1 . 5 ( 44 ) . Regression analysis was based on residual mean squares , using a rooted ML tree ( PhyML , GTR substitution model , 100 bootstraps ) , R-square and adjusted p-value are reported . For the comparison of different Bayesian phylogenetic models , we used path sampling with an alpha of 0 . 3 , 50% burn-in and 15 million iterations ( resulting in mean ESS values > 100 ) , marginal likelihood estimates were calculated with BEAST v2 . 4 . 2 ( 45 ) , and Δ marginal L estimates are reported relative to the best model . First , we employed a strict molecular clock fixed to 1 × 10−7 substitutions per site per year as reported previously ( Ford et al . , 2011; Walker et al . , 2013; Roetzer et al . , 2013 ) without tip dating , a strict molecular clock with tip dating and a relaxed molecular clock with tip dating . BEAST templates were created with BEAUti v2 applying a coalescent constant size demographic model , a GTR nucleotide substitution model , a chain length of 300 million ( 10% burn-in ) and sampling of 5000 traces/trees . Second , we ran different demographic models ( i . e . coalescent constant size , exponential , and Bayesian skyline ) under a relaxed molecular clock using tip dates and the same parameters for the site model and Markov-Chain-Monte-Carlo ( MCMC ) as described above . Third , we tested and compared the best models for the Karakalpakstan CAO-clade under a strict molecular clock prior including the upper and lower 95% HPD interval ( Appendix—table 2 ) . Inspection of BEAST log files with Tracer v1 . 6 showed an adequate mixing of the Markov chains and all parameters were observed with an effective sample size ( ESS ) >200 for the combined dataset ( n = 220 ) and in the thousands for the Karakalpakstan CAO clade ( n = 173 ) , suggesting an adequate number of effectively independent draws from the posterior sample and thus sufficient statistical support . Other priors between the model comparisons were not changed . Changes of the effective population size of the CAO clade in Karakalpakstan over the last four decades were calculated with a Bayesian skyline plot using BEAST v2 . 4 . 2 ( 45 ) using a tip date approach with a strict molecular clock model of 0 . 94 × 10−7 substitutions per site per year ( best model according to path sampling results , see above ) , and a GTR nucleotide substitution model . We further used a random starting tree , a chain length of 300 million ( 10% burn-in ) and collected 5000 traces/trees . Again adequate mixing of the Markov chains and ESS values in the hundreds were observed . A maximum clade credibility genealogy was calculated with TreeAnnotator v2 . We used multiple linear regression to examine the respective contributions of antimicrobial resistance and putative fitness cost-compensating mutations to the transmission success of tuberculosis . To take transmission duration into account , we computed , for each isolate and each period length T in years ( from 1 to 40y before sampling ) , a transmission success score defined as the number of isolates distant of less than T SNPs , divided by T . This approach relied on the following rationale: based on MTBC evolution rate of 0 . 5 mutation per genome per year , the relation between evolution time and SNP divergence is such that a cluster with at most N SNPs of difference is expected to have evolved for approximately N years . Thus , transmission success score over T years could be interpreted as the size of the transmission network divided by its evolution time , hence as the average yearly increase of the network size . For each period T , the transmission success score was regressed on the number of resistance mutations and on the presence of putative compensatory mutations . The regression coefficients with 95% confidence intervals were computed and plotted against T to identify maxima , that is , time periods when the transmission success was maximally influenced by either resistance-conferring or –compensating mutations . These analyses were conducted independently on outbreak isolates of the Beijing-CAO clade in the Karakalpakstan cohort and of the Beijing-A clade in the Samara cohort . Differences between cohorts and numbers of sampled isolates per year category were performed using Chi-squared analysis ( mid-P exact ) or Fisher’s exact test , while comparison of median age was performed using the Mann-Whitney test . p-Values for pairwise comparisons of groups regarding pairwise genetic distances , number of resistant DST results and number of resistance related mutations were calculated with an unpaired t-test ( Welch correction ) or a t-test according to the result of the variances comparison using a F-test . Boxplot , bubble plots and density plots have been performed in R . | Multidrug-resistant tuberculosis , often shortened to MDR-TB , is a public health crisis with close to half a million patients falling ill each year globally . Some strains of the bacterium Mycobacterium tuberculosis , which causes tuberculosis disease , are resistant to the two most effective drugs used to treat the infection . As a result , patients with MDR-TB require a longer treatment of up to two years , often with severe side effects and a low chance of cure . Resistant strains of the bacteria are usually weaker than drug-susceptible strains . So , for a long time , large MDR-TB epidemics were considered to be unlikely and outbreaks of MDR-TB were often regarded as locally contained phenomenona . Recent research has shown that MDR-TB strains are often just as likely as drug-susceptible strains to be transmitted and therefore just as likely to cause large country-wide outbreaks . It has also become clear that the resistant bacteria acquire additional mutations over time to compensate for any weakness . However , a lack of detailed history of outbreaks has meant the role of the genetics of MDR-TB bacteria has not been fully understood . Without this knowledge , prevention of future outbreaks and containment of the most successful strains in areas with a high burden of disease is difficult . To address this , Merker , Barbier et al . reconstructed the evolutionary history of MDR-TB strains obtained in 2001–2006 from an outbreak in Uzbekistan . Whole genome sequencing followed by statistical analysis highlighted one predomininant strain that likely emerged in the mid-1970s , when the country was part of the former Soviet Union . This strain has since acquired mutations that make it resistant to eight different drugs . The most successful bacterial strains found also had compensatory mutations that seem to aid their survival . In 1998 , the health authorities implemented a TB treatment program in the region without knowing the true extent of the MDR-TB outbreak at that time . Testing for drug resistance was not routinely available , and Merker , Barbier et al . saw that MDR-TB strains resistant to the drugs used spread in the study region and were later also found independently in Russia . A lack of routine testing for drug resistance in TB remains common in many countries with high burdens of the disease . These findings emphasize the need for universal access to tests for TB drug resistance , therapies tailored for individual patients , and access to new and repurposed drugs to reduce the risk of future outbreaks of drug-resistant TB . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology"
] | 2018 | Compensatory evolution drives multidrug-resistant tuberculosis in Central Asia |
Bacteria often live in biofilms , which are microbial communities surrounded by a secreted extracellular matrix . Here , we demonstrate that hydrodynamic flow and matrix organization interact to shape competitive dynamics in Pseudomonas aeruginosa biofilms . Irrespective of initial frequency , in competition with matrix mutants , wild-type cells always increase in relative abundance in planar microfluidic devices under simple flow regimes . By contrast , in microenvironments with complex , irregular flow profiles – which are common in natural environments – wild-type matrix-producing and isogenic non-producing strains can coexist . This result stems from local obstruction of flow by wild-type matrix producers , which generates regions of near-zero shear that allow matrix mutants to locally accumulate . Our findings connect the evolutionary stability of matrix production with the hydrodynamics and spatial structure of the surrounding environment , providing a potential explanation for the variation in biofilm matrix secretion observed among bacteria in natural environments .
In nature , bacteria predominantly exist in biofilms , which are surface-attached or free-floating communities of cells held together by a secreted matrix ( Hall-Stoodley et al . , 2004; Nadell et al . , 2009; Flemming et al . , 2016 ) . The extracellular matrix defines biofilm structure , promotes cell-cell and cell-surface adhesion , and confers resistance to chemical and physical insults ( Flemming and Wingender , 2010; Hobley et al . , 2015; Teschler et al . , 2015; Tseng et al . , 2013; Doroshenko et al . , 2014; Landry et al . , 2006 ) . The matrix also plays a role in the social evolution and population dynamics of biofilm-dwelling bacteria ( Nadell et al . , 2009; Steenackers et al . , 2016; Nadell et al . , 2016; Ghoul and Mitri , 2016; Mitri et al . , 2016 , 2013 , 2011 ) . In some species , such as the soil bacterium Bacillus subtilis , matrix materials are readily shared among cells , leading to public goods dilemmas in which non-producing strains can outcompete producing strains ( van Gestel et al . , 2015 , 2014; Kovács , 2014 ) . In other species , including the pathogens Vibrio cholerae , Pseudomonas fluorescens , and P . aeruginosa , matrix-secreting cell lineages privatize most matrix components , allowing them to smother or laterally displace other cell lineages and , in so doing , outcompete non-producing cells ( Xavier and Foster , 2007; Nadell and Bassler , 2011; Nadell et al . , 2015; Kim et al . , 2014c; Schluter et al . , 2015; Irie et al . , 2016; Drescher et al . , 2016; Yan et al . , 2016; Madsen et al . , 2015; Oliveira et al . , 2015 ) . Interestingly , not all wild and clinical isolates of these species produce a biofilm matrix , despite the clear ecological and competitive benefits of possessing a matrix ( Mann and Wozniak , 2012; Yawata et al . , 2014; Chowdhury et al . , 2016 ) . Theory and experiments investigating bacterial colonies on agar show that constrained movement can promote coexistence of different strains and species ( Levin , 1974; Levin and Paine , 1974; Durrett and Levin , 1994 , 1998; Kerr et al . , 2002; Kim et al . , 2008; Poltak and Cooper , 2011 ) . Fitness trade-offs between the benefits of being adhered to surfaces and the ability to disperse to new locations can cause variability in matrix production ( Nadell and Bassler , 2011; Yawata et al . , 2014; Levin , 1974; Cohen and Levin , 1991 ) , but it is not well understood how selective forces within the biofilm environment itself might drive the coexistence of strains that make matrix with strains that do not . Here , we explore how selection for matrix production occurs within biofilms on different surface geometries and under different flow regimes , including those that are relevant inside host organisms and in abiotic environments such as soil . The local hydrodynamics associated with natural environments can have dramatic effects on biofilm matrix organization . This phenomenon has been particularly well established for P . aeruginosa , a common soil bacterium ( Green et al . , 1974 ) and opportunistic pathogen that thrives in open wounds ( Fazli et al . , 2009; Burmølle et al . , 2010 ) , on sub-epithelial medical devices ( Guaglianone et al . , 2010 ) , and in the lungs of cystic fibrosis patients ( Harmsen et al . , 2010; Ciofu et al . , 2013; Folkesson et al . , 2012; Stacy et al . , 2015; McNally et al . , 2014 ) . Under steady laminar flow in simple microfluidic channels , P . aeruginosa forms biofilms with intermittent mushroom-shaped tower structures ( Harmsen et al . , 2010; Friedman and Kolter , 2004; Miller et al . , 2012; Parsek and Tolker-Nielsen , 2008 ) . Under irregular flow regimes in more complex environments , however , P . aeruginosa also produces sieve-like biofilm streamers that protrude into the liquid phase above the substratum ( Persat et al . , 2015; Kim et al . , 2014a , 2014b; Rusconi et al . , 2010 ) . These streamers – whose structure depends on the secreted matrix – are proficient at catching cells , nutrients , and debris that pass by , leading to clogging and termination of local flow ( Drescher et al . , 2013 ) . The spatial and temporal characteristics of flow thus combine to alter matrix morphology , which , in turn , feeds back to alter local hydrodynamics and nutrient advection and diffusion ( Nadell et al . , 2016; Hellweger et al . , 2016; Stewart and Franklin , 2008 ) . By modifying community structure and solute transport in and around biofilms ( Stewart , 2012 ) , this feedback could have a significant influence on the evolutionary dynamics of matrix secretion in natural environments ( Coyte et al . , 2016 ) . Here , we study within-biofilm competition as a function of flow regime using strains of P . aeruginosa PA14 that differ only in their production of Pel , a viscoelastic matrix polysaccharide that serves as the primary structural element for biofilm and streamer formation ( Friedman and Kolter , 2004; Drescher et al . , 2013; Chew et al . , 2014; Jennings et al . , 2015 ) . Using a combination of fluid flow visualization and population dynamics analyses , we reveal a novel interaction between hydrodynamic conditions , biofilm architecture , and competition within bacterial communities .
Several approaches are available to study how competitive dynamics differ in particular flow environments . Most commonly , one would monitor biofilm co-cultures of wild-type and ∆pelA PA14 cells over time until their strain compositions reached steady state . Performing such time-series experiments was not possible here due to a combination of low-fluorescence output in the early phases of biofilm growth coupled with phototoxicity incurred by cells during epifluorescence imaging . To circumvent this issue , we measured the change in frequency of wild-type and ∆pelA cells over a fixed 72 hr time period as a function of their initial ratio in both planar chambers and soil-mimicking chambers containing column obstacles . Population composition was quantified in all cases using microscopy , as described in the Materials and methods section . From these measurements , we could infer the final stable states of Pel-producing and non-producing cells as a function of surface topography and flow conditions . This method is commonly used to evaluate the behavior of dynamical systems , and it has been employed in a variety of related experimental applications ( Nadell and Bassler , 2011; Nadell et al . , 2015; Madsen et al . , 2015; Chuang et al . , 2009; Sanchez and Gore , 2013; Drescher et al . , 2014 ) . In planar chambers with simple parabolic flows , wild-type PA14 increased in relative abundance regardless of initial population composition , indicating uniform positive selection for Pel secretion ( Figure 1A , B ) . This result is consistent with recent studies of V . cholerae and Pseudomonas spp . demonstrating that – in these species – core structural polysaccharides of the secreted matrix cannot be readily exploited by non-producing mutants ( Nadell and Bassler , 2011; Nadell et al . , 2015; Kim et al . , 2014c; Schluter et al . , 2015; Irie et al . , 2016; Yan et al . , 2016; Madsen et al . , 2015 ) . Confocal microscopy revealed that Pel-producers mostly excluded non-Pel-producing cells from biofilm clusters in planar chambers ( Figure 2A , B ) , although some ∆pelA mutants resided on the periphery of wild-type biofilms . The liquid effluent from these chambers contained an over-representation of the ∆pelA mutant relative to wild type , consistent with the interpretation that ∆pelA strain was displaced from the substratum over time ( Figure 2C ) . When wild-type and ∆pelA cells competed in microfluidic devices simulating porous microenvironments , by contrast , there was a pronounced shift to negative frequency-dependent selection for Pel production ( Figure 1A , C ) . Wild-type PA14 was selectively favored at initial frequencies below ~0 . 6 . Above this critical frequency , the ∆pelA mutant was favored . From this result , we can infer that in this porous environment , ∆pelA null mutants can grow and stably coexist with wild type Pel-producers . 10 . 7554/eLife . 21855 . 003Figure 1 . Wild-type P . aeruginosa PA14 outcompetes the ΔpelA mutant under simple flow conditions , but the two strains coexist under complex flow conditions . ( A ) Wild-type and ∆pelA strains were co-cultured at a range of initial frequencies in the simple planar ( black data ) or column-containing ( blue data ) microfluidic chambers . In both cases , fresh minimal M9 medium with 0 . 5% glucose was introduced at flow rates adjusted to equalize the volume of medium flowing through each chamber per unit time across all experiments . The diagonal gray line denotes the maximum possible increase in wild-type frequency for a given initial condition . Each data point is an independent biological replicate . ( B ) A maximum intensity projection ( top-down view ) of a confocal z-stack of wild type ( green ) and ∆pelA ( red ) biofilms in simple flow chambers . ( C ) An epifluorescence micrograph ( top-down view ) of wild type ( green ) and ∆pelA ( red ) biofilms after 72 hr growth in a flow chamber containing column obstacles to simulate a porous environment with irregular flows . Images in ( B ) and ( C ) were taken from chambers in which the wild type was inoculated at a frequency of 0 . 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 00310 . 7554/eLife . 21855 . 004Figure 1—source data 1 . Change in WT frequency as a function of initial frequency in two flow conditionsDOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 00410 . 7554/eLife . 21855 . 005Figure 1—figure supplement 1 . The maximum growth rates of P . aeruginosa PA14 wild type and ∆pelA cells in mixed liquid culture . n = 4 biological x 6 technical replicates ( different overnight cultures and microtiter plate wells , respectively ( bars denote means ± S . E . ) . Growth rates were measured as the maximum slope of growth curves of cultures whose optical density at 600 nm was taken every 30 min until stationary phase . Cultures were grown in minimal M9 medium with 0 . 5% glucose at ambient room temperature ( 24°C ) . The maximum growth rates of wild-type cells expressing GFP and wild-type cells expressing mCherry were not different , indicating that the fluorescent protein constructs did not introduce growth rate bias ( two-sample t = 0 . 235 , df = 6 , p=0 . 822 ) . A two-sample t-test comparing the maximum growth rate of the ∆pelA strain against the pooled data for wild type ( t = 2 . 31 , df = 10 , p=0 . 0432 ) suggests a growth rate decrement on the part of wild type relative to ∆pelA , but was not statistically significant at a critical p<0 . 05 threshold with Bonferroni correction for two pairwise comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 00510 . 7554/eLife . 21855 . 006Figure 1—figure supplement 1—source data 1 . Maximum liquid culture growth rates of study strains . DOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 00610 . 7554/eLife . 21855 . 007Figure 2 . Matrix production confers a competitive advantage to wild-type P . aeruginosa PA14 in biofilms under simple flow conditions . ( A ) Absolute abundances of wild-type and ΔpelA strains in monoculture and co-culture in planar microfluidic flow-chambers ( bars denote means ± S . D . for n = 3–6 ) . The two strains were inoculated alone ( left two bars ) or together at a 1:1 ratio ( right two bars ) . ( B ) Single optical plane 3 μm from the surface , and z-projections at right and bottom respectively , of the wild-type ( green ) and ΔpelA ( red ) strains grown in co-culture for 48 hr after inoculation at a 1:1 ratio . ( C ) Relative abundance of wild-type and ∆pelA strains in the liquid effluent of planar microfluidic devices ( points denote means ± S . D . for n = 3 ) . Wild-type and ∆pelA cells were combined at a 1:1 initial ratio and co-inoculated on the glass substratum of simple flow chambers . At 0 , 24 , and 48 hr , 5 µL of effluent was collected from chamber outlets . Wild-type frequency was calculated within biofilms and within the liquid effluent for each time point . DOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 00710 . 7554/eLife . 21855 . 008Figure 2—source data 1 . Biofilm production of WT and Pel-deficient P . aeruginosa in mono-culture and co-culture; cell counts in chamber effluents . ( A ) WT and ΔpelA biomass accumulation . ( C ) WT and ΔpelA cell counts in effluent . DOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 008 Previous work has shown that in environments with flow and with corners , wild-type P . aeruginosa produces Pel-dependent biofilm streamers that extrude from the surface into the passing liquid ( Kim et al . , 2014a; Rusconi et al . , 2010; Drescher et al . , 2013 ) . In our experiments , streamers were produced by wild-type cells and could be readily detected via microscopy throughout column-containing chambers , but not planar chambers . Streamers are known to catch cells and debris that pass by ( Drescher et al . , 2013 ) , and although ∆pelA cells could be found in the streamers in our experiments , they were not abundant . Cell capture by streamers therefore cannot account for the observed coexistence of the two strains ( Figure 3—figure supplement 1 ) . This result suggests that streamers do indeed catch debris , consistent with prior studies ( Kim et al . , 2014b; Drescher et al . , 2013 ) , but that in our system , ∆pelA cells are not present at high enough density in the passing liquid phase to accumulate substantial population sizes by this mechanism . Our microscopy-based observation of chambers containing column obstacles suggested that wild-type biofilms gradually obstructed some of the regions located between columns over time . We hypothesized that partial clogging could render those portions of the chambers more suitable for growth of the ∆pelA strain , which was previously shown to be sensitive to removal by shear ( Colvin et al . , 2012 , 2011 ) . This hypothesis predicts that ∆pelA cells should be found predominantly in regions of the chamber that have been clogged by wild type biofilms . To test this prediction , we repeated our co-culture competition experiment with wild-type and ∆pelA cells in chambers containing columns , and we measured the distribution of each strain as above . We next introduced fluorescent beads into the chambers by connecting new influent syringes to the inflow tubing . By tracking the beads with high frame-rate microscopy , we could distinguish areas in which flow was present from areas in which flow was absent or very low , and then we could superimpose this information onto the spatial distributions of wild-type and ∆pelA mutant cells ( Figure 3—figure supplement 2 ) . Wild-type biofilms accumulated intermittently , often with clusters of ∆pelA cells in close proximity . Importantly , and in support of our prediction , ∆pelA biofilm clusters occurred significantly more often in regions in which flow was blocked by wild-type biofilms than in regions in which flow was not interrupted ( Figure 3A , B ) . As shown previously ( Drescher et al . , 2013 ) , ∆pelA cells did not clog chambers when grown in isolation , supporting the interpretation that ∆pelA accumulation relies on and only occurs after clogging by the wild type . This result is consistent with prior indications that Pel-producers have higher shear tolerance than Pel-deficient cells ( Colvin et al . , 2012 , 2011 ) , which we confirmed in our system by growing each strain in isolation under varying shear stress in planar chambers ( Figure 3C ) . Further supporting our interpretation , the ∆pelA strain can outcompete the wild type when the two are grown together in planar microfluidic chambers in the absence of flow ( Figure 3—figure supplement 3 ) . This experiment approximates the clogged areas of complex flow environments , and the results explain how – on spatial scales encompassing areas of high flow and low or no flow – the wild-type and the ∆pelA strains can coexist . 10 . 7554/eLife . 21855 . 009Figure 3 . Pel-deficient mutants occupy locations protected from flow due to local clogging by wild-type P . aeruginosa biofilms . ( A ) Wild-type ( green ) and ∆pelA ( red ) P . aeruginosa strain mixtures were inoculated into complex flow chambers with irregularly-spaced column obstacles . Biofilms were imaged using confocal microscopy , after which fluorescent beads were flowed through the chamber . The presence or absence of flow was monitored through averaging successive exposures of bead tracks ( white lines are bead tracks; blue arrows highlight flow trajectories ) . ( B ) Analysis of co-occurrence of flow and wild-type or ∆pelA cell growth at the end of 1:1 competition experiments in complex flow chambers with column obstacles , as illustrated by the micrograph in ( A ) . The occurrence of wild-type ( gray ) and ∆pelA ( white ) cell clusters are shown as a function of whether local flow has been blocked or remained open after 72 hr of competition ( bars denote means ± S . E . for n = 3 ) . These occurrence frequency data are normalized to the total area of blocked versus open flow in the microfluidic devices , as determined by the presence or absence of fluorescent bead tracks . There is no significant difference in wild type occurrence in regions in which flow is unobstructed and in regions in which flow is blocked ( two-sample t = 0 . 995 , df = 4 , p=0 . 376 ) , but the ∆pelA strain is significantly more likely to occur in regions in which flow is blocked at a p<0 . 05 threshold with Bonferroni correction for two pairwise comparisons ( two-sample t = 3 . 60 , df = 4 , p=0 . 0227 ) . ( C ) Biofilm growth of wild-type P . aeruginosa PA14 ( gray ) and the ΔpelA mutant ( white ) in monoculture in planar flow chambers under different shear stress exposure treatments ( bars denote means ± S . D . for n = 5–10 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 00910 . 7554/eLife . 21855 . 010Figure 3—source data 1 . Occurrence of WT and Pel-deficient P . aeruginosa in areas with flow blocked versus areas with flow open . Biomass accumulation of WT and Pel-deficient P . aeruginosa under different degrees of shear stress . ( B ) WT and ΔpelA biomass accumulation . ( C ) WT and ΔpelA growth under variable shearDOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 01010 . 7554/eLife . 21855 . 011Figure 3—figure supplement 1 . Streamer structures produced by wild-type P . aeruginosa PA14 ( green ) in microfluidic chambers with complex flow profiles do not capture large numbers of co-cultured ∆pelA mutants ( red ) over 72 hr of biofilm growth ( black circles are column obstacles ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 01110 . 7554/eLife . 21855 . 012Figure 3—figure supplement 2 . Analysis procedure for correlating local flow and accumulation of ∆pelA and wild-type cells . Liquid enters the chamber on the left and exits on the right . ( A ) First , the positions of the columns were identified from the fluorescence image . The centers of the columns served as nodes to divide the spaces into triangular regions . Within each region , the area covered by the column was removed . ( B ) The corresponding image with fluorescent beads ( white ) to track local flows . Typically , 10 to 15 images were taken and integrated to cover the flow regions sampled by the beads . Overlaid are triangular regions showing ∆pelA cell accumulation ( from ( C ) ) . ( C ) Averaged intensity within each sampling region was calculated , and a threshold was set to determine if biofilm accumulation occurred in each region . White triangles correspond to regions identified as containing biofilms using this method . ∆pelA cells accumulated primarily in regions lacking flow due to upstream clogging or in regions immediately downstream of obstructed areas . ( D ) Wild-type cells accumulated both in areas of high flow and areas of low flow . DOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 01210 . 7554/eLife . 21855 . 013Figure 3—figure supplement 3 . Change in frequency of WT cells from a 1:1 starting population with ∆pelA with and without flow . P . aeruginosa PA14 and the ∆pelA strain were inoculated at a 1:1 ratio and allowed to form biofilms in planar microfluidic devices with or without flow for 72 hr . At that time , the biofilms were imaged , and the change in relative frequency of the wild-type strain was calculated ( as in Figure 1 in the main text ) . When flow was present , the WT outcompeted the ∆pelA strain , and thus , the WT strain tended to show a positive change in frequency within biofilms ( gray ) . When flow was absent , the WT strain tended to show a negative change in frequency ( white ) . Bars denote means ± S . D . for n = 3–4 . DOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 01310 . 7554/eLife . 21855 . 014Figure 3—figure supplement 3—source data 1 . Comparison of competition in simple chambers with or without flow . DOI: http://dx . doi . org/10 . 7554/eLife . 21855 . 014 Biofilm growth is ubiquitous in porous microenvironments and often causes clogging in natural and industrial contexts , including soil beds and water filtration systems ( Knowles et al . , 2011 ) . Here , we have shown that the clogging process can dramatically influence population dynamics within growing biofilms by generating a feedback between hydrodynamic flow , biofilm spatial architecture , and competition ( Coyte et al . , 2016 ) . Our findings suggest that when P . aeruginosa wild type and ΔpelA mutants experience irregular flow in heterogeneous environments , wild-type biofilm formation causes partial clogging , regionally reducing local flow speed . The lack of flow generates favorable conditions for the ∆pelA strain , whose biofilms would otherwise be removed by shear forces , presumably enabling it to proliferate locally if sufficient nutrients for growth diffuse from other areas of the chamber in which medium continues to flow ( Bottero et al . , 2013 ) . P . aeruginosa is notorious as an opportunistic pathogen of plants and animals , including humans ( Xavier , 2016 ) . It also thrives outside of hosts , for example , in porous niches such as soil ( Fierer et al . , 2007 ) . Despite the well-documented ecological benefits of matrix secretion during biofilm formation , environmental and clinical isolates of P . aeruginosa exhibit considerable variation in their production of matrix components , including loss or overexpression of Pel ( Mann and Wozniak , 2012; Chew et al . , 2014 ) . Our results offer an explanation for natural variation in the ability of P . aeruginosa to produce extracellular matrix , particularly among bacteria in porous microhabitats: the evolutionary stable states of extracellular matrix secretion vary with the topographical complexity of the flow environment in which the bacteria reside .
All strains are derivatives of Pseudomonas aeruginosa PA14 ( RRID:WB_PA14 ) . Wild-type PA14 strains constitutively producing fluorescent proteins ( Drescher et al . , 2013 ) were provided by Albert Siryaporn ( UC Irvine ) , and they harbor genes encoding either EGFP or mCherry under the control of the PA1/04/03 promoter in single copy on the chromosome ( Choi and Schweizer , 2006 ) . The ∆pelA strain was constructed using the lambda red system modified for P . aeruginosa ( Lesic and Rahme , 2008 ) . To determine maximum growth rates and the potential for fluorescent protein production to cause fitness differences , bacterial strains were grown overnight in M9 minimal medium with 0 . 5% glucose at 37°C . Overnight cultures were back-diluted into minimal M9 medium with 0 . 5% glucose at room temperature and monitored until their optical densities at 600 nm were ~0 . 2 , corresponding to logarithmic phase . Cultures were back diluted again into minimal M9 medium with 0 . 5% glucose and transferred to 96-well plates at room temperature . This experiment was repeated for four biological replicates ( different overnight inoculation cultures ) , each repeated for six technical replicates ( different wells within a 96-well plate ) . Measurements of culture optical density at 600 nm were taken once per 10 min until saturation , corresponding to stationary phase . Matlab ( Natick , MA ) curve fitting software was used to calculate the maximum growth rate of each strain ( wild type [GFP]: 0 . 00716 h−1 , wild type [mCherry]: 0 . 00733 h−1 , ∆pelA [mCherry]: 0 . 00842 h−1 ) . These experiments confirmed that the fluorescent protein markers had no measurable effect on growth rates and thus did not contribute to competitive outcomes in our experiments . Microfluidic devices consisting of poly ( dimethylsiloxane ) ( PDMS ) bonded to 36 mm x 60 mm glass slides were constructed using standard soft photolithography techniques ( Sia and Whitesides , 2003 ) . We used planar microfluidic devices with no obstacles to simulate environments with simple parabolic flow profiles , and we used devices with PDMS pillars interspersed throughout the chamber volume to simulate environments with complex ( i . e . irregular , non-parabolic ) flow profiles . The size and spatial distributions of these column obstacles were determined by taking a cross-section through a simulated volume of packed beads mimicking a simple soil environment . Flow rates through these two chamber types were adjusted to equalize the average initial flow velocities , although the local flow velocity within each chamber varied as biofilms grew during experiments ( see main text ) . For all competition experiments , bacterial strains were grown overnight . The following morning , aliquots of the overnight cultures were added to Eppendorf ( Hamburg , Germany ) tubes , and their optical densities were equalized prior to preparation of defined mixtures of wild-type and ∆pelA cells . 100 µL volumes of the wild-type strain alone , the ∆pelA strain alone , or mixtures of the two strains ( for competition experiments ) , were introduced into microfluidic chambers using 1 mL syringes and Cole-Parmer ( Vernon Hills , IL ) polytetrafluoroethylene tubing ( inner diameter = 0 . 30 mm; outer diameter = 0 . 76 mm ) . After 3 hr , fresh tubing connected to syringes containing fresh minimal M9 medium with 0 . 5% glucose were inserted into the inlet channels . The syringes ( 3 mL BD Syringe , 27G; Becton , Dickinson and Co . ; Franklin Lakes , NJ ) were mounted onto high-precision syringe pumps ( Harvard Apparatus; Holliston , MA ) , which were used to tune flow speeds according to empirical measurements of flow speeds in soil ( Heath , 1983 ) . In our experiments , the average flow speed was 150–200 µm/s , unless noted otherwise . In Figure 3C , to alter shear , we varied the average flow speed; shear was estimated using standard calculations for surface shear stress under fluid flow: τ ( y ) =α∂u∂y= ΔpL H2 where τ is the shear stress , y is the height above the surface ( evaluated in this case for y=0 ) , α is the dynamic viscosity of the fluid , and u is the fluid flow velocity field , calculated for a rectangular channel in terms of the pressure decrease Δp/L across the length L and height H of the channel . The pressure decrease was calculated for our channel dimensions and flow rates using previously published results ( Fuerstman et al . , 2007 ) . Biofilms were grown at room temperature . It should be noted that microfluidic experiments in the obstacle-containing chambers experience a high failure rate , in which no biofilms appear to grow after the 72 hr period of the experiment . No data could be extracted from such chambers , which were omitted from analysis . This problem was overcome by performing the experiment at high replication . Sufficient data were thus collected to populate the relevant panels in Figures 1 and 3 of the main text . In the case of competition experiments , one replicate was defined as the output from one independently inoculated microfluidic chamber ( e . g . Figure 1A ) . For experiments in which biofilm growth was measured as a function of flow-mediated shear stress ( Figure 3C ) , one replicate was defined as the output from one imaging location within a microfluidic chamber , with two to three locations per chamber being sampled . To obtain spatial patterns of column obstacles that mimic soil or sand , we first generated a 3D model of packed spheres . The centers of the spheres were positioned such that they had equal radii of 1 ( arbitrary units ) , in a close-packed arrangement . Soil grains , however , are not all the same size . To include heterogeneity in sphere size in our model , we adjusted each sphere’s radius using uniformly distributed random numbers to generate a range of sphere radii varying from 0 . 4 to 1 . 0 . For a plane that is oblique to any of the symmetry planes defined by the centers of the spheres , we generated a cross-section through the 3D packed-sphere model . This cross-section of the spheres was used to define the borders of the columns in our soil-mimicking microfluidic devices . To convert the arbitrarily sized spheres from the 3D model to the actual sizes of physical columns in our microfluidic chambers , we chose column radii that varied from 80 to 200 μm , corresponding to particle sizes of fine- and medium-grain sand . Mature biofilms were imaged using a Nikon ( Tokyo , Japan ) Ti-E inverted microscope via a widefield epifluorescence light path ( using a 10x objective ) or a Borealis-modified Yokogawa CSU-X1 ( Tokyo , Japan ) spinning disk confocal scanner ( using a 60x TIRF objective ) . A 488-nm laser line was used to excite EGFP , and a 594-nm laser line was used to excite mCherry . Quantification of biofilm composition was performed using Matlab and Nikon NIS Elements analysis software ( Drescher et al . , 2014 ) . Imaging of biofilms could only be performed once for each experiment , precluding time-series analyses , due to phototoxicity effects after multiple rounds of imaging . Phototoxicity was a particularly notable issue here due to dimness of the fluorescent proteins in P . aeruginosa , which made long exposures necessary to capture images of sufficient quality for later analysis . For this reason , we opted for inferential population dynamics analysis as described in the main text . To measure strain frequencies in the biofilm effluent of planar chambers ( Figure 2C ) , 1:1 strain mixtures of wild-type and ∆pelA cells were prepared and inoculated into simple flow chambers according to the procedure outlined above for competition experiments . At 0 , 24 , and 48 hr , 5 µL samples were collected from the microfluidic chamber outlet tubing , mixed vigorously by vortex , and plated onto agar in serial dilution . After overnight growth at 37°C , plates were imaged with an Image Quant LAS 4000 ( GE Healthcare Bio-Sciences; Pittsburgh , PA ) . Cy3 and Cy5 fluorescence settings were used for EGFP and mCherry excitation , respectively . Image Quant TL Colony Counting software was used to measure the relative abundance of each strain . 1:1 mixtures of the wild-type and the ∆pelA mutant were prepared and introduced into obstacle-containing flow chambers according to the procedure described above . Minimal M9 medium with 0 . 5% glucose was introduced into the chambers for 72 hr as described above . The entire chamber was then imaged using widefield epifluorescence microscopy to document the locations of wild-type and ∆pelA cell clusters . Subsequently , the influent syringes were replaced with syringes containing yellow-green fluorescent beads ( sulfate-modified , diameter = 2 µm; Invitrogen; Carlsbad , CA ) at a concentration of 0 . 3% , and bead suspensions were flowed into the microfluidic chambers . To determine the presence or absence of flow with respect to the spatial distributions of wild-type and ∆pelA cells , and to obtain large images for statistics , the entire chamber was imaged with a 1 s exposure time , over which traveling beads were captured as streaks . It should be noted that this experiment also has a high failure rate due to the sensitivity of the microfluidic chambers to removal and re-insertion of syringes , and required optimization to execute successfully . Custom Matlab code was written to correlate the presence or absence of fluid flow with the accumulation of wild-type and ∆pelA cells . In brief , the positions of the columns were first identified and used to divide the chamber into triangular sampling areas using a network structure in which columns served as nodes and straight lines between column centers served as edges . Within each sampling triangle , the area covered by columns was first removed , and subsequently , the averaged ∆pelA and wild-type fluorescence intensities in the remaining area were used to determine if a region had wild type and/or ∆pelA accumulation . In parallel , each sampling area was scored for the presence of flow in the corresponding bead tracking images ( Figure 3—figure supplement 2 ) . In all cases where displayed , bars denote the mean values of the measurements taken , and with the exception of Figure 3B and Figure 1—figure supplement 1 , the error bars denote standard deviations . In Figure 3B and Figure 1—figure supplement 1 , error bars denote standard errors . In Figure 3B , we report the results of a two-tailed t-test comparing the wild type occurrence frequency in regions of soil-mimicking chambers where flow was obstructed , versus the wild type occurrence frequency in regions where flow was unobstructed . A second t-test was performed to make the same comparison for the ∆pelA cells . The p-values from these tests were evaluated against a critical threshold of p<0 . 05 adjusted by Bonferroni correction for two pairwise comparisons . Two t-tests were also performed on the data in Figure 1—figure supplement 1 measuring the maximum growth rates of our strains in liquid culture . | Bacteria often live together – attached to surfaces like river rocks , water pipes , the lining of the gut and catheters – in communities called biofilms . These groups of bacteria are small-scale ecosystems in which cells cooperate and compete with one another to obtain resources , such as food and space to grow . Within a biofilm , a sticky glue-like substance called the matrix binds the cells to each other and to the surface . Cells that make the matrix typically have an advantage over those that do not because they can better resist the shearing forces experienced when liquid flows over the surface . The matrix also helps cells to capture nutrients from the passing liquid . Nevertheless , not all strains of bacteria make matrix , despite its advantages . Because of where they can grow , biofilms are fundamentally important in the environment , in industry and in infections . Resolving why some bacteria make matrix while others do not could therefore allow scientists and engineers to re-design the surfaces involved in these settings to discourage harmful biofilms or to encourage beneficial ones . Nadell , Ricaurte et al . have now used a bacterium called Pseudomonas aeruginosa to explore how the properties of the surface and the flowing liquid affect matrix production among cells in biofilms . P . aeruginosa typically lives in soil and can cause infections in people , especially in hospital patients and people who have weakened immune systems . Nadell , Ricaurte et al . studied normal P . aeruginosa bacteria and a mutant strain that is unable to make matrix . The strains were labeled with fluorescent markers and put into special chambers that simulated different environments . The proportion of each strain was measured after three days of biofilm growth . When biofilms were grown under flowing liquid in simple environments with flat surfaces , matrix producers always outcompeted non-producers . However , the two strains coexisted in more complex and porous environments , like those found in soil . Nadell , Ricaurte et al . went on to show that the strains could co-exist because the matrix producers made biofilms that created areas within the environment where the liquid flows very slowly or not at all . In these regions , non-producing cells could compete successfully because resistance to shearing forces is less important when flow is weak or absent , and so the non-producing cells were not washed away . The results begin to explain why matrix production among cells in environmental settings is diverse and highlight that the environment is important in the evolution of bacterial biofilms . | [
"Abstract",
"Introduction",
"Results/discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"short",
"report",
"microbiology",
"and",
"infectious",
"disease"
] | 2017 | Flow environment and matrix structure interact to determine spatial competition in Pseudomonas aeruginosa biofilms |
PP2C phosphatases control biological processes including stress responses , development , and cell division in all kingdoms of life . Diverse regulatory domains adapt PP2C phosphatases to specific functions , but how these domains control phosphatase activity was unknown . We present structures representing active and inactive states of the PP2C phosphatase SpoIIE from Bacillus subtilis . Based on structural analyses and genetic and biochemical experiments , we identify an α-helical switch that shifts a carbonyl oxygen into the active site to coordinate a metal cofactor . Our analysis indicates that this switch is widely conserved among PP2C family members , serving as a platform to control phosphatase activity in response to diverse inputs . Remarkably , the switch is shared with proteasomal proteases , which we identify as evolutionary and structural relatives of PP2C phosphatases . Although these proteases use an unrelated catalytic mechanism , rotation of equivalent helices controls protease activity by movement of the equivalent carbonyl oxygen into the active site .
Reversible protein phosphorylation is widely used in biological systems to control the activity of enzymes or the association of proteins with other proteins . Kinases and phosphatases control the phosphorylation state of target proteins in response to specific cellular or environmental cues , making reversible phosphorylation a flexible mechanism to control diverse biological systems ( Huse and Kuriyan , 2002; Shi , 2009; Taylor and Kornev , 2011 ) . Here we address the question of how members of the PP2C family of serine/threonine phosphatases are regulated to control processes such as cell growth and death , development , and responses to stress in all kingdoms of life ( Kerk et al . , 2015; Lammers and Lavi , 2007; Shi , 2009 ) . Among serine/threonine phosphatases , a distinctive feature of the PP2C family is that the activity of a conserved catalytic domain is controlled by diverse regulatory domains that are often linked in cis to the catalytic domain ( Shi , 2009; Zhang and Shi , 2004 ) . We investigated the PP2C family member SpoIIE , which controls the activation of the cell-specific transcription factor σF during the developmental process of sporulation in the bacterium Bacillus subtilis . Sporulation involves the formation of an asymmetrically-positioned septum that divides the developing cell into large and small cellular compartments ( Stragier and Losick , 1996 ) . SpoIIE is the most upstream member of a three-protein pathway that activates σF in the small cell ( Figure 1A ) . It does so by dephosphorylating the phosphoprotein SpoIIAA-P ( Duncan et al . , 1995 ) . Dephosphorylated SpoIIAA , in turn , displaces σF from the anti-sigma factor SpoIIAB to release the free and active transcription factor ( Figure 1A ) ( Diederich et al . , 1994 ) . A long-standing mystery is how SpoIIE is regulated to generate dephosphorylated SpoIIAA selectively in the small cell . Recent work indicates that SpoIIE initially associates with the asymmetrically-positioned cytokinetic ring and then during cytokinesis is handed off to the adjacent cell pole , which will become the small cell ( Bradshaw and Losick , 2015 ) . Cell-specific activation is mediated by the self-association of SpoIIE molecules in the small cell , which protects the protein from proteolysis and activates the phosphatase ( Bradshaw and Losick , 2015 ) . Here we focus on the molecular mechanism of phosphatase activation . 10 . 7554/eLife . 26111 . 003Figure 1 . The structure of SpoIIE with its regulatory domain . A is a diagram of the three-protein pathway controlling σF . B is a schematic diagram of the SpoIIE primary structure with its N-terminal cytoplasmic degradation tag in black , the 10 transmembrane segments in dark grey , the regulatory domain in blue , and the PP2C phosphatase domain shown in light grey . Also shown are the switch helices in orange and the metal-coordinating residues within the active site in red . The black box identifies the SpoIIE457-827 fragment that was crystallized . C is a ribbon diagram of a single molecule of SpoIIE457-827 with front and side views using the same color scheme as the diagram in panel B . The inset shows the putative metal coordinating sidechains of the active site ( from top to bottom: D795 , D746 , and D628 ) and the backbone carbonyl of G629 . Figure 1—figure supplement 1 shows the 2Fo-Fc electron density map and a stereo representation of the SpoIIE457-827 structure . D shows the dimer observed in the crystal structure of SpoIIE457-827 ( chains A and B ) with the two protomers in darker and lighter shades ( buried surface area 1500–2000 Å2 per monomer ) . The two and a half dimers in the asymmetric unit are shown in Figure 1—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 00310 . 7554/eLife . 26111 . 004Figure 1—figure supplement 1 . Validation of the SpoIIE457–827 structure . A shows the long α-helix from the regulatory domain of chain A from the SpoIIE457–827 structure . The 2Fo–Fc electron density map is shown with a 4 Å carve radius around the α-helix in grey mesh contoured to 1 . 0 σ , and the anomalous difference map from seleno-methionine derivatized crystals is shown in yellow mesh contoured to 4 . 0 σ . B shows a stereo representation of chain A from the SpoIIE457–827 structure with the 2Fo–Fc electron density map shown in grey mesh contoured to 1 . 5 σ with a 2 . 5 Å carve radius around chain A . C shows a stereo representation of chain A from the SpoIIE457–827 structure . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 00410 . 7554/eLife . 26111 . 005Figure 1—figure supplement 2 . The crystal lattice contains three similar SpoIIE dimers . Each panel shows the asymmetric unit from the crystals of SpoIIE457–827 on the left with a single dimer circled in red and presented in isolation on the right . The chains are labeled as in the PDB file with the exception of Esym , which is a crystallographic symmetry mate of chain E . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 005 Like other PP2C family phosphatases , the catalytic center of SpoIIE uses two divalent cations ( manganese in the case of SpoIIE ) to deprotonate a water molecule that serves as the nucleophile for dephosphorylation ( Arigoni et al . , 1996; Schroeter et al . , 1999 ) . This active site is embedded in the conserved fold of the PP2C domain , which is shared by all PP2C family members ( Shi , 2009 ) . The PP2C domain is paired with diverse regulatory modules ( over 1500 unique domain architectures have been identified in the InterPro database ) ( Mitchell et al . , 2015 ) , but how these regulatory modules control phosphatase activity was not understood . Here we identify a pair of α-helices at the heart of the regulatory mechanism that rotate to position a carbonyl oxygen to bind an active site Mn2+ ion and activate SpoIIE . We present evidence that this mechanism is widely conserved among PP2C family members . Remarkably , rotation of equivalent α-helices is also used to control an unrelated catalytic mechanism in the structurally similar family of enzymes that form the catalytic core of the proteasome ( Arciniega et al . , 2014; Ruschak and Kay , 2012; Shi and Kay , 2014; Sousa et al . , 2000 ) . This raises the possibility that PP2C phosphatases and proteasome proteases have a common evolutionary history that is linked by a shared regulatory mechanism .
To investigate how PP2C phosphatase activity is regulated , we sought to determine X-ray crystal structures of SpoIIE with the phosphatase in the active and inactive states . We present a structure of a fragment that includes the entire PP2C phosphatase domain and a portion of the adjacent regulatory domain . This structure shows that the regulatory domain mediates the formation of dimers between SpoIIE molecules , and evidence indicates that dimerization is needed to activate the phosphatase . We also present a structure of the phosphatase domain alone . A comparison of the structures reveals that dimerization rotates two α-helices of the PP2C fold ( α1 and α2 of the conserved PP2C fold ) ( Das et al . , 1996 ) relative to the phosphatase core . We refer to these helices as switch helices and present evidence that this shift in position switches the phosphatase from the inactive to active state . To determine how SpoIIE is regulated , we first sought to determine the structure of the molecule in an active , self-associated state . The entire , 270-residue-long regulatory domain mediated the formation of heterogeneous multimers that were refractory to crystallization ( Bradshaw and Losick , 2015 ) . Using bioinformatic analysis , we devised a construct ( SpoIIE457–827 ) that included the C-terminal half of the regulatory domain and the PP2C phosphatase domain ( Figure 1B; information on the design of the construct is presented in the Materials and methods ) . This construct produced monodisperse protein that yielded crystals . Despite limited ( 3 . 9 Å ) resolution of the diffraction data , the overall secondary structure elements were well-defined in electron density maps for both the regulatory and the phosphatase domains ( Figure 1 , Figure 1—figure supplement 1 , and Table 1 ) . The most striking feature of the regulatory domain was an N-terminal 45-residue long α-helix ( residues 473–518 ) that makes intramolecular contacts with the switch helices ( α1 and α2 ) of the phosphatase domain ( Figure 1C ) . 10 . 7554/eLife . 26111 . 006Table 1 . Data collection and refinement statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 006SpoIIE457-827 ( 5UCG ) SpoIIE590-827 ( 5MQH ) Data collectionBeam sourceAPS 24-ID-CDiamond , I02Wavelength ( Å ) 0 . 97920 . 97950Space groupP43212C2221Cell dimensions a , b , c ( Å ) 125 . 62 , 125 . 62 , 330 . 7056 . 29 , 122 . 51 , 81 . 62 α , β , γ ( ° ) 90 , 90 , 9090 , 90 , 90Resolution ( Å ) *60–3 . 9 ( 3 . 97–3 . 9 ) 61 . 34–2 . 44 ( 2 . 48–2 . 44 ) Total reflections*284918 ( 8031 ) 60359 ( 4228 ) Unique reflections*24917 ( 1181 ) 10961 ( 681 ) Rsym†*0 . 102 ( 1 . 448 ) 0 . 057 ( 0 . 631 ) CC1/20 . 999 ( 0 . 847 ) 0 . 999 ( 0 . 874 ) CC*1 . 00 ( 0 . 958 ) -I / σI*24 . 7 ( 0 . 8 ) 20 . 1 ( 2 . 8 ) Completeness ( % ) *99 . 7 ( 97 . 4 ) 99 . 7 ( 99 . 8 ) Redundancy*11 . 4 ( 6 . 8 ) 6 . 3 ( 6 . 2 ) RefinementResolution ( Å ) *50–3 . 9 ( 4 . 1–3 . 9 ) 50–2 . 45 ( 2 . 51–2 . 45 ) No . reflections2155810187Rwork / Rfree‡*0 . 28/0 . 320 . 21/0 . 28No . atoms Protein131661783B-factors Protein93 . 068 . 0R . m . s . deviations Bond lengths ( Å ) 0 . 0020 . 010 Bond angles ( ° ) 0 . 5251 . 545Ramachandran plot Favored ( % ) 92 . 4896 . 9 Allowed ( % ) 7 . 403 . 1 Outliers ( % ) 0 . 120Rotamer outliers ( % ) 6 . 4415 . 4*Values in parentheses are for highest-resolution shell . †Rsym = ∑hkl∑i|Ii - <I> |/∑hkl∑i <I> where Ii is the intensity of the ith measurement of a reflection with indexes hkl and <I> is the statistically weighted average reflection intensity . ‡Rwork = ∑||Fo| - |Fc||/∑|Fo| where Fo and Fc are the observed and calculated structure factor amplitudes , respectively . Rfree is the R-factor calculated with 5% of the reflections chosen at random and omitted from refinement . The five molecules of SpoIIE457-827 in the asymmetric unit were paired in similar dimers; two dimers were formed within the asymmetric unit and the fifth molecule dimerized across a crystallographic two-fold axis ( Figure 1D , Figure 1—figure supplement 2 ) . The core of the dimer interface ( 1500–2000 Å2 buried surface per monomer ) was formed from antiparallel contacts between the long α-helices from the regulatory domains of adjacent molecules . Additionally , the switch helices at the base of each phosphatase domain contact each other across the dimer interface ( Figure 1 , Figure 2A and B , shown in orange ) . 10 . 7554/eLife . 26111 . 007Figure 2 . Dimerization activates the phosphatase . A is a surface representation of the SpoIIE457–827 dimer with the phosphatase domain , the switch , and the regulatory domain color coded as indicated in the associated schematic . Chain A is colored with darker shades and Chain B is colored with lighter shades . B is an open-book view of the SpoIIE457–827 dimer with the interface ( defined as residues within 4 . 5 Å of the adjacent molecule ) outlined in black . Red circles mark positions of amino-acid substitutions that blocked stabilization , localization , and activation ( V480K , L484K , V487K , M491K , F494K , I498K , L646K , I650K , and T663K ) , whereas white circles mark positions of substitutions that blocked activation ( as judged by σF activity ) but not stabilization and localization ( E639K , E642K , and I667K ) . Figure 2—figure supplement 1 presents the analysis of the behavior of the SpoIIE mutants in vivo . C is a surface representation of Chain A of SpoIIE457-827 rotated approximately 180° relative to the dimeric view in A . White circles indicate positions of substitutions that led to defects in activation ( but not localization ) of SpoIIE in vivo ( Q483A , G486K , V490K , and E497K ) . The box outlines the section of the long α-helix of the regulatory domain that is represented as a helical wheel in D . Figure 2—figure supplement 1 presents the analysis of the behavior of the SpoIIE mutants in vivo . D is a helical wheel representation of residues 480 to 498 from the long α-helix of the regulatory domain . Positions at which substitutions led to defects in σF activation are indicated by circles colored as in B and C . Black text ( A481K , S488K , D493K , and S495K ) indicates positions where substitutions did not lead to a phenotype , grey text represents positions that were not tested . Figure 2—figure supplement 1 presents the analysis of the behavior of the SpoIIE mutants in vivo . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 00710 . 7554/eLife . 26111 . 008Figure 2—figure supplement 1 . Functional analysis of the dimer interface . A shows the intracellular localization of the SpoIIE mutants described in Figure 2 . Images of SpoIIE-YFP fluorescence from representative sporulating cells that had completed asymmetric division are shown for wild-type SpoIIE ( top center ) , the mis-localized mutant SpoIIEV480K ( left ) , and the forespore-localized mutant SpoIIEQ483A ( right ) . Average fluorescence intensity profiles of SpoIIE-YFP are plotted for each mutant . Plots from mutants with defects in localization and activation of σF are in red ( left ) , and plots from variants that are defective only in σF activation are in white ( right ) . The blue bar at the right indicates substituted residues that reside in the long α-helix of the regulatory domain , whereas the orange bar marks substitutions in the PP2C phosphatase domain switch region . Each trace represents an average intensity profile normalized to the membrane dye FM4-64 from hundreds of asymmetrically divided cells aligned at the forespore pole . A reference plot from wild-type SpoIIE is in grey , and the dashed line represents the approximate position of the asymmetric septum . After σF activation , SpoIIE is recruited back to the forespore face of the asymmetric septum and then moves along with the engulfing membrane to encompass the forespore . Thus , mutants with the most severe defects in σF activation haves fluorescence profiles that are slightly shifted towards the forespore pole relative to that of wild-type cells . B-D are immunoblots showing the levels of SpoIIE-YFP , and CFP produced from a σF dependent promoter ( detected using an α-GFP antibody ) for the mutants in panel A . B shows immunoblots for the mutants that exhibited reduced SpoIIE levels and abnormal localization ( red ) . C shows immunoblots for the mutants shown in panel B in which the FtsH degradation tag of SpoIIE had been removed to stabilize SpoIIE protein ( the immunoblot for intact SpoIIE is shown in the left lane ) . D shows immunoblots for the properly localized mutants ( white ) . The colored bars and mutant labels are color coded as in panel A . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 008 To investigate the role of dimerization in stabilization , localization and phosphatase activation , we systematically created substitutions of residues that make up the dimer interface and investigated the ability of these mutants to function during sporulation . We substituted the native amino-acids with lysine because the positive charge and the long side chain would be expected to impair dimerization . The effect of these substitutions on stabilization and subcellular localization was investigated by use of a SpoIIE-YFP fusion and the effect on phosphatase activity was judged by use of a σF-dependent reporter ( Figure 2B , red circles , Figure 2—figure supplement 1A , B and C ) . The results revealed that a continuous region of the dimer interface ( marked with red circles in Figure 2B ) composed of six residues from the long α-helix of the regulatory domain ( V480 , L484 , V487 , M491 , F494 , and I498 ) and three residues from the switch helices ( L646 , I650 , and T663 ) were needed for all three aspects of SpoIIE function . These findings are consistent with the hypothesis that the dimers observed in our structure represent the active state of the phosphatase . To investigate how dimerization activates phosphatase activity , we sought to compare the active dimeric structure of SpoIIE457–827 to inactive SpoIIE . Previously , we determined the structure of SpoIIE590–827 , a fragment that included the PP2C phosphatase domain but lacked the adjacent regulatory domain ( Levdikov et al . , 2012 ) . We hypothesized that this structure represented the inactive state because it lacked the dimeric interface of the SpoIIE457–827 structure . Although monomeric in solution under physiological conditions , SpoIIE590–827 had undergone a domain-swap dimerization during crystallization ( Levdikov et al . , 2012 ) . Here , we solved an additional structure for SpoIIE590–827 ( with an amino acid substitution A624I that was designed to block domain swapping ) that was in a different crystal form and was not domain-swapped ( Figure 3A ) . Importantly , the only significant differences between the two SpoIIE590–827 structures were at the site of the domain-swap ( Figure 3—figure supplement 1 ) . Also , contacts between the phosphatase domains observed in the SpoIIE457–827 dimer were not present in either of the SpoIIE590–827 structures . 10 . 7554/eLife . 26111 . 009Figure 3 . Repositioning the switch activates the phosphatase . A is a ribbon diagram of the structure of SpoIIE590–827 , which is the phosphatase domain of SpoIIE lacking the regulatory domain . The region of the protein that was crystallized is diagramed above . The switch region and Mn2+-coordinating residues are color-coded as in Figure 1A . Figure 3—figure supplement 1 shows a comparison with the previously published domain swapped SpoIIE590–827 structure . B compares the conformations of the phosphatase domain in the dimeric SpoIIE457–827 structure ( switch helices in dark orange ) and the isolated phosphatase domain of SpoIIE590–827 ( switch helices in light orange ) . The structures were aligned based on the core of the phosphatase domain excluding the switch region ( residues 590–628 and 678–827 ) with an RMSD = 0 . 952 Å ( 970 to 970 atoms ) . The major conformational change upon dimerization corresponds to a rotation and upward movement of the switch helices . Figure 3—figure supplement 2 shows how gain of function mutants may promote the conformational change . C is a model for how rotation of the switch helices leads to phosphatase activation . In the inactive state ( left ) G629 is not positioned to coordinate the M2 metal . We propose that dimerization ( right ) leads to rotation of the switch helices ( orange ) , which repositions G629 to recruit manganese and complete the active site . We note that an additional glycine of RsbX ( G47 ) , corresponding to G631 of SpoIIE , also coordinates M2 . Thus , it is possible that G631 also coordinates M2 in place of the lower right-hand water molecule depicted in the schematic diagram ( Teh et al . , 2015 ) . Figure 3—figure supplement 3 shows details of the active site in the SpoIIE457–827 structure . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 00910 . 7554/eLife . 26111 . 010Figure 3—figure supplement 1 . Comparison of the SpoIIE590–827 structures . A shows the domain swapped structure of SpoIIE590–827 ( PDB ID 3T91 ) . The PP2C domain of Chain B is grey and the switch helices are colored orange and Chain A is colored blue . B shows an overlay of the unswapped SpoIIE590–827 structure ( light shades ) , and the domain swapped SpoIIE 590–827 ( colored as in panel A ) . The overlay was done using residues 678–800 ( RMSD 0 . 72 Å ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 01010 . 7554/eLife . 26111 . 011Figure 3—figure supplement 2 . Gain-of-function alleles activate the phosphatase . A shows the side-chains that surround V697 ( green ) in the inactive ( SpoIIE590–827 left , switch helices in light orange ) and active ( SpoIIE457–827 right , switch helices in orange ) conformations . Residues depicted as sticks are L647 , I661 , I664 , N665 , L668 , I676 , L680 , L695 , L718 , F726 , and V728 . B is a head-on ribbon representation of SpoIIE457–827 as in Figure 1 with spheres indicating the position of residues substituted in gain-of-function mutants that were isolated as suppressors of the spoIIE48 mutation . The residues are in three clusters: those that contact the switch helices from the PP2C phosphatase domain ( I684 , L695 , V697 , and V728 ) , those on the switch helices that make contacts across the dimer interface ( K649 and I650 ) , and those that point up towards the switch helix from the long α-helix of the regulatory domain ( L479 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 01110 . 7554/eLife . 26111 . 012Figure 3—figure supplement 3 . Manganese binding in the SpoIIE active site . A shows an anomalous difference map calculated from an X-ray dataset collected from manganese-soaked crystals overlaid on the ribbon diagram of SpoIIE457–827 as in Figure 1 . The side view of the SpoIIE457–827 structure is shown for chain A , and the inset panels show the active site regions with the putative metal-coordinating side-chains for each of the five chains in the asymmetric unit . The purple spheres represent the manganese ions from superimposed RsbX ( PDB ID 3W43 , see panel C below ) , displayed here for reference . The maps are shown with a 4 Å carve radius around the indicated chain and are contoured at 4 . 0 σ for chains A and B and 3 . 5 σ for chains C , D , and E . B shows the 2Fo–Fc electron density map from the X-ray data in grey mesh contoured to 1 . 0 σ with a 2 . 5 Å carve radius around the active site loop residues 628–635 of SpoIIE457–827 . Residues 628–635 are shown as sticks . C shows an overlay of SpoIIE457–827 and RsbX ( PDB ID 3W43 ) aligned based on residues 590–628 . SpoIIE457–827 is shown as a darker shade , and RsbX is shown as a lighter shade , and the putative metal-coordinating side-chains of the active sites are shown as sticks . The purple spheres represent the manganese ions from the RsbX structure . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 012 Comparison of the SpoIIE590–827 structures with SpoIIE457–827 revealed that dimerization rotated the switch helices ( α1 and α2 of the PP2C fold , corresponding to SpoIIE residues 630–678 ) approximately 45° as a rigid body relative to the phosphatase core ( Figure 3B , Video 1 ) . We hypothesized that this conformational change of the switch helices is responsible for activation of the SpoIIE phosphatase . 10 . 7554/eLife . 26111 . 013Video 1 . The PP2C phosphatase domain of SpoIIE changes conformation upon dimerization . Shown is the PP2C phosphatase domain of SpoIIE ( switch helices in orange ) morphing from the structure of SpoIIE590–827 to the structure of SpoIIE457–827 . The structures were aligned based on the core of the phosphatase domain excluding the switch region ( residues 590–628 and 678–827 ) as in Figure 3B . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 013 To evaluate whether repositioning of the switch region is responsible for phosphatase activation , we returned to our genetic analysis of the contacts made in the SpoIIE457–827 structure . In the dimer , the switch helices are held in position by intramolecular contacts with the long α-helix of the regulatory domain and intermolecular contacts between switch helices across the dimer interface ( Figures 1D , 2B and C ) . We found that single-amino acid substitutions at either of these contact sites blocked phosphatase activity but not stabilization or localization to the small cell . Phosphatase activity was assessed by σF-directed gene expression and stabilization and localization by use of a SpoIIE-YFP fusion ( white circles in Figure 2C and D and Figure 2—figure supplement 1A and D ) . This result defines two roles for the long α-helix: one face of the helix mediates dimerization and is required for all three aspects of SpoIIE function ( stabilization , localization and phosphatase activity ) ( Figure 2B and D red circles ) , and the other face , which makes intramolecular contacts with the switch region , is specifically required for phosphatase activity ( Figure 2B–D white circles ) . Additionally , these results are consistent with the idea that dimerization stimulates phosphatase activity by repositioning the switch helices . Replacement of valine at position 697 with alanine causes a gain-of-function mutant phenotype in which σF is activated constitutively ( Carniol et al . , 2004; Hilbert and Piggot , 2003 ) . The V697A substitution also enhanced phosphatase activity as measured in vitro ( Bradshaw and Losick , 2015 ) . But how this substitution acts had been unclear . Our structure of SpoIIE590–827 reveals that in the monomeric state , V697 packs in a hydrophobic pocket between the β strands at the base of the PP2C domain and the switch ( Figure 3—figure supplement 2A ) . In contrast , N665 from the switch packs near V697 in the structure of SpoIIE457–827 in the dimeric state . We therefore hypothesize that in the wild type , V697 stabilizes the conformation of the switch helices in the inactive monomeric state and that truncating V697 to alanine stabilizes the active conformation by promoting solvation of the polar residue N665 ( Figure 3—figure supplement 2A ) . Thus , and according to our hypothesis , replacing V697 with alanine destabilizes the inactive state by removing hydrophobic contacts and favors the active conformation by eliminating a repulsive interaction . Reinforcing this hypothesis , substitution of V697 with a bulky hydrophobic residue ( phenylalanine ) , which could similarly destabilize the inactive conformation , also causes constitutive activity ( Bradshaw and Losick , 2015 ) . Other gain-of-function mutants that stimulate phosphatase activity in the context of a loss-of-function mutant support this hypothesis ( Carniol et al . , 2004 ) . The amino acid substitutions in these mutants ( L479F , K649T , I650L , I684V , L695W , and V728M ) were all located at positions in the structure that could contribute to positioning the switch helices ( Figure 3—figure supplement 2B ) . I684 and L695 project down from the β-strands at the base of the phosphatase domain to contact the switch . K649 and I650 are themselves part of the switch helices and project across the dimer interface . V728 projects towards the switch from the loop implicated in substrate binding in other PP2C phosphatases . Finally , L479 projects up towards the switch from the long α-helix of the regulatory domain . We conclude that , like V697A , these amino-acid substitutions bias the phosphatase domain to the active conformation of the switch region . How does repositioning the switch region activate phosphatase activity ? All PP2C phosphatases coordinate 2–3 divalent metals ( usually manganese ) in their active sites ( Das et al . , 1996; Shi , 2009 ) . The two core metal ions , known as M1 and M2 , directly participate in catalysis by deprotonating a water molecule that serves as the nucleophile for hydrolysis ( Das et al . , 1996 ) . Based on the universally conserved architecture of the catalytic center , the M2 metal of SpoIIE is predicted to be coordinated by the side-chain of D628 and the carbonyl oxygen of G629 ( Schroeter et al . , 1999 ) ( Figure 1C and Figure 3C ) . G629 is at the junction between the switch helices and the β strands at the base of the phosphatase domain , such that movement of the switch helices could be coupled with bringing G629 into position to recruit M2 . In support of this idea , G629 is not in position to coordinate M2 in our isolated phosphatase domain structures , which we thus conclude represent an inactive state . This is supported by the fact that although our previously published structures included manganese in the crystallization conditions , the M2 site was unoccupied and the active site contained only a single manganese ( Levdikov et al . , 2012 ) . While soaking SpoIIE457–827 crystals with manganese degraded the diffraction , an anomalous difference map provided evidence that manganese was bound in the active site ( Figure 3—figure supplement 3A and Table 2 ) . Due to the low ( 5 . 4 Å ) resolution of the data for the manganese-soaked crystals , the number of bound metal ions and their position in the active site could not be established . In the dimeric SpoIIE457–827 structure and in contrast to the SpoIIE590–827 structure , the loop connecting the switch helices to G629 was ordered ( Figure 3—figure supplement 3B ) and overlaid well with M2-containing structures of closely related phosphatases such as B . subtilis RsbX ( Teh et al . , 2015 ) , M . tuberculosis Rv1364c ( King-Scott et al . , 2011 ) , and S . thermophilus Sthe_0969 ( Nocek et al . , 2010 ) ( Figure 3—figure supplement 3C ) . We propose that the shift of the switch helices activates the phosphatase by repositioning G629 to recruit M2 and complete the active site ( Figure 3C ) . 10 . 7554/eLife . 26111 . 014Table 2 . Data collection statistics for anomalous datasets . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 014SpoIIE457-827 MnSpoIIE457-827 SeMetData collectionBeam sourceAPS 24-ID-CAPS 24-ID-ESpace groupP43212P43212Cell dimensions a , b , c ( Å ) 124 . 783 , 124 . 783 , 329 . 787123 . 081 , 123 . 081 , 329 . 556 α , β , γ ( ° ) 90 , 90 , 9090 , 90 , 90InflectionInflectionWavelength ( Å ) 1 . 893500 . 97920Resolution ( Å ) *50–5 . 4 ( 5 . 49 — 5 . 4 ) 50–5 . 7 ( 5 . 8–5 . 7 ) Total reflections*40325 ( 318 ) 51233 ( 4024 ) Unique reflections*8598 ( 187 ) 8071 ( 706 ) Rsym*0 . 145 ( 0 . 535 ) 0 . 175 ( 1 . 475 ) CC1/2*0 . 99 ( 0 . 75 ) 0 . 996 ( 0 . 459 ) CC* *0 . 997 ( 0 . 926 ) 0 . 999 ( 0 . 793 ) Mean I / σI*9 . 14 ( 1 . 00 ) 7 . 86 ( 1 . 13 ) Completeness ( % ) *90 . 1 ( 41 . 6 ) 99 . 0 ( 97 . 4 ) Redundancy*4 . 7 ( 1 . 7 ) 6 . 3 ( 5 . 7 ) *Values in parentheses are for highest-resolution shell . A prediction of the hypothesis that movement of the helices allows recruitment of M2 is that binding of metal to the active site of SpoIIE should be coupled to dimerization and activation . Whereas in cells , cues in the forespore promote self-association of SpoIIE to induce phosphatase activity ( Bradshaw and Losick , 2015 ) , we reasoned that in vitro , in the absence of cellular cues , addition of high concentrations of manganese should drive dimerization and activation by mass action ( Figure 4A ) . We used size exclusion chromatography coupled to multi angle laser light scattering ( SEC-MALLS ) to monitor SpoIIE dimerization over a range of manganese concentrations . In the absence of manganese , SpoIIE457–827 eluted as a single monodisperse peak with a calculated molecular weight of 42 kDa , consistent with the calculated molecular weight of a monomer ( Figure 4B ) . Addition of 250 µM and 1 mM MnCl2 induced dimerization of SpoIIE457–827 , shifting and broadening the peak in concert with an increase in molecular mass ( Figure 4B ) . In support of the idea that this dimerization uses the interface found in the SpoIIE457–827 structure , substitution of a residue from the interface ( L484 ) with lysine blocked dimerization even after addition of 1 mM MnCl2 ( Figure 2 , Figure 4—figure supplement 1 ) . Additionally , substitution of the M2 coordinating residue D628 with alanine partially impaired dimerization in the presence of 1 mM MnCl2 , suggesting that manganese binding in the active site promoted dimerization ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 26111 . 015Figure 4 . The switch promotes manganese binding in the phosphatase active site . A is a model for phosphatase activation . During sporulation , cellular cues induce dimerization of SpoIIE molecules , rotating the switch helices and leading to Mn2+ binding in the active site . A prediction of this model is that high concentrations of Mn2+ would drive SpoIIE to become activated and form dimers . B shows SEC-MALLS ( size exclusion chromatography coupled to multi angle laser light scattering ) results for the SpoIIE457-827 fragment to assess complex formation at various concentrations of Mn2+ . The top plot shows molecular weights calculated from light scattering and the bottom plot shows the corresponding UV absorbance traces for both wild-type SpoIIE ( left-hand side ) and the gain-of-function mutant SpoIIEV697A ( right ) . The experiments were performed in the absence of Mn2+ ( grey ) , with 0 . 25 mM MnCl2 ( light purple ) , and at 1 mM MnCl2 ( purple ) . All experiments were performed in triplicate and data from representative runs are shown . Figure 4—figure supplement 1 shows size exclusion chromatography analysis of additional SpoIIE mutants . The source data are included as Figure 4—source data 1 . C is a plot of phosphatase activity ( initial rates , vobs ) for the wild-type ( black ) and V697A mutant ( purple ) SpoIIE457–827 fragments as a function of MnCl2 concentration using SpoIIAA-P as the substrate . The data were fit with the equation vobs=Vmax*[MnCl2]h/ ( K+[MnCl2]h ) where h is the Hill coefficient calculated from the inset panel [Vmax = 4 . 15 ± 0 . 04 min–1 ( 2 . 28 ± 0 . 04 min–1 for SpoIIEV697A ) and K = 0 . 32 ± 0 . 02 mM ( 0 . 020 ± 0 . 002 mM for SpoIIEV697A ) ] . The K1/2 values reported in the text were calculated from this equation and represent the concentration of MnCl2 at which SpoIIE has half maximal activity . Inset is a Hill plot for data points representing 10–90% activity . Lines are linear fits to the data using the equation log ( vobs/ ( Vmax–vobs ) ) =h*log[MnCl2]–logK [h = 2 . 0 ± 0 . 1 ( 1 . 92 ± 0 . 1 for SpoIIEV697A ) and K = 0 . 31 ± 0 . 04 mM ( 0 . 022 ± 0 . 008 mM for SpoIIEV697A ) ] . The reported error is the error of the fit to the data . Experiments were repeated at least three times and data from a representative experiment are shown . The source data are included as Figure 4—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 01510 . 7554/eLife . 26111 . 016Figure 4—source data 1 . Source data for Figure 4B and C . Provided is a spreadsheet with source data for Figure 4 panels B and C . Data for size-exclusion chromatography ( panel B ) , molecular weight determined by MALLS ( panel B ) and phosphatase activity ( panel C ) are provided as separate sheets . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 01610 . 7554/eLife . 26111 . 017Figure 4—figure supplement 1 . Manganese-induced dimerization requires the active site and dimer interface . Shown is size exclusion chromatography analysis in the absence of manganese ( grey ) or in the presence of 1 mM MnCl2 ( purple ) for 200 µM wild type ( Panel A ) , SpoIIED628A ( Panel B ) , and SpoIIEL484K ( Panel C ) . The fact that some dimerization still occurs for the D628A variant was not unexpected because the other metal-coordinating residues are still present ( including G629 ) and high concentrations of protein and MnCl2 were used in the experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 017 To test whether manganese-induced dimerization correlated with phosphatase activation , we measured the dependence of phosphatase activity on manganese concentration using an assay for dephosphorylation of SpoIIAA-P , the native substrate of SpoIIE . By varying the manganese concentration in the presence of saturating substrate , we determined that SpoIIE was cooperatively activated ( h = 2 . 0 ) with a K1/2 for manganese of 0 . 56 mM ( Figure 4C ) . This correlates well with the manganese dependence of dimerization ( Figure 4B left panel ) . Additionally , cooperative activation with a Hill coefficient of two indicates that at least two manganese ions bind in the active site of SpoIIE , consistent with the proposed catalytic mechanism ( Figure 3C ) . Our hypothesis also predicts that the V697A substitution would reduce the manganese concentration required for dimerization and phosphatase activity by favoring the active conformation of the switch . Indeed , the K1/2 for manganese was reduced from 0 . 56 mM to 0 . 13 mM for SpoIIEV697A ( Figure 4C ) , and SEC-MALLS revealed that the V697A substitution similarly reduced the concentrations of manganese required to promote dimer formation ( Figure 4B right panel ) . Together these experiments provide biochemical evidence that SpoIIE dimerization is coupled to phosphatase activity by rotation of the switch region and coordination of manganese in the active site .
The following illustrative examples highlight the conservation and adaptability of the allosteric regulatory mechanism among PP2C phosphatases ( Figure 5 ) :10 . 7554/eLife . 26111 . 018Figure 5 . Evidence that the switch mechanism is broadly conserved among phosphatases . The structure of the active SpoIIE457–827 phosphatase domain is shown in the center . The SpoIIE dimerization interface that mediates activation is indicated with an orange arc . Similarly , additional arcs indicate regions where regulatory inputs impinge on the PP2C phosphatase domain for RsbP ( brown , Figure 5—figure supplement 1 ) , Pdp1 ( phosphorylation is shown in red , and lipoyl binding is shown in teal , Figure 5—figure supplement 2 ) , Fem-2 ( blue , Figure 5—figure supplement 3 ) , Hab1 ( grey , Figure 5—figure supplement 4 ) , and RssB ( green , Figure 5—figure supplement 5 ) . The diagram is based on structures illustrated in Figure 5—figure supplements 1–5 . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 01810 . 7554/eLife . 26111 . 019Figure 5—figure supplement 1 . Gain-of-function bypass suppressors suggest that the switch controls the energy stress response PP2C phosphatase RsbP . A shows the regulatory cascade in which the PP2C phosphatase RsbP activates the transcription factor σB . In response to energy stress RsbP is activated as a phosphatase and dephosphorylates RsbV-P , which binds to the anti-sigma factor RsbW and displaces σB , which then is competent to activate transcription . B shows the position of bypass suppressor substitutions that activate RsbP in the absence of RsbQ , the stress-responsive activator of RsbP . The positions of these substitutions are mapped onto the structure of RssB ( PDB ID 3F7A ) by homology ( no structure of RsbP is available ) and are indicated with spheres in green . Residues shown are RsbP 173 , 181 , 211 , 230 , 233 , 241 , 242 , 244 , and 246 ( corresponding to RssB residues 159 , 166 , 199 , 215 , 218 , 228 , 229 , 231 , and 233 ) . The α-helix corresponding to the predicted α0 helix of the RsbP regulatory domain is blue , and the switch helices are orange . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 01910 . 7554/eLife . 26111 . 020Figure 5—figure supplement 2 . Structural evidence that the switch is used to control pyruvate dehydrogenase phosphatase . A diagrams how the PP2C phosphatase pyruvate dehydrogenase phosphatase ( Pdp1 ) promotes flux through the TCA cycle by activating pyruvate dehydrogenase . B is a ribbon diagram of Pdp1 ( PDB ID 2PNQ ) . The PP2C phosphatase domain is in grey with tyrosine-94 , which is phosphorylated to inhibit the activity of Pdp1 , in red sticks , and the switch helices in orange . The metal-coordinating residues of the active site are shown in stick representation . A Pdp1-specific insertion , colored in blue , contains the predicted activating binding site for lipoyl groups from the E2 subunit of pyruvate dehydrogenase , as indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 02010 . 7554/eLife . 26111 . 021Figure 5—figure supplement 3 . Structural evidence that the switch is used to control the sex-determining PP2C phosphatase FEM-2 . A diagrams how the PP2C phosphatase Fem-2 , together with FEM-1 and FEM-3 , promotes sex determination in C . elegans . B is a ribbon diagram of FEM-2 ( PDB ID 4JND ) . The PP2C phosphatase domain is grey , the switch helices are orange , and the N-terminal domain that binds FEM-1 and FEM-3 is blue . The metal-coordinating sidechains of the active site are shown as sticks , and the magnesium ions are shown as green spheres . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 02110 . 7554/eLife . 26111 . 022Figure 5—figure supplement 4 . The switch for the drought responsive PP2C phosphatase Hab1 switch could coordinate phosphatase activity and substrate binding . A diagrams how the PP2C phosphatase Hab1 promotes drought tolerance in plants . B shows a ribbon diagram of Hab1 ( PDB ID 3UJG ) . The PP2C phosphatase domain is grey , the switch helices are orange , and the ‘flap’ region is blue . The metal-coordinating sidechains of the active site are shown , and the magnesium ions are shown as spheres . C shows a surface representation of Hab1 as in panel B . The contact surface with the kinase SnRK2 ( defined as residues within 4 Å ) is indicated with a black outline and the ‘lock’ residue W385 , which is critical for binding of Hab1 to both its substrate and to the PYR/PYL/RCAR family of abscisic acid binding receptors that inhibit Hab1 activity , is indicated with a red circle . D is an overlay of the structures of dimeric SpoIIE457–827 ( switch helices colored dark orange ) and monomeric SpoIIE590–827 ( switch helices colored light orange ) as in Figure 3 . The ‘flap’ region of SpoIIE is blue ( SpoIIE457–827 ) or light blue ( SpoIIE590–827 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 02210 . 7554/eLife . 26111 . 023Figure 5—figure supplement 5 . The switch for the pseudo-PP2C phosphatase RssB controls protease adapter activity . A diagrams how the pseudo-PP2C phosphatase RssB destabilizes σS by acting as an adapter protein for ClpXP proteolysis . B is a ribbon diagram of RssB ( PDB ID 3F7A ) . The PP2C phosphatase domain is grey , the switch helices are orange , and the regulatory region is blue . The positions of bypass suppressor amino acid substitutions ( residues 149 , 156 , 160 , 164 , 220 , 222 , 224 , 227 , 228 , 260 , 261 , and 263 from P . aeruginosa corresponding to residues 146 , 150 , 154 , 158 , 214 , 216 , 218 , 221 , 222 , 254 , 255 , and 257 from E . coli ) that render RssB active in the absence of stress are indicated with spheres . C shows a comparison of the RssB dimer ( above ) and the SpoIIE457-827 dimer ( below ) . The structures are colored as in panel A . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 023 One of the most striking discoveries of our investigation is that the PP2C regulatory switch strongly resembles the allosteric switch that regulates the family of proteases that form the catalytic core of the proteasome ( Arciniega et al . , 2014; Ruschak and Kay , 2012; Shi and Kay , 2014; Sousa et al . , 2000 ) . These proteases are the most structurally similar family to PP2C phosphatases as revealed using the DALI server ( Holm and Rosenström , 2010 ) and the ECOD database ( Cheng et al . , 2014 ) , and like PP2C phosphatases their catalytic activity is subject to allosteric regulation . Specifically , the proteasome proteases and PP2C phosphatases have a conserved core fold ( Figure 6A and B ) , which includes the switch helices , and the active sites are positioned in the same overall part of the structure . Although the proteases use different functional groups to mediate catalysis , the carbonyl oxygen of a conserved glycine ( G629 of SpoIIE ) at the junction of the core domain and the switch helices is used by both enzyme families for catalytic activity ( Sousa et al . , 2000 ) . 10 . 7554/eLife . 26111 . 024Figure 6 . The switch mechanism is shared with proteasome proteases . A is a secondary structure topology diagram for SpoIIE ( left ) and for HslV ( the E . coli homolog of the proteasome protease; right ) . β strands are shown as arrows pointing from N to C terminus and α-helices as circles in cross section . Conserved features are dark grey , whereas variable features are light grey . The conserved glycine that moves to activate each protein is indicated with a red circle . The switch helices of SpoIIE and the corresponding α-helices of HslV are colored orange . B shows ribbon diagrams of SpoIIE and HslV ( PDB ID 1G3I ) colored as in A . The position of the conserved regulatory glycine ( G649 in SpoIIE , and G69 in HslV ) is shown with a red sphere and the insertions specific to each protein are indicated by brackets . C is a schematic of how the regulatory particle ( blue ) activates the proteasome proteases ( grey ) . D shows an overlay of the active ( PDB ID 1G3I ) and inactive ( PDB ID 1G3K ) states of HslV following superimposition of the regions in grey . The switch helices are color-coded orange and light orange for the active and inactive states , respectively . The active site residues T1 , K33 , and the carbonyl oxygen of G69 are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 024 Association with the regulatory cap activates the proteases , ensuring that the proteolytic active sites are sequestered prior to activation ( Seol et al . , 1997 ) ( Figure 6C ) . Early studies on HslV , the E . coli homologue of the proteasome proteases , revealed that allosteric activation by the HslU cap takes place by rotation of the switch helices to position the active site glycine ( Figure 6D ) ( Sousa et al . , 2000 ) . This mechanism is remarkably similar to the regulatory mechanism we proposed for PP2C phosphatases; docking of a regulatory module repositions the structurally homologous region in the same way to position the same functional group to achieve catalytic activity ( Video 2 ) . 10 . 7554/eLife . 26111 . 025Video 2 . PP2C phosphatases and proteasomal proteases share a common conformational switch . Shown are side-by-side displays of SpoIIE and HslV morphing from the inactive to active states . Shown on the left is the PP2C phosphatase domain of SpoIIE morphing from SpoIIE590–827 ( inactive , PDB ID 5MQH ) to SpoIIE457–827 ( active , PDB ID 5UCG ) as in Figure 3B . Shown on the right is HslV morphing from the HslU free structure ( inactive , PDB ID 1G3K ) to the HslU bound structure ( active , PDB ID 1G3I ) as in Figure 6D . The switch helices are colored orange and the active site residues of each protein are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 025 This mechanism is also conserved for the archaeal proteasome , which like the eukaryotic proteasome includes an additional layer of related , but catalytically inactive α subunits; docking of the cap displaces the switch helices of the α subunits , which directly contact and reposition the switch helices of the catalytic β subunits ( Ruschak and Kay , 2012 ) . Several lines of evidence suggest that this mechanism is conserved for the eukaryotic proteasome ( Arciniega et al . , 2014 ) and is additionally used by chaperones that promote proteasome maturation ( Wani et al . , 2015 ) . Finally , comparative studies of the constitutive proteasome and the immune proteasome suggested that differences in the conformational flexibility of the switch underlies their differences in activity ( Arciniega et al . , 2014 ) . Thus , our identification of the PP2C switch demonstrates that PP2C phosphatases and the proteasome use the same allosteric regulatory module , revealing an unexpected link between two fundamental signaling systems – reversible phosphorylation and regulated proteolysis . Independent analysis of structural and sequence similarity suggest that this is a result of common evolutionary ancestry . Structural comparison by the ECOD database , which classifies the evolutionary relationships of protein folds places PP2C phosphatases and the proteasome proteases in the same ‘X-group’ , which is consistent with homology ( Cheng et al . , 2014 ) . Independently , sequence-based searches using HHPRED ( Söding et al . , 2005 ) detected weak sequence similarity between phosphatases and the broad family of NTN-hydrolases that includes the proteasome proteases . For example , using the SpoIIE phosphatase domain sequence to search hidden Markov model alignments for B . subtilis proteins identified weak sequence similarity to D-fructose-6-phosphate amidotransferase , an NTN hydrolase . Notably , the region of possible sequence similarity maps to the switch helices and the β strands that follow and pack with the switch ( although it is not known whether the switch helices play a regulatory role in amidotransferases ) . What evolutionary path might connect proteasomal proteases and PP2C phosphatases ? Acquisition of a new catalytic mechanism requires that the ancestral protein retain function while acquiring the changes necessary for the new catalytic mechanism . However , conversion between the catalytic mechanisms of the proteasomal proteases and PP2C phosphatases would require multiple changes that would individually inactivate both activities ( including circular permutation of the enzyme , loss/gain of metal binding , and charge swaps of residues at essential positions for catalysis ) . The conservation of the allosteric regulatory switch suggests a possible solution to this dilemma: namely , that the intermediate was a noncatalytic pseudoenzyme that retained the allosteric regulatory switch . RssB is an example of this sort of hypothetical pseudoenzyme; RssB uses the PP2C switch to regulate protease adapter function without functioning as a phosphatase ( Battesti et al . , 2013 ) . An RssB-like intermediate would provide evolutionary pressure to preserve the regulatory mechanism , while creating a condition of neutrality to other mutations that would allow the new chemistry to evolve . Indeed , E . coli RssB lacks the C-terminal β strand of PP2C phosphatases that is substituted by the N-terminus of the proteases ( Figure 6A ) , suggesting a pathway for how a gene fusion event could produce the topological change required to evolve protease activity . Allosteric regulatory modules have facilitated the evolutionary diversification of enzyme families to respond to new regulatory inputs , and the regulatory mechanism we have described for PP2C phosphatases may have similarly facilitated phosphatase diversification . A recent investigation of the evolution of ligand specificity in PDZ domains proposed that allostery produces conformational flexibility and thus may arise as a consequence of evolutionary history ( Raman et al . , 2016 ) . Here we propose a mechanism whereby pre-existing allosteric regulatory modules such as we have identified for PP2C phosphatases facilitated the evolution of new enzymatic activities by transition through a pseudoenzyme intermediate that is pre-programmed for regulation . Pseudoenzymes are abundant ( for example , 10% of kinase family members are pseudoenzymes ) ( Leslie , 2013 ) and thus may be important for their evolutionary potential in addition to their current biological functions .
The SpoIIE457–827 construct was designed based on a putative sub-domain immediately N-terminal to the conserved PP2C phosphatase domain that we identified using HHPRED RRID:SCR_010276 ( Söding et al . , 2005 ) . This region exhibited weak similarity to several proteins including another sporulation protein SpoIIIAH . Analysis of the regulatory domain ( the newly determined portion of the structure ) using the DALI server RRID:SCR_013433 ( Holm and Rosenström , 2010 ) identified similarity to GpsL , a component of the type II secretion system , and structural alignment of the regulatory domain with SpoIIIAH matched the alignment predicted by HHPRED ( Söding et al . , 2005 ) . The SpoIIE457–827 coding sequence was inserted into pET47b vector that had been digested with XmaI and XhoI using isothermal assembly . SpoIIE amino acid residue substitutions were introduced to this construct by Quikchange site directed mutagenesis . These constructs were introduced to E . coli BL21 ( DE3 ) cells for protein expression . Cells were grown at room temperature to an OD600 of 0 . 4 , then were shifted to 14°C and expression was induced for 14–18 hr with 1 mM IPTG . Cells were harvested and pellets were resuspended in 5 ml/L of cell culture of 50 mM K•HEPES pH 8 , 200 mM NaCl , 20 mM Imidazole , 10% Glycerol , 0 . 5 mM DTT , and 1 mM PMSF . Cells were lysed using a cell disruptor in one-shot mode ( Constant Systems , Daventry , United Kingdom ) and lysates were clarified by spinning for 30 min at 16 , 000 RPM in a Sorvall SS-34 rotor at 4°C . Lysates were loaded to a HisTrap-HP column on an AKTA FPLC and eluted with a gradient of imidazole to 200 mM . The 6His tag was cleaved overnight with PreScission protease during dialysis to 50 mM K•HEPES pH 8 , 200 mM NaCl , 20 mM Imidazole , 10% Glycerol , 0 . 5 mM DTT at 4°C . The PreScission protease was removed by flowing the dialyzed protein over a Ni-NTA resin , and the flowthrough was loaded to a Resource Q column that had been pre-equilibrated in 50 mM K•HEPES pH 8 , 100 mM NaCl , 2 mM EDTA , 2 mM DTT . Protein was eluted using a gradient to 500 mM NaCl . Fractions containing SpoIIE were concentrated on Amicon Ultra centrifugal filters and loaded to a Superdex 200 column equilibrated with 20 mM K•HEPES pH 8 , 50 mM NaCl , 2 mM DTT . Fractions containing SpoIIE were concentrated and immediately used to set up crystallization trials or were flash frozen in liquid nitrogen after addition of 10% glycerol . Seleno-Methionine derivatized SpoIIE457–827 protein was grown in fully supplemented M9 media . Fifteen minutes before induction , 100 mg/L L-Phenylalanine , 50 mg/L L-Isoleucine , 100 mg/L L-Lysine , 50 mg/L L-Leucine , 100 mg/L L-Threonine , 50 mg/L L-Valine , and 60 mg/L L-Selenomethionine were added . Otherwise induction and purification were identical to the un-derivatized protein . Recombinant SpoIIE590–827 ( with an amino acid substitution A624I that was designed to block domain swapping ) was overproduced from E . coli BL21 ( DE3 ) harboring a pET-YSBLIC derivative plasmid . Cultures were grown at 37°C and induced at OD600 = 0 . 6–0 . 7 by addition of IPTG to 1 mM followed by overnight growth at 16°C . Cells were harvested and the pellets resuspended in 20 mM sodium phosphate ( pH 7 . 5 ) , 0 . 5 M NaCl , 20 mM imidazole ( Buffer A ) . The supernatant was loaded onto a HiTrap Ni-NTA column equilibrated with buffer A and eluted with a 20–500 mM imidazole gradient in buffer A . Fractions containing SpoIIE were concentrated before loading on to a Superdex S200 column equilibrated with 20 mM Tris pH 8 . 5 , 150 mM NaCl . SpoIIE457–827 crystals were grown in sitting drops using Swissci 3 well 96 well plates ( Hampton , Aliso Viejo , CA ) with 40 µl well solution ( 0 . 5 mM LiSO4 , 8% PEG8000 , 0 . 05 mM NaF , 6% glycerol ) . SpoIIE457–827 ( 11 mg/mL ) in 20 mM K•HEPES pH 8 , 50 mM NaCl , 2 mM DTT was supplemented with 0 . 05 mM NaF and mixed at a 2:1 ratio with well solution in 300 nL drops using an NT8 robot ( Formulatrix , Bedford , MA ) . Crystals grew over two weeks at room temperature . Crystals were cryoprotected by serial transfer to well solution supplemented with 10% and then 15% glycerol and plunged in liquid nitrogen . Data were collected at the Advanced Photon Source at Argonne National Laboratory on NE-CAT beamlines 24ID-C and 24ID-E . Data were processed using HKL-2000 ( Otwinowski and Minor , 1997 ) and initial phases were determined by molecular replacement using MR-PHASER RRID:SCR_014219 ( McCoy et al . , 2007 ) and an unswapped model from the published structure of SpoIIE590–827 as the search model ( Levdikov et al . , 2012 ) . Iterative model building and refinement was done in COOT RRID:SCR_014222 ( Emsley et al . , 2010 ) and refinement in PHENIX RRID:SCR_014224 ( Adams et al . , 2010 ) . Non-crystallographic symmetry was initially enforced for the five chains in the asymmetric unit , then released first for chain B and finally for all chains . In later stages of refinement NCS was again imposed on regions where the chains differed by less than 4 Å . Model restraints were used based on the structure of SpoIIE590–827 published here during an intermediate stage of refinement . The model for SpoIIE457–827 was additionally validated using anomalous signal from crystals grown with seleno-methionine derivatized protein ( Table 2 ) . With the exception of M557 , signals were observed for all methionines at the expected sites in the anomalous difference map ( an example is shown in Figure 1—figure supplement 1A ) . Crystallization experiments with SpoIIE590–827 consistently led to crystals of the domain-swapped dimeric form of the protein , even though SEC-MALLS analysis showed that SpoIIE590–827 is predominantly monomeric ( Levdikov et al . , 2012 ) . To stabilize the PP2C domain and slow down/prevent domain-swapping during crystallization , we introduced residue substitutions to reinforce the interface involved in domain-swapping . One such SpoIIE590–827 mutant , A624I , constructed by Quikchange mutagenesis ( changing the GCA codon to ATA ) , led to the crystallization of SpoIIE590–827 without domain swapping . Residues with bulkier aliphatic side-chains ( L , I , V or M ) are found at the position corresponding to A624 in many SpoIIE orthologues . Crystals of SpoIIE590–827 ( A624I ) were grown from hanging drops formed by mixing 1 µl of 38 mg/mL protein with 1 µL of 2 M sodium formate , 100 mM sodium acetate , pH 4 . 6 . The crystals were cryo-protected in mother liquor containing 4 M sodium formate for X-ray data collection on beamline I02 at the DIAMOND Light Source . Data extending to 2 . 44 Å spacing were collected and processed using HKL-2000 ( Otwinowski and Minor , 1997 ) . Initial phases were determined by molecular replacement using MOLREP ( Vagin and Teplyakov , 2010 ) , and a coordinate set derived from PDB ID 3T91 as the search model . The structure was rebuilt and refined using iterative cycles of COOT RRID:SCR_014222 ( Emsley et al . , 2010 ) and REFMAC RRID:SCR_014225 ( Murshudov et al . , 1997 ) respectively . Data collection and refinement statistics are given in Table 1 . SEC-MALLS was performed by loading 100 µL of 200 µM SpoIIE457–827 to a Wyatt WTC-030S5 column using an Agilent HPLC in line with Wyatt DAWN-HELIOS and Optilab rEX detectors . Before running SpoIIE457–827 was exchanged to 25 mM K•HEPES pH 8 , 100 mM NaCl using a Superdex 200 column . The SEC-MALLS instrument was equilibrated in 25 mM K•HEPES pH 8 , 100 mM NaCl , supplemented with MnCl2 as appropriate . SpoIIE457–827 samples were supplemented with MnCl2 shortly before running on the SEC-MALLS . Analysis was conducted using the ASTRA software . All SEC-MALLS samples were run in at least triplicate . SEC experiments shown in Figure 4—figure supplement 1 were conducted similarly , loading 200 µL of 200 µM SpoIIE457–827 on a 20 mL Superdex 200 column on an AKTA FPLC . Phosphatase assays were performed as reported in Bradshaw and Losick , 2015 . SpoIIAA , SpoIIAA-P , and SpoIIAB were produced and purified as described previously . SpoIIAA-P was produced by overexpression of 6H-SpoIIAA in an E . coli strain that also expressed SpoIIAB ( Levdikov et al . , 2012 ) . To produce 32P labeled SpoIIAA-P , 75 µM SpoIIAA , 5 µM SpoIIAB and 50 µCi of γ-32P ATP were incubated overnight in 50 mM K•HEPES pH 7 . 5 , 50 mM KCl , 750 µM MgCl2 , 2 mM DTT . The protein was exchanged to 20 mM K•HEPES pH 7 . 5 , 200 mM NaCl , 2 mM DTT over a Zeba spin column ( Pierce ) to remove unincorporated nucleotide and then flowed over Q sepharose resin to remove SpoIIAB . Phosphatase assays were performed in 25 mM K•HEPES pH 8 , 100 mM NaCl , 100 µg/ml BSA ( supplemented with MnCl2 as appropriate ) with 2 . 5 µM SpoIIE and 200 µM SpoIIAA-P . Reactions were started by adding SpoIIE to a mixture containing SpoIIAA-P and MnCl2 . Reactions were stopped in 1 M KPO4 pH 3 . 3 , 2% Triton X-100 and run on PEI-Cellulose TLC plates developed in 1 M LiCl2 , 0 . 8 M Acetic Acid , and imaged on a Typhoon ( GE Life Sciences , Pittsburgh , PA ) . Phosphatase assays were performed more than three independent times as separate experiments . B . subtilis strains were constructed using standard molecular genetic techniques ( Harwood and Cutting , 2010 ) in the PY79 strain background ( Youngman et al . , 1984; Zeigler et al . , 2008 ) and were validated to contain the correct constructs by double-crossover recombination at the correct insertion site . All strains used in this study are described in Table 3 . For imaging , cells were grown at 37°C in 25% LB to OD 0 . 6 , resuspended in minimal sporulation resuspension medium , and grown for 2 . 5 hr . Cells were immobilized on 2 . 5% agarose pads made with the sporulation resuspension medium and imaged on an Olympus BX-61 upright microscope with a 100X objective . Cells were segmented using SuperSegger ( Stylianidou et al . , 2016 ) and analyzed with custom MatLab scripts ( Bradshaw and Losick , 2015 ) . Samples were taken from the same cultures for western blot analysis; cells were lysed using a FastPrep ( MP-BIO , Santa Ana , CA ) and blots were probed with polyclonal α-GFP antibody . 10 . 7554/eLife . 26111 . 026Table 3 . Table of strains . B . subtilis strains ( all strains are in the background of PY79-RL3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26111 . 026Strain #GenotypeReferenceRL3prototrophicYoungman et al . , 1984RL5874spoIIE::kan yxiD::spoIIE-yfp spc amyE::PspoIIE-cfp cmBradshaw and Losick , 2015RL5902spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp L646K spcBradshaw and Losick , 2015RL5904spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp Q483A spcBradshaw and Losick , 2015RL5905spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp G486K spcBradshaw and Losick , 2015RL5907spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp E639K spcBradshaw and Losick , 2015RL6198spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp V480K spcthis studyRL6199spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp A481K spcthis studyRL6200spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp L484K spcthis studyRL6201spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp V487K spcthis studyRL6202spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp S488K spcthis studyRL6203spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp V490K spcthis studyRL6204spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp M491K spcthis studyRL6205spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp D493K spcthis studyRL6206spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp F494K spcthis studyRL6207spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp S495K spcthis studyRL6208spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp E497K spcthis studyRL6209spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp I498K spcthis studyRL6210spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp E642K spcthis studyRL6211spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp I650K spcthis studyRL6212spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp T663K spcthis studyRL6213spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-yfp I667K spcthis studyRL5915spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-∆tag-yfp L646K spcBradshaw and Losick , 2015RL6246spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-∆tag-yfp V480K spcthis studyRL6247spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-∆tag-yfp L484K spcthis studyRL6248spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-∆tag-yfp V487K spcthis studyRL6249spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-∆tag-yfp F494K spcthis studyRL6250spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-∆tag-yfp I498K spcthis studyRL6251spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-∆tag-yfp I650K spcthis studyRL6252spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-∆tag-yfp T663K spcthis studyRL6253spoIIE::kan yhdGH::PspoIIQ-cfp tet amyE::spoIIE-∆tag-yfp M491K spcthis studyE . coli strainsRL6214BL21 ( DE3 ) pET47b H6-3C-spoIIE 457–827This studyRL6215BL21 ( DE3 ) pET47b H6-3C-spoIIE 457–827 V697Athis studyRL6216BL21 ( DE3 ) pET47b H6-3C-spoIIE 457–827 D628Athis studyRL6217BL21 ( DE3 ) pET47b H6-3C-spoIIE 457–827 L484Kthis studyRL6218BL21 ( DE3 ) pET23a H6-sumo-spoIIAABradshaw and Losick , 2015RL6219BL21 ( DE3 ) pET23a H6-sumo-spoIIABBradshaw and Losick , 2015AW2001BL21 ( DE3 ) pET-YSBLIC H6-3C-spoIIE 590–827 A624ILevdikov et al . , 2012AW2002BL21 ( DE3 ) pET-YSBLIC H6-3C-spoIIAA spoIIABLevdikov et al . , 2012 | To regulate the activity of proteins , cells often modify them by adding or removing chemical groups called phosphates . Therefore , the enzymes that add or remove these phosphate groups must be tightly regulated so that they are active at the right time and place . Enzymes known as phosphatases remove phosphate groups from proteins and the PP2Cs are one such family of enzymes that are found in bacteria , plants and animals . Despite their broad importance , it was not clear how cells control the PP2Cs . One way to understand how an enzyme is controlled is to compare the three-dimensional structures of the enzyme when it is active and when it is inactive . Bradshaw et al . used a PP2C enzyme from bacteria as a model to understand how the cell regulates other PP2Cs . The experiments reveal that the bacterial enzyme has a structural element that acts as a switch to control its activity . The phosphatase needs to bind metal ions to be active , and movement of the switch promotes binding of the metal ions to activate the phosphatase . The switch is also found in other members of the PP2C family . Furthermore , members of a seemingly unrelated family of enzymes called the proteasomal proteases , which degrade proteins , also have a similar architecture and are controlled by a similar switch . Thus , the phosphatase and protease families may have a common evolutionary history . Multiple members of the PP2C family are involved in cancer and other diseases . The discovery of a regulatory switch provides new opportunities to use drugs to control phosphatase activity in patients . Many cancer drugs that are currently in use or are under development target enzymes that add phosphate groups to proteins , but efforts to target the phosphatases have largely been unsuccessful . Bradshaw et al . ’s findings may enable the development of new drugs that target protein phosphatases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2017 | A widespread family of serine/threonine protein phosphatases shares a common regulatory switch with proteasomal proteases |
Mechanisms enabling positional identity re-establishment are likely critical for tissue regeneration . Planarians use Wnt/beta-catenin signaling to polarize the termini of their anteroposterior axis , but little is known about how regeneration signaling restores regionalization along body or organ axes . We identify three genes expressed constitutively in overlapping body-wide transcriptional gradients that control trunk-tail positional identity in regeneration . ptk7 encodes a trunk-expressed kinase-dead Wnt co-receptor , wntP-2 encodes a posterior-expressed Wnt ligand , and ndl-3 encodes an anterior-expressed homolog of conserved FGFRL/nou-darake decoy receptors . ptk7 and wntP-2 maintain and allow appropriate regeneration of trunk tissue position independently of canonical Wnt signaling and with suppression of ndl-3 expression in the posterior . These results suggest that restoration of regional identity in regeneration involves the interpretation and re-establishment of axis-wide transcriptional gradients of signaling molecules .
Robust pattern control is a central but poorly understood feature of regenerative abilities ( Wolpert , 1969; French et al . , 1976 ) . Animals cannot anticipate how a given injury will alter tissue composition , so regeneration likely depends critically on the re-establishment of missing tissue identity . Planarians use pluripotent stem cells to regenerate from nearly any amputation to restore a complete set of regionalized tissues , including cephalic ganglia in the anterior and a pharynx and mouth in the trunk , and are a model of positional restoration after amputation ( Reddien , 2011; Adler and Sánchez Alvarado , 2015 ) . Canonical Wnt signaling controls anterior-versus-posterior pole identity in planarian regeneration , with principal upstream determinants wnt1 expressed at the posterior pole ( Petersen and Reddien , 2009; Gurley et al . , 2010 ) and the secreted Wnt inhibitor notum expressed at the anterior pole ( Petersen and Reddien , 2011 ) , both activated transcriptionally early after wounding . Inhibition of Wnt signaling components β-catenin-1 , wnt1 , Evi/wntless , Dvl-1/2 and teashirt causes regeneration of ectopic heads ( Gurley et al . , 2008; Iglesias et al . , 2008; Petersen and Reddien , 2008; Petersen and Reddien , 2009; Owen et al . , 2015; Reuter et al . , 2015 ) ; conversely , inhibition of APC or notum can cause regeneration of ectopic tails ( Gurley et al . , 2008; Petersen and Reddien , 2011 ) . Other pathways participate in head or tail regeneration , with Hedgehog signaling influencing injury-induced wnt1 expression ( Rink et al . , 2009 ) , several transcription factors required for head formation ( prep , foxD , zic-1/zicA , pbx , egr-4 ) ( Felix and Aboobaker , 2010; Blassberg et al . , 2013; Chen et al . , 2013; Fraguas et al . , 2014; Scimone et al . , 2014; Vásquez-Doorman and Petersen , 2014; Vogg et al . , 2014 ) and/or tail formation ( junli-1 , pitx , pbx ) ( Blassberg et al . , 2013; Chen et al . , 2013; Currie and Pearson , 2013; Marz et al . , 2013; Tejada-Romero et al . , 2015 ) , and influenced by gap junction and calcium signaling ( Oviedo et al . , 2010; Zhang et al . , 2011 ) . However , comparatively little is known about the restoration of positional information along the head-to-tail body axis through regeneration . Expression profiling and homology searching have identified a cohort of factors related to Wnt , Hox , and FGF signaling expressed regionally in domains along the anteroposterior ( A-P ) axis in planarians ( Cebrià et al . , 2002; Petersen and Reddien , 2008; Reddien , 2011; Owen et al . , 2015; Reuter et al . , 2015 ) . These factors could in principle form a molecular coordinate system that re-specifies axis identity in regeneration and have been termed 'positional control genes' ( PCGs ) ( Witchley et al . , 2013 ) . However , few phenotypes of positional displacement have been reported from perturbation of PCGs ( Cebrià et al . , 2002; Kobayashi et al . , 2007 ) , so it remains unclear how the majority of these genes participate in regeneration .
To uncover programs responsible for patterning along the body axis , we examined PCGs as defined by prior homology and expression profiling studies ( Petersen and Reddien , 2008; Reddien , 2011; Witchley et al . , 2013; Owen et al . , 2015; Reuter et al . , 2015 ) and found a unique expression pattern for a planarian homolog of ptk7 , expressed in an animal-wide graded fashion maximally in the trunk , and also the CNS and pharynx ( Figure 1A , Figure 1—figure supplement 1 ) . Ptk7 proteins encode cell-surface Wnt co-receptors with a predicted intracellular pseudokinase domain that participate in noncanonical , β-catenin-independent Wnt signaling ( Lu et al . , 2004 ) , and can either weakly suppress or activate canonical β-catenin-dependent Wnt signaling in a context-dependent manner ( Peradziryi et al . , 2011; Puppo et al . , 2011; Hayes et al . , 2013; Bin-Nun et al . , 2014 ) . Like other described PCGs , planarian ptk7 was expressed in collagen+ cells of the body-wall musculature ( Witchley et al . , 2013 ) and aspects of its expression domain could become re-established after amputation even in irradiated animals lacking neoblasts and the ability to form new tissues ( Figure 1A–C ) . In ptk7 ( RNAi ) animals amputated to remove both head and tail , regeneration produced a normal head and tail ( 35/35 ) but caused formation of an ectopic posterior mouth at a high penetrance ( 83% , n=35 ) and , more rarely , formation of an ectopic posterior pharynx with broadly normal orientation with respect to the primary body axis ( 14% , n=35 ) ( Figure 1D–E ) . Thus , ptk7 limits trunk identity in planarian regeneration . 10 . 7554/eLife . 12850 . 003Figure 1 . ptk7 is a positional control gene that suppresses trunk identity in regenerating planarians . ( A ) Left panel , Double FISH to detect coexpression of ptk7 within collagen+ cells of the body-wall musculature in a trunk-centered gradient ( 116/125 collagen+ cells were ptk7+ and 113/125 ptk7+ cells were collagen+ , scored in ventral prepharyngeal subepidermal region ) . Right panels , higher magnification of collagen+ptk7+ cells . ( B , upper panels ) Freshly amputated head fragments have ptk7 expression in the CNS but minimal levels in subepidermal cells but by 48–96 hr expression appears at a region within the new anterior of the fragment ( arrows , anterior extent of ptk7 expression ) . ( B , lower panels ) Animals treated with lethal doses of gamma irradiation ( 6000 Rads ) three days prior to amputation undergo a similar re-establishment of a ptk7 expression domain along the A-P axis . Images represent at least 3/3 animals probed . ( C ) Irradiation controls showing elimination of smedwi-1-expressing neoblasts in animals from the same cohort as ( B ) . ( D ) Animals were injected with control or ptk7 dsRNA three times over three days , amputated to remove heads and tails , allowed to regenerate , fixed at 14 days and stained with a laminin riboprobe detecting both the mouth and pharynx ( left panels , dotted line indicates amputation plane , red box shows enlarged region of ptk7 ( RNAi ) animals ) or stained with Hoechst dye to label nuclei and visualize the pharynx ( right ) . ptk7 RNAi caused formation of an ectopic posterior mouth in regenerating trunk fragments ( 28/35 animals ) , but not in regenerating head or tail fragments ( 35/35 animals each ) . ( D , right ) More rarely , ptk7 inhibition caused formation of an ectopic posterior pharynx . ( E ) Control or ptk7 ( RNAi ) animals stained with a fluorescent lectin Concanavalin A to visualize the epidermis and the pre-existing or ectopic mouth ( red arrow ) . Bars , 25 ( A ) , or 200 ( B ) , or 400 microns ( D–E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 00310 . 7554/eLife . 12850 . 004Figure 1—figure supplement 1 . Sequence alignment of Smed-ptk7 . Alignment of predicted protein sequence of Smed-ptk7 with human ( HsPtk7 , NP_002812 . 2 ) , mouse ( MmPtk7 , AAH76578 . 1 ) , Xenopus laevis ( XlPtk7 , NP_001083315 . 1 ) , zebrafish ( DrPtk7 , AGT63300 . 1 ) and Drosophila melanogaster ( otk , AAF58596 . 1 ) Ptk7 proteins . SMED-PTK7 is predicted to have 7 extracellular Ig-family domains and an intracellular tyrosine kinase domain ( http://smart . embl-heidelberg . de ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 00410 . 7554/eLife . 12850 . 005Figure 1—figure supplement 2 . RNAi enhancement screen identifies modulators of ptk7 activity involved in trunk patterning . ( Upper ) Plot of ectopic posterior pharynx phenotype penetrance after administration of dsRNA targeting ptk7 and each of 22 previously identified positional control genes or Wnt-related factors expressed regionally along the body axis ( wnt11-1 , wnt11-2 , wnt11-4 , wntP-2 , wnt2-1 , and wnt5 ) , general intracellular components of Wnt signaling ( frizzled receptors expressed in the posterior: fzd4-1/fzd4 , fzd4-2 , and fzd1/2/7; Disheveled proteins: Dvl-1 and Dvl-2 ) , downstream components of Wnt signaling that act positively ( teashirt ) or negatively ( axin-B ) to influence β-catenin-1 signaling , secreted Wnt inhibitors ( sFRP-1 , sFRP-2 , sFRP-3 , notum ) , Wnt-PCP pathways ( DAAM1 , ROCK , vangl2 ) , and FGFR-like/nou-darake family factors ( ndl-3 , ndl-4 ) . At least 5 animals were examined in each condition except for wntP-2 , DAAM1 , ROCK , vangl2 RNAi treatments in which at least 24 animals were examined . Asterisks indicate p < 0 . 05 from Fisher’s exact test versus control + ptk7 dsRNA treatment after Benjamini-Hochberg correction for false discovery . Inhibition of ptk7 along with either wntP-2 ( p=5 . 16E-14 ) , ndl-3 ( p=1 . 24E-05 ) , fzd-1/2/7 ( p=0 . 046 ) , or Dvl-2 ( p=4 . 18E-06 ) caused the strongest occurrence of the ectopic pharynx phenotype . ( Lower ) in situ hybridizations showing regionalized expression of selected genes from the screen . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 005 We next used RNAi and a phenotypic enhancement assay to identify other PCGs that participate with ptk7 in trunk identity regulation ( Figure 1—figure supplement 2 ) . Co-inhibition of ptk7 with either wntP-2/wnt11-5/wnt4b ( hereafter referred to as wntP-2 ) , ndl-3 , Dvl-2 ( Almuedo-Castillo et al . , 2011 ) , or fzd-1/2/7 caused the strongest enhancement of the ectopic pharynx phenotype , resulting in 100% of animals affected , whereas co-inhibition with other PCGs affected this phenotype more weakly and below statistical significance under these conditions . wntP-2 encodes a Wnt gene expressed in a graded fashion from the posterior ( Figure 1—figure supplement 2 ) and based on phylogenetic analyses has been proposed to be either a Wnt11 ( Gurley et al . , 2010 ) or Wnt4 ( Riddiford and Olson , 2011 ) family member . Planarian ndl-3 is expressed in a graded fashion from the anterior ( Rink et al . , 2009 ) ( Figure 1—figure supplement 2 ) , and encodes a member of the conserved FGFRL/nou-darake class of cell-surface molecules that possess an FGF-receptor-like extracellular domain but lacks a tyrosine kinase intracellular domain and are thus proposed to function as FGF signaling inhibitors ( Cebrià et al . , 2002; Gerber et al . , 2009 ) . fzd-1/2/7 encodes a predicted Wnt receptor expressed broadly ( Figure 1—figure supplement 2 ) . Individual inhibition of ptk7 , wntP-2 , and ndl-3 caused formation of ectopic posterior mouths ( 30/35 ptk7 ( RNAi ) animals , 22/33 wntP-2 ( RNAi ) animals , 14/23 ndl-3 ( RNAi ) animals ) , and ectopic posterior pharynges ( 5/35 ptk7 ( RNAi ) animals , 8/33 wntP-2 ( RNAi ) animals , 2/23 ndl-3 ( RNAi ) animals ) in amputated animals regenerating both their heads and tails compared to controls ( 0/46 animals ) ( Figure 2A ) . We verified knockdown of ptk7 , wntP-2 , and ndl-3 using in situ hybridizations and qPCR ( Figure 2—figure supplement 1A–B ) . Inhibition of any pairwise combinations of the three genes enhanced the trunk expansion phenotypes ( Figure 2A–B ) ( animals with ectopic pharynges: 45/55 wntP-2 ( RNAi ) ;ptk7 ( RNAi ) , 29/32 ptk7 ( RNAi ) ;ndl-3 ( RNAi ) , 24/40 wntP-2 ( RNAi ) ;ndl-3 ( RNAi ) ) , with double-RNAi animals also frequently forming two ectopic pharynges ( Figure 2A ) and co-inhibition of ptk7 and wntP-2 producing highest penetrance and expressivity of trunk duplication phenotypes . The orientation of ectopic pharynges was generally oblique , and sometimes inverted , but the majority in all conditions pointed toward the posterior rather than anterior direction ( Figure 2C ) . Live animals with ectopic pharynges were observed during feeding , and the duplicated pharynx could obtain food ( Figure 2D , Video 1 ) , suggesting normal functionality of this organ . Taken together , ptk7 , wntP-2 and ndl-3 form a core group of regionally expressed PCGs that jointly suppress trunk identity during posterior regeneration . 10 . 7554/eLife . 12850 . 006Figure 2 . ptk7 acts with wntP-2 and ndl-3 to suppress trunk identity in a context-dependent manner . ( A ) ptk7 , wntP-2 , and ndl-3 dsRNAs were fed to animals individually or in pair-wise combinations prior to amputation to remove heads and tails , fixation 25 days later , staining with a laminin riboprobe and Hoechst dye . ( B ) Scoring information for pharynx and mouth duplication phenotypes . Animals were scored for presence of ectopic mouth ( defined as a superficial circle of laminin+ cells which was always present posterior to the original mouth , arrow ) , and ectopic pharynx ( defined as having pharynx morphology by laminin+ and Hoechst+ staining , double arrows ) and its orientation with respect to the A-P body axis . Animals with an ectopic mouth but not a fully formed ectopic pharynx often had varying degrees of internal laminin expression suggestive of a growing pharynx primordium and were scored as having an ectopic mouth only . Co-inhibition of any pairwise combination of the three genes enhanced the penetrance and expressivity of the ectopic pharynx phenotypes . Note that combined pairwise inhibition of ptk7 , wntP-2 and ndl-3 enhanced the trunk duplication phenotype and that dual inhibition of ptk7 and wntP-2 produced the strongest effects . ( C ) Analysis of ectopic pharynx orientation , measured at the proximal end of the ectopic pharynx . In many cases , the ectopic pharynx was oriented at an oblique angle with respect to the body axis , perhaps as a result of ectopic mouth placement nearby the original mouth , and ectopic pharynges were observed with fully inverted polarity . In all animals that formed 2 ectopic pharynges ( derived from pairwise combinations of dsRNAs ) , both structures were oriented toward a common ectopic mouth located along the posterior midline . ( D ) Images of live animals showing the ectopic pharynx in ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animals ( red arrow ) can be functional for feeding . ( E ) Prolonged inhibition of ptk7 and wntP-2 in uninjured animals for at least 36 days ( 64 days shown ) caused formation of an ectopic pharynx ( 6/8 animals ) and multiple posterior mouths ( 8/8 animals ) . ( F ) Inhibition of ptk7 and wntP-2 or ptk7 and ndl-3 caused head and tail fragments to regenerate only a single pharynx like control animals . Therefore , the effects of ptk7 , ndl-3 and wntP-2 in body patterning are context dependent . Anterior , left . Bars , 300 ( A , E ) or 500 ( F ) microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 00610 . 7554/eLife . 12850 . 007Figure 2—figure supplement 1 . Verification of RNAi knockdown for ptk7 , wntP-2 and ndl-3 . ( A ) In situ hybridizations verifying that dsRNA to ptk7 , wntP-2 , and ndl-3 individually reduced the expression of each gene after 12 days of RNAi in the absence of injury ( ptk7 ) or 20 days regeneration after head and tail amputation ( wntP-2 and ndl-3 ) of RNAi . Images represent 100% of animals stained , n > 4 . Bars , 500 microns . ( B ) qPCR showing knockdown of ptk7 , wntP-2 and ndl-3 mRNA . Asterisks indicate p<0 . 05 by a 2-tailed t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 00710 . 7554/eLife . 12850 . 008Video 1 . Ectopic pharynges in ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animals can be functional for feeding , related to Figure 2 . Movie of a live ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animal showing both the original and ectopic pharynges in a 21 day regenerating trunk fragment feeding on liver paste . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 008 wntP-2 and ptk7 were expressed regionally in the absence of injury , so we examined whether they act only in regeneration or constitutively to regulate trunk regionalization . Prolonged co-inhibition of wntP-2 and ptk7 produced animals with an ectopic pharynx , indicating that these genes together restrict trunk identity homeostatically in the absence of injury ( Figure 2E ) . Furthermore , the fact that such animals ultimately formed multiple ectopic mouths extending toward the posterior suggests that graded activities of wntP-2 and/or ptk7 , rather than their control of a binary switch in developmental outcomes , could pattern the tail and trunk regions . We next examined whether trunk suppression mediated by wntP-2 , ptk7 , and ndl-3 was operational in all regeneration contexts , as is the case for several characterized patterning genes in planarians . By contrast , under RNAi conditions that produced an 80–100% penetrant pharynx duplication in regenerating trunk fragments , head and tail fragments from wntP-2 ( RNAi ) ;ptk7 ( RNAi ) animals or ndl-3 ( RNAi ) ;ptk7 ( RNAi ) animals formed only a single laminin+ pharynx as did control animals ( Figure 2F ) . Eventually , after prolonged dsRNA feeding after regeneration , such animals could form ectopic pharynges ( 76 days of RNAi , n=3 of 12 animals examined ) , consistent with homeostatic functioning of the three genes . However , these results indicate that ptk7 , wntP-2 and ndl-3 suppress trunk expansion in a context-dependent manner and suggest they may provide information about trunk absence or presence during regeneration . We investigated the anatomy of ptk7 ( RNAi ) ;wntP-2 ( RNAi ) and ptk7 ( RNAi ) ;ndl-3 ( RNAi ) regenerating trunk fragments to determine the extent of the axis under control of the three genes . We first examined the influence of ptk7 , wntP-2 , or ndl-3 inhibition on expression of PCGs and tissue-specific genes marking A-P axial identity ( Figure 3A–B ) . Such ptk7 ( RNAi ) ;wntP-2 ( RNAi ) or ptk7 ( RNAi ) ;ndl-3 ( RNAi ) regenerating animals had normal anterior pole and brain regions ( marked by notum , ndk ) and a normal pre-pharyngeal region anterior to the original pharynx ( marked by wnt2 and novel gene SMU15014980 ) , expansion of trunk-related peripharyngeal cells ( expressing mmp1 , FoxA , and SMU15007112 ) , a reduced domain of PCGs expressed in the posterior ( wnt11-1 , fzd-4-1 and Abd-Ba ) , and normal expression of wnt1 at the posterior pole . We performed additional examinations of the brain ( marked by chat and cintillo ) and far posterior ( marked by wnt11-2 ) in ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animals and found no apparent differences compared to control animals ( Figure 3—figure supplement 1A–B ) . Thus , ptk7 , wntP-2 , and ndl-3 normally promote anterior tail identity at the expense of the trunk and do not strongly affect head or tail formation . Both the pre-existing and ectopic pharynx were capable of regeneration after amputation by sodium azide treatment ( Adler et al . , 2014 ) , suggesting that wntP-2/ptk7 signaling acts in part to limit the size of the trunk region within the posterior rather than only functioning to position the anterior extent of newly made trunk tissue ( Figure 3C ) . We next inhibited ptk7 and wntP-2 in a regenerating sexual strain of S . mediterranea that forms reproductive organs posterior to the pharynx upon attainment of appropriate size . Such animals formed both an ectopic pharynx and ectopic reproductive organs marked by laminin and dmd-1 expression respectively ( Chong et al . , 2013 ) , indicating ptk7 and wntP-2 regulate trunk regionalization beyond only control of pharynx and mouth formation ( Figure 3—figure supplement 1C ) . 10 . 7554/eLife . 12850 . 009Figure 3 . ptk7 , wntP-2 and ndl-3 control tail-versus-trunk identity . ( A ) In situ hybridizations to detect A-P tissue regionalization in control and ptk7 ( RNAi ) ;wntP-2 ( RNAi ) and ptk7 ( RNAi ) ;ndl-3 ( RNAi ) regenerating trunk fragments fixed 21 days after head and tail amputation , marking the anterior and head region ( ndk , wnt2 ) , prepharyngeal region ( novel gene SMU15014980 ) , trunk ( novel gene SMU15007112 , mmp1 , foxA ) , posterior ( wnt11-1 , fzd4-1 , Abd-Ba ) , and the anterior and posterior poles ( wnt1 , notum ) . All panels represent 100% of at least 6 animals stained . Arrow , ectopic trunk gene expression . Brackets , decrease in size of tail domain . ( B ) Quantitation of domain size changes from experiments described in ( A ) , measured as length of domain normalized to body length . ptk7 ( RNAi ) ;wntP-2 ( RNAi ) and ptk7 ( RNAi ) ;ndl-3 ( RNAi ) regenerating trunk fragments had increased sizes of trunk domains marked by expression of mmp1 , foxA and SMU15007112 , and decreased sizes of tail domains marked by expression of wnt11-1 and fzd4-1 with little to no change to other domains . Asterisks , p<0 . 05 by 2-tailed t-test . ( C ) Both the pre-existing and ectopic pharynx in wntP-2 ( RNAi ) ;ptk7 ( RNAi ) animals regenerated ( 4/4 animals ) after amputation with brief sodium azide treatment , using FISH to mark the gut ( porcupine , green ) and mouth and pharynx ( laminin , red ) . Asterisk , pre-existing pharynx; arrows , ectopic pharynx , arrowhead , ectopic mouth . ( C ) ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animals form ectopic FoxA+ cells by day 7 of regeneration . Bars , 100 ( D ) , 200 ( A ) , or 300 ( C ) microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 00910 . 7554/eLife . 12850 . 010Figure 3—figure supplement 1 . Additional histological analysis of ptk7 ( RNAi ) ;wntP-2 ( RNAi ) and animals . ( A ) In situ hybridizations as in 2A showing normal expression of chat , mag-1 , and wntl11-2 , but reduced expression domain of fzd-4-1 in ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animals 18 days after head and tail amputation . ( B , left ) Co-inhibition of ptk7 and wntP-2 did not affect numbers of cintillo+ neurons of the lateral brain region in day 21 regenerating head fragments or trunk fragments , ( B , right ) with quantifications of cell number normalized to animal area . Anterior , top . ( C ) Control sexual strain animals regenerate to have a single pharynx ( asterisk , laminin expression ) and a penis papilla ( double arrow , dmd-1 expression ) within the trunk ( 8/8 animals ) , whereas ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animals regenerated to form an ectopic laminin+ pharynx ( arrow , 7/7 animals ) and ectopic dmd-1-expressing tissues ( 3/7 animals ) . Bars , 400 microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 010 The pharynx is formed from FoxA+ precursors derived from smedwi-1+ neoblasts . Because inhibition of trunk identity genes produced an ectopic pharynx , we reasoned this structure likely arose from FoxA+ progenitor cells . Indeed , ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animals regenerating an ectopic pharynx produced an ectopic domain of FoxA+ cells at a time ( 7 days of regeneration ) prior to appearance of a fully formed ectopic pharynx ( 14 days of regeneration ) ( Figure 3C ) . Expression domains of ptk7 ( Figure 1B ) and wntP-2 ( Petersen and Reddien , 2009; Gurley et al . , 2010 ) can be altered by amputation independently of neoblasts , so we suggest these genes likely function to regulate axis organization upstream of controlling neoblast fates . The unidirectional nature of these axis patterning phenotypes as expansion of trunk at the expense of tail identity prompted us to determine more precisely the nature of the regionalized expression of ptk7 , wntP-2 and ndl-3 mRNAs . Visual inspection of colorimetric in situ hybridizations suggested that ndl-3 , ptk7 , and wntP-2 are expressed in overlapping domains along the anteroposterior axis ( Figure 4A ) . We verified this interpretation by quantifying the staining intensity of colorimetric in situ hybridizations along a lateral region running from head to tail ( Figure 4B ) . ndl-3 staining intensity was maximal in the anterior in a region of graded ptk7 staining . ptk7 staining was maximal in the trunk region in which wntP-2 and ndl-3 staining form opposing gradients . wntP-2 staining was maximal in the posterior tail in a region of relatively less ptk7 expression . The graded expression detected in this manner could arise from regional differences in the abundance of cells that uniquely express each factor or in regulation of cells that express combinations of the three genes . To test these models , we performed triple FISH to simultaneously detect expression of all three genes in eight domains along the head-tail axis , which broadly confirmed the regionalized expression behavior of anterior/pre-pharyngeal maximal ndl-3 expression , trunk maximal ptk7 expression , and tail maximal wntP-2 expression ( Figure 4C ) . We sought to verify this trend quantitatively and segmented the images to examine cells expressing any of the three genes ( see Materials and Methods ) then determined the mean FISH intensity for each gene per cell . This demonstrated that regions of maximal expression of ndl-3 , ptk7 , and wntP-2 are comprised of cells with high expression of these genes ( Figure 4D ) . We next examined pairwise comparisons of expression of each gene across the body axis ( Figure 4E ) . This approach identified cells that co-express wntP-2 and ndl-3 ( in particular in head-to trunk proximal regions R3-R5 , Figure 4E ) , ptk7 and ndl-3 ( head-to-trunk regions R2-R4 ) , and ptk7 and wntP-2 ( trunk-to-tail regions R5-R7 ) . We explored whether this approach could identify the spatial distribution of discrete states of cells expressing all combinations of the three genes . We pooled the cell-based expression data across all regions to determine an approximate threshold to define higher versus lower expression for each gene , then assigned each measured cell into one of eight expression classes representing each combination of high versus low expression of each ptk7 , wntP-2 , and ndl-3 and determined their distribution across the head-tail axis ( Figure 4—figure supplement1A–C ) . This approach identified domains enriched for each expression state , finding a cohort with high wntP-2 expression and low ptk7 and ndl-3 expression in the far posterior , a cohort co-expressing ptk7 and wntP-2 in the anterior tail and trunk regions , a cohort only expressing ptk7 and not wntP-2 or ndl-3 centered in the trunk , cohorts of triple-positive cells and ptk7+ndl-3+ cells in the pre-pharyngeal region , and cohort of ndl-3+ cells in the anterior . These results suggest that a complexity of PCG cell expression states populate regions of the body axis and that a region of high wntP-2 and ptk7 expression exists in the anterior tail at a location where these two genes act together to suppress trunk identity . 10 . 7554/eLife . 12850 . 011Figure 4 . ndl-3 , ptk7 , and wntP-2 are expressed in a graded fashion in domains along the anteroposterior axis . ( A ) In situ hybridizations showing body-wide graded expression of ptk7 centered in the trunk , wntP-2 expression in a gradient from the posterior and ndl-3 expression in a graded fashion from the anterior . ( B ) Quantitation of colorimetric in situ hybridization staining across the body axis . 4–6 planarians stained as in ( A ) were imaged on a dissecting microscope , the images were inverted and then analyzed for position-specific staining intensity along a lateral domain depicted in the diagram ( dotted line with arrow showing directionality ) . To compare animals of different lengths , position was normalized to length of this domain and signal intensity was normalized such that the minimum and maximum values across each animal were 0 and 1 , respectively , and average intensity at each region was determined for animals stained with each probe treatment computed followed by obtaining average intensity , with bars showing standard deviations . ( C ) Triple FISH showing expression of ndl-3 ( red ) , ptk7 ( blue ) , and wntP-2 ( green ) mRNA . Panels are maximum projections from a stack of seven 1-micron thick confocal images taken at 40x along the body axis at the regions represented in the cartoon , then adjusted for brightness and contrast uniformly for each channel across the image series . m , mouth . Bars , 100 microns . ( D–E ) Quantification of FISH signal intensity for cells identified in images shown in ( C ) . 3-color images were segmented by merging all three channels to define a set of cells in each region with wntP-2 , ndl-3 and/or ptk7 expression and this mask used to measure mean FISH signal intensity for each cell . ( D ) Scatter and box plots showing expression of ndl-3 highest in the anterior , expression of ptk7 highest in the trunk and tail , and highest wntP-2 expression in the posterior . ( E ) Plots comparing pairwise FISH signal intensity between the indicated genes across eight body axis regions ( R1-R8 ) as in ( C ) . Note the existence of cells expressing both ndl-3 and wntP-2 ( R3-R5 ) , ptk7 and ndl-3 ( R3-R4 ) , and wntP-2 and ptk7 ( R5-R7 ) . Bars , 100 ( C ) or 200 ( A ) microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 01110 . 7554/eLife . 12850 . 012Figure 4—figure supplement 1 . Distribution of ptk7 , wntP-2 and ndl-3 expression states across the body axis . ( A ) Mean fluorescence intensity of cells from all regions in Figure 4C were combined and density histograms plotted to determine a cutoff ( dotted line=55 in arbitrary units ) for higher versus lower expression of ptk7 , wntP-2 , and ndl-3 . Cells within each region shown in Figure 4C were assigned membership in one of eight classes defined by high ( 'hi' ) or low ( unstated ) expression of each of the three genes ( 1: wntP-2hi only , 2: ptk7hiwntP-2hi , 3: ptk7hi only , 4: ptk7hiwntP-2hindl-3hi , 5: ptk7hindl-3hi , 6: wntP-2hindl-3hi , 7: ndl-3hi , 8: all 3 genes low ) . Class membership was plotted as a fraction of all cells measured in each region ( B ) and as a scatterplot of all cells examined ( C ) . By these criteria , several classes of cells expressing combinations of ptk7 , wntP-2 , and ndl-3 exist and are distributed in domains along the body axis . The tail tip has the highest frequency of wntP-2+ only cells , whereas anterior tail and trunk regions have a comparatively greater fraction of wntP-2+ptk7+ cells . The above analysis was repeated for a range of high/low expression cutoffs between 30 and 75 , resulting in similar the same trends , with lower thresholds resulting in fewer cells assigned as ptk7lowntP-2londl3loand more cells as ptk7hiwntP-2hindl-3hi . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 012 We examined the involvement of canonical Wnt signaling on expression of these factors , as this pathway has multiple functions in patterning the primary body axis ( Petersen and Reddien , 2009 ) . β-catenin-1 inhibition in uninjured animals severely reduced the expression of ptk7 , wntP-2 and ndl-3 ( Figure 5A ) , consistent with prior analyses of their expression requirements in regeneration ( Owen et al . , 2015; Reuter et al . , 2015 ) . We additionally inhibited APC , encoding an intracellular negative regulator of beta-catenin stability and examined regenerating animals for expression of the three trunk regulatory genes ( Figure 5—figure supplement 1 ) . Such animals regenerated anterior tails that expressed wntP-2 throughout , that expressed ptk7 strongly in a region near the amputation plane and away from the terminus , and that lacked ndl-3 expression . These results suggest that beta-catenin upregulation can be sufficient for tail axis formation in conjunction with wntP-2 and ptk7 expression . Taken together , normal levels of beta-catenin signaling are important for the normal expression of pkt7 , wntP-2 and ndl-3 . 10 . 7554/eLife . 12850 . 013Figure 5 . Trunk control genes likely signal independently of β-catenin-1 . ( A ) In situ hybridizations show reduced expression of ptk7 ( 11/11 animals ) , wntP-2 ( 6/6 animals ) , and ndl-3 ( 6/6 animals ) after 8 days ( wntP-2 , ndl-3 ) or 19 days ( ptk7 ) of β-catenin-1 RNAi in uninjured animals . ( B ) In situ hybridizations showing reduction of axin-B expression after 11 days of β-catenin-1 RNAi ( 14/14 animals ) but not after inhibition of wntP-2 and ptk7 ( 14/14 animals ) or ndl-3 and ptk7 ( 14/14 animals ) . Bars , 400 microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 01310 . 7554/eLife . 12850 . 014Figure 5—figure supplement 1 . Examining the effect of APC RNAi on expression of ptk7 , wntP-2 , and ndl-3 . APC ( RNAi ) regenerating animals formed a domain of ectopic ptk7 ( 8/8 animals ) and wntP-2 ( 9/9 animals ) expression in the anterior tail . The ectopic tail appears to have a domain expressing wntP-2 and not ptk7 at the terminus and a domain expressing both ptk7 and wntP-2 near the amputation site . Such animals form an ectopic anterior pharynx likely as the consequence of tail formation , and this expressed ndl-3 , ptk7 , and wntP-2 . Bars , 400 microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 01410 . 7554/eLife . 12850 . 015Figure 5—figure supplement 2 . fzd1/2/7 and dvl-2 inhibition causes ectopic pharynx and mouth formation in the posterior . The indicated dsRNA was delivered to animals by injection prior to amputation , 23 days of regeneration and staining for laminin expression to detect pharyngeal tissues . fzd1/2/7 inhibition ( 4/6 animals ) caused formation of an ectopic pharynx similar to ptk7 , wntP-2 and/or ndl-3 inhibition . dvl-2 dsRNA enhanced the ptk7 ectopic pharynx phenotype ( 9/9 animals ) . Bars , 300 microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 01510 . 7554/eLife . 12850 . 016Figure 5—figure supplement 3 . Testing planar cell polarity genes for involvement in trunk patterning . Regenerating trunk fragments undergoing RNAi as indicated and stained by FISH with a laminin riboprobe to detect the pharynx and mouth . Inhibition of ptk7 along with vangl1 ( 8/8 animals ) , vangl2 ( 4/4 animals ) , DAAM1 ( 6/6 animals ) and ROCK ( 6/6 animals ) did not suppress or enhance defects due to ptk7 inhibition alone . Bars , 300 microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 01610 . 7554/eLife . 12850 . 017Figure 5—figure supplement 4 . ptk7 , wntP-2 , and ndl-3 inhibition do not influence axin-B expression and are not modified by APC inhibition . ( A ) qPCR detecting expression of axin-B normalized to ubiquilin on RNA purified from day 10 regenerating animals after the indicated dsRNA treatments . β-catenin-1 inhibition reduced relative axin-B mRNA abundance , but ptk7+wntP-2 dsRNA and ptk7 + ndl-3 dsRNA had no effect . ( B ) ptk7+ wntP-2 RNAi did not alter expression of the β-catenin-1 target gene teashirt in animals 18 days after head and tail amputation ( 3/3 animals ) . ( C ) APC inhibition did not detectably affect the frequency of ectopic pharynx formation due to ptk7 and wntP-2 RNAi ( p=1 . 000 , Fisher’s exact test ) ( animals that regenerated an ectopic pharynx: 0/10 control fragments , 0/10 APC ( RNAi ) fragments , 7/9 ptk7 ( RNAi ) ;wntP-2 ( RNAi ) ;control ( RNAi ) fragments and 7/10 ptk7 ( RNAi ) ;wntP-2 ( RNAi ) ;APC ( RNAi ) fragments ) . Bars , 200 microns . Anterior , top . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 017 We next examined candidates for signaling that could occur downstream of wntP-2 , ptk7 , and ndl-3 . Ptk7 proteins can signal through several pathways , including as a coreceptor for Wnt/Frizzled signaling ( Lhoumeau et al . , 2011 ) . The results of our ptk7 RNAi enhancement screen suggested a candidate molecular pathway in which wntP-2 signals through ptk7 and fzd1/2/7 to suppress trunk identity . We verified that RNAi of fzd1/2/7 alone resulted in ectopic mouth and pharynx formation similar to ptk7 and/or wntP-2 inhibition ( Figure 5—figure supplement 2A ) . Similarly , we verified that Dvl-2 but not Dvl-1 could interact with ptk7 genetically , suggesting the involvement of Dvl-2 in trunk suppression ( Figure 5—figure supplement 2B ) . Ptk7 proteins can act in planar cell polarization so we tested for possible functional interactions between ptk7 and components of the Planar Cell Polarity pathway vangl1 , vangl2 , DAAM1 and ROCK but inhibition of these genes did not increase or decrease the occurrence of ectopic mouth or pharynx phenotypes ( Figure 5—figure supplement 3 ) . We additionally tested whether a downstream step in trunk suppression could be promotion or inhibition of beta-catenin activity . However , co-inhibition of ptk7 and wntP-2 or ptk7 and ndl-3 had no detectable effect on expression of beta-catenin-dependent transcripts axin-B and teashirt ( Figure 5B , Figure 5—figure supplement 4A–B ) ( Owen et al . , 2015; Reuter et al . , 2015 ) . β-catenin-1 ( RNAi ) animals lose trunk regional identity and their pharynx while becoming completely anteriorized ( Iglesias et al . , 2008 ) , so we tested for possible functional interactions between beta-catenin signaling and trunk control genes by inhibiting ptk7/wntP-2 along with APC . We could not detect any enhancement or suppression of ectopic pharynx formation in that assay , further suggesting independence between ptk7/wntP-2 and APC/β-catenin-1 signaling ( Figure 5—figure supplement 4C ) . These observations are consistent with the clear distinction between the β-catenin-1 RNAi phenotype of ectopic head production ( Gurley et al . , 2008; Iglesias et al . , 2008; Petersen and Reddien , 2008 ) as compared to the ptk7 ( RNAi ) ;wntP-2 ( RNAi ) and ptk7 ( RNAi ) ;ndl-3 ( RNAi ) phenotypes of ectopic trunk formation without affecting the identity of the anterior and posterior poles . Together , these results strongly suggest that trunk regionalization can be separable from pole identity and that ptk7 and wntP-2 likely do not operate exclusively through β-catenin-1 to pattern the trunk and tail . Regeneration can involve the re-definition of positional identity within pre-existing tissues , but the mechanisms controlling this process are unclear . We examined the expression and activities of ptk7 , wntP-2 , and ndl-3 in forming a mouth and pharynx ( expressing laminin ) within the pre-existing tail tissue ( Figure 6B ) . In amputated tail fragments , the mouth and pharynx were formed during the first 5 days of regeneration , with expression of laminin evident as early as 3 days . wntP-2 was initially expressed throughout the amputated tail , but its expression restricted posteriorly starting around day 2 , reaching a minimum around day 4 , and re-establishing to intact proportions around day 7 ( Petersen and Reddien , 2009; Gurley et al . , 2010 ) . ndl-3 expression was initially absent in the tail fragments , emerged at day 1 in the far anterior then spread posterior to occupy the anterior half of the regenerating fragments by 4 days . ptk7 expression remained broad in tail fragments through early times in regeneration and re-established a trunk-proximal domain evident by 7 days . Notably , the position of the newly regenerated pharynx correlated with a domain in which wntP-2 expression was reduced and ndl-3 was elevated . 10 . 7554/eLife . 12850 . 018Figure 6 . ptk7 acts with wntP-2 and ndl-3 to specify trunk position in regeneration . ( A ) Cartoon shows regions of trunk control gene expression and in uninjured animals . In situ hybridizations of regenerating tail fragments showing that pharynx formation ( marked by laminin expression ) coincides with early reduction of wntP-2 expression and increase in ndl-3 expression . ptk7 is expressed broadly and re-establishes a trunk-centered gradient by 7 days . All images represent at least 4/4 animals probed except laminin ( d0 ) and ptk7 ( d0 , d3 and d4 ) representing 3/3 animals probed . ( B ) ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animals form a single pharynx located too far posteriorly ( 7/8 animals , graph shows average distance between posterior pole and laminin expression domain normalized to animal length as in Figure 6—figure supplement 1A–B , error bars are standard deviations and asterisks shows p<0 . 05 by a 2-tailed t-test . ( C ) wntP-2 expression is reduced in ndl-3 ( RNAi ) regenerating tail fragments ( 10/14 animals ) . ( D-E ) ndl-3 expression is expanded posteriorly in ( D ) ptk7 ( RNAi ) ;wntP-2 ( RNAi ) regenerating tail fragments by 7 days after amputation ( 4/5 animals ) and in ( E ) intact animals after 12 or 17 days of RNAi ( 25/28 animals ) . ( F ) Control or ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animals were stained for ndl-3 and collagen expression after 17 days of RNAi and optical sections were imaged of the body-wall musculature in the region posterior to the pre-existing pharynx . Simultaneous inhibition of ptk7 and wntP-2 increased the frequency of ndl-3+collagen+ cells versus total ndl-3+ cells found in the tail region ( 41 of 63 cells in ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animals versus 7 of 32 cells scored in control animals , p<0 . 0001 by Fisher’s exact test ) . Bars , 200 ( A–D ) or 400 ( E ) microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 01810 . 7554/eLife . 12850 . 019Figure 6—figure supplement 1 . ptk7 , wntP-2 and ndl-3 participate in positioning the pharynx during tail fragment regeneration . ( A ) Assay design used to measure pharynx position in ( B–C ) . Figure 4B ( B ) Quantification of pharynx placement phenotypes in 7 day regenerating tail fragments after inhibition of wntP-2 , ptk7 and ndl-3 singly or in pairwise combinations as indicated . The strongest effects were observed after simultaneous inhibition of ptk7 and wntP-2 or ptk7 and ndl-3 . Bars , averages; asterisks indicate p<0 . 05 by 2-tailed t-test . ( C ) Dilution of ptk7+wntP-2 dsRNA with control dsRNA caused a progressively weaker pharynx placement phenotype , suggesting a gradation of their activities affect trunk position . ( D , left ) ptk7 ( RNAi ) ;wntP-2 ( RNAi ) animals stained with riboprobes to several regionally expressed genes shows such animals form normal anterior and prepharyngeal regions , but have posteriorly shifted trunk and tail gene expression domains quantified in ( D , right ) . Graphs show averages of at least 7 samples , error bars show standard deviations , and asterisks indicate p<0 . 05 by two-tailed t-test . Bars , 200 microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 01910 . 7554/eLife . 12850 . 020Figure 6—figure supplement 2 . Determining critical period for ptk7/wntP-2 signaling in pharynx placement . ( A ) Measurements of pharynx position as in Figure 4—figure supplement 1A at selected days after amputation in regenerating control or ptk7 ( RNAi ) regenerating tail fragments treated with dsRNA for three days prior to amputation or ( B ) injected with control or both wntP-2 and ptk7 dsRNA only at the indicated days with respect to the day of amputation ( day 0 ) . Injection of wntP-2+ptk7 dsRNA as late as 1–2 days after amputation can posteriorize the position of the regenerated pharynx . Graphs show averages of at least 4 samples , error bars show standard deviations , and asterisks indicate p<0 . 05 by two-tailed t-test . Bars , 200 microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 02010 . 7554/eLife . 12850 . 021Figure 6—figure supplement 3 . Examining the influence of ptk7 , wntP-2 and ndl-3 on each other’s expression . ( A–C ) Examining intact animals for regulatory interactions among trunk control genes . Uninjured animals fed with the indicated dsRNAs for two weeks prior to fixation and staining by ( A ) WISH or ( B ) FISH to examine effects on expression of ptk7 , wntP-2 or ndl-3 . No treatment caused strongly increased or reduced expression except for ptk7+wntP-2 RNAi which caused ectopic expression of ndl-3 in the anterior tail region ( A , 5/8 animals; B , 12/15 animals ) . ( C ) Animals were treated with dsRNA as in A and B , then analyzed by qPCR for expression changes to ptk7 , wntP-2 , and ndl-3 following RNAi . Inhibition of each gene caused similar knockdown in either single or double-RNAi contexts . wntP-2 RNAi caused a weak but statistically significant reduction to ptk7 transcript levels , similar to WISH detection ( A ) . fzd1/2/7 RNAi did not significantly alter expression of ptk7 , wntP-2 , or ndl-3 . Bars , 400 microns ( A ) , 300 microns ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 02110 . 7554/eLife . 12850 . 022Figure 6—figure supplement 4 . Measurements of the influence of trunk control genes on wntP-2 and ndl-3 expression in tail fragment regeneration . ( A–B ) Animals were treated with the indicated dsRNAs for three days prior to amputation , regenerating tail fragments were fixed 7 days later , and stained for ( A ) wntP-2 expression or ( B ) ndl-3 expression . Measurements were made from the posterior tip of the regenerating animal to ( A ) the anterior edge of wntP-2 expression or ( B ) the posterior edge of ndl-3 expression then normalized to animal length . Graphs show averages and error bars are standard deviations . Asterisks indicate p<0 . 05 by a 2-tailed t-test . Bars , 200 microns . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 022 We next examined the functions of ptk7 , wntP-2 and ndl-3 in positional information control within regenerating tail fragments . In tail fragments , inhibition of ptk7 caused a posterior shift to the location of the mouth and pharynx ( Figure 6A , Figure 6—figure supplement 1A ) , an effect enhanced by co-inhibition of wntP-2 or ndl-3 though not caused by wntP-2 or ndl-3 inhibition alone ( Figure 6—figure supplement 1B ) . Reduced doses of wntP-2 and ptk7 dsRNA resulted in intermediate placement phenotypes , and the distributions of placement phenotypes were not biphasic ( Figure 6—figure supplement 1C ) , suggesting the activity of these genes could regulate pharynx position in a graded fashion rather than controlling a switch between two alternate organ locations . ptk7 ( RNAi ) ;wntP-2 ( RNAi ) tail fragments regenerated with posteriorly restricted expression of fzd4-1 and wnt11-1 and posteriorly expanded expression of sFRP-2 , without strongly affecting the positioning of the brain ( ndk ) or pre-pharyngeal regions ( SMU15014980 ) ( Figure 6—figure supplement 2D ) . Together , these experiments suggest that ptk7 , along with wntP-2 and ndl-3 , has a primary activity in controlling the positioning of trunk/tail tissues in regeneration . To examine the relationship between the expression and function of ptk7 and wntP-2 signaling in regional identity re-establishment , we analyzed the time of emergence and critical time for pharynx placement phenotypes in regenerating tail fragments . RNAi of ptk7 for three days prior to tail amputation resulted in a posterior shift in the location of the newly formed laminin expression domain detected as early as 3 days of regeneration ( Figure 6—figure supplement 2A ) . These results indicate that ptk7 acts before day 3 of regeneration . To determine a lower bound for the latest time of action for trunk control genes , we performed timed delivery of dsRNA via injections performed at successive two-day intervals during regeneration then examined the position of the laminin expression domain 7 days after amputation . Injections of wntP-2 and ptk7 dsRNA only prior to amputation , or as late as day 1 and day 2 , were capable of posteriorly shifting the laminin expression domain ( Figure 6—figure supplement 2B ) . Taken together , these results suggest that ptk7/wntP-2 signaling has functions in pharynx positioning after day 1 and before day 3 , coinciding with the timing of wntP-2’s posterior restriction and ndl-3’s anterior expression . This occurs prior to the ultimate reestablishment of ptk7 , wntP-2 and ndl-3 expression into finalized trunk , posterior , and anterior domains by 7–14 days , suggesting a separation between processes that control initial organ placement and ultimate proportion restoration ( Figure 6A ) . The coordinate regulation of wntP-2 and ndl-3 in regeneration led us to examine candidate transcriptional interactions among wntP-2 , ndl-3 , and ptk7 . We did not detect strongly apparent changes to these expression domains after single inhibition of the other two genes by WISH , FISH or qPCR ( Figure 6—figure supplement 3A–C ) . We tested for possible mutual expression requirements in tail fragments , reasoning this might provide a sensitized context in which the expression domains normally undergo regulation . wntP-2 expression was reduced in ndl-3 ( RNAi ) tail fragments , particularly along the lateral edges ( Figure 6C ) , and posteriorly restricted in ptk7 ( RNAi ) regenerating tail fragments ( Figure 6—figure supplement 4A ) , suggesting that these genes can normally promote expression of wntP-2 in regenerating tail fragments . By contrast , ndl-3 expression was expanded posteriorly in ptk7 ( RNAi ) and ptk7 ( RNAi ) ;wntP-2 ( RNAi ) tail fragments ( Figure 6D , Figure 6—figure supplement 4B ) . We inhibited ptk7 and wntP-2 in uninjured animals to determine whether this effect was specific for regenerating tail fragments . Such animals expressed ndl-3 ectopically posterior to the pharynx at a time ( by 12 days of homeostatic RNAi ) prior to significant pharynx formation ( Figure 6E–F , Figure 6—figure supplement 3A–B ) . Analysis of these ectopic ndl-3+ cells by double-FISH revealed that the majority ( 41/63 ) co-expressed collagen and were located within the plane of the body-wall musculature ( Figure 6F ) . These observations argue for a specific regulatory relationship in which ptk7 and wntP-2 together suppress expression of ndl-3 in the posterior . Alternatively , the nascent pharynx could exert influence over the expression of ndl-3 in a manner indirectly controlled by ptk7 and wntP-2’s suppression of pharynx identity . Together these experiments indicate ptk7 , wntP-2 and ndl-3 can directly or indirectly engage in regulatory interactions in homeostatic maintenance of tissue pattern and also in re-establishment of pattern in regeneration .
These experiments suggest a molecular model in which trunk identity and axis position are determined by low wntP-2 activity signaling through the co-receptor ptk7 and receptor fzd1/2/7 , which together could provide competence for beta-catenin-independent outputs necessary for trunk and anterior tail patterning ( Figure 7A ) . ndl-3 is expressed in the trunk region yet acts with the same sign as wntP-2 and ptk7 to suppress posterior trunk expansion , either through a parallel or downstream process engaged in trunk suppression ( Figure 7A ) or perhaps due to its ability under some circumstances to promote robust expression of wntP-2 ( Figure 6C ) . In the tail region of uninjured animals and regenerating trunk fragments , high levels of wntP-2/ptk7 activity prevent the acquisition of trunk identity and enable the formation of anterior tail tissue ( Figure 7B ) . By contrast , head fragments initially have lower expression of wntP-2 and ptk7 , which could facilitate their formation of trunk tissue through regeneration . Amputated tail fragments would be expected to initially possess high levels of wntP-2 and ptk7 activity , but a regeneration expression regulatory program reduces wntP-2 mRNA from the anterior to allow trunk regionalization to occur at an appropriate location . These models suggest that the expression status of wntP-2 and ptk7 could provide information about the presence or absence of pre-existing tissues used in determining regeneration outcomes . A previously proposed animal-wide gradient of β-catenin-1 activity ( Reuter et al . , 2015 ) could set up the axis into distinct ptk7 , wntP-2 and ndl-3 expression domains refined by wntP-2 and ptk7 repression of ndl-3 expression . However , our experiments argue that a β-catenin-1-independent signaling output downstream of wntP-2 and ptk7 likely act to suppress trunk regional identity and thereby control placement of a trunk/tail boundary along the axis according to a gradation of their activities within the posterior . The identification of trunk expansion phenotypes independent of head/tail identity transformations suggests that whole-body regeneration involves a regulatory hierarchy of anterior/posterior pole formation followed by subsequent regional subdivision . 10 . 7554/eLife . 12850 . 023Figure 7 . Model for wntP-2 , ptk7 and ndl-3 in control of patterning . ( A ) A candidate molecular pathway of action in which wntP-2 signals through ptk7 and fzd1/2/7 and Dvl-2 to suppress trunk identity within the anterior tail region . The FGFRL ndl-3 acts with the same sign as these components to suppress trunk regionalization and could act in a parallel pathway or modify the activity of the pathway through an unknown mechanism . wntP-2 and ptk7 can inhibit ndl-3 expression in the posterior of regenerating tail fragments and intact animals and ndl-3 promotes wntP-2 expression in regenerating tail fragments , suggesting the potential for feedback regulation within this pathway ( not shown ) . β-catenin-1 signaling is required upstream for expression of ptk7 , wntP-2 and ndl-3 ( not shown ) . ( B ) Model relating expression of pathway components to patterning functions . The highest region of expression co-expression of wntP-2 and ptk7 occurs in the anterior tail and trunk at a location where these genes prevent trunk fates in animals undergoing tissue homeostatic maintenance and in regenerating trunk fragments that form new tail tissues . By contrast , regenerating head fragments lack abundant co-expression of wntP-2 and ptk7 , which we suggest is important for enabling the normal formation of trunk regional identity and regeneration of associated structures . Regenerating tail fragments would initially possess high levels of wntP-2 and ptk7 predicted to suppress trunk identity , but undergo a regeneration expression regulatory program that reduces wntP-2 mRNA in this region , enabling trunk regeneration at a position that we suggest could be defined by a particular A-P location of ptk7 and wntP-2 activity present at an appropriate time in regeneration . According to this model , wntP-2/ptk7 signaling provides information about the presence/absence of the trunk region used to control regeneration outcomes . DOI: http://dx . doi . org/10 . 7554/eLife . 12850 . 023 Our results thus establish a link between Wnt , Ptk7 and FGFRL proteins in regeneration patterning and axis formation . In mice and zebrafish , Ptk7 deletion causes defects in axis formation including a lack of convergent extension within the trunk and tail and a mispolarized auditory epithelium in mice , similar to disruption of core planar cell polarity components that signal independently of beta-catenin ( Lu et al . , 2004; Hayes et al . , 2013 ) . However , studies in Drosophila , zebrafish , Xenopus and mammalian tissue culture have found conflicting evidence that Ptk7 can also either promote ( Puppo et al . , 2011; Bin-Nun et al . , 2014 ) or inhibit ( Peradziryi et al . , 2011; Hayes et al . , 2013 ) beta-catenin-dependent signaling in a context-dependent manner . The regional identity defects we observe after ptk7 RNAi in planarians are not obviously the result of defective planar cell orientations , are phenocopied by inhibition of a Wnt gene , and do not globally affect beta-catenin transcriptional targets or beta-catenin-dependent processes . We suggest that Ptk7 proteins can control tissue fate through an alternate mechanism , perhaps by coordinating the activities of cell cohorts within a field . Planarians and most other animals have expression of multiple Wnts in posterior domains , pointing to their ancient use in organizing the primary body axis ( Petersen and Reddien , 2009 ) . The use of Ptk7 proteins for trunk/tail regionalization could therefore have an ancient origin and allow posterior Wnts to produce distinct signaling outcomes for combinatorial pattern control . We also find with wntP-2 and ndl-3 a second example in planarians of shared patterning regulation between Wnt and FGFRL factors . Whereas wntP-2/wnt11-5/wnt4b and ndl-3 restrict the domain of the trunk , wntA/wnt11-6/wnt4a ( Kobayashi et al . , 2007 ) and nou darake ( Cebrià et al . , 2002 ) restrict the domain of the head and brain . Intriguingly , in both cases , the FGFRL genes have most prominent expression in the anterior but act with the same sign as the Wnt genes to promote more posterior identity outside of the domain of their own expression . In mammals , FGFRL1 and Wnt4 are each required for formation of the metanephric kidney from intermediate mesoderm , suggesting this positive regulatory relationship is conserved ( Stark et al . , 1994; Gerber et al . , 2009 ) . Pattern restoration in regeneration has been proposed to involve intercalation ( French et al . , 1976 ) or progressive specification of adjacent regions ( Roensch et al . , 2013 ) to restore positional values across a field disrupted by amputation . In planarians , asymmetric wound-induced expression of notum provides information about wound site directionality used to program anterior-versus-posterior pole fates ( Petersen and Reddien , 2011 ) . By contrast , we find genes encoding signaling molecules that are constitutively expressed in body-wide mRNA gradients and are used for control of positional information in regeneration . Graded expression of paracrine factors across fields of cells could enable patterning over the large length-scales necessary for adult regeneration . The interactions between this tissue-wide positional information present at the time of injury , combined with wound-induced directional cues , could account for robust pattern control in regeneration . Note added in proof: while this manuscript was under review , Sureda-Gómez et al . reported that wnt11-5 RNAi causes a pharynx duplication phenotype ( Sureda-Gómez et al . , 2015 ) .
Asexual and sexual strain Schmidtea mediterranea were maintained in 1x Montjuic salts between 18–20°C . Gamma irradiation ( 6000 Rads ) was performed with a Cesium-137 source irradiator at least 24 hr prior to amputation to eliminate all dividing cells . Smed-protein tyrosine kinase-7 ( ptk7 ) was identified through homology searches through BLAST on a planarian transcriptome database ( Planmine , http://planmine . mpi-cbg . de ) identifying a Schmidtea mediterranea ptk7 homolog dd_Smed_v6_6999_0_1 . Smed-wntP-2 was described previously and cloned using primers 5’-TTAAATGTTCTAAGCCAAAACAACA-3’ and 5’-AAAACTTTTATGATCAATCTGAATGC-3’ ( NCBI accession number: EU29663 ) ( Petersen and Reddien , 2009 ) . Smed-ndl-3 was cloned using primers 5’-TTATTGACAGTAGGAACCAAAGCC-3’ and 5’-ATCCTGAATCAAGTCAACGCCA-3’ for dsRNA and riboprobe synthesis as described ( Rink et al . , 2009 ) . Unless otherwise noted , riboprobes for a 4313-bp fragment of ptk7 were made from a PCR product cloned by RT-PCR into pGEM-T-easy vector using the primers 5’-GTACTACCTGCCGAAAGTATACA-3’ , 5’-GCGCATATTCTATTGTGTAACGC-3’ . In Figure S7 , ptk7 riboprobe was synthesized from a 2068-bp fragment using the primers 5’-CGACTGTTAGTTGGTTTATGGAC-3’ , 5’-ACTTGCCTTCTCTTTGAGCG-3’ . SMU15014980 and SMU15007112 ( Robb et al . , 2015 ) ( http://smedgd . stowers . org ) expression patterns were identified by in situ hybridization screening from genes defined as BPKG22168 and BPKG1900 by prior RNA-seq studies ( Labbé et al . , 2012 ) . SMU15014980 was cloned using the primers 5’-GGATGCTTTTGCATTTTGCT -3’ and 5’-ATTGGCAAGAAAGCCATGAG -3’ . SMU15007112 was cloned using the primers 5’-CCCCGTGTGGATATTTCAGT -3’ , 5’-AGCAAAATCGGTTCTCCGTA -3’ . For inhibition of ptk7 , dsRNA was synthesized from a 1412-bp cDNA fragment cloned by RT-PCR into pGEM vector using the primers 5’-TGCTGGAAATAGTCTGTTGCAT-3’ , 5’-AAGATGGAACCCCAATAGCC-3’ . Control dsRNA was generated from a 1527-bp fragment of Photinus pyralis luciferase from the pGL3-control vector ( Promega , Fitchburg , WI USA ) ( primers 5’-TATCCGCTGGAAGATGGAAC-3′ , 5′-CGGTACTTCGTCCACAAACA- 3′ ) . wntP-2 ( Petersen and Reddien , 2009 ) , ndl-3 ( Rink et al . , 2009 ) , laminin ( Adler et al . , 2014 ) , porcupine ( Rink et al . , 2009 ) , dmd-1 , FoxA ( Adler et al . , 2014 ) , smedwi-1 , ndk , wnt1 ( Petersen and Reddien , 2009 ) , collagen ( Witchley et al . , 2013 ) , chat , mag-1 , cintillo , β-catenin-1 ( Petersen and Reddien , 2008 ) , APC ( Gurley et al . , 2008 ) , wnt11-1 ( Petersen and Reddien , 2008 ) , wnt11-2 ( Petersen and Reddien , 2008 ) , wnt11-4/wntP-3 ( Petersen and Reddien , 2008 ) , wnt2 ( Petersen and Reddien , 2008 ) , wnt5 ( Gurley et al . , 2010 ) , fzd4-1 ( Petersen and Reddien , 2008 ) , fzd4-2 , fzd-1/2/7 , Dvl-1 ( Gurley et al . , 2008 ) , Dvl-2 ( Gurley et al . , 2008 ) , axin-B , teashirt ( Owen et al . , 2015 ) , notum ( Petersen and Reddien , 2011 ) , sfrp-1 ( Gurley et al . , 2010 ) , sfrp-2 ( Gurley et al . , 2010 ) , sfrp-3 ( Gurley et al . , 2010 ) , DAAM1 ( Beane et al . , 2012 ) , ROCK ( Beane et al . , 2012 ) , vangl1 , vangl2 ( Almuedo-Castillo et al . , 2011; Beane et al . , 2012 ) , and ndl-4 ( Rink et al . , 2009 ) riboprobes and dsRNAs were described previously . Animals were fixed and stained as previously described ( Pearson et al . , 2009; King and Newmark , 2013 ) . In brief , animals were killed in 5% N-acetyl-cysteine in 1xPBS for 5 min and then fixed in formaldehyde for 20 min at room temperature . Subsequently , animals were bleached overnight ( ~16 hr ) in 6% hydrogen peroxide in methanol on a light box . Digoxigenin- or fluorescein-labeled riboprobes were synthesized as described previously ( Pearson et al . , 2009 ) . Colorimetric ( NBT/BCIP ) or fluorescence in situ hybridizations were performed as previously described ( Pearson et al . , 2009; King and Newmark , 2013 ) . For FISH , blocking solution was MABT with 10% horse serum and 10% western blot blocking reagent ( Roche ) ( King and Newmark , 2013 ) . Riboprobes were detected using anti-Dig-HRP ( 1:2000 ) , anti-FL-HRP ( 1:2000 ) , anti-DNP ( 1:2000 ) , or anti-Dig-AP ( 1:4000 ) antibodies . Hoechst 33 , 342 ( Invitrogen ) was used at 1:500 as counterstain . Images of colorimetric staining were acquired using a Leica M210F scope with a Leica DFC295 camera and adjusted for brightness and contrast . Fluorescence imaging was performed on a Leica DM5500B compound microscope with Optigrid structured illumination system for optical sectioning or a Leica SPE confocal microscope . Images are maximum projections of a z-series with adjustments to brightness and contrast using Photoshop or ImageJ . Concanavalin A conjugated to AlexaFluor 488 ( Invitrogen ) was used to stain the epidermis as described ( Zayas et al . , 2010 ) . For Figure 4B , animals stained with NBT/BCIP for detection of ptk7 , wntP-2 , and ndl-3 were visualized with a Leica M210F dissecting microscope . Images were inverted and an intensity profile obtained from a segmented line drawn along a lateral region running from the anterior to posterior of the animal with a width approximately 1/6 of the animal width using ImageJ ( 'plot profile' ) . To make comparisons of intensity plots across animals of different sizes , position along the segmented line was normalized to its length . Background , taken as the minimum intensity along this profile , was subtracted , and the resulting values were normalized to the maximum intensity along this profile . This axis was divided into 100 equal sized bins to allow comparing intensity measurements across animals of different sizes , and the average normalized intensity was determined for each bin . These plots were compared for 4–6 animals stained with each riboprobe to generate an average and standard deviation of position- and global intensity-normalized in situ hybridization signal and plotted in Figure 4B . In Figure 4C , Triple FISH was used to simultaneously detect expression of ptk7 , wntP-2 , and ndl-3 by obtaining maximum projections of 1-micron thick confocal images taken at 40x at 8 regions . The resulting three-color images were processed in ImageJ to convert to a merged grayscale image used for automated threshholding ( 'Auto Threshhold' Li's Minimum Cross Entropy method [Li , 1998] ) and segmentation ( 'Analyze particles' ) using empirically optimized parameters ( size=0 . 06–0 . 40 inch^2 in the image , corresponding to 24–166 microns^2 in the sample , and circularity 0–1 . 00 ) . The resulting segmented areas were manually inspected to verify that the majority surrounded individual cells . Mean intensity per segmented area was determined for each channel and plotted using R ( ggplot2 ) in Figure 4D and Figure 4E . In Figure 4—figure supplement 1A , density histograms were plotted to determine a cutoff for estimating high and low expression of each gene within each analyzed cell and used to classify cells as expression positive or negative . This assigned each cell to one of 8 classes plotted as a fraction of total cells in each region in a histogram in Figure 4—figure supplement 1B and as a scatterplot in Figure 4—figure supplement 1C . Similar trends were observed when threshold was taken to be any of 10 different threshold values between intensities of 30 and 75 . RNAi treatments were performed either by dsRNA injection or feeding . dsRNA was synthesized as described previously ( Petersen and Reddien , 2011 ) . Unless otherwise noted , RNAi by injection was performed using a Drummond microinjector to deliver 5 x 32 nL dsRNA on three consecutive days , followed by transverse amputations and regeneration for the indicated number of days ( Figures 1D , E , Figure 1—figure supplement 2 , Figure 2D , E , Figure 3A , D , Figure 3—figure supplement 1A , B , Figures 5—figure supplement 1–3 , Figure 6A–D , Figure 6—figure supplement 1 , 2A , 4 ) . For experiments involving RNAi by feeding , animals were given a mixture of liver paste and in vitro transcribed dsRNA , as described ( Rouhana et al . , 2013 ) . In brief , animals were fed every 2–3 days for either 1 week ( Figure 2—figure supplement 1 , Figure 5 , Figure 5—figure supplement 4 , Figure 6E , F ) or 2 weeks ( Figure 2A , Figure 6—figure supplement 3 ) and were maintained homeostatically or allowed to regenerate for the indicated number of days prior to fixation . For long-term RNAi treatment in the absence of injury , animals were fed a mixture of liver paste and dsRNA every 2–3 days for 2 weeks , followed by one dsRNA feeding every subsequent week ( Figure 2F ) . For RNAi treatment in sexual S . mediterranea , animals were fed dsRNA every 2–3 days for one week and amputated transversely to create trunk fragments containing the pharynx and reproductive structures including the copulatory apparatus then fed dsRNA once per week for 62 days ( Figure 3—figure supplement 1C ) . For timed dsRNA delivery experiments ( Figure 6—figure supplement 2B ) , RNAi-treated animals on days -2 , -1 , 0 were injected with dsRNA on three consecutive days , amputated transversely to create tail fragments on day 0 , and animals fixed on day 7 of regeneration . For subsequent timed delivery of dsRNA , animals were amputated transversely to create tail fragments on day 0 , followed by dsRNA injection for two consecutive days as indicated , and fixed on day 7 of regeneration . Animals were fed an equal mixture of ptk7 and wntP-2 dsRNA every 2–3 days for one week , amputated to remove heads and tails and allowed to regenerate for 20 days to produce an ectopic pharynx , then both the pre-existing and ectopic pharynges were removed by treatment with 100 mM sodium azide as described previously ( Adler et al . , 2014 ) . Animals were then fixed either 2 days or 19 days later and stained with laminin and porcupine riboprobes to assess regeneration of the ectopic pharynx ( Figure 3C ) . Total RNA was extracted by mechanical homogenization in Trizol ( Invitrogen ) from three RNAi-treated intact animals , and purified in three biological replicates for each treatment . RNA samples were DNased-treated ( TURBO DNase , Ambion ) and cDNA was synthesized using SuperScript III reverse transcriptase ( Invitrogen ) . qPCR was performed using SYBR Green PCR Master Mix ( Applied Biosystems ) . axin-B mRNA was detected using the primers 5’-TTCCAGTTCAGGTCACATCG-3’ and 5’-CATTGACACCTTCCGAACCT-3’ , ubiquilin mRNA was detected using 5’-AAATTCGCCTGCCTGTTGGG-3’ and 5’-CCGGTGGCATTAATCCATCTGT-3’ , clathrin was detected using 5’-GACTGCGGGCTTCTATTGAG-3’ and 5’-GCGGCAATTCTTCTGAACTC-3’ , wntP-2 was detected using 5’- TGCTAAATCAACACCAGAATCAGCT-3’ and 5’- CACATCCACAATTACTATGCACCCC-3’ , ndl-3 was detected using 5’- CTCCCACAATTTATGAGTGCGGT-3’ and 5’- TCTTGGGCCAATTTTGAGTTTTGATCTA-3’ , and ptk7 was detected using 5’- GATCAAATCCCAAATCCAGTTC-3’ and 5’-GGGTTTCTGGGAGTTTATATCGTA-3’ . Relative RNA abundance was calculated using the delta-Ct method after verification of primer amplification efficiency . p-values were computed from a 2-tailed t-test . | Some animals can regrow tissues that have been amputated . A group of flatworms called planarians are often used as a model to study the regeneration process because they are able to restore any lost tissue or even an entire animal from even tiny pieces of the body . For regeneration to be successful , it is critical to identify which tissues or regions of the body need to be replaced . The planarian body is divided into three main parts: head , trunk and tail . Several genes involved in specifying what tissues regenerate are very active ( or “highly expressed” ) in muscle cells in different regions of the planarian body . Some of the genes are involved in mechanisms that allow cells to communicate with each other , such as the Wnt and FGF signaling pathways . These genes could form a coordinated system to control regeneration , but their precise roles remain poorly understood . Two groups of researchers have now independently identified genes that provide cells with information about their location in the flatworm body . Lander and Petersen identified three genes that are expressed in an overlapping manner along the body of uninjured animals . One of the genes – known as ptk7 – is mainly produced in the trunk region , while the second gene ( wntP-2 ) is produced from the tail and the third ( ndl-3 ) is produced from the head region . The wntP-2 and ptk7 encode components of the Wnt signaling pathway , while ndl-3 encodes a protein involved in FGF signaling . Lander and Petersen used a technique called RNAi to lower the activity of the three genes individually or in pairs , and then examined whether this affected the ability of the worms to regenerate . Inhibition of any of the three genes resulted in an expansion of the trunk tissues into the tail region , indicating that the normal role for these genes is to stop cells adopting the trunk “identity” . Another study by Scimone , Cote et al . found that two separate sets of genes – including wntP-2 and ndl-3 – are needed to correctly position tissues in the head and trunk of planarians . Together these findings suggest that the Wnt and FGFRL pathways act in a body-wide system that co-ordinates where and which new tissues form during regeneration . A future challenge will be to decipher the complete network of genes that provides the positional information needed for regeneration . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"short",
"report"
] | 2016 | Wnt, Ptk7, and FGFRL expression gradients control trunk positional identity in planarian regeneration |
Estimating local surface orientation ( slant and tilt ) is fundamental to recovering the three-dimensional structure of the environment . It is unknown how well humans perform this task in natural scenes . Here , with a database of natural stereo-images having groundtruth surface orientation at each pixel , we find dramatic differences in human tilt estimation with natural and artificial stimuli . Estimates are precise and unbiased with artificial stimuli and imprecise and strongly biased with natural stimuli . An image-computable Bayes optimal model grounded in natural scene statistics predicts human bias , precision , and trial-by-trial errors without fitting parameters to the human data . The similarities between human and model performance suggest that the complex human performance patterns with natural stimuli are lawful , and that human visual systems have internalized local image and scene statistics to optimally infer the three-dimensional structure of the environment . These results generalize our understanding of vision from the lab to the real world .
Understanding how vision works in natural conditions is a primary goal of vision research . One measure of success is the degree to which performance in a fundamental visual task can be predicted directly from image data . Estimating the 3D structure of the environment from 2D retinal images is just such a task . However , relatively little is known about how the human visual system estimates 3D surface orientation from images of natural scenes . 3D surface orientation is typically parameterized by slant and tilt . Slant is the amount by which a surface is rotated away from an observer; tilt is the direction in which the surface is rotated ( Figure 1A ) . Compared to slant , tilt has received little attention , even though both are critically important for successful interaction with the 3D environment . For example , even if slant has been accurately estimated , humans must estimate tilt to determine where they can walk . Surface with tilts of 90° , like the ground plane , can sometimes be walked on . Surfaces with tilts of 0° or 180° , like the sides of tree trunks , can never be walked on . Numerous psychophysical , computational , and neurophysiological studies have probed the human ability to estimate surface slant , surface tilt , and 3D shape . Systematic performance has been observed , and models have been developed that nicely describe performance . Most previous studies have used stimuli having planar ( Stevens , 1983;Knill , 1998a , 1998b; Hillis et al . , 2004; Burge et al . , 2010a; Rosenholtz and Malik , 1997; Rosenberg et al . , 2013; Murphy et al . , 2013; Velisavljević and Elder , 2006; Saunders and Knill , 2001; Welchman et al . , 2005; Sanada et al . , 2012; Tsutsui et al . , 2001 ) or smoothly curved ( Todd et al . , 1996; Fleming et al . , 2011; Todd , 2004;Marlow et al . , 2015; Li and Zaidi , 2000 , 2004; Norman et al . , 2006 ) surface shapes and regular ( Knill , 1998a , 1998b; Hillis et al . , 2004; Watt et al . , 2005; Rosenholtz and Malik , 1997; Rosenberg et al . , 2013; Murphy et al . , 2013; Velisavljević and Elder , 2006; Li and Zaidi , 2000 , 2004; Welchman et al . , 2005 ) or random-patterned ( Burge et al . , 2010a; Fleming et al . , 2011 ) surface markings . These stimuli are not representative of the variety of surface shapes and markings encountered in natural viewing . Surfaces in natural scenes often have complex surface geometries and are marked by complicated surface textures . Thus , performance with simple artificial scenes may not be representative of performance in natural scenes . Also , models developed with artificial scenes often generalize poorly ( or cannot even be applied ) to natural scenes . These issues concern not just studies of 3D surface orientation perception but vision and visual neuroscience at large . Few studies have examined the human ability to estimate 3D surface orientation using natural photographic images , the stimuli that our visual systems evolved to process . None , to our knowledge , have done so with high-resolution groundtruth surface orientation information . There are good reasons for this gap in the literature . Natural images are complex and difficult to characterize mathematically , and groundtruth data about natural scenes are notoriously difficult to collect . Research with natural stimuli has often been criticized ( justifiably ) on the grounds that natural stimuli are too complicated or too poorly controlled to allow strong conclusions to be drawn from the results . The challenge , then , is to develop experimental methods and computational models that can be used with natural stimuli without sacrificing rigor and interpretability . Here , we report an extensive examination of human 3D tilt estimation from local image information with natural stimuli . We sampled thousands of natural image patches from a recently collected stereo-image database of natural scenes with precisely co-registered distance data ( Figure 1B ) ( Burge et al . , 2016 ) . Groundtruth surface orientation was computed directly from the distance data ( see Materials and methods ) . Human observers binocularly viewed the natural patches and estimated the tilt at the center of each patch . The same human observers also viewed artificially-textured planar stimuli matched to the groundtruth tilt , slant , distance , and luminance contrast of the natural stimuli . First , we compared human performance with natural and matched artificial stimuli . Then , we compared human performance to the predictions of an image-computable normative model , a Bayes’ optimal observer , that makes the best possible use of the available image information for the task . This experimental design enables direct , meaningful comparison of human performance across stimulus types , allowing the isolation of important stimulus differences and the interpretation of human response patterns with respect to principled predictions provided by the model . A rich set of results emerges . First , tilt estimation in natural scenes is hard; compared to performance with artificial stimuli , performance with natural stimuli is poor . Second , with natural stimuli , human tilt estimates cluster at the cardinal tilts ( 0° , 90° , 180° and 270° ) , echoing the prior distribution of tilts in natural scenes ( Figure 1C ) ( Burge et al . , 2016; Yang and Purves , 2003a;Yang and Purves , 2003b; Adams et al . , 2016 ) . Third , human estimates tend to be more biased and variable when the groundtruth tilts are oblique ( e . g . , 45° ) . Fourth , at each groundtruth tilt , the distributions of human and model errors tend to be very similar , even though the error distributions themselves are highly irregular . Fifth , human and model observer trial-by-trial errors are correlated , suggesting that similar ( or strongly correlated ) stimulus properties drive both human and ideal performance . Together , these results represent an important step towards the goal of being able to predict human percepts of 3D structure directly from photographic images in a fundamental natural task .
We asked whether the complicated pattern of human performance with natural stimuli is consistent with optimal information processing . To answer this question , we compared human performance to the performance of a normative model , a Bayes optimal observer that optimizes 3D tilt estimation in natural scenes given a squared error cost function ( Burge et al . , 2016 ) . The model takes three local image cues C as input — luminance , texture , and disparity gradients — and returns the minimum mean squared error ( MMSE ) tilt estimate τ^MMSE as output . ( The MMSE estimate is the mean of the posterior probability distribution over groundtruth tilt given the measured image cues . ) To determine the optimal estimate for each possible triplet of cue values , we use the natural scene database . At each pixel in the database , the image cues are computed directly from the photographic images within a local area , and the groundtruth tilt is computed directly from the distance data ( see Materials and methods; [Burge et al . , 2016] ) . In other words , the model is ‘image-computable’: the model computes the image cues from image pixels and produces tilt estimates as outputs . We approximate the posterior mean E[τ|C]=∑ττp ( τ|C ) by computing the sample mean of the groundtruth tilt conditional on each unique image cue triplet ( Figure 5A ) . The result is a table , or ‘estimate cube , ’ where each cell stores the optimal estimate τ^MMSE=E[τ|C] for a particular combination of image cues ( Figure 5B ) . In the cue-combination literature , cues are commonly assumed to be statistically independent ( Ernst and Banks , 2002 ) . In natural scenes , it is not clear whether this assumption holds . Fortunately , the normative model used here is free of assumptions about statistical independence and the form of the joint probability distribution ( see Discussion ) . Thus , our normative model provides a principled benchmark , grounded in natural scene statistics , against which to compare human performance . We tested the model observer on the exact same set of natural stimuli used to test human observers ( Figure 5C ) . The model observer predicts the overall pattern of raw human responses ( see also Figure 5—figure supplement 1 ) . More impressively , the model observer predicts the counts , means , and variances of the human tilt estimates ( Figure 2D–F ) , the conditional error distributions ( Figure 3 ) , and the conditional groundtruth tilt distributions ( Figure 4 ) . The model explains a large proportion of the variance for all of these performance measures ( Figure 5D ) . These results indicate that human visual system estimates tilt in accordance with optimal processes that minimize error in natural scenes . We conclude that the biased and imprecise human tilt estimates with natural stimuli are nevertheless lawful . Two points are worth emphasizing . First , this model observer had no free parameters that were fit to the human data ( Burge et al . , 2016 ) ; instead , the model observer was designed to perform the task optimally given the three image cues . Second , the close agreement between human and model performance suggests that humans use the same cues ( or cues that strongly correlate with those ) used by the normative model ( see Discussion ) . If human and model observers use the same cues in natural stimuli to estimate tilt , variation in the stimuli should cause similar variation in performance . Are human performance and model observer performance similar in individual trials ? The same set of natural stimuli was presented to all observers . Thus , it is possible to make direct , trial-by-trial comparisons of the estimation errors that each observer made . If the properties of individual natural stimuli influence estimates similarly across observers , then observer errors across trials should be correlated . Accounting for trial-by-trial errors is one of the most stringent comparisons that can be made between model and human performance . Natural stimuli do elicit similar trial-by-trial errors from human and model observers ( Figure 6A ) . The model predicts trial-by-trial human errors far better than chance . We quantify the model-human similarity with the circular correlation coefficients of the trial-by-trial model and human estimates ( Figure 6B ) . The correlation coefficients are significant . This result implies that the errors are systematically and reliably dependent on the properties of natural stimuli and that these properties affect human and model observers similarly . However , because both human and model observers produced biased estimates with natural stimuli ( Figure 2E , Figure 2—figure supplement 2 ) , it is possible that the biases are responsible for the error correlations . To remove the possible influence of bias , we computed the bias-corrected error . On each trial , we subtracted the observer bias at each groundtruth tilt e∗= ( τ^−τ ) ⏞error−E ( τ^−τ|τ⏞bias ) from the raw error . Human and model bias-corrected errors are also significantly correlated ( Figure 6C , D ) . The human-human correlation ( dashed line in Figure 6B , D; see Figure 6—figure supplement 1 ) sets an upper bound for the model-human correlation . The model-human correlation approaches this bound in some cases . Other measures of trial-by-trial similarity ( e . g . , choice probability; Figure 6—figure supplement 2C ) yield similar conclusions . These results show that natural stimulus variation at a given groundtruth tilt causes similar response variation in human observers and the model observer . To ensure that the predictive power of the model observer is not trivial , we developed multiple alternative models . All other models predict human performance more poorly ( Figure 6—figure supplement 2 ) . Our results do not rule out the possibility that another model could predict human performance better , but the current MMSE estimator establishes a strong benchmark against which other models must be compared . Thus , the normative model , without fitting to the human data , accounts for human tilt estimates at the level of the summary statistics ( Figure 2D–F ) , the conditional distributions ( Figure 3 and Figure 4 ) , and the trial-by-trial errors ( Figure 6 ) . Together , this evidence suggests that the human visual system’s perceptual processes and the normative model’s computations are making similar use of similar information . We conclude that the human visual system makes near-optimal use of the available information in natural stimuli for estimating 3D surface tilt . In our experiment , natural and artificial stimuli were matched on many dimensions: tilt , slant , distance , and luminance contrast . These stimulus factors are commonly controlled in perceptual experiments . Consistent with previous reports , slant and distance had a substantial impact on estimation error ( Watt et al . , 2005 ) with both natural and artificial stimuli ( Figure 7 ) . ( Luminance contrast had little impact on performance . ) Even after controlling for these stimulus dimensions , tilt estimation with natural stimuli is considerably poorer than tilt estimation with artificial stimuli . Other factors must therefore account for the differences . What are they ? In our experiment , each artificial scene consisted of a single planar surface . Natural scenes contain natural depth variation ( i . e . , complex surface structure ) ; some surfaces are approximately planar , some are curved or bumpy . How are differences in surface planarity related to differences in performance with natural and artificial scenes ? To quantify the departure of surface structure from planarity , we defined local tilt variance as the circular variance of the groundtruth tilt values in the central 1° area of each stimulus . Then , we examined how estimation error changes with tilt variance . First , we found that estimation error increases linearly with tilt variance for both human and model observers ( Figure 8A ) . Unfortunately , tilt variance co-varies with groundtruth tilt — cardinal tilts tend to have lower tilt variance than oblique tilts , presumably because of the ground plane ( Figure 8B ) — which means that the effect of groundtruth tilt could be misattributed to tilt variance . Hence , we repeated the analysis of overall error separately for cardinal tilts alone and for oblique tilts alone . We found that the effect of tilt variance is independent of groundtruth tilt ( Figure 8C ) . Thus , like slant and distance , tilt variance ( i . e . , departure from surface planarity ) is one of several key stimulus factors that impacts tilt estimation performance . Second , we found that for near-planar natural stimuli , average estimation error with natural and artificial stimuli are closely matched ( left-most points in Figure 8A ) . Does this result mean that tilt variance accounts for all performance differences between natural and artificial stimuli ? No . Performance with near-planar natural stimuli is still substantially different from performance with artificial stimuli ( Figure 8—figure supplement 1 ) . In addition , individual human and model trial-by-trial estimation errors are still correlated for the near-planar natural stimuli . Furthermore , the patterns of human performance with natural stimuli are robust across a wide range of tilt variance . Figure 9 shows the summary statistics ( estimate counts , means , and variances; cf . Figure 2D–F ) for multiple different tilt variances of human observers . Model performance is also similarly robust to tilt variance ( Figure 9—figure supplement 1 ) . We conclude that although tilt variance is an important performance-modulating factor , it is not the only factor responsible for performance differences with natural and artificial stimuli . Other factors must be responsible . Understanding these other factors is an important direction for future work .
The standard approach to modeling cue-combination , sometimes known as maximum likelihood estimation , includes a number of assumptions: a squared error cost function , cue independence , unbiased Gaussian-distributed single cue estimates , and a flat or uninformative prior ( Ernst and Banks , 2002 ) ( but see [Oruç et al . , 2003] ) . The approach used here ( normative model; see Figure 5 ) assumes only a squared error cost function , and is guaranteed to produce the Bayes optimal estimate given the image cues , regardless of the common assumptions . In natural scenes , it is often unclear whether the common assumptions hold . Methods with relatively few assumptions can therefore be powerful tools for establishing principled predictions . We have not yet fully investigated how the image cues are combined in tilt estimation , but we have conducted some preliminarily analyses . For example , a simple average of the single-cue estimates ( each based on luminance , texture , or disparity alone ) underperforms the three-cue normative model . This result is not surprising given that the individual cues are not independent , that the single cue estimates do not follow Gaussian distribution , and that the tilt prior is not flat . However , the current study is not specifically designed to examine the details of cue combination in tilt estimation . To examine cue-combination in this task rigorously , a parametric stimulus-sampling paradigm should be employed , a topic that will be explored in future work . A grand problem in perception and neuroscience research is to understand how local estimates are grouped into more accurate global estimates . We showed that local tilt estimates are unbiased predictors of groundtruth tilt and have nearly equal reliability ( Figure 4 ) . This result implies that optimal spatial pooling of the local estimates may be relatively simple . Assuming statistical independence ( i . e . , naïve Bayes ) , optimal spatial pooling is identical to a simple linear combination of the local estimates: the straight average of N local estimates τ^global=1N∑iNτ^ilocal . Of course , local groundtruth tilts and estimates are spatially correlated , so the independence assumption will not be strictly correct . However , the spatial correlations could be estimated from the database and incorporated into the computations . Our work thus lays a strong empirically grounded foundation for the investigation of local-global processing in surface orientation estimation . In classic studies of surface orientation perception , stimuli are usually limited in at least one of two important respects . If the stimuli are artificial ( e . g . , computer-graphics generated ) , groundtruth surface orientation is known but lighting conditions and textures are artificial , and it is uncertain whether results obtained with artificial stimuli will generalize to natural stimuli . If the stimuli are natural ( e . g . , photographs of real scenes ) , groundtruth surface orientation is typically unknown which complicates the evaluation of the results . The experiments reported here used natural stereo-images with laser-based measurements of groundtruth surface orientation , and artificial stimuli with tilt , slant , distance , and contrast matched to the natural stimuli . This novel design allows us to relate our results to the classic literature , to determine the generality of results with both natural and artificial stimuli and to isolate performance-controlling differences between the stimuli . In particular , we found that tilt variance is a pervasive performance-altering feature of natural scenes that is not explicitly considered in most investigations . The human visual system must nevertheless contend with tilt variance in natural viewing . We speculate that characterizing its impact is likely to be fundamental for understanding 3D surface orientation estimation in the real-world , just as characterizing the impact of local luminance contrast has been important for understanding how humans detect spatial patterns in noise ( Burgess et al . , 1981 ) . The current study is the latest in a series of reports that have attempted , with ever increasing rigor , to link properties of perception to the statistics of natural images and scenes . Our contribution extends previous work in several respects . First , previous work demonstrated similarity between human and model performance only at the level of summary statistics ( Girshick et al . , 2011; Burge et al . , 2010b; Weiss et al . , 2002; Stocker and Simoncelli , 2006 ) . We demonstrate that a principled model , operating directly on image data , predicts the summary statistics , the distribution of estimates , and the trial-by-trial errors . Second , previous work showed that human observers behave as if their visual systems have encoded the task-relevant statistics of 2D natural images ( Girshick et al . , 2011 ) . We show that human observers behave as if they have properly encoded the task-relevant joint statistics of 2D natural images and the 3D properties of natural scenes ( also see ( Burge et al . , 2010b ) ) . Third , previous work tested and modeled human performance with artificial stimuli only ( Girshick et al . , 2011; Burge et al . , 2010b; Weiss et al . , 2002; Stocker and Simoncelli , 2006 ) . We test human performance with both natural and artificial stimuli . The dramatic , but lawful , changes in performance with natural stimuli highlight the importance of studies with the stimuli that visual systems evolved to process .
The stereo images were presented with a ViewPixx Technologies ProPixx projector fitted with a 3D polarization filter . Left and right images were presented sequentially at a refresh rate of 120 Hz ( 60 Hz per eye ) and with the same resolution of the two images ( 1920 × 1080 pixel ) . The observer was positioned 3 . 0 m from a 2 . 0 × 1 . 2 m Harkness Clarus 140 XC polarization maintaining projection screen . This viewing distance minimizes the potential influence of screen cues to flatness ( e . g . , blur ) . Human observers wore glasses with passive ( linear ) polarized filters to isolate the image for the left and right eyes . The observer’s head was stabilized with a chin- and forehead-rest . From this viewing position , the projection screen subtended 36° x 21° of visual angle . The disparity-specified distance created by this projection system matched to the distances measured in the original natural scenes . The projection display was linearized over 10 bits of gray level . The maximum luminance was 84 cd/m2 . The mean luminance was set to 40% of the projection system’s maximum luminance . Three human observers participated in the experiment; two were authors , and one was naïve about the purpose of the experiment . Informed consent was obtained from participants before the experiment . The research protocol was approved by the Institutional Review Board of the University of Pennsylvania and is in accordance with the Declaration of Helsinki . Human observers binocularly viewed a small region of a natural scene through a circular aperture ( 1° or 3° diameter ) positioned 5 arcmin of disparity in front of the scene point along the cyclopean line of sight . Observers communicated their tilt estimate with a mouse-controlled probe . Each observer viewed 3600 unique natural stimuli ( 150 stimuli per tilt bin x 24 tilt bins ) presented with each of two apertures in the experiment ( 7200 total ) . Natural stimuli were constrained to be binocularly visible ( no half-occlusions ) , to have slants larger than 30° , to have distances between 5 m and 50 m , and to have contrasts between 5% and 40% . Each observer also viewed 1440 unique artificial stimuli ( 60 stimuli per tilt bin x 24 tilt bins ) with two apertures ( 2880 total ) . Artificial stimuli ( 1/f noise and phase- and orientation-randomized plaids ) were matched to the natural stimuli on multiple additional dimensions ( tilt , slant , distance , and contrast ) . Natural stimuli were presented in 48 blocks of 150 trials each , and artificial stimuli were presented in 12 blocks of 240 trials each , with interleaved blocks using small and large apertures . Tilt is a circular ( angular ) variable . We computed the mean , variance , and error using standard circular statistics . The circular mean is defined as τ¯=arg[R] where R=[∑τexp[jτ]]/N is the complex mean resultant vector . The circular variance is defined as var ( τ ) =1−|R| . Estimation error e=arg[exp[j ( τ^−τ ) ]] is the circular distance between the tilt estimate and groundtruth . Groundtruth tilt τ is computed from the distance data ( range map r ) co-registered to each natural image in the database . We defined groundtruth tilt tan−1 ( ∇yr/∇xr ) as the orientation of the normalized range gradient ( Marr , 1982 ) . The range gradient was computed by convolving the distance data with a 2D Gaussian kernel having space constant σ and then taking the partial derivatives in the x and y image directions ( Burge et al . , 2016 ) . For the results presented in this manuscript , groundtruth tilt was computed using a space constant of σ=3 arcmin; doubling this space constant does not change the qualitative results . The space constants correspond to kernel sizes of ~0 . 25°−0 . 50° . Image cues to tilt ( disparity , luminance , and texture cues ) were computed directly from the images . Like groundtruth tilt , image cues were defined as the orientation tan−1 ( ∇ycue/∇xcue ) of the local disparity and luminance gradients . The local disparity gradient is computed from the disparity image , which is obtained from the left and right eye luminance images via standard local windowed cross-correlation ( Burge et al . , 2016; Tyler and Julesz , 1978; Banks et al . , 2004 ) . The window for cross-correlation had the same space constant as the derivative operator that was used to compute the gradient ( see below ) . The texture cue to tilt is defined as the orientation of the major axis of the local amplitude spectrum of the luminance image . This texture cue is non-standard ( but see [Fleming et al . , 2011] ) . However , this texture cue is more accurate in natural scenes than traditional texture cues ( Burge et al . , 2016; Clerc and Mallat , 2002; Galasso and Lasenby , 2007; Malik and Rosenholtz , 1997; Massot and Hérault , 2008 ) . For the main results presented in this manuscript , image cues were computed from the gradients using a space constant of σ= 6 arcmin; using the space constants to σ=3 , 6 , 9 , or 12 arcmin does not change the qualitative results . The space constants correspond to kernel sizes of ~0 . 25°−1 . 0° . Luminance contrast was defined as the root-mean-squared luminance values within a local area weighted by a cosine window . Specifically , luminance contrast is C=[∑x∈A ( ( I ( x ) −I¯ ) /I¯ ) 2W ( x ) ]/∑x∈AW ( x ) where x is the spatial location , W is a cosine window with area A , and I¯=[∑x∈AI ( x ) W ( x ) ]/∑x∈AW ( x ) is the local mean intensity . On each trial , human observers communicated their perceptual estimate τ^ by making a response τ^rsp with a mouse-controlled probe . Unfortunately , the responses are not guaranteed to equal the perceptual estimates . An output-mapping function τ^rsp=g ( τ^ ) relates the response to the perceptual estimate , and an estimation function τ^=f ( τ ) relates the estimate to the groundtruth tilt of each stimulus . When responses are biased , it is hard to conclude whether the biases are due to the output-mapping function or to the estimation function . When responses are unbiased , a stronger case can be made that the human responses equal the perceptual estimates . To obtain unbiased responses τ^rsp=τ from biased estimates τ^≠τ , the output mapping function would have to equal exactly the inverse of a biased estimation function: g ( . ) =f−1 ( . ) ; this possibility seems unlikely and has no explanatory power . Thus , by Occam’s razor , unbiased responses imply unbiased output-mapping and estimation functions: τ^rsp=τ^=τ . Human responses to artificial stimuli were unbiased ( Figure 2E ) , implying an unbiased output-mapping function . Assuming that the output-mapping function is stable across stimulus types , we conclude that the biased responses to natural stimuli accurately reflect biased perceptual estimates . To determine whether the model predictions are representative of randomly sampled natural stimuli , we simulated 1000 repeats of the experiment . On each repeat , we obtained a different sample of 3600 natural stimuli ( 150 in each tilt bin ) from which we obtained 3600 optimal estimates . The samples are used to compute 95% confidence intervals on the model predictions , which are shown as the shaded regions in Figure 3A and Figure 4A . | The ability to assess how tilted a surface is , or its ‘surface orientation’ , is critical for interacting productively with the environment . For example , it helps organisms to determine whether a particular surface is better suited for walking or climbing . Humans and other animals estimate 3-dimensional ( 3D ) surface orientations from 2-dimensional ( 2D ) images on their retinas . But exactly how they calculate the tilt of a surface from the retinal images is not well understood . Scientists have studied how humans estimate surface orientation by showing them smooth ( often planar ) surfaces with artificial markings . These studies suggested that humans very accurately estimate the direction in which a surface is tilted . But whether humans are as good at estimating surface tilt in the real world , where scenes are more complex than those tested in experiments , is unknown . Now , Kim and Burge show that human tilt estimation in natural scenes is often inaccurate and imprecise . To better understand humans’ successes and failures in estimating tilt , Kim and Burge developed an optimal computational model , grounded in natural scene statistics , that estimates tilt from natural images . Kim and Burge found that the model accurately predicted how humans estimate tilt in natural scenes . This suggests that the imprecise human estimates are not the result of a poorly designed visual system . Rather , humans , like the computational model , make the best possible use of the information images provide to perform an estimation task that is very difficult in natural scenes . The study takes an important step towards generalizing our understanding of human perception from the lab to the real world . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | The lawful imprecision of human surface tilt estimation in natural scenes |
Annual changes in the environment threaten survival , and numerous biological processes in mammals adjust to this challenge via seasonal encoding by the suprachiasmatic nucleus ( SCN ) . To tune behavior according to day length , SCN neurons display unified rhythms with synchronous phasing when days are short , but will divide into two sub-clusters when days are long . The transition between SCN states is critical for maintaining behavioral responses to seasonal change , but the mechanisms regulating this form of neuroplasticity remain unclear . Here we identify that a switch in chloride transport and GABAA signaling is critical for maintaining state plasticity in the SCN network . Further , we reveal that blocking excitatory GABAA signaling locks the SCN into its long day state . Collectively , these data demonstrate that plasticity in GABAA signaling dictates how clock neurons interact to maintain environmental encoding . Further , this work highlights factors that may influence susceptibility to seasonal disorders in humans .
Daily rhythms generated by the circadian system serve to anticipate changes in the environment caused by the Earth’s rotation ( Hut and Beersma , 2011; Pittendrigh , 1960 ) . In mammals , the circadian system is a hierarchical collection of biological clocks orchestrated by a master pacemaker in the suprachiasmatic nucleus ( SCN ) of the anterior hypothalamus ( Mohawk et al . , 2012 ) . The SCN itself is a network of clock cells that interact with one another to regulate emergent circuit properties , process photic cues , and provide daily signals to downstream tissues ( Evans , 2016; Hastings et al . , 2018 ) . At the cellular level , SCN neurons display ca . 24 hr rhythms in electrical and molecular activity ( Hastings et al . , 2018 ) generated by genetic feedback loops that control daily expression of clock proteins ( Buhr and Takahashi , 2013 ) and the clock-controlled genes that regulate cellular physiology ( Zhang et al . , 2014 ) . Although the circadian molecular clock drives cellular rhythms , intercellular interactions among SCN neurons synchronize , amplify , and stabilize rhythms of cells within the network ( Evans , 2016; Hastings et al . , 2018 ) . In addition , the SCN adjusts to the environment , effectively matching the 24 hr period of the solar cycle , the phase of the local time zone , and the waveform of the prevailing day length as it changes over the year . Thus , the SCN network serves as both daily clock and annual calendar , and defining the circuit-level mechanisms by which SCN neurons interact is paramount for understanding the temporal regulation of behavior and physiology . Seasonal changes in the environment threaten the survival and well being of organisms living on this planet . To cope with this challenge , many biological processes are modulated on a seasonal basis . Although daily timekeeping is a cellular property , seasonal encoding is accomplished via changes in the spatiotemporal relationships of SCN clock cells ( Meijer et al . , 2010; Evans and Gorman , 2016 ) . When days are short , SCN neurons form a highly synchronized population of cellular clocks with similar times of daily protein expression and electrical activity . In contrast , when days are long , the SCN network switches to an alternate state characterized by two subclusters of clock neurons that cycle in anti-phase ( Evans et al . , 2013; Inagaki et al . , 2007 ) . The ability of the SCN network to encode season is critical for the regulation of a large variety of biological processes , including reproduction , immune function , and metabolism . In humans , seasonality likewise influences a wide range of functions such as sleep , metabolism , and neurotransmission ( Garbazza and Benedetti , 2018 ) , which can produce annually recurrent pathology in 3–10% of people ( Wehr et al . , 2001; Lewy et al . , 2006 ) . Although seasonal affective disorder is closely linked to photic modulation of circadian clock function ( Wirz-Justice , 2018 ) , the mechanisms that regulate sensitivity to day length remain ill-defined . GABA is the primary neurotransmitter expressed by SCN neurons ( Albers et al . , 2017 ) , and recent work indicates that day length modulates GABA signaling in the SCN network ( Evans et al . , 2013; Myung et al . , 2015; Farajnia et al . , 2014 ) . Although GABA is typically classified as an inhibitory neurotransmitter , the GABAA receptor is a heteropentameric ligand-gated ion channel permeable to chloride and bicarbonate that can elicit either hyperpolarization or depolarization depending on the electrochemical gradient of chloride ( Kaila et al . , 2014 ) . In mature neurons , levels of intracellular chloride are regulated by the Cl- extruder KCC2 ( K+ Cl- co-transporter 2 ) and the Cl- importer NKCC1 ( Na+ K+ Cl- co-transporter 1 ) . Changes in the relative expression and/or function of these two chloride co-transporters cause ionic plasticity: short- and long-term modulation that alters neuronal responses to GABAA signaling ( Kaila et al . , 2014 ) . The most well documented example of this neuroplasticity occurs during development ( Rivera et al . , 1999 ) , when NKCC1 drives a depolarizing chloride reversal potential in immature neurons , but up-regulation of KCC2 causes a switch to the hyperpolarizing GABA response in adulthood . However , this process is not immutable; with downregulation of KCC2 in mature neurons causing a switch to excitatory GABAA signaling in several distinct circuits in the adult brain ( Chung , 2012; Hewitt et al . , 2009; Lee et al . , 2011; Ostroumov et al . , 2016; Sarkar et al . , 2011 ) . In the SCN , recent work has shown that exposure to long days increases the number of neurons that respond to GABA with depolarization ( Farajnia et al . , 2014 ) and elevates intracellular chloride ( Myung et al . , 2015 ) . Further , alterations in GABAA signaling are associated with changes in circadian behavior ( Myung et al . , 2015; Farajnia et al . , 2014; DeWoskin et al . , 2015 ) , and the role of GABAA signaling in coupling SCN neurons depends on seasonal state ( Evans et al . , 2013 ) . Collectively , this work suggests that GABAA signaling in the SCN varies with day length , but it remains unclear how plasticity in GABA circuits regulates seasonal encoding by the master clock network . Here we provide insight into the mechanisms and functional significance of seasonal plasticity in GABAA signaling in the SCN network . First , we show how changes in day length modulate molecular responses of the SCN clock to GABA stimulation . Next , we provide insight into mechanisms underlying seasonal plasticity in GABAA signaling by demonstrating that long days decrease KCC2 expression in the specific SCN compartment that processes light . Lastly , we test the functional significance of photoperiodic changes in chloride transport and GABAA signaling for SCN network dynamics . Our results demonstrate that the seasonal switch in GABAA signaling accelerates recovery of the SCN network back to its synchronous state . Through pharmacological interrogation , we define the regional locus of this neuroadaptation , identify consequences for network encoding , and shed new light on how this form of plasticity may modulate daily rhythms in behavior . Overall , these results reveal a novel mechanism by which seasonal adjustments in GABAA circuits of the master clock serve to maintain sensitivity to changes in day length .
To gain greater insight into the functional consequences of seasonal changes in GABAA signaling , we first tested how day length modulates molecular responses of the SCN clock to GABA . Previous work has predicted that the seasonal switch to excitatory GABAA signaling will alter how the SCN clock is reset by tonic GABA stimulation ( DeWoskin et al . , 2015 ) . To test this prediction , we examined photoperiodic changes in GABA-induced resetting using PER2::LUC rhythms as a readout of the SCN molecular clock . mPer2Luc mice were exposed to either 12 hr of light per day ( L12 ) or a long-day with 20 hr of light per day ( L20 ) for at least 8 weeks . SCN slices were collected from mice housed under each photoperiod , and PER2::LUC rhythms were monitored for three cycles in vitro before treatment with either 200 μM GABA or vehicle ( Figure 1—figure supplement 1 ) . GABA was not washed out following application to provide a tonic stimulus , and its concentration was stable in culture for at least 24 hr ( Figure 1—figure supplement 1 ) . As in previous work ( Evans et al . , 2013 ) , the SCN clock shifted to a later phase when GABA was applied to L12 slices at the trough of the PER2::LUC rhythm ( Figure 1—figure supplement 1 ) . Importantly , this resetting response was blocked by the GABAA receptor antagonist bicuculline ( Figure 1—figure supplement 1 ) , consistent with a prior study demonstrating that GABA-induced resetting is mediated by GABAA signaling and not GABAB signaling ( Liu and Reppert , 2000 ) . Next we examined whether long days altered SCN responses to GABA by measuring resetting responses at times spanning the circadian cycle ( Figure 1A ) . In L12 slices , GABA shifted the PER2::LUC rhythm in a phase-dependent manner , eliciting phase delays during early subjective night and phase advances during other times of the circadian cycle ( Figure 1A–C ) . Notably , the shape of this phase-resetting rhythm was similar to that reported previously for GABA-induced phase shifts of electrical rhythms in dissociated SCN neurons ( Liu and Reppert , 2000 ) . In L20 slices , GABA elicited phase delays for a larger proportion of the circadian cycle and produced phase advances at later times relative to L12 slices ( Figure 1A–C ) . These results indicate that the waveform of the GABA-induced resetting rhythm is altered by photoperiod , as illustrated by either a phase response curve ( Figure 1A ) or polar plot ( Figure 1B ) . On the other hand , long day exposure did not alter systematically the magnitude of resetting to this dose of GABA ( Figure 1C ) or vehicle ( Figure 1—figure supplement 1 ) , which distinguishes it from photoperiodic suppression of NMDA- ( Figure 1—figure supplement 1 ) and light-induced resetting ( Pittendrigh et al . , 1984; vanderLeest et al . , 2009 ) . Overall , these results confirm that the seasonal switch in GABAA signaling modulates the intrinsic responses of the SCN molecular clock ( DeWoskin et al . , 2015 ) . Notably , the effect on waveform is of particular interest because this parameter often reflects changes in SCN organization ( Pittendrigh et al . , 1984; vanderLeest et al . , 2009 ) . Next we tested cellular mechanisms that could alter GABA responses of the SCN network . Long days increase mRNA encoding NKCC1 relative to KCC2 in the SCN ( Myung et al . , 2015 ) , but the switch to excitatory GABAA signaling in other neural circuits of the adult brain is often driven by post-translational downregulation of KCC2 ( Chung , 2012; Hewitt et al . , 2009; Lee et al . , 2011; Ostroumov et al . , 2016; Sarkar et al . , 2011 ) . To test how day length modulates chloride co-transporter expression at the protein level , we measured KCC2 and NKCC1 in SCN slices collected from L12 and L20 mice at time points spanning the light:dark cycle . Based on previous reports documenting regional differences in GABA responses and chloride co-transporter expression in the rat SCN ( Belenky et al . , 2010; Albus et al . , 2005 ) , we analyzed KCC2 and NKCC1 separately in the two major subdivisions of the SCN network ( i . e . , shell and core; Abrahamson and Moore , 2001 ) using arginine vasopressin ( AVP ) expression to demarcate the SCN shell ( Figure 2A , Figure 2—figure supplement 1 ) . Lastly , we also used ratiometric analyses of KCC2/NKCC1 expression ( Figure 2B ) given that relative co-transporter expression influences intracellular chloride concentration ( Kaila et al . , 2014; Lee et al . , 2011 ) . Importantly , cellular patterns of KCC2 and NKCC1 immunoreactivity were consistent with previous studies ( Kaila et al . , 2014 ) , with KCC2 immunoreactivity largely confined to the plasma membrane and NKCC1 immunoreactivity evident in both membrane and cytoplasm ( Figure 2—figure supplement 1 ) . First we examined chloride co-transporter expression under L12 because there is scarce information on spatiotemporal patterns in the murine SCN . Under L12 , KCC2/NKCC1 fluctuated over the day ( Figure 2B–C ) in the specific SCN compartment receiving dense retinal innervation ( Abrahamson and Moore , 2001 ) . Specifically , KCC2/NKCC1 varied in the SCN core ( Figure 2C , Circwave cosinor rhythmicity test: p<0 . 05 ) , but not in the SCN shell ( Figure 2C , Circwave cosinor rhythmicity test: p>0 . 5 ) due to region-specific expression of both chloride co-transporters ( Figure 2D , Figure 2—figure supplement 1 ) . First for KCC2 , expression was lower in AVP+ regions than in non-AVP-expressing regions of not only the SCN , but also the SON and PVN ( Figure 2—figure supplement 1 ) , consistent with work in the rat ( Belenky et al . , 2010; Haam et al . , 2012 ) . In the SCN , KCC2 expression was 91-fold lower in the AVP+ shell versus the non-AVP-expressing core ( Student’s t: p<0 . 005 ) , but KCC2 levels did not vary over the day in either region ( Figure 2D , Circwave cosinor rhythmicity test: p>0 . 2 ) . In contrast , NKCC1 expression fluctuated over the day in both SCN compartments ( Figure 2D , Circwave cosinor rhythmicity test: p<0 . 005 for each region ) , which is associated with daily KCC2/NKCC1 fluctuations in the SCN core ( Figure 2C ) . These results establish that KCC2 and NKCC1 in the SCN are regulated in a spatiotemporal manner under standard lighting conditions , which may contribute to regional differences in clock neuronal excitability and GABAA responses ( Albers et al . , 2017; Albus et al . , 2005; Klett and Allen , 2017; Choi et al . , 2008 ) . Notably , L20 eliminated the daily rhythm of KCC2/NKCC1 in the SCN core ( Figure 2B–C , Circwave cosinor rhythmicity test: p>0 . 4 ) by decreasing the mean daily expression of KCC2 ( Figure 2D–E , LSM contrasts: p<0 . 01 ) . Also in the SCN core , L20 attenuated the daily rhythm in NKCC1 expression ( Figure 2D , Circwave cosinor rhythmicity test: p=0 . 1 ) , but did not alter its levels across the day ( Figure 2F , LSM contrasts: p>0 . 8 ) or at individual time points ( Figure 2D , Full Factorial ANOVA: p=0 . 09 ) . In the SCN shell , L20 increased NKCC1 expression ( Figure 2F , LSM contrasts: p<0 . 01 ) and disrupted the daily rhythm of NKCC1 ( Circwave cosinor rhythmicity test: p>0 . 1 ) due to elevation at specific times of the light:dark cycle ( Figure 2D , LSM contrasts: p<0 . 01 ) . This produced a slight decrease in overall KCC2/NKCC1 levels ( L20: 14 . 83/124 . 61 = 0 . 12 , L12: 13 . 16/93 . 97 = 0 . 14 ) but did not markedly alter daily expression of KCC2/NKCC1 in the SCN shell ( Figure 2C ) . These results indicate that long days regulate the expression of both chloride co-transporters in the SCN , with decreased KCC2 in the SCN core , increased NKCC1 in the SCN shell , and elimination of the NKCC1 rhythm in both compartments . Given that small changes in chloride co-transporter expression can alter the electrochemical gradient of chloride ( Rivera et al . , 2002; Woodin et al . , 2003 ) , these collective changes would be predicted to increase the probability of depolarizing responses to GABA in each SCN compartment . This is consistent with previous work demonstrating that long days elevate intracellular chloride , depolarize the chloride equilibrium potential , and increase the number of SCN neurons that exhibit excitatory responses to GABA ( Myung et al . , 2015; Farajnia et al . , 2014 ) . To directly test the functional significance of photoperiodic changes in chloride co-transporter expression , we next examined whether inhibition of KCC2 or NKCC1 activity would modulate SCN network dynamics using selective antagonists of each chloride co-transporter ( Haam et al . , 2012; Choi et al . , 2008 ) . To test effects of chloride transport on network encoding , SCN slices were collected from L12 and L20 mice , then cultured with the NKCC1 antagonist bumetanide ( BU , 80 μM ) , the KCC2 antagonist VU0240551 ( VU , 80 μM ) , or vehicle-treated medium . In L12 slices , we found that both BU and VU increased SCN period by 0 . 35 hr ( Figure 3—figure supplement 1 ) , which was blocked by co-culture with the GABAA receptor antagonists bicuculline or gabazine ( Figure 3—figure supplement 1 ) . Importantly , these results indicate that inhibition of KCC2 and NKCC1 modulates the function of the SCN molecular clock by altering GABAA signaling , which is consistent with previous work demonstrating these compounds alter chloride transport , neuronal excitability , and GABA-evoked calcium responses ( Farajnia et al . , 2014; Haam et al . , 2012; Deisz et al . , 2014 ) . At the cellular level , VU or BU increased the period of both SCN shell and core neurons in L12 slices ( Figure 3—figure supplement 1 ) , but did not markedly alter SCN organization ( Figure 3—figure supplement 1 ) or damping of cellular rhythms ( Figure 3—figure supplement 1 ) . These data indicate that network and cell function was not adversely affected by drug treatment . Consistent with previous work ( Evans et al . , 2013 ) , in vivo exposure to L20 reorganized the SCN by imposing a large phase difference between the SCN shell and core on the first cycle in vitro ( Figure 3A ) . This phase difference diminished over time in vehicle-treated slices ( Figure 3A ) , which is driven by tetrodotoxin-sensitive signaling mechanisms ( Evans et al . , 2013 ) . To quantify the dynamic process of network recovery , we measured changes in the phase relationship of SCN shell and core neurons over time in vitro ( Figure 3B , see Materials and methods ) . As predicted for a coupled oscillator system ( Hansel et al . , 1995 ) , the magnitude and direction of the cellular responses depended on the core-shell relationship at the start of the recording ( Figure 3B ) . Specifically , the shell-core phase difference was reduced in the ‘negative’ direction when SCN core neurons phase-led SCN shell neurons by 2–6 hr ( Figure 3B ) , whereas it was reduced in the opposite ‘positive’ direction when SCN core neurons phase-led SCN shell neurons by 8–16 hr ( Figure 3B ) . Analysis of cellular period indicated that this coupling response is driven largely by changes in the cellular rhythms of SCN core neurons , which adopted a longer period when the network re-synchronized in the negative direction ( Figure 3C , Figure 4—figure supplement 1 ) and a shorter period when re-synchronization occurred in the opposite direction ( Figure 3C , Figure 4—figure supplement 1 ) . Period modulation in SCN shell neurons was less affected by phase relationship ( Figure 3C , Figure 4—figure supplement 1 ) , suggesting that this specific neuronal subpopulation did not shift much under vehicle conditions . Overall , these results indicate that the SCN network recovers from its reorganized state in vitro largely due to intercellular signals that modulate the phase and period of SCN core neurons ( Figure 3D ) . Notably , the dynamics of this coupling response were modulated by KCC2 and NKCC1 inhibition , with complementary changes depending on which chloride co-transporter was targeted ( Figure 4A , Figure 4—figure supplement 1 ) . When L20 SCN slices were cultured with the KCC2 inhibitor VU , the process of network resynchronization was accelerated in the negative direction ( Figure 4A–B , Figure 4—figure supplement 1 ) due to altered period responses of SCN core neurons ( Figure 4C–D , Figure 4—figure supplement 1 ) . In contrast , BU inhibition of NKCC1 slowed network recovery ( Figure 4A–B , Figure 4—figure supplement 1 ) by affecting the period of both SCN shell and core neurons ( Figure 4C–D , Figure 4—figure supplement 1 ) . Interestingly , BU blocked the period responses of SCN core neurons markedly when the network was reorganized by L20 ( Figure 4—figure supplement 1 ) , which is consistent with the decreased KCC2 in this region under this photoperiod ( Figure 2D ) . Collectively , these data suggest that photoperiodic changes in KCC2/NKCC1 activity determine the rate and dynamics of network recovery by modulating GABAA signaling in the SCN core . To further test the role of decreased KCC2 in modulating network coupling after L20 , we treated SCN slices with the KCC2 activator CLP290 ( Gagnon et al . , 2013 ) . CLP290 is a carbamate prodrug that increases KCC2 surface expression , reduces intracellular chloride , and restores EGABA in mature neurons with diminished KCC2 expression ( Ostroumov et al . , 2016; Gagnon et al . , 2013; Ferrini et al . , 2017; Chen et al . , 2017 ) . Based on the ratiometric model of chloride flux , we predicted that enhancing KCC2 function would mimic the results of the NKCC1 antagonist and likewise slow SCN recovery . To test whether restoration of KCC2 function would adversely affect SCN coupling , we treated L12 and L20 slices with 100 μM CLP290 , as in Chen et al . ( 2017 ) . Similar to VU and BU , CLP290 increased the period of L12 SCN slices by lengthening period in both shell and core neurons ( Figure 3—figure supplement 1 ) , but did not markedly alter network organization ( Figure 3—figure supplement 1 ) or cellular damping over time in vitro ( Figure 3—figure supplement 1 ) . Consistent with predictions , CLP290 attenuated network recovery in L20 slices ( Figure 4A–B ) due to abrogation of coupling responses in both SCN core and shell neurons ( Figure 4C–D , Figure 3—figure supplement 1 ) . Overall , the effects of CLP290 in L20 slices were qualitatively similar to those induced by BU ( Figure 4E ) , providing complementary evidence that the photoperiodic decrease in KCC2-mediated chloride transport is a critical modulator of intercellular coupling during network recovery from the long day state . Given that BU delayed network recovery in vitro , we next tested whether it would likewise modulate circadian behavior in vivo . Bumetanide is an FDA-approved diuretic used in humans and rodent models , which has been shown to influence neural function when administered orally ( Tyzio et al . , 2014 ) . To test the effects of BU on circadian behavior , L12 and L20 mice were singly housed in wheel-running cages and provided BU ( 5 mg/kg ) or vehicle in the drinking water . After administration in drinking water , BU concentration reached 5 μM in plasma and 40 nM in the hypothalamus , the latter of which approaches the half-maximal inhibitory concentration of BU for NKCC1 ( Russell , 2000 ) . After 8 weeks of photoperiodic pre-treatment with or without BU , L12 and L20 mice were released into constant darkness ( DD , Figure 5A , Figure 5—figure supplement 1 ) to monitor changes in circadian behavior reflective of SCN coupling ( Evans and Gorman , 2016; Pittendrigh and Daan , 1976a ) . Specifically , exposure to long days alters circadian waveform in nocturnal rodents by compressing the duration of the active phase , which is confirmed by decompression after release into DD . Importantly , the waveform of daily rhythms is closely associated with changes in SCN state , and decompression of the active phase manifests as the SCN returns to its synchronous state ( Evans et al . , 2013; Margraf et al . , 1991; Jagota et al . , 2000 ) . Long days also shorten free-running period , but this aftereffect persists after SCN organization has recovered in vivo ( Evans et al . , 2013; Myung et al . , 2015; Pittendrigh and Daan , 1976b ) . Consistent with results obtained in vitro , BU slowed the recovery of circadian waveform after release from L20 into DD in vivo ( Figure 5A–C , Figure 5—figure supplement 1 ) . L20 mice treated with BU displayed slower recovery of circadian waveform over time in DD due to decreased rate of activity decompression ( Repeated Measures ANOVA: Drug*Time p<0 . 05 ) , but BU did not significantly alter behavior after release from L12 ( Repeated Measures ANOVA: Drug*Time p>0 . 5 ) . Further , BU did not alter wheel running rhythms under entrained conditions , ( Figure 5B , Repeated Measures ANOVA: p>0 . 5 ) , SCN organization , body weight , water intake , activity levels , free-running period , phase angle of entrainment , or behavioral adjustment to simulated jetlag ( Figure 5—figure supplement 1 ) . This pattern of results suggests that photoperiodic changes in the relative role of NKCC1 regulate the rate at which behavior recovers from exposure to long days . Our results indicate that exposure to long days modulates chloride transport in the SCN by decreasing KCC2 in the specific SCN region that processes light input . Because downregulation of KCC2 alters the sign of GABAA signaling in other regions of the adult brain ( Hewitt et al . , 2009; Lee et al . , 2011; Ostroumov et al . , 2016; Sarkar et al . , 2011 ) , we hypothesized that non-canonical GABA responses ( i . e . , depolarization ) modulate the dynamics of SCN coupling after exposure to long days . Depolarizing responses to GABA are driven by anionic gradient shifts in the flux of both chloride and bicarbonate ( Staley et al . , 1995; Rivera et al . , 2005 ) , and bicarbonate regeneration is necessary for excitatory GABAA responses ( Ostroumov et al . , 2016; Staley et al . , 1995 ) . Thus , we predicted that if SCN recovery from L20 requires excitatory GABAA responses , then network resynchronization would be attenuated when bicarbonate regeneration is inhibited by the carbonic anhydrase antagonist , acetazolamide ( Ostroumov et al . , 2016; Staley et al . , 1995 ) . Mice refused to drink water treated with acetazolamide , thus we returned to our in vitro assay to track SCN coupling in real time . SCN slices from L12 and L20 mice were cultured with either acetazolamide ( ACE , 100 μM ) or vehicle ( <0 . 1% DMSO ) . Similar to VU , BU , and CLP290 , ACE increased L12 SCN period in a GABAA-dependent manner ( Figure 6—figure supplement 1 ) . ACE also eliminated the period difference between SCN core and shell neurons in L12 SCN slices by specifically increasing the period of SCN shell neurons ( Figure 6—figure supplement 1 ) , consistent with low KCC2/NKCC1 expression in this region ( Figure 2C ) . Interestingly , equalization of cellular period reduced the shell-core phase difference that typically manifests in L12 slices over time in vitro ( Figure 6—figure supplement 1 ) or after release into DD ( Evans et al . , 2013; Evans et al . , 2011 ) . This suggests that excitatory GABAA signaling in SCN shell neurons imposes the shell-core phase difference , consistent with previous work using the non-selective chloride transport inhibitor furosemide ( Myung et al . , 2015 ) . Because L20 decreased KCC2 in the SCN core , we predicted that ACE would exert more pronounced effects under this condition . Remarkably , ACE treatment locked the SCN network in the reorganized L20 state ( Figure 6A–B ) by specifically blocking the period response of SCN core neurons ( Figure 6C–D , Figure 6—figure supplement 1 ) . This is of interest because L20 decreased KCC2 expression ( Figure 2E ) and increased BU responsiveness ( Figure 4—figure supplement 1 ) in this specific SCN compartment . Similar to effects produced by chloride transport inhibition , effects of ACE were region-specific and state-dependent because the coupling responses of SCN shell neurons were not altered when the network was in the reorganized state ( Figure 6—figure supplement 1 ) . Collectively , these data reveal that the photoperiodic switch in GABAA signaling is necessary for the functional restoration of the SCN network and that responses of SCN core neurons are those most influenced by the photoperiodic changes in GABAA signaling .
Light is the most salient cue for the SCN network , and seasonal changes in photic conditions are encoded via the spatiotemporal relationships of SCN cellular clocks ( Meijer et al . , 2010; Evans and Gorman , 2016 ) . Given the ever-changing nature of the photic environment , there is a premium on maintaining plasticity of SCN encoding . Here we provide novel insight into seasonal encoding by revealing that the photoperiodic switch in GABAA signaling is necessary for restoring SCN function after it is disrupted by light . By examining dynamic changes in cellular synchrony , we have uncovered that plasticity in GABAA signaling is a critical neuroadaptation that modifies how SCN clock neurons communicate with one another to regulate circuit function in accordance with an ever-changing environment . This switch corresponds with KCC2 downregulation in the specific SCN compartment that receives and processes light input , which is the same region in L20 slices displaying altered cellular responses during pharmacological modulation of chloride transport and non-canonical GABAA signaling . Collectively , these results indicate that downregulation of KCC2 in this specific region is critical for network recovery after exposure to long days , and that this process is likely driven by mechanisms similar to those regulating GABAA signaling in other important networks controlling behavior . Here we find that seasonal changes in GABAA signaling in the SCN network regulate the transition back to its synchronous state . In this manner , seasonal plasticity in GABAA signaling may be viewed as a homeostatic mechanism because it acts to restore circuit function after it is perturbed by light . Homeostasis is not a term commonly employed in a circadian context , but its applicability to seasonal encoding is not without precedent . For instance , formal assays have shown that photoperiodic adjustments in behavior and physiology return to steady-state when the long day stimulus is removed in vivo ( Pittendrigh and Daan , 1976a ) . Further , photoperiodic alterations in SCN cellular relationships return to steady-state after DD release in vivo and culture ex vivo ( Evans et al . , 2013 ) . Homeostatic recalibration is critical for most biological systems , but the underlying mechanisms remain poorly understood in many neural networks because circuit transitions are difficult to study in real-time . By tracking this dynamic process ex vivo , we first demonstrated an unexpected role for GABAA signaling in network recovery ( Evans et al . , 2013 ) . Here we show this form of neuroplasticity requires non-canonical GABAA signaling by demonstrating that ACE locks the SCN network into its reorganized state by modulating the coupling response of SCN core neurons . Interestingly , our data suggest that non-canonical GABAA signaling also modulates SCN dynamics under steady state by acting on a different cellular target . Specifically , ACE influenced cellular relationships in L12 slices due to modulation of the molecular clock in SCN shell neurons . This equalized cellular period across the network and attenuated the shell-core phase difference that manifests over time in culture or after release into DD in vivo ( Evans et al . , 2013; Evans et al . , 2011 ) . This indicates that non-canonical GABAA signaling acts to impose phase differences during steady state by modulating the period of SCN shell neurons , but facilitates synchrony in the polarized state by acting on SCN core neurons . Recent reports indicate that GABAA signaling is involved in other types of network behaviors not examined directly here ( Freeman et al . , 2013; Azzi et al . , 2017 ) , thus it would be of interest to determine whether non-canonical responses to GABA are likewise involved in these other SCN processes . Further work investigating how state-dependent GABA signaling regulates cellular physiology in different subclasses of SCN neurons is warranted given that this is the major neurotransmitter expressed by SCN neurons ( Albers et al . , 2017 ) . In addition to providing novel insight into the functional significance of seasonal changes in GABAA signaling , the present results expand understanding of the cellular mechanisms underlying this form of neuroplasticity . Neuronal responses to GABAA signaling are influenced by the ratio of KCC2 and NKCC1 expression , and changes in either co-transporter can invert ion flux because the equilibrium potential for chloride is close to the resting membrane potential ( Rivera et al . , 2002; Woodin et al . , 2003 ) . In other adult networks , KCC2 downregulation is the key event modulating the sign of GABAA signaling because it alters the co-transporter ratio to favor NKCC1-mediated Cl- influx ( Hewitt et al . , 2009; Lee et al . , 2011; Ostroumov et al . , 2016; Sarkar et al . , 2011 ) . Here we show that long days decrease KCC2 in the SCN core and increase NKCC1 in the SCN shell , consistent with previous work demonstrating that long days increase depolarizing responses to GABA in both SCN compartments ( Farajnia et al . , 2014 ) . Given the limits of drawing inferences from expression alone , we next tested the functional implications of photoperiodic changes in KCC2/NKCC1 using two different approaches ( i . e . , NKCC1 antagonism with bumetanide , KCC2 agonism with CLP290 ) . Both drugs attenuated the ability of the long-day SCN network to return to its basal state ex vivo in a qualitatively similar manner , suggesting that photoperiodic adjustments of chloride transport serve to speed SCN recovery after disruption by light . Notably , long days decreased KCC2 in the SCN core during the daytime , which matches the phase when SCN neuronal responses to GABA are affected most by photoperiod ( Farajnia et al . , 2014 ) . Further , KCC2 antagonism accelerated SCN recovery in vitro by changing the coupling responses of SCN core neurons , suggesting that further suppression of KCC2 activity beyond that achieved by light may have potential benefits . Lastly , we find that ACE blocked network recovery by specifically modulating the coupling responses of SCN core neurons . The convergent nature of these results is consistent with the known effects of these compounds on chloride transport , neuronal excitability , and GABA-evoked calcium responses ( Farajnia et al . , 2014; Haam et al . , 2012; Deisz et al . , 2014 ) . Collectively , these results indicate that KCC2 downregulation in the SCN core is the critical cellular adaptation driving network recovery in the master clock . It remains unclear how light modulates KCC2 expression , but it is likely that this photoperiodic change is achieved via post-translational regulation of protein expression and/or activity because long days do not increase Kcc2 transcription ( Myung et al . , 2015 ) . Indeed , KCC2 function can be controlled by its phosphorylation state ( Kaila et al . , 2014 ) , which is modulated by intercellular signaling mechanisms known to be involved in photic processing , including glutamate , BDNF , and extrasynaptic GABA signaling ( Albers et al . , 2017 ) . Given that KCC2 is downregulated in the SCN compartment that receives and processes photic input , identifying the specific pathways by which light represses KCC2 in this region represents an important area for future work . Consistent with its effects in vitro , we found that bumetanide reduced circadian plasticity in vivo . Although the systemic approach used here did not target a specific locus , bumetanide specifically decreased plasticity of circadian waveform , which is a rhythmic property closely linked to spatiotemporal encoding by the SCN network ( Evans and Gorman , 2016; Pittendrigh and Daan , 1976a; Margraf et al . , 1991; VanderLeest et al . , 2007 ) . Slower recovery of behavioral state following systemic bumetanide administration in vivo is similar to the reduced rate of SCN recovery in vitro . In contrast , in vivo bumetanide administration did not prevent long day encoding itself or the aftereffect on free-running period , which may reflect that brain levels of BU were insufficient to block these effects . Hypothalamic levels of bumetanide in the current study approached the half-maximal inhibitory concentration for NKCC1 ( Russell , 2000; Löscher et al . , 2013; Hampel et al . , 2018 ) . Systemically administered bumetanide has been shown to modulate hippocampal synaptic plasticity and memory in mouse models of disease ( Marguet et al . , 2015; Deidda et al . , 2015 ) , and the efficacy of this compound in previous and current work may relate to its prolonged administration over many days to mice , a species in which the plasma half-life of bumetanide is >45 min ( Töpfer et al . , 2014 ) . Given that the SCN regulates both circadian period and waveform , future work targeting this specific locus may provide additional insight into the processes controlling seasonal encoding and aftereffects . Although pharmacological manipulations may be limited by the need for continued drug efficacy over weeks without daily interference , the current results may be used to develop and optimize genetic approaches to modulate chloride transport and/or GABAA signaling specifically in SCN neurons . The current results indicate that cell identity and phasing may be important considerations in the design of these future studies . In agreement with previous work suggesting daily SCN changes in GABAA signaling correspond with altered NKCC1 expression ( Belenky et al . , 2010; Choi et al . , 2008 ) , our results demonstrate that NKCC1 and KCC2/NKCC1 are regulated over the day under a standard lighting condition . Thus , the adult SCN is likely a network in which neuroplasticity of GABAA signaling influences circuit function on both a seasonal and circadian basis ( Albers et al . , 2017 ) . The regional differences in KCC2/NKCC1 expression described here reflect disparities in both KCC2 and NKCC1 expression across SCN compartments . Similar to other AVP+ structures in the hypothalamus ( Haam et al . , 2012 ) , we detected stark regional differences in KCC2 expression across the SCN shell and core . Further , the timing of peak NKCC1 expression appeared to differ across SCN compartment , consistent with Western blot analyses of NKCC1 in the mouse SCN conducted at two times of day ( Choi et al . , 2008 ) . Interestingly , low NKCC1 expression in the SCN core corresponds with the phase of low intracellular chloride in this specific region ( DeWoskin et al . , 2015 ) . Thus , daily variation in chloride co-transporter expression may contribute to documented regional differences in neuronal excitability and GABAA responses of SCN neurons ( Albers et al . , 2017; Albus et al . , 2005; Klett and Allen , 2017; Choi et al . , 2008 ) . Considering that changes in KCC2 and NKCC1 can impact myriad other cellular processes ( Kaila et al . , 2014 ) , further investigation into their role in the SCN network is warranted . Seasonal changes in the SCN circuit also alter molecular responses to GABA stimulation . Phase resetting responses elicited here are consistent with findings that tonic GABA stimulation effectively shifts the phase of the SCN molecular clock ( DeWoskin et al . , 2015 ) . One caveat to longer-term treatment is the inadvertent modulation of additional targets . With this in mind , it is notable that present responses to tonic GABA treatment appear similar in several respects to those elicited by more acute treatment ( Liu and Reppert , 2000 ) . For instance , GABA-induced resetting was occluded by GABAA antagonism , replicating previous work that also directly excluded a contribution of GABAB signaling ( Liu and Reppert , 2000 ) . In addition , it has been found that 1 hr and 6 hr pulses of GABA elicit phase shifts of similar magnitude ( Liu and Reppert , 2000 ) , suggesting that SCN resetting responses are not fundamentally altered by longer-term treatment . Consistent with this idea , the waveform of GABA-induced resetting displayed by L12 SCN slices replicates that elicited in previous work ( Liu and Reppert , 2000 ) . Although the absolute magnitude and precise phasing of resetting responses vary across studies , this may reflect other important methodological differences ( e . g . , phase via PER2::LUC versus electrical rhythms [Liu and Reppert , 2000] , SCN slice from adult mouse versus dissociated SCN neurons from early postnatal pups [Liu and Reppert , 2000] ) . Going forward , the assay developed here may be useful to examine whether photoperiod modulates sensitivity to GABA signaling across the circadian cycle . Further , imaging may add insight into cell-type specific responses since the signal recorded with luminometry is a population-level rhythm that largely reflects the brightest regions of the slice ( i . e . , shell ) ( Azzi et al . , 2017 ) . The results of further studies such as these may provide additional insight into seasonal plasticity of GABA circuits and how these intersect with the intracellular circadian clock . Last , the current results may have implications for annual changes in human health and physiology . Seasonal disorders manifest in 3–10% of people , with the greatest incidence occurring at high latitudes where seasonal variation in photoperiod is most pronounced ( Wirz-Justice , 2018 ) . Evidence supports a light-based , circadian basis for seasonality ( Wehr et al . , 2001; Wirz-Justice , 2018 ) , but the neurobiological basis of human variation in sensitivity to seasonal disorders remains unknown . Like humans , many rodent species are characterized by two seasonal morphs: short day responders and short day non-responders ( Nelson , 1987 ) . The Siberian hamster , a classic model for interrogating photoperiodic circuits , typically displays seasonal alterations in reproduction , metabolism , immune function , affective state , and cognition ( Goldman , 1999 ) . However , a subset of Siberian hamsters fail to transition to the winter phenotype after exposure to long days due to a genetic pre-disposition that locks their SCN into a summer state ( Margraf et al . , 1991; Puchalski and Lynch , 1991; Goldman et al . , 2000 ) . It has been theorized that short day non-responsiveness reflects individual differences in SCN coupling that hinders network plasticity . Here we demonstrate that GABAA signaling regulates seasonal plasticity in the state of the SCN network , raising the possibility that individual differences in SCN GABA circuits may contribute to natural variation in seasonal responsiveness . Thus , further investigation of mechanisms and factors that regulate photoperiodic remodeling of GABA circuits in the SCN may help to better understand the neural basis of annually reoccurring disease .
All procedures involving mice were conducted according to the NIH Guide for the Care and Use of Animals and were approved by the Institutional Animal Care and Use Committees at Marquette University . Homozygous mPer2Luc mice ( Yoo et al . , 2004 ) were bred and raised under a standard 24 hr light:dark cycle with 12 hr light and 12 hr darkness ( LD12:12 , lights-off: 1800 CST ) . At 10–12 weeks of age , male mPer2Luc mice were individually housed with either 12 hr of light per day ( L12 , lights-off 1800 CST ) or 20 hr of light per day ( L20 , lights-off: 1800 CST ) for a minimum of 8 weeks . Ambient room temperature was maintained at 22 C ± 2 C under both colony and experimental conditions , and mice had ab libitum access to water and food ( Teklad Rodent Diet #8604 ) . Coronal SCN slices ( 150 μm ) were collected from mPer2Luc mice and cultured as described previously ( Evans et al . , 2013 ) . Briefly , SCN slices were collected 4–6 hr before lights off , since dissections during late subjective day do not markedly reset the phase of the SCN ( Davidson et al . , 2009 ) . Each SCN was cultured on a membrane insert in a dish containing 1 . 2 mL of air-buffered Dulbecco’s modified explant medium ( DMEM , Sigma D2902 ) supplemented with 0 . 1 mM beetle luciferin , 0 . 02% B27 ( Gibco 17504 ) , 0 . 01% HEPES ( Gibco 15630 ) , 0 . 005% NaCHO3 ( Gibco 25080 ) , 0 . 004% Dextrose ( Sigma G7021 ) , and 0 . 01% penicillin/streptomycin ( Gibco 15140 ) . For resetting experiments , PER2::LUC rhythms were monitored using an Actimetrics luminometer housed within a incubator set to 37°C . At times spanning the circadian cycle , GABA ( 200 μM ) or the equivalent volume of DMEM was added to the culture medium after 3 days in vitro . The direction and magnitude of the resetting response was quantified using Lumicycle analyses software ( Actimetrics ) by measuring the difference between the predicted and actual time of peak PER2::LUC expression on the cycle following drug/vehicle application . Derivatized GABA concentration over time in culture was quantified using high-performance liquid chromatography ( HPLC ) coupled to fluorescence detection . Briefly , media samples were diluted 200-fold with HPLC grade water prior to pre-column derivatization with ortho-phthalaldehdye ( OPA ) in the presence of 2-mercaptoethanol using a Shimadzu LC10AD VP autosampler . Chromatographic separation was achieved using a Kinetex XB C-18 ( 50 × 4 . 6 mm , 2 . 6 μm; Phenomenex ) and a mobile phase consisting of 100 mM Na2HPO4 , 0 . 1 mM EDTA , and 10% acetonitrile ( pH of 6 . 04 ) . GABA was detected using a Shimadzu 10RF-AXL fluorescence detector with an excitation and emission wavelength of 320 and 400 nm , respectively . Brains were collected at eight timepoints spanning the circadian cycle ( n = 4/timepoint/photoperiod ) , and protein expression was evaluated in free-floating slices with triple immunohistochemistry using methods described previously ( Evans et al . , 2015 ) . Briefly , SCN slices ( 40 μm ) were washed six times in PBS , blocked in normal donkey serum , and incubated for 48 hr at 4°C with primary antibodies for AVP ( 1:1K , Peninsula Laboratories , Cat# T-5048 , RRID: AB_2313978 ) , NKCC1 ( 1:500 , Abcam , Cat#ab99558 , RRID:AB_10675276 ) and KCC2 ( 1:500 , EMD Millipore , Cat# 07–432 , RRID:AB_310611 ) . Following primary antibody incubation , SCN slices were washed six times in PBS , incubated for 2 hr at room temperature with secondary antibodies ( 1:500 for each: RRID:AB_10893040 , Alexa Fluor 488; RRID:AB_10894526 , Alexa Fluor 555; RRID:AB_10895029 , Alexa Fluor 647 ) , washed six times in PBS , and then mounted onto microscope slides using Promount anti-fade gold reagent . Fluorescence images were obtained with a Nikon A1R+ confocal microscope using identical settings for all samples . Images were analyzed in ImageJ . Using AVP immunoreactivity , regions of interest ( ROIs ) were selected for each SCN slice by placing a circular ROI on the AVP+ dorsomedial shell and a second ROI of identical size and shape in the AVP-negative ventral core . To calculate chloride co-transporter expression , each ROI was transferred onto KCC2 and NKCC1 images thresholded for background subtraction . For each sample , the KCC2/NKCC1 ratio was calculated and averaged across the two SCN lobes . Similar results were obtained with free-form ROIs encompassing each SCN compartment , but this approach was discarded because AVP expression in the SCN core was higher and more variable across samples . Lastly , mean daily expression of each chloride co-transporter was calculated by collapsing data from all samples collected at different timepoints . SCN PER2::LUC rhythms were imaged as in Evans et al . ( 2013 ) . Briefly , PER2::LUC was collected using a Stanford Photonics CCD camera within a light-tight incubator set to 37°C . For drug treatments , pharmacological agents were added at the start of the experiment directly to the culture medium ( always <0 . 05% total volume of medium ) and remained for the duration of the recording . SCN slices were treated with the NKCC1 antagonist Bumetanide ( 80 μM , BU ) , the KCC2 antagonist VU0240551 ( 80 μM , VU ) , the KCC2 agonist CLP290 ( 100 μM ) , the carbonic anhydrase antagonist Acetazolamide ( ACE , 100 μM ) , or vehicle ( DMEM with DMSO <0 . 1% ) . To maintain drug efficacy for long-term recordings in this interface-culture preparation , drug concentration was increased slightly relative to previous work employing bath application of drugs for short-term electrophysiological recordings ( i . e . , 50 μM BU [Choi et al . , 2008] , 75 μM VU [Haam et al . , 2012] , 50 μM ACE [Ostroumov et al . , 2016] ) . To assess effects of drug treatment on whole SCN rhythms , the time series for the entire SCN was extracted using ImageJ and analyzed with Lumicycle analyses software . To analyze SCN function at the network and cellular level , Matlab-based computational analyses were used as described previously ( Evans et al . , 2013; Evans et al . , 2011 ) . Briefly , a time series was generated for each 12-pixel diameter ROI of the image using a uniform grid with 2-pixel spacing , for which the linear trend was eliminated and a Butterworth filter was applied to remove high- and low-frequency interference . To generate composite phase maps illustrating results for an entire group , samples were aligned to the same X-Y coordinates by minimizing the sum of squared difference of the 24 h-summed bioluminescence profiles . To evaluate cellular rhythms , an iterative process was employed to locate and extract data from cell-like ROIs after background and local noise subtraction . To measure interactions among SCN shell and core neurons , we calculated the phase difference between peak PER2::LUC in SCN core ROIs and the average peak time of SCN shell ROIs over Cycles 2–4 in vitro , as in Evans et al . ( 2013 ) . The change in the phase relationship over time in vitro was used to quantify the magnitude and direction of coupling responses . Cellular period was calculated as the average peak-to-peak cycle length over the same interval in vitro . For both coupling and period responses , the relative peak time was determined subtracting the peak time of each cell on Cycle two from the average peak time of the complementary cell type . For drug treatments in vivo , mice received bumetanide ( BU , 5 mg/kg ) or vehicle ( DMSO ) in their drinking water for 8 weeks of photo-entrainment ( L12 or L20 ) and 5 weeks of constant darkness ( DD ) . Bumetanide levels were tested in plasma and hypothalamus with a bumetanide ELISA Kit ( Neogen Corporation , Cat#103719–1 ) according to manufacturer’s instructions . Drug concentration was assessed with a VersaMax Microplate Reader ( Molecular Devices ) and compared to a standard curve . To assess effects of bumetanide on circadian behavior , daily locomotor rhythms were monitored via wheel-running cages and the Clocklab acquisition system ( Actimetrics , Evanston , IL ) . Using ClockLab software ( Actimetrics , Evanston , IL ) , the time of activity onset , activity offset , activity duration , and the number of wheel revolutions was determined for each day of the experiment . The time of activity onset was identified each day as the first bin above a threshold of 15 counts , preceded by at least 6 hr of inactivity and followed within 30 min by at least two more bins likewise above threshold . Activity offset was determined by a similar but opposite rule . The difference between the time of activity offset and onset was used to calculate activity duration . Free-running period was measured from the slope of a regression line fit to activity onsets over the four weeks in DD . Lastly , effects of BU on jetlag behavior was assessed in a subset of mice by abruptly advancing the L12 light:dark cycle by 6 hr . Rate of re-entrainment was quantified by calculating the number of days required for each mouse to shift activity onset by 6 hr , at which point the activity rhythm was re-aligned to the new LD cycle . Data are represented in figures as Mean ± SEM . Statistical analyses were performed with JMP software ( SAS Institute , Cary , NC ) or CircWave software ( Oster et al . , 2006 ) . Phase , coupling , and period response curves were generated using a 4 hr running average , then collapsed into time bins for statistical analyses . Full Factorial ANOVA was used to assess effects of photoperiod and circadian time ( or drug condition or cell type ) , as well as the interaction of the two main variables . Post hoc pairwise comparisons were performed with Least Square Means Contrasts to correct for family wise error . Changes in circadian behavior over time and its interaction with photoperiod and drug condition were analyzed with Full Factorial Repeated Measures ANOVA , followed by post hoc Least Square Means Contrasts . Statistical significance was set at p≤0 . 05 in all cases . | In winter , as the days become shorter , millions of people find that their mood and energy levels start to drop . They crave carbohydrates , struggle with their weight , and find it harder to get out of bed in the mornings . These individuals are suffering from the ‘winter blues’ or seasonal affective disorder ( SAD ) , and most find that their symptoms spontaneously improve in the spring when the days become longer again . Many also benefit from bright light therapy during the winter months , but not everyone responds fully to this treatment , so additional options are needed . The winter blues occur when the brain adjusts to changes in day length with the onset of winter . The brain region responsible for making this adjustment is the suprachiasmatic nucleus ( SCN ) . The SCN is the master clock of the brain that coordinates the body’s circadian rhythms – the daily fluctuations in things like appetite , body temperature , sleep and wakefulness . But as well as being the brain’s clock , the SCN is also the brain’s calendar . In winter , when the days are short , SCN neurons coordinate their activity and fire in synchrony . But in summer , when the days are long , SCN neurons divide into two clusters , which fire at different times . By transitioning between these two states , the SCN helps the body adjust to seasonal changes in day length . Rohr , Pancholi et al . now provide new insight into the mechanism behind this process by showing that light alters the neurochemistry of the SCN . Exposing mice to long days causes a brain chemical called GABA to switch from inhibiting neurons in the SCN to activating them . Blocking this switch from inhibition to activation locks the SCN into its 'summer state' . Rohr , Pancholi et al . propose that this failure to transition to the winter state may be an interesting way to prevent the winter blues . While much remains to be learned about this process , these findings pave the way for better understanding the neurobiology of winter depression and how best to treat it . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2019 | Seasonal plasticity in GABAA signaling is necessary for restoring phase synchrony in the master circadian clock network |
Sodium/proton antiporters maintain intracellular pH and sodium levels . Detailed structures of antiporters with bound substrate ions are essential for understanding how they work . We have resolved the substrate ion in the dimeric , electroneutral sodium/proton antiporter PaNhaP from Pyrococcus abyssi at 3 . 2 Å , and have determined its structure in two different conformations at pH 8 and pH 4 . The ion is coordinated by three acidic sidechains , a water molecule , a serine and a main-chain carbonyl in the unwound stretch of trans-membrane helix 5 at the deepest point of a negatively charged cytoplasmic funnel . A second narrow polar channel may facilitate proton uptake from the cytoplasm . Transport activity of PaNhaP is cooperative at pH 6 but not at pH 5 . Cooperativity is due to pH-dependent allosteric coupling of protomers through two histidines at the dimer interface . Combined with comprehensive transport studies , the structures of PaNhaP offer unique new insights into the transport mechanism of sodium/proton antiporters .
The Na+/H+ antiporter NhaP from Pyrococcus abyssi ( PaNhaP ) exchanges protons against sodium ions across the cell membrane . PaNhaP is a functional homologue of the human Na+/H+ exchanger NHE1 , which controls intracellular pH and Na+ concentration . NHE1 is an important drug target ( Karmazyn et al . , 1999 ) , but its structure and detailed mode of action are unknown . Transport mechanisms of eukaryotic membrane proteins are conserved in the more robust prokaryotic transporters from thermophilic bacteria and archaea ( Yamashita et al . , 2005; Boudker et al . , 2007; Lee et al . , 2013 ) . High-resolution structures of such homologues are of great value for understanding the mechanisms of cation/proton antiport , provided that ( i ) the transported substrate ions are resolved , ( ii ) structures of the same transporter are available in different conformations , and ( iii ) kinetic data of substrate binding and transport are available . In this paper we report the structure of the electroneutral Na+/H+ antiporter PaNhaP from the hyperthermophilic archaeon P . abyssi in two different conformations at pH 4 and pH 8 , with the substrate ion resolved at pH 8 . We show that , like NHE1 , transport by PaNhaP is cooperative in a pH-dependent manner , indicating a pH-dependent allosteric interaction of protomers in the dimer . The first structure of a cation-proton antiporter ( CPA ) revealed that Escherichia coli Na+/H+ NhaA ( EcNhaA ) is a dimer in the membrane ( Williams et al . , 1999 ) . The 6 Å map of EcNhaA resolved 12 trans-membrane helices ( TMH ) in the protomer , arranged in a 6-helix bundle , plus a row of six TMHs at the dimer interface ( Williams , 2000 ) . A membrane dimer was also found for the NhaP1 antiporter from Methanocaldococcus jannaschii ( MjNhaP1 ) ( Vinothkumar et al . , 2005; Goswami et al . , 2011; Paulino and Kühlbrandt , 2014 ) . MjNhaP1 , PaNhaP and the medically important NHE1 belong to the CPA1 subfamily ( Brett et al . , 2005 ) of antiporters , which exchange Na+ and protons with 1:1 stoichiometry and are thus electroneutral . By contrast , EcNhaA and TtNapA from Thermus thermophilus , as well as the eukaryotic NHA1-2 and AtChx1 ( Brett et al . , 2005 ) , belong to the CPA2 subfamily of electrogenic antiporters , which exchange one Na+ against two protons . Neither the x-ray structure of EcNhaA ( Hunte et al . , 2005 ) nor that of TtNapA ( Lee et al . , 2013 ) resolved the substrate ion .
Crystals of seleno-methionine derivatized PaNhaP grown at pH 8 diffracted isotropically to 3 . 15 Å resolution . The structure was solved by SAD ( Tables 1 and 2 ) . Twelve out of the 14 SeMet positions in the asymmetric unit containing one PaNhaP dimer were identified ( Figure 1—figure supplement 1 ) . Seen from the cytoplasm , the PaNhaP dimer is roughly rectangular , with a long axis of 90 Å and a short axis of 53 Å ( Figure 1A , Figure 1—figure supplement 2A ) . Each protomer has 13 TMHs ( H1-H13 ) connected by short loops or helices on the membrane surface . H4-6 and H11-13 form the 6-helix bundle , while H1-3 and H7-10 form the dimer interface . H1-6 and H8-13 are two halves of an inverted 6-helix repeat , connected by H7 . Several helices are highly tilted , especially H7 and H8 , which include angles of more than 45° with the membrane normal while others , in particular H6 and H10 , are bent . H5 and H12 in the 6-helix bundle are discontinuous . Their cytoplasmic and extracellular halves ( referred to as H5C , H5E and H12C , H12E respectively ) are each connected by unwound stretches with antiparallel orientation , which cross one another in the centre of the protomer ( Figure 1 , Figure 1—figure supplement 2 ) . The membrane surfaces are marked by three short amphipathic helices connecting H3 to H4 on the cytoplasmic side , H6 to H7 and H10 to H11 on the extracellular side . H10 protrudes by 11 Å on the cytoplasmic surface , and the helix hairpin connecting H12 to H13 protrudes by about 7 Å on the extracellular side . The loops connecting helices H1 to H2 and H8 to H9 are ∼10 Å below the cytoplasmic or extracellular surface ( Figure 1B , Figure 1—figure supplement 2B ) . 10 . 7554/eLife . 03579 . 003Table 1 . Data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 003SeMet @ pH 8Thallium @ pH 8Native @ pH 4Data collectionSLS PXII Wavelength0 . 9790 . 9790 . 978 Space groupP21P21P64Cell dimensions a , b , c ( Å ) 54 . 5 , 107 . 9 , 107 . 954 . 1 , 107 . 4 , 99 . 8109 . 6 , 109 . 6 , 209 . 6 α , β , γ ( ° ) 90 . 0 , 95 . 2 , 90 . 090 . 0 , 96 . 4 , 90 . 090 . 0 , 90 . 0 , 120 . 0 Resolution ( Å ) 48 . 5–3 . 15 ( 3 . 35–3 . 15 ) 49 . 6–3 . 20 ( 3 . 40–3 . 20 ) 48 . 6–3 . 50 ( 3 . 72–3 . 50 ) Rpim0 . 033 ( 0 . 503 ) 0 . 038 ( 0 . 622 ) 0 . 021 ( 0 . 486 ) I / σI11 . 9 ( 1 . 5 ) 13 . 4 ( 1 . 8 ) 19 . 9 ( 1 . 9 ) CC*1 . 000 ( 0 . 943 ) 1 . 000 ( 0 . 936 ) 1 . 000 ( 0 . 906 ) Completeness ( % ) 99 . 5 ( 99 . 2 ) 99 . 6 ( 99 . 4 ) 100 . 0 ( 100 . 0 ) Multiplicity10 . 8 ( 10 . 4 ) 17 . 1 ( 17 . 4 ) 9 . 2 ( 9 . 1 ) Refinement Resolution ( Å ) 48 . 5–3 . 15 ( 3 . 35–3 . 15 ) 49 . 6–3 . 20 ( 3 . 40–3 . 20 ) 48 . 6–3 . 5 ( 3 . 72–3 . 5 ) Unique reflections38 , 95234 , 76333 , 232 Reflections in test set211118841782 Rwork/Rfree ( % ) 23 . 8/27 . 8 ( 31 . 8/39 . 9 ) 24 . 8/29 . 5 ( 35 . 9/43 . 4 ) 24 . 1/26 . 4 ( 31 . 8/35 . 6 ) CC ( work ) /CC ( free ) 0 . 843/0 . 898 ( 0 . 842/0 . 760 ) 0 . 861/0 . 754 ( 0 . 813/0 . 713 ) 0 . 791/0 . 935 ( 0 . 749/0 . 617 ) Wilson B-Factor ( Å2 ) 13381146 No . atoms in AU671566516592 Protein658265606560 Ligands1298131 Water4101 r . m . s . deviations: Bond lengths ( Å ) 0 . 0030 . 0030 . 009 Bond angles ( ° ) 0 . 7580 . 7141 . 00210 . 7554/eLife . 03579 . 004Table 2 . Data collection and phasing statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 004Dataset 1Dataset 2MergeData collectionSLS PXII Wavelength0 . 9790 . 9790 . 979 Space groupP21P21P21 Cell dimensions a , b , c ( Å ) 54 . 7 , 109 . 0 , 110 . 854 . 6 , 108 . 3 , 110 . 554 . 7 , 108 . 9 , 110 . 7 α , β , γ ( ° ) 90 . 0 , 94 . 6 , 90 . 090 . 0 , 95 . 0 , 90 . 090 . 0 , 94 . 7 , 90 . 0 Resolution ( Å ) 49 . 3–3 . 8 ( 3 . 97–3 . 8 ) 49 . 0–3 . 8 ( 3 . 97–3 . 8 ) 49 . 2–3 . 8 ( 3 . 97–3 . 8 ) Rpim0 . 029 ( 0 . 470 ) 0 . 034 ( 0 . 228 ) 0 . 036 ( 0 . 315 ) I / σI13 . 8 ( 2 . 1 ) 12 . 4 ( 3 . 9 ) 14 . 5 ( 2 . 9 ) CC*1 . 000 ( 0 . 929 ) 0 . 996 ( 0 . 985 ) 1 . 000 ( 0 . 981 ) Completeness ( % ) 99 . 7 ( 99 . 7 ) 99 . 7 ( 99 . 6 ) 100 ( 100 ) Multiplicity24 . 5 ( 16 . 5 ) 9 . 1 ( 9 . 5 ) 33 . 0 ( 25 . 8 ) Phasing CCanom0 . 348 Anom slope1 . 061 FOM after Phasing ( Refmac ) 0 . 230 FOM after DM ( Parrot ) 0 . 59410 . 7554/eLife . 03579 . 005Figure 1 . PaNhaP at pH 8 . ( A ) Cytoplasmic view of the PaNhaP dimer . Helices H1 to H13 are color-coded and numbered in one protomer . In the other protomer only the partly unwound helices H5 and H12 are coloured . ( B ) Side view with the C-terminus of helix H13 on the cytoplasmic side . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 00510 . 7554/eLife . 03579 . 006Figure 1—figure supplement 1 . Experimental electron density map of PaNhaP . The structure was solved at 3 . 15 Å by SeMet-SAD . Twelve of the 14 SeMet positions ( grey spheres ) in the asymmetric unit were resolved in the anomalous difference map ( purple densities at 4σ ) . SeMet positions werfe used to generate an initial map ( blue density at 1σ ) and to trace the polypeptide chain . ( A ) Cytoplasmic view of PaNhaP dimer . ( B ) Side view with the extracellular side above . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 00610 . 7554/eLife . 03579 . 007Figure 1—figure supplement 2 . X-ray structure of PaNhaP . Cartoon representation of PaNhaP with helices shown as cylinders . ( A ) Cytoplasmic and ( B ) side view of the dimer colour-coded as in the main figure . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 00710 . 7554/eLife . 03579 . 008Figure 1—figure supplement 3 . Hydrophilic cavities in PaNhaP . Side view of the dimer with hydrophilic cavities coloured by surface potential . The substrate-binding site is accessible from the cytoplasmic side through the cytoplasmic funnel and a narrow polar channel . Access from the extracellular side is blocked . One water molecule is trapped in an enclosed polar cavity near the ion-binding site . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 008 On the cytoplasmic side of the protomer , a solvent-filled ∼16 Å-deep funnel , lined by H3 , H5C , H6 , and H10 , penetrates to the centre of the protomer between the 6-helix bundle and the dimer interface ( Figure 1—figure supplement 3 , Video 1 ) . A second , narrow polar channel , lined by the unwound stretches of H5C , H12C and the cytoplasmic halves of H6 and H13 , extends from the cytoplasmic surface to the region near the deepest point of the funnel ( Figure 1—figure supplement 3 , Video 1 ) . On the extracellular side , a deep cavity on the twofold axis of the dimer , lined by interface helices H1 , H3 , H8 and H10 , extends ∼27 Å into the hydrophobic protein interior . Electron density in this cavity indicated bound lipid ( Figure 2 ) , which was identified by thin-layer chromatography as phosphatidyl ethanolamine ( PE ) , carried over from the E . coli expression host . The lipid stretches from the 6-helix bundle of one protomer to the interface helices H3 and H10 of the other , providing a hydrophobic link between them . The cavity is large enough to accommodate two lipids , only one of which was resolved in the dimer ( Figure 2 ) . The surface potential of the dimer indicates clusters of charged residues on both sides of the membrane ( Figure 2—figure supplement 1 ) . The cytoplasmic ends of H10 and H7 are positively charged , carrying a total of seven lysine and arginine residues . The deep funnel on the cytoplasmic surface is lined by negative charges , which would attract positively charged substrate ions . 10 . 7554/eLife . 03579 . 009Video 1 . Movie of PaNhaP monomer with hydrophilic cavities . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 00910 . 7554/eLife . 03579 . 010Figure 2 . Hydrophobic extracellular cavity with bound lipid . ( A ) One lipid molecule ( PE , green ) in the cavity between the two protomers in the dimer contributes to the hydrophobic contacts across the dimer interface . The extracellular surface is slightly negatively charged . ( B ) The alkyl chain of the lipid extends to the center of the molecule . ( C ) The lipid-facing surface of the central cavity is mainly hydrophobic . The surface potential was calculated at pH 7 . 0 by APBS . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 01010 . 7554/eLife . 03579 . 011Figure 2—figure supplement 1 . pH-dependent charge distribution . The surface potential of the PaNhaP dimer was calculated at pH 4 , pH 6 , and pH 8 by APBS and visualized in PyMOL . At pH 4 the cytoplasmic surface is strongly positively charged . At pH 6 , the extracellular surface and the cytoplasmic funnel are largely neutral . At pH 8 the extracellular surface is predominantly negatively charged and the cytoplasmic funnel is strongly negatively charged , inhibiting the release of substrate ions . The ion-binding site is located at the bottom of the cytoplasmic funnel . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 011 Crystals of PaNhaP grown at pH 8 soaked with thallium acetate diffracted to 3 . 2 Å ( Table 1 ) . Two thallium ions were identified in the dimer by anomalous scattering , one each near the deepest point of the cytoplasmic funnel in the two protomers ( Figure 1—figure supplement 3 , Video 1 , Figure 3A ) . The Tl+ ions were located ∼14 Å below the cytoplasmic surface and ∼22 Å from the extracellular surface . The ion-binding site is accessible from the cytoplasm but not from the extracellular side , so that the structure shows the inward-open conformation of PaNhaP ( Figure 1—figure supplement 3 , Video 1 ) . Like Na+ and Li+ , but unlike K+ , Tl+ is a substrate of PaNhaP ( Figure 4 ) . The thallium ions and their surroundings provide a unique view of the ion-binding site and substrate ion coordination in sodium-proton antiporters ( Figure 3B , C ) . Three acidic side chains in three different TMHs contribute to substrate ion-binding . The carboxyl groups of Glu73 in H3 and Asp159 in H6 coordinate the substrate ion directly . Asp130 in the unwound stretch of H5 interacts with the ion via a bound water molecule ( Figure 3 ) . The main-chain carbonyl of Thr129 , likewise in the unwound stretch of H5 , and the hydroxyl side chain of Ser155 in H6 provide two additional ligands , bringing the total up to five . The ion coordination geometry is that of a distorted trigonal bipyramid , with Asp159 , Ser155 and the water molecule forming a triangle around the central substrate ion , and the Thr129 main chain carbonyl and Glu73 at the tips of the bipyramid ( Figure 3B , C ) . 10 . 7554/eLife . 03579 . 012Figure 3 . Substrate ion coordination in PaNhaP . ( A ) Section view of the ion-binding site and interface region of PaNhaP from the cytoplasmic side . Interface helices of the two protomers are shown in blue and beige , respectively . The acidic side chains of Glu73 , Asp159 , a water molecule held by Asp130 , the hydroxyl group of Ser155 and the main-chain carbonyl of Thr129 coordinate the substrate ion . The anomalous density for the Tl+ ion ( grey sphere ) in the substrate-binding site between helix H3 , H6 and the unwound stretch of H5 is shown in magenta at 4σ . The 3σ omit map for the H2O molecule next to Tl+ is green . The water molecule near Glu154 and Asn158 is not directly involved in ion coordination . ( B , C ) Detailed views of the substrate-coordinating residues from the extracellular and cytoplasmic side , respectively . ( D ) Side view of core helices and substrate-binding residues in the 6-helix bundle . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 01210 . 7554/eLife . 03579 . 013Figure 4 . Ion selectivity of PaNhaP . Ion selectivity was determined by acridine orange fluorescence at pH 6 . Na+ , Li+ , Tl+ are transported by PaNhaP , K+ is not . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 013 The second , narrow polar channel next to the cytoplasmic funnel ( Figure 1—figure supplement 3 , Video 1 ) leads to an enclosed polar cavity near Asp93 , Thr129 , Asn158 and the ion pair Glu154/Arg337 , which are highly conserved in the CPA1 antiporters ( Goswami et al . , 2011 ) . A water molecule in the enclosed cavity links the functionally important groups that surround it . The Glu154/Arg337 ion bridge and Thr129 separate the cavity from the narrow polar channel . The ion-binding site at the end of the cytoplasmic funnel is accessible from both the polar cavity and the narrow polar channel via Thr129 , Ser155 , and Asn158 ( Figure 1—figure supplement 3 , Video 1 , Figure 3 ) . The structure of PaNhaP crystals grown at pH 4 was determined at 3 . 5 Å ( Table 1 ) by molecular replacement . As at pH 8 , the ion-binding site is accessible from the cytoplasm via the cytoplasmic funnel , but not from the extracellular side . Both structures therefore show an inward-open state . In contrast to the pH 8 structure , the second narrow polar channel is blocked at pH 4 by rearrangements of the surrounding residues Ile151 , Phe355 , Gly359 . The most conspicuous differences to the pH 8 structure are observed near the dimer interface . At pH 8 , the His292 sidechains in H10 of the two protomers form a 15 Å chain of hydrogen bonds with the Glu233 residues near the cytoplasmic ends of H8 ( Figure 5 , Video 2 , Figure 5—figure supplement 1 ) . At pH 4 , each of the two histidines moves by 6–8 Å , apparently due to electrostatic repulsion upon protonation at acidic pH ( Figure 2—figure supplement 1 ) . This pH-induced conformational change disrupts the chain of hydrogen bonds linking the two protomers ( Figure 5—figure supplement 1 ) . Other major pH-induced changes are found in the ion bridges linking the protomers across the dimer interface ( Figure 5—figure supplement 1 ) . At pH 8 , Arg25/Glu228 and Arg26/Asp231 connect the cytoplasmic ends of H8 and H1 , while the Glu8/Arg249 bridge links the extracellular ends of these helices . At pH 4 , all six ion pairs break , apparently due to partial protonation of the acidic sidechains , so that each protomer tilts away from the dimer interface ( Video 2 , Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 03579 . 014Figure 5 . pH-induced conformational changes in the PaNhaP dimer . ( A ) cytoplasmic view , ( B ) side view as in Figure 1 . At pH 4 ( red ) , helix H4 moves towards the cytoplasm by 1 . 5 Å . Within the 6-helix bundle , the extracellular ends of helix H5E and H6 move towards H12 by ∼1 . 5 Å . Helix H11 and H13 tilt by about 2–3° each , such that the cytoplasmic end of helix H11 moves towards H12C , which shifts by a similar amount in the same direction . The extracellular end of helix H12E moves towards helix H3 by ∼3 Å . The rmsd between the structures at pH 8 and pH 4 is 1 . 57 Å . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 01410 . 7554/eLife . 03579 . 015Figure 5—figure supplement 1 . pH-induced conformational changes at the dimer interface . The pH 8 structure ( transparent ) is superposed on the pH 4 structure . Red and black arrows indicate mainchain and sidechain movements , respectively . ( A ) At pH 4 , His292 at the dimer interface moves by 6–8 Å from its pH 8 position towards the centre of H3 . At pH 8 , the two His292 in the dimer form a line of hydrogen bonds with glutamates Glu233 on either side ( dashed lines ) . Protonation at pH 4 disrupts the hydrogen bond network , so that the His sidechains and H10 move towards the helix bundle . ( B ) Six salt bridges between the cytoplasmic end of H1 and H7/8 present at pH 8 break at pH 4 . The N-terminus of H1 on the extracellular side becomes more ordered at pH 4 and moves towards H8 . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 01510 . 7554/eLife . 03579 . 016Figure 5—figure supplement 2 . pH-induced conformational changes in the substrate binding site . ( A ) In protomer A , Asp130 in the unwound stretch of H5 and Glu73 in H3 move towards the ion-coordinating Asp159 . The ion-binding site in protomer B changes only slightly from pH 8 to pH 4 . ( B ) At pH 8 , Tyr31 is within hydrogen bonding distance of Asp130 in the substrate-binding site . At pH 4 , the two residues do not interact . The rmsd between the pH 8 and pH 4 structures is 1 . 6 Å . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 01610 . 7554/eLife . 03579 . 017Video 2 . pH-induced conformational changes in PaNhaP . A morph between the pH 4 and pH 8 structures reveals only small changes in the 6-helix bundle , but significant rearrangements at the dimer interface . Six ion bridges that lock the two protomers together at pH 8 break at pH 4 . As a result , the two protomers tilt away from each other at lower pH . His292 has a pivotal role in the allosteric pH-dependent protomer interaction . At pH 4 , the protonated His292 side chains on the cytoplasmic side of the dimer interface repel one another by electrostatic repulsion , resulting in a ∼7 Å movement that disrupts the hydrogen-bonding network with Glu233 . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 017 Other significant pH-induced differences occur at the N-terminus of the protomer , where residues 3–6 become ordered at pH 4 , so that H1 extends by one turn , and shifts by 3 Å towards the cytoplasmic side ( Figure 5—figure supplement 1 ) . In the ion-binding site itself , the sidechain of Asp130 in protomer A moves by 2 . 7 Å into the space that is occupied by the substrate ion at pH 8 ( Figure 5—figure supplement 2A ) . This movement , which is not observed in the other protomer ( Figure 5—figure supplement 2B ) , could displace a bound substrate ion or prevent ion binding . At pH 4 , the conserved Asn158 that interacts with Asp93 at pH 8 moves by ∼2 . 5 Å towards the ion-coordinating Asp159 in protomer A , forming a H-bond network with Thr129 and the main-chain carbonyls of Glu154 and Ser155 . In this way , the reorientation of Asn158 may regulate access to the ion-binding site through the narrow polar channel ( Figure 1—figure supplement 3 , Video 1 ) . A chain of hydrogen bonds stretches from Glu290 in H10 via His75 near the cytoplasmic end of H3 to Glu73 , the only ion-coordinating sidechain from one of the interface helices ( Figure 3A ) . This residue most likely relays allosteric changes from the dimer interface to the ion-binding site . An opening of the ion bridges that link H1 and H8 and the movement of the adjacent H2 in the pH 8 to pH 4 transition is likely to affect substrate binding via Tyr31 , which is within H-bonding distance of the substrate-coordinating Asp130 ( Figure 5—figure supplement 2 ) . In this way , the conformational changes caused by repulsion of the protonated histidines 292 at the dimer interface are relayed to the ion-binding site to modulate the Na+ binding affinity in a pH-dependent manner ( Figure 6 ) . 10 . 7554/eLife . 03579 . 018Figure 6 . Transport activity of PaNhaP . ( A ) pH dependence of transport activity determined by 22Na+ uptake with inside-acidic PaNhaP proteoliposomes . The antiporter is active at pH 5 and pH 6; at pH 4 and pH 7 transport drops to background level . ( B ) Concentration-dependent 22Na+-uptake by inside-acidic PaNhaP proteoliposomes at pH 5 gives a vmax of 87 . 9 ± 7 . 5 nmol · min−1 · mg−1 at room temperature , indicating a transport rate of 4 . 4 Na+ ions per protomer per minute . At pH 5 the Hill coefficient ( nh ) is 1 . 1 ± 0 . 30 , indicating non-cooperative transport . ( C ) At pH 6 , transport is cooperative , with a Hill coefficient of 1 . 9 ± 0 . 26 , indicating allosteric coupling of the two ion-binding sites in the dimer . vmax at room temperature decreases to 16 . 5 ± 0 . 5 nmol · min−1 · mg−1 . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 01810 . 7554/eLife . 03579 . 019Figure 6—figure supplement 1 . Sodium efflux measurements . ( A ) Sodium efflux was measured under symmetrical pH conditions by acridine orange fluorescence with PaNhaP reconstituted into proteoliposomes . Transport activity of PaNhaP drops towards pH 7 , consistent with 22Na uptake measurements ( Figure 4 ) . Transport was not affected by 100 nM valinomycin ( B , red curve ) , indicating that PaNhaP is electroneutral . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 01910 . 7554/eLife . 03579 . 020Figure 6—figure supplement 2 . Eadie-Hofstee plots . ( A ) Eadie–Hofstee transformation of kinetic data at pH 5 shows a linear correlation typical for Michaelis–Menten kinetics . The data point at 5 µM sodium concentration was omitted from the linear regression . ( B ) Eadie–Hofstee transformation of pH 6 data results in a concave curve , indicating homotropic activation . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 020 22Na uptake into reconstituted PaNhaP proteoliposomes is strongly pH-dependent ( Figure 6A ) . Transport activity was highest at pH 5 , dropping to about 75 % at pH 6 , 20 % at pH 7 , and to background level at pH 8 . At pH 4 , the activity was about 5 % of the peak value at pH 5 , resulting in a roughly bell-shaped pH profile . Sodium uptake measurements performed with reconstituted , inside-acidic proteoliposomes ( Figure 6B , C ) or sodium efflux measurements under symmetrical pH conditions ( Figure 6—figure supplement 1A ) showed comparable transport behaviour at basic pH . Valinomycin had no effect on the transport rate ( Figure 6—figure supplement 1B ) , demonstrating that PaNhaP is electroneutral . Measurements of 22Na+ uptake at pH 6 revealed clear positive cooperativity , with a Hill coefficient of 1 . 9 ( Figure 6C , Figure 6—figure supplement 2B ) . Since PaNhaP forms stable dimers in detergent solution and each protomer binds only one substrate ion at a time , this indicates that the interaction of protomers across the dimer interface is allosteric , such that at pH 6 , an ion binding to one protomer increases the binding affinity of the other , as indicated by the K0 . 5 value of 25 µM ( Figure 6C ) , compared to the Km of 506 µM at pH 5 ( Figure 6C ) . At the pH 5 activity maximum the Hill coefficient was ∼1 , indicating non-cooperative transport ( Figure 6B , Figure 6—figure supplement 2 ) . Note that the pH-dependent allosteric change of the dimer is different from the inside-open to outside-open transition in the transport cycle of the protomer . At room temperature , vmax of PaNhaP at the pH 5 activity maximum was 87 . 9 nmol · min−1 · mg−1 , giving a transport rate of 4 . 4 ± 0 . 4 Na+ ions per minute for each protomer . Between 20°C and 45°C , vmax grew exponentially by a factor of 2 . 1 for every 5°C rise in temperature ( Figure 7A , B ) according to the Arrhenius equation . Extrapolation to 100°C , the physiological temperature for P . abyssi , suggests a rate of about 5000 ions per second . Note that temperature affects the transport rate but not substrate binding ( Figure 7C , D ) . 10 . 7554/eLife . 03579 . 021Figure 7 . Temperature dependence of PaNhaP . ( A , B ) At pH 6 transport activity increases by a factor of 2 . 1 for every 5°C rise in temperature , as measured by sodium efflux under symmetrical pH . The slight rise in fluorescence towards longer times at 40°C and above in A is due to increasing proton leakage of the proteoliposomes . ( C , D ) Effect of temperature on substrate affinity at 25°C ( empty dots ) and 30°C ( filled squares ) measured by ΔpH-driven sodium uptake in proteoliposomes using Acridine orange fluorescence . In contrast to vmax , Km does not change much with increasing temperature ( 1 . 56 ± 0 . 11 mM at 25°C; 1 . 85 ± 0 . 49 mM at 30°C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 021 Residues involved in substrate binding were replaced and the transport activity of mutant proteins was measured in proteoliposomes . Replacement of both ion-coordinating aspartates ( Asp130 and Asp159 ) by serine abolished transport completely ( Figure 8A ) , whereas mutation of Glu73 to alanine increased the activity ( Figure 8B ) , most likely because the substrate ion is released more readily from the binding site . Changing Ser155 to alanine had no significant effect , but mutation of Thr129 to valine that takes this position in eukaryotic CPA1 transporters ( Goswami et al . , 2011 ) , reduced the activity significantly . This was surprising , because Thr129 coordinates the substrate ion not by its sidechain but by its main-chain carbonyl . However the Thr129 sidechain is a potential interaction partner of the conserved Asn158 that may control access to the ion-binding site through the narrow polar channel . A hydrophobic valine in place of Thr129 would interrupt the local network of hydrogen bonds , which could affect ion binding or proton translocation . A mutant in which His292 was replaced by cysteine migrates as a dimer under oxidizing conditions in SDS-PAGE ( Figure 8—figure supplement 1A ) . The activity of the crosslinked dimer was 35 % of wildtype ( Figure 8—figure supplement 1B ) . Under reducing conditions , when the disulfide bridge between the protomers is broken , activity increases to 150 % of wildtype , highlighting the importance of this position for the regulation of transport . 10 . 7554/eLife . 03579 . 022Figure 8 . Transport activity of binding site mutants . Sodium efflux from proteoliposomes at pH 6 was measured to investigate PaNhaP mutants . Antiport activity establishes a ΔpH across the membrane , which results in acridine orange fluorescence quenching . ( A ) Mutation of Asp130 or Asp159 to serine abolishes transport activity . ( B ) Replacement of Thr129 by valine , as in eukaryotic antiporters , reduces the transport activity . Replacement of Glu73 by alanine increases activity significantly , whereas exchanging Ser155 against alanine has no effect compared to wildtype . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 02210 . 7554/eLife . 03579 . 023Figure 8—figure supplement 1 . Interface crosslinks . ( A ) Mutation of His292 to cysteine results in a prominent dimer band under oxidizing conditions , as protomers are crosslinked by covalent disulfide bonds across the dimer interface both in detergent and in proteoliposomes ( PL ) . Addition of reducing agent ( 10 mM DTT ) breaks the disulfide bonds between crosslinked protomers . Asterisks mark the PaNhaP monomer and dimer on SDS-PAGE . ( B ) Transport measurements indicate a 60 % drop in activity of cross-linked PaNhaP dimers compared to wildtype . Under reducing conditions the activity of the H292C mutant is 50 % higher than wildtype , due to weaker protomer interactions . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 023
The trigonal bipyramidal coordination geometry of sodium ions observed in PaNhaP is not uncommon in membrane transporters ( Penmatsa et al . , 2013 ) . The same geometry is found in c-rings of Na+-translocating F-type ATPase of I . tartaricus and F . nucleatum ( Meier et al . , 2009; Schulz et al . , 2013 ) , which , like the archaeal CPA1 antiporters , bind and release Na+ in rapid exchange . Although the ion radius of monovalent Tl+ ( 1 . 5 Å ) is similar to that of K+ ( 1 . 44 Å ) and larger than that of Na+ ( 1 . 12 Å ) ( Shannon , 1976; Cotton and Wilkinson , 1988 ) , Tl+ is able to replace Na+ in PaNhaP . The same has been found for the Na+-dependent aspartate transporter GltPh ( Boudker et al . , 2007 ) , the mammalian glutamate transporter EAAC1 ( Tao et al . , 2008 ) and fructose-1 , 6-biphosphatase ( Villeret et al . , 1995 ) . In GltPh , Na+ but not K+ competes for Tl+ binding , and Tl+ inhibits Na+-driven aspartate transport ( Boudker et al . , 2007 ) . Coordination geometry and ligand distances for Tl+ in PaNhaP are similar to those typically found for protein-bound Na+ in the PDB ( Harding , 2002 ) . The larger ion radius of Tl+ may account for the lower transport rate in PaNhaP . However , Tl+ is a much better substrate than K+ , which is not transported at all ( Figure 4 ) . The selectivity for Na+ over K+ is reminiscent of the striking selectivity of sodium channels , which is thought to be related to ion solvation ( Roux et al . , 2011 ) . Presumably , the same principle applies to the Na+/H+ antiporters . The water molecule between the sidechain of Asp130 and Tl+ indicates that the bound substrate ion retains part of its hydration shell , as complete dehydration is energetically unfavourable . In PaNhaP , all ion-binding residues are found in the first half of the inverted repeat . Interestingly , the structure and interaction of the ion-coordinating Asp159 and Ser155 in H6 resemble those of the inversely oriented Glu408 and Ser404 in H13 in the second half of the inverted repeat ( Figure 3A , D ) . This may imply that an early form of the CPA1 antiporters , which must have arisen by gene duplication of an unknown precursor , had a second , symmetrical ion-binding site that has been lost in the course of evolution . Arg362 in the unwound stretch of H12 , which is essential in MjNhaP1 ( Hellmer et al . , 2003 ) and completely conserved in the CPA1 antiporters ( Hellmer et al . , 2003; Goswami et al . , 2011 ) , may be a tethered positive charge that takes the place of the Na+ ion in the second half of the inverted repeat , in a way similar to the arginine that replaces the co-transported Na+ in the sodium-independent substrate/product antiporter CaiT ( Kalayil et al . , 2013 ) . The transport activity of PaNhaP is highest at pH 5 and declines at higher or lower pH . The resulting bell-shaped pH profile is explained in terms of the Na+ affinity of the acidic residues in the substrate-binding pocket . The protonation state of these is likely to affect the affinity of the binding site for Na+ . At low pH , most if not all of the ion-coordinating carboxyl sidechains ( Glu73 , Asp130 , and Asp159 ) would be protonated , resulting in reduced affinity for Na+ , as has been shown for MjNhaP1 by electrophysiological measurements on solid-supported membranes ( Calinescu et al . , 2014 ) . At pH 5–7 these carboxyl sidechains would be increasingly deprotonated and able to bind and release Na+ ions , as is necessary for transport . At pH > 7 , the ion-binding site is predominantly deprotonated and negatively charged ( Figure 2—figure supplement 1 ) , resulting in an increased Na+ affinity . As a result , the transport rate would decrease , as the ions are bound more tightly . This is consistent with the increased transport rate of the E73A mutant , which has one less carboxyl in the binding site , hence releases Na+ more easily ( Figure 8B ) . In addition , the propagation of the pH-induced conformational changes at the dimer interface via Glu73 or Tyr31 would modulate the binding site by changing the coordination geometry for the ions ( Figure 5—figure supplement 1A , Figure 5—figure supplement 2 ) . Future structure-based molecular dynamics simulations should show how the protonation state of each of these residues influence the affinity of the binding site for Na+ in a pH-dependent manner . The pH-dependent transport activity of PaNhaP suggests a self-regulatory mechanism for the binding site rather than regulation by a separate pH sensor as proposed for EcNhaA ( Herz et al . , 2010; Diab et al . , 2011; Schushan et al . , 2012 ) . At the pH 5 activity maximum of PaNhaP , transport is not limited by Na+ affinity . Under these conditions , substrate binding of the PaNhaP dimer is non-cooperative , but unexpectedly it becomes cooperative at pH 6 . Cooperative ion binding is most likely mediated by Glu73 and may be important for controlling the intracellular pH at neutral or basic pH , where a cooperative increase in Na+ affinity would gradually inhibit substrate release and slow down transport . This may be a safety mechanism to protect the organism against excessive influx of Na+ , and hence efflux of protons , at rising pH , which may be critical for survival . The medically relevant but elusive human Na+/H+ exchanger NHE1 is a dimer ( Fafournoux et al . , 1994 ) like PaNhaP . Several other common features , including high sequence homology ( Goswami et al . , 2011 ) especially of the unwound stretches in H5 and H12 , key residues in the ion binding site such as Ser155 , Asp130 and the ND motif , the functionally important Arg337 and Arg362 ( Hellmer et al . , 2003 ) , as well as pH profiles and transport kinetics suggest that the archaeal and mammalian CPA1 antiporters ( Fuster et al . , 2008 ) work essentially in the same way . Remarkably , NHE1 also shows pH-dependent Na+ cooperativity , with a Hill coefficient of 1 . 8 at pH 6 . 8 that drops to ∼1 at pH 6 ( Fuster et al . , 2008 ) . The PaNhaP structure thus serves as an excellent model for the membrane part of NHE1 . Molecular details of allosteric regulation in NHE1 are likely to be different , as the His292 that reorients in response to pH in PaNhaP is not conserved ( Goswami et al . , 2011 ) . The electrogenic CPA2 antiporters , such as EcNhaA or TtNapA , which exchange two protons against one Na+ , have two conserved aspartates in place of the ND motif in H6 . In terms of its overall structure , TtNapA ( Lee et al . , 2013 ) is more similar to PaNhaP than to EcNhaA ( Hunte et al . , 2005 ) , especially with respect to the dimer interface . The tertiary structure of the CPA antiporters is thus not a diagnostic of electroneutral or electrogenic transport . In Pyrococcus , the Na+ gradient required for ATP synthesis is maintained by specific antiporters ( McTernan et al . , 2014 ) . We therefore assume that PaNhaP , like human NHE1 , utilizes the Na+ gradient across the membrane ( Cohen et al . , 2003 ) for pH homeostasis . Protons , probably in the form of hydronium ions ( H3O+ ) , can reach the binding pocket either through the cytoplasmic funnel or through the narrow polar channel ( Figure 9 ) . Only small rearrangements of the residues lining this channel would be required for H3O+ to pass . Using the second narrow polar channel for proton translocation would physically separate the routes for Na+ and H3O+ on the cytoplasmic side , which may be an advantage as the two ion currents flow in opposite directions . It would also explain why residues that line this channel , in particular the Glu154/Arg337 ion bridge and Asn158 , which do not participate in ion coordination , are so highly conserved in the family . Molecular dynamics simulations and functional analysis of suitable mutants will be required to differentiate between the two proton paths , which both appear equally likely on the basis of the x-ray structures . 10 . 7554/eLife . 03579 . 024Figure 9 . Substrate ion exchange on the cytoplasmic side . The substrate-binding site of PaNhaP is located between the unwound stretches in the six-helix-bundle and the interface domain . The substrate ion is bound by acidic sidechains and polar groups in the bundle helices H5 and H6 , and a glutamate in the interface helix H3 at the deepest point of the cytoplasmic funnel . While the funnel extends between the six-helix bundle and the dimer interface , the narrow polar channel is defined by the bundle helices H5C , H12C , H6 and H13 . Protons may approach the binding site either through the cytoplasmic funnel , or through the narrow polar channel ( red arrows ) . A proton displaces the bound substrate ion , which escapes to the cytoplasm ( black arrow ) . Employing the narrow polar channel as the proton path would separate the Na+ ion and proton currents on the cytoplasmic side , which may be advantageous at high transport rates . DOI: http://dx . doi . org/10 . 7554/eLife . 03579 . 024
A codon-optimized synthetic gene for the Na+/H+ antiporter from Pyrococcus abyssi ( WP_010868413 . 1 ) was cloned into a vector with a C-terminal cysteine protease domain fusion as described previously for soluble proteins ( Shen et al . , 2009 ) . Mutations were introduced by site-directed mutagenesis ( Braman et al . , 1996 ) . The resulting plasmids were used to transform E . coli C41- ( DE3 ) cells . The protein was expressed for 10 hr at 37°C in ZYM-5052 autoinduction medium ( Studier , 2005 ) . Membranes were isolated from a 12 l culture and resuspended at 15 mg/ml protein in 20 mM Tris pH 7 . 4 , 250 mM sucrose , 140 mM choline chloride . The suspension was diluted 1:3 in solubilization buffer ( 20 mM Tris pH 7 . 4 , 150 mM NaCl , 30 % Glycerol and 1 . 5 % Cymal-5 ) . After solubilization overnight at 4°C the solution was clarified by centrifugation at 100 , 000×g for 1 hr . The supernatant was supplemented with 5 mM imidazole , incubated for 2 hr with Talon resin ( Clontech , Mountain View , CA ) at 4°C and loaded on a Biorad column . Unspecifically bound proteins were eluted with washing buffer ( 20 mM Tris pH 7 . 4 , 300 mM NaCl , 10 mM imidazole and 0 . 15 % Cymal-5 ) . PaNhaP was cleaved off the column by incubating the beads in elution buffer ( 20 mM Tris pH 7 . 4 , 300 mM NaCl , 0 . 15 % Cymal-5 , 20 µM inositol-hexaphosphate ) for 30 min . The eluted protein was concentrated to 5 mg/ml using a concentrator with 50 kDa cutoff and applied to a Superdex 200 size exclusion column equilibrated with 10 mM Na-Citrate pH 4 . 0 , 300 mM NaCl and 0 . 15 % Cymal-5 . Antiporter-containing fractions were pooled and concentrated to 5 mg/ml . The concentrated protein solution was diluted 1:4 with the same buffer containing 100 mM NaCl and re-concentrated as above . Selenomethionine ( SeMet ) labeled protein was expressed in PASM-5052 autoinduction medium ( Studier , 2005 ) and purified as described for the native protein in the presence of 5 mM β-mercaptoethanol throughout all purification steps . β-mercaptoethanol was exchanged to 1 mM TCEP ( Tris- ( 2-carboxyethyl ) phosphine ) in the final size-exclusion chromatography step . E . coli polar lipids ( EPL , Avanti Polar Lipids , Inc . , Alabaster , AL ) were dried under nitrogen and resuspended in reconstitution buffer . Unilamellar vesicles were prepared by extruding the resuspended lipids using an extruder ( Avestin , Inc . , Canada ) with 400 nm polycarbonate filters . Vesicles were destabilized by stepwise addition of n-octyl-β-D-glucoside ( OG ) . The process was monitored at 540 nm . Addition of OG was stopped at around 1 % final concentration when the absorbance decreased rapidly . Protein was added to the destabilized liposomes at a lipid-to-protein ratio ( LPR ) of 100:1 and incubated for 1 hr at room temperature . The solution was dialyzed ( 14 kDa cutoff ) overnight at room temperature against detergent-free reconstitution buffer . Biobeads ( SM2 , Biorad , Hercules , CA ) were added to the dialysis buffer to ensure complete removal of the detergent . Proteoliposomes were centrifuged at 300 , 000×g for 20 min and washed once with reconstitution buffer . Washed proteoliposomes were pelleted again and resuspended at ∼60 mg/ml lipid in reconstitution buffer for further use . PaNhaP was reconstituted into proteoliposomes in 10 mM choline citrate/Tris pH 6–8 , 200 mM NaCl and 5 mM KCl . To start the reaction 2 µl of proteoliposome suspension were diluted into 2 ml reaction buffer ( 10 mM choline-citrate/Tris at same pH , 5 mM KCl , 2 µM acridine orange ) . Emission of acridine orange ( excitation: 495 nm ) was monitored at 530 nm . To determine ion selectivity 5 mM NaAc , LiAc , KAc or TlAc were added to the reaction mixture after the initial sodium efflux reached equilibrium . Addition of substrates for PaNhaP to the reaction buffer results in proton efflux and fluorescence dequenching . Finally , the remaining proton gradient was dissipated by adding 25 mM ( NH4 ) 2SO4 in all experiments as a control . Electrogenic transport was investigated by addition of 100 nM valinomycin to the reaction buffer . The temperature was kept constant in a water bath during each experiment . Temperature dependence of transport was measured ( triplicates ) between 20°C and 45°C by correlating the speed of fluorescence quenching in the mid of the curve drop . PaNhaP was reconstituted in 20 mM choline citrate/Tris pH 4–8 , 10 mM ( NH4 ) 2SO4 . The reaction mixture contained 20 mM of the same buffer , 10 mM choline chloride , 1 µCi/ml 22Na+ and NaCl concentrations ranging from 1 µM to 5 mM . The pH-profile was determined at 200 µM NaCl . For each reaction 2 µl proteoliposomes were diluted in 200 µl reaction buffer to initiate the reaction . The addition of proteoliposomes to the reaction buffer results in NH3 efflux , producing an outward-directed proton gradient ( Dibrov and Taglicht , 1993 ) . At the time points indicated , 200 µl samples of the reaction mixture were applied to a 0 . 2 µm millipore nitrocellulose ( Millipore , Billerica , MA ) filter and washed with 3 ml 22Na+-free reaction buffer . Filters were transferred to counting tubes and 4 ml liquid scintillation cocktail ( Rotiszint , Germany ) was added . All measurements were performed at room temperature and repeated at least three times . Prior to crystallization , the buffer for native protein was exchanged in the final concentrating step to 10 mM Tris/HCl pH 7 . 4 , 100 mM NaCl , 0 . 15 % Cymal-5 . Crystallization was performed in 24-well plates in hanging drops at 18°C . SeMet protein was supplemented with 1 % OG and native protein with 1 % n-octyl-β-thio-maltoside ( OTM ) . The protein solutions were mixed 1:1 with reservoir buffer ( native protein: 40 mM Na-Citrate pH 4 . 0 , 100 mM NaCl , 28–33 % PEG 350 MME; SeMet protein: 100 mM Tris/HCl pH 8 . 0 , 100 mM CaCl2/MgCl2 , 35–40 % PEG 400 ) . Trapezoidal pH 4 crystals grew up to 200 µm within 7 days . At pH 8 , long needle-like crystals grew to full size within 3 months . Crystals were vitrified directly in liquid nitrogen for data collection . For thallium soaks , crystals grown at pH 8 were transferred into a buffer containing 100 mM Tris/acetate , 100 mM MgAc2 , 40 % PEG 400 , 2 mM K-citrate , 0 . 15 % Cymal-5 and 1 % OG . After five minutes the crystals were transferred to another drop of the same solution containing 25 mM TlAc . Crystals were incubated overnight and vitrified directly in liquid nitrogen . All diffraction data were collected with crystals kept at 100 K at the beamline X10SA of the Swiss Light Source in Villigen , Switzerland . Datasets were processed with XDS ( Kabsch , 1993 ) and scaled with AIMLESS in the CCP4 package ( Collaborative Computational Project 4 , 1994 ) . Resolution cut-offs were chosen based on CC1/2 ( cross correlation of half datasets ) , completeness and I/σ ( I ) -values in high resolution shells ( Karplus and Diederichs , 2012 ) . Coot ( Emsley and Cowtan , 2004 ) was used for model building and the PHENIX package ( Adams et al . , 2004 ) for refinement . Phases were obtained by single-wavelength anomalous dispersion ( SAD ) using SeMet crystals . Datasets from two crystals were merged to achieve a high multiplicity and to increase the anomalous signal ( Liu et al . , 2011 ) . The Selenium substructure containing 11 out of 14 possible positions was determined at 5 . 7 Å using SHELXD ( Sheldrick , 2010 ) . Phasing , hand determination , density modification with Parrot ( Zhang et al . , 1997 ) and initial model building with Buccaneer ( Cowtan , 2006 ) was performed with a beta version of CRANK2 ( Pannu et al . , 2011 ) . The resulting electron density map was used for manual building of an initial backbone model . Selenium positions were used to assign side chains in initial refinement rounds . Molecular replacement was performed using PHASER ( McCoy , 2007 ) with the assigned dimer model to extend the resolution to 3 . 15 Å . The final pH 8 model was used for molecular replacement to phase the pH 4 structure and the thallium bound structure at pH 8 . Superimpositions were performed using secondary structure superimposition ( Krissinel and Henrick , 2004 ) within Coot ( Emsley and Cowtan , 2004 ) . Figures were prepared with PyMOL ( DeLano and Lam , 2005 ) . The potential surface was calculated with pdb2pqr ( Dolinsky et al . , 2007 ) and APBS ( Baker et al . , 2001 ) . Analysis of transport pathways , channels and cavities was performed with Hollow ( Ho and Gruswitz , 2008 ) and visualized within PyMOL . Atomic coordinates and structure factors have been deposited with the PDB under accession codes: 4cz8 for the pH 8 SeMet structure , 4cz9 for the pH 4 structure and 4cza for the thallium-bound structure at pH 8 . | Although the membrane that surrounds a cell is effective at separating the inside of a cell from the outside environment , certain molecules must enter or leave the cell for it to work correctly . One way this transport can occur is via proteins embedded in the cell membrane , called transporters . Transporters that are found in all organisms include the sodium/proton antiporters , which exchange protons from inside the cell with sodium ions from outside . However , exactly how the antiporter works was unknown . Previous work suggested that the structure and activity of the sodium/proton antiporter changes as the acidity of its environment changes , but the precise details of how this occurs were unclear . Wöhlert et al . have now crystallised a sodium/proton antiporter from a single-celled organism called Pyrococcus abyssi , a species of archaea that has been found living in hydrothermal vents deep in the Pacific Ocean . The structures the protein takes on in different functional states were then deduced from these crystals using a technique called X-ray crystallography . Using heavy thallium ions instead of sodium ions , which are less visible to X-rays , Wöhlert et al . found the site in the antiporter where the transported ion binds as it moves through the membrane . The antiporter has a funnel-shaped cavity that faces inwards ( into the cell ) in both acidic and alkaline conditions , although a second narrow channel that is open in alkaline conditions is blocked in acidic conditions by small protein rearrangements . Wöhlert et al . suggest that the differences between both structures explain how the antiporter tunes its ability to bind to the ions it transports . Wöhlert et al . further measured the activity of the antiporter and observed that the transport of ions was most rapid under slightly acidic conditions . In more acidic conditions , the sodium ion cannot bind to the antiporter , and in an alkaline environment , the sodium ions bind too strongly to the antiporter; in both cases , the ions cannot be transported . Comparing the findings presented here with separate work that uncovers the structure of the sodium/proton antiporter in a different species of archaea revealed very similar structures . Related transporters are also found in mammals , and defects in these transporters can lead to problems with the heart and kidneys . A better understanding of the sodium/proton antiporter structure could therefore help to develop new treatments for these conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | Structure and substrate ion binding in the sodium/proton antiporter PaNhaP |
The ability of a neuron to regenerate its axon after injury depends in part on its intrinsic regenerative potential . Here , we identify novel intrinsic regulators of axon regeneration: poly ( ADP-ribose ) glycohodrolases ( PARGs ) and poly ( ADP-ribose ) polymerases ( PARPs ) . PARGs , which remove poly ( ADP-ribose ) from proteins , act in injured C . elegans GABA motor neurons to enhance axon regeneration . PARG expression is regulated by DLK signaling , and PARGs mediate DLK function in enhancing axon regeneration . Conversely , PARPs , which add poly ( ADP-ribose ) to proteins , inhibit axon regeneration of both C . elegans GABA neurons and mammalian cortical neurons . Furthermore , chemical PARP inhibitors improve axon regeneration when administered after injury . Our results indicate that regulation of poly ( ADP-ribose ) levels is a critical function of the DLK regeneration pathway , that poly- ( ADP ribosylation ) inhibits axon regeneration across species , and that chemical inhibition of PARPs can elicit axon regeneration .
Unlike damaged peripheral nerves , the central nervous system does not successfully regenerate after injury . Failure to regenerate has been attributed to two components of the regeneration response: intrinsic and extrinsic factors . While extrinsic inhibitory factors such as the glial microenvironment can be modulated with some success , regeneration potential is still substantially hindered , providing evidence that intrinsic factors play a significant role in modulating the ability of an axon to regenerate ( Richardson et al . , 1980; Neumann and Woolf , 1999; GrandPré et al . , 2000; Fournier et al . , 2001; Qiu et al . , 2002; Yiu and He , 2006; Park et al . , 2008; Smith et al . , 2009; Wang et al . , 2011 ) . Developing an understanding of the intrinsic mechanisms that regulate regeneration will provide insight into the treatment of neurological injury and disease . DLK-1 ( Dual Leucine Zipper Kinase ) is a mitogen activated protein kinase kinase kinase ( MAPKKK ) identified in C . elegans that functions intrinsically to regulate regeneration of adult axons in the central and peripheral nervous systems across species , including flies and mammals ( Hammarlund et al . , 2009; Yan et al . , 2009; Xiong et al . , 2010; Shin et al . , 2012; Wang et al . , 2013; Watkins et al . , 2013; Byrne et al . , 2014 ) . Activation of dlk-1 enhances axon regeneration and loss of dlk-1 function inhibits axon regeneration in young and aged animals ( Hammarlund et al . , 2009; Yan et al . , 2009; Byrne et al . , 2014 ) . In worms , flies , and mice , the function of DLK signaling in regeneration depends on gene transcription ( Xiong et al . , 2010; Shin et al . , 2012; Yan and Jin , 2012; Watkins et al . , 2013; Stone et al . , 2014 ) . These data suggest that specific targets of DLK transcriptional regulation may mediate the ability of DLK signaling to promote regeneration . Further , these targets may identify novel aspects of the cell biology of axon regeneration . Finally , modulation of these targets might increase the intrinsic regenerative potential of injured axons .
To identify targets of DLK transcriptional regulation in neurons , we took advantage of a recently developed method that uses FACS to isolate C . elegans neurons and compare their gene expression profiles ( Spencer et al . , 2014 ) . We sorted GABA motor neurons from animals with activated DLK signaling ( dlk-1 ( OE ) , conferred by overexpression of DLK-1L [Hammarlund et al . , 2009; Yan and Jin , 2012] ) and compared them to wild-type GABA neurons . To control for potential off-target effects of DLK activation , we also analyzed neurons that contained both dlk-1 ( OE ) and a loss of function mutation in pmk-3 , the MAP kinase at the end of the canonical DLK signaling pathway ( Nakata et al . , 2005; Hammarlund et al . , 2009; Yan and Jin , 2012 ) . RNA sequencing and analysis suggested the parg genes as candidates for further evaluation . The gene parg-2 ( poly ( ADP-ribose ) glycohydrolase-2 ) was significantly upregulated in neurons with activated DLK signaling ( 187-fold upregulated in dlk-1 ( OE ) vs wild type , p<0 . 01 , two-way ANOVA , Bonferroni post-test ) ( Figure 1A ) . Further , examination of RNA-Seq results for the parg-2 paralog , parg-1 , detected a 2 . 5-fold increase in the dlk-1 ( OE ) background compared to wild type ( p<0 . 05 , two-way ANOVA , Bonferroni post-test , Figure 1A ) ( See Materials and methods ) . Elevated expression of parg-1 and parg-2 by DLK signaling depended on the canonical DLK MAP kinase pathway since up-regulation was eliminated in neurons that over-expressed dlk-1 but lacked its downstream effector pmk-3 ( Figure 1A ) . These data suggested that regulation of PARG function might be a major effect of DLK signaling . Overall , up-regulation ( >two fold , p<0 . 05 ) ( See Materials and methods ) of gene expression by DLK signaling was observed for only 1 . 9% of coding genes ( 386 out of 20 , 375 protein coding genes assayed ) ; expression of most genes was not affected by DLK signaling . For example , expression levels of the pan-neuronal control gene rgef-1 ( a ras nucleotide exchange factor ) were not altered in either mutant background ( Figure 1A ) . Together , our data indicate parg expression is regulated by DLK signaling in GABA neurons . 10 . 7554/eLife . 12734 . 003Figure 1 . PARG genes regulate axon regeneration . ( A ) dlk-1 overexpression upregulates parg-1 and parg-2 expression levels in GABA neurons . The upregulation is suppressed by loss of pmk-3 function . rgef-1 ( a pan-neuronal Ras nucleotide exchange factor ) expression levels are not affected by manipulations of the dlk-1 pathway ( *p<0 . 05 , ***p<0 . 01 , two-way ANOVA , Bonferroni post-test ) . ( B–C ) dlk-1 regulates expression of nuclear-localized mCherry driven by the parg-2 promoter . Pparg-2::NLS::mCherry::NLS was observed ( arrows ) in 90% of nuclei of GABA neurons in dlk-1 ( OE ) animals and in 19% of GABA neurons in wild type animals ( asterisks ) . GABA neurons express GABA neuron-specific GFP marker , Punc-47::GFP . ( C ) parg-2 expression was significantly increased in both severed axons and neighboring uncut axons relative to axons in uninjured wild type animals . parg-2 was not expressed in dlk-1 ( lf ) axons , whether severed or intact . ( *p<0 . 05 , Fisher’s exact test , relative to wild type , n = 111 , 20 , 115 , 34 , 18 , 69 , 36 ) . ( D ) The GABA motor nervous system of C . elegans . GFP-labeled axons were severed with a pulsed laser at the midline ( dark brown line ) and scored for regeneration . ( E ) Representative micrographs of uninjured wild type , severed wild type , severed parg-1 ( - ) , and severed parg-2 ( - ) GABA axons . Each carry the oxIs12 transgene which drives GFP expression in GABA neurons . Arrowheads and arrows indicate proximal and distal stumps , respectively . ( F ) Axon regeneration is significantly reduced in parg-1 ( - ) , parg-2 ( - ) and parg-1 ( - ) parg-2 ( - ) mutants compared to wild type animals ( *p<0 . 05 , Fisher’s exact test , relative to wild type , n = 50 , 39 , 67 , 21 ) . ( G ) parg-1 and parg-2 are closely linked on chromosome IV , making construction of a double mutant difficult . To create a double parg-1 parg-2 mutant , parg-1 was mutated with CRISPR in a parg-2 ( lf ) background . The resulting frameshift mutation ( wp20 ) truncates PARP-1 earlier than the canonical gk120 deletion allele . ( H ) Expression of parg-1 or parg-2 in GABA motor neurons rescued axon regeneration in parg-1 ( lf ) and parg-2 ( lf ) mutants , respectively . ( I ) Model of PARG function in axon regeneration . DOI: http://dx . doi . org/10 . 7554/eLife . 12734 . 003 To further test whether DLK regulates parg expression in GABA neurons , we built a reporter construct that expresses nuclear-localized mCherry driven by the parg-2 promoter . At low magnification ( 4X ) , mCherry was only detected in dlk-1 ( OE ) animals ( 20/20 dlk-1OE animals , 0/20 wild type animals ) . At high magnification ( 40X ) , mCherry was seen in 90% of GABA neurons in dlk-1 ( OE ) animals and 19% of GABA neurons in control animals ( Figure 1B , C ) ( p<0 . 0001 , Fisher’s exact test ) . mCherry was not detected in dlk-1 ( lf ) animals , which lack DLK-1 . Therefore , parg-2 expression is dependent on dlk-1 , even in intact , uninjured axons . Next , we tested the effect of axon injury on dlk-1-dependent parg expression ( Figure 1B , C ) . Approximately 10 hr post-axotomy , parg-2 expression was significantly elevated in cut GABA axons relative to parg-2 expression in GABA axons of uninjured animals ( 47% vs 19% , p=0 . 0028 , Fisher’s exact test ) . Further , by examining neighboring uncut GABA axons , we found that parg-2 expression also increased in uninjured neurons to equivalent levels ( 47% vs 41% , p=0 . 6722 , Fisher’s exact test ) . The increase in parg-2 expression in response to injury is entirely dependent on dlk-1 , as parg-2 expression was not seen in cut or uncut axons in injured dlk-1 ( lf ) animals ( p=0 . 0003 and p=0 . 0001 , relative to cut axons and uncut axons in injured wild type animals ) . Thus , parg-2 expression is upregulated in injured neurons and their neighbors , and dependent on dlk-1 . Poly ( ADP-ribose ) glycohydrolases ( PARGs ) catalyze dePARylation: the removal of the post-translational modification poly ( ADP-ribose ) ( PAR ) from target proteins ( Miwa and Sugimura , 1971; Althaus and Richter , 1987 ) . The parg-1 and parg-2 genes encode the only two PARGs in the C . elegans genome ( Gagnon et al . , 2002 ) . We determined the function of parg-1 and parg-2 in axon regeneration by assessing regrowth after single neuron laser axotomy in GABA neurons ( Byrne et al . , 2011 ) ( Figure 1D ) . Loss of either parg-1 or parg-2 reduced axon regeneration to approximately half of normal levels: only 39% and 36% of axons regenerated in parg-1 and parg-2 mutants , respectively , while 70% of axons regenerated in control animals ( Figure 1E , F ) . Therefore , parg-1 and parg-2 regulate axon regeneration . The parg-1 and parg-2 genes are closely linked on chromosome IV , complicating generation of a double mutant . To assess whether complete elimination of PARG activity could further reduce regeneration , we used a CRISPR-Cas9 approach ( Friedland et al . , 2013 ) to mutate parg-1 in the parg-2 ( lf ) background ( Figure 1G ) . The resulting parg-1 ( lf ) ; parg-2 ( lf ) double mutant was viable and displayed wild-type morphology and behavior , indicating PARG function is not essential . Axon regeneration in parg-1 ( lf ) ; parg-2 ( lf ) animals was similar to axon regeneration in either parg-1 or parg-2 mutant animals ( Figure 1F ) . Thus , PARG activity is required for normal axon regeneration , but some regeneration occurs even in animals that completely lack parg . To test whether the parg genes act within GABA neurons to regulate axon regeneration , we reintroduced parg-1 or parg-2 specifically in GABA neurons ( using the unc-47 promoter ) of parg-1 ( lf ) or parg-2 ( lf ) mutants , respectively , and assessed regeneration . We found that 87% and 85% of injured axons regenerated in parg-1 and parg-2 worms whose GABA neurons had restored PARG expression ( Figure 1H ) . We conclude cell-intrinsic PARG function is required for axon regeneration of GABA neurons ( Figure 1I ) . Cellular levels of PARylation are determined by the balance between the activity of PARGs , which remove PAR , and the activity of poly ( ADP-ribose ) polymerases ( PARPs ) , which transfer PAR onto target proteins ( Schreiber et al . , 2006; Gibson and Kraus , 2012 ) . Thus , axon regeneration defects in parg-1 and parg-2 mutants ( Figure 1F ) could be due to accumulation of PAR . To test this hypothesis , we analyzed regeneration in animals with reduced PAR . The C . elegans genome contains two PARP homologs , parp-1 and parp-2 ( Gagnon et al . , 2002 ) . We found that mutation of either parp-1 or parp-2 increased axon regeneration relative to control animals: 92% and 90% of axons regenerated in parp-1 and parp-2 mutants , respectively , while 76% of axons in controls regenerated ( Figure 2A , B ) . Regenerating axons in these assays include all those that initiate a migrating growth cone after injury . To determine whether axons in PARP mutants are capable of sustained growth toward their original target , we assessed ability to extend towards the dorsal nerve cord ( Figure 2C ) . We found that 56% and 53% of regenerating axons in parp-1 and parp-2 mutants , respectively , regrew at least 3/4 of the distance between the ventral and the dorsal nerve cords compared to only 26% of regenerating axons in controls ( Figure 2D ) . Moreover , some parp-2 mutants sprouted new axons from the cell body ( Figure 2E ) . The opposing effects of PARGs and PARPs on poly ( ADP ) -ribose and on axon regeneration indicate that PARylation is a critical determinant of regenerative potential . 10 . 7554/eLife . 12734 . 004Figure 2 . PARPs inhibit axon regeneration . ( A ) Representative micrographs of severed wild type , parp-1 ( lf ) , and parp-2 ( lf ) GABA motor neurons . ( B ) Axon regeneration in parp-1 ( lf ) and parp-2 ( lf ) mutants compared to wild type animals ( *p<0 . 05 , Fisher’s exact test , n = 84 , 26 , 61 ) . ( C ) Cut axons ( 1 ) are scored for the distance they extend towards their targets in the dorsal nerve cord ( 2 , 3 , 4 , 5 ) . ( D ) Axon regeneration to at least 3/4 of the distance to the dorsal cord ( 4 ) is significantly increased in parp-1 ( lf ) and parp-2 ( lf ) mutants relative to wild type animals ( *p<0 . 01 , Fisher’s exact test , n = 61 , 43 , 38 ) . ( E ) Axon regeneration from the cell body ( 2 ) is seen in parp-2 ( lf ) mutants ( n = 47 , 32 , 34 ) . ( F ) Representative micrographs of injured cortical neurons exposed to negative control shRNA or PARP1 shRNA . ( G ) Axon regeneration is increased in murine cortical neurons lacking PARP1 ( ***p<0 . 001 , ****p<0 . 0001 , Anova with Tukey’s multiple comparisons test , n = 108 , 8 , 8 ) . Axon regeneration was measured in injured cortical neurons exposed to non-coding negative control shRNA ( shNC ) or either of two unique PARP1 shRNAs . ( H , I ) Exposure to either shPARP significantly reduced PARP levels in cortical neurons relative to PARP levels in cortical neurons exposed to negative control ( shNC ) lentivirus ( *p<0 . 05 , **p<0 . 005 , Anova with Tukey’s multiple comparisons test ) . ( J ) parg-1 and parg-2 loss of function incompletely suppress the increase in regeneration conferred by dlk-1 ( OE ) ( *p<0 . 05 , relative to wild type , §p<0 . 05 , relative to dlk-1 ( OE ) , Fisher’s exact test , n = 24 , 21 , 48 , 62 ) , indicating the PARGs regulate regeneration downstream of dlk-1 with at least one parallel pathway . DOI: http://dx . doi . org/10 . 7554/eLife . 12734 . 00410 . 7554/eLife . 12734 . 005Figure 2—figure supplement 1 . Detailed characterization of regeneration in dlk-1 ( OE ) and parg-2 ( - ) parg-2 ( - ) mutants . For each severed axon represented in Figure 2H , we determined whether a given axon regenerated below the midline of the worm ( M- ) , beyond the midline ( M+ ) , at least 3/4 of the distance to the dorsal cord ( M++ ) . The greatest difference in regenerative ability between the genotypes is in overall regeneration , represented by combining the blue bars for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 12734 . 005 PARG and PARP function are well-conserved between C . elegans and mammals ( Gagnon et al . , 2002; St-Laurent et al . , 2007 ) . Therefore , we hypothesized that blocking PARP function might be sufficient to improve regeneration of mammalian CNS neurons . To test this hypothesis , we assessed the effect of PARP knockdown on mouse cortical neuron regeneration . We cultured primary cortical mouse neurons in 96-well plates ( Huebner et al . , 2011 ) . We subsequently added lentiviral control or one of two unique PARP1 shRNAs at three days in vitro ( DIV ) , and injured the neurons with a custom pin-replicator five days later ( Huebner et al . , 2011 ) . Three days after injury , we fixed the neurons and assessed regeneration . We found that axons exposed to PARP1 shRNA regenerated significantly better than axons exposed to control shRNA ( Figure 2F , G ) . To confirm the shRNA clones targeted PARP1 , we performed western blots on cortical neurons exposed to the negative control shRNA or to the two unique shRNA that target PARP1 . In each case PARP was detected in the insoluble fraction and not in the lysate , in agreement with previously reported localization to the nucleus ( reviewed in Bai , 2015 ) . PARP was significantly reduced in neurons exposed to either of the PARP1-targeting shRNAs compared to negative control shRNA ( Figure 2H , I ) . PARP levels were normalized to actin levels in each sample of neurons . Therefore , PARP-1 and PARylation are conserved inhibitors of axon regeneration after injury , and reducing their function improves axon regeneration across species . Having established that PARGs are novel regulators of axon regeneration , we sought to determine the extent to which DLK function is mediated by PARGs . We assessed regeneration in animals with activated DLK signaling ( dlk-1 ( OE ) ) , but lacking both parg-1 and parg-2 . We found that loss of both parg-1 and parg-2 function reduced regeneration in the dlk-1 ( OE ) background , just as loss of parg-1 and parg-2 reduces regeneration in animals with wild type levels of DLK signaling ( Figure 2J and Figure 2—figure supplement 1 ) . In both cases , regeneration is reduced but not eliminated , and the amount of regeneration that remains is higher than complete loss of dlk-1 signaling ( 0% regeneration ) or loss of the downstream pmk-3 in the dlk-1 ( OE ) background ( 7% regeneration ) ( Nakata et al . , 2005; Hammarlund et al . , 2009 ) . Thus , DLK-dependent regeneration depends in part on parg-1 and parg-2 . In addition to controlling axon regeneration , DLK signaling regulates presynaptic development ( Nakata et al . , 2005 ) . To test whether presynaptic development is regulated by PARylation , we quantified synapses at the GABA neuromuscular junction with the pre-synaptic reporter hpIs3 . The hpIs3 reporter expresses GFP tagged SYD-2 ( alpha-liprin ) in presynaptic active zones of GABA motor neurons ( Zhen and Jin , 1999 ) . In control animals , SYD-2::GFP is distributed in a punctate pattern at regularly interspaced intervals along the dorsal nerve cord ( Figure 3 ) ( Yeh et al . , 2005 ) . Loss of all PARG activity did not affect synapse morphology . Increased DLK signaling in dlk-1 ( OE ) animals causes synapse morphology defects ( Figure 3 ) ( Nakata et al . , 2005 ) . In dlk-1 ( OE ) animals , SYD-2::GFP is diffuse along the dorsal nerve cord , which increases the average baseline fluorescence along a line scan ( Figure 3 ) . However , loss of PARG activity did not suppress these defects . Together , these data indicate that in contrast to its role in axon regeneration , PARylation does not regulate synapse formation , even when DLK signaling is activated . 10 . 7554/eLife . 12734 . 006Figure 3 . Loss of parg-1 and parg-2 function does not suppress mislocalizion of presynaptic active zones caused by dlk-1 overexpression . ( A ) Dorsal nerve cords of wild type , dlk-1 ( OE ) , parg-1 ( lf ) parg-2 ( lf ) , and dlk-1 ( OE ) ; parg-1 ( lf ) parg-2 ( lf ) animals . All animals express the presynaptic active zone marker SYD-2::GFP in their GABA neurons . SYD-2::GFP is expressed in discrete puncta ( arrowhead ) in wild type animals and is not expressed continuously along the dorsal cord ( asterisk ) . Conversely , SYD-2::GFP is expressed in a diffuse pattern ( arrow ) in dlk-1 ( OE ) animals . ( B ) Average maxima and average baseline fluorescence are calculated along line scans of each dorsal cord and represented in ( C ) . ( C ) dlk-1 overexpression disrupts SYD-2::GFP expression in GABA neurons and results in higher baseline fluorescence compared to wild type . Loss of parg function does not affect localization of SYD-2::GFP nor does it suppress the mislocalization caused by dlk-1 ( OE ) ( *p<0 . 05 , ***p<0 . 01 , multiple ANOVA , Bonferroni post-test ) . ( D ) There are no significant differences in average maxima between genotypes . Sample size is 15 , 8 , 8 , and 15 animals for wild type , dlk-1 ( OE ) , parg-1 ( lf ) parg-2 ( lf ) , and dlk-1 ( OE ) ; parg-1 ( lf ) parg-2 ( lf ) animals , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 12734 . 006 Axon injury triggers an acute response that includes activation of DLK signaling ( Yan and Jin , 2012 ) . PARylation is a short-lived modification , and PAR levels are normally maintained by the continuous activity of PARP and PARG proteins ( Schreiber et al . , 2006; Gibson and Kraus , 2012 ) . These data suggest a model in which increased PARG expression downstream of DLK signaling acutely reduces PAR levels in response to axon injury , thereby facilitating regeneration . We hypothesized that acute reduction of PAR levels by inhibition of PARP might also increase axon regeneration after injury , potentially similar to increased regeneration in PARP mutants ( Figure 2 ) . Multiple chemical PARP inhibitors are currently in preclinical and clinical trials for indications including cancer therapy and stroke ( Ford and Lee , 2011; Anwar et al . , 2015 ) . We found that treatment with chemical PARP inhibitors after injury resulted in significantly enhanced axon regeneration in vivo in C . elegans GABA neurons and in vitro in murine cortical neurons ( Figure 4A–C , and Figure 4—figure supplement 1 ) . Drug treatment post-injury also improved behavioral recovery , demonstrating that enhanced regeneration after PARP inhibition results in functional reconnection ( Figure 4D , E ) . Thus , acute poly ( ADP-ribose ) levels determine the response of neurons to axon injury , and inhibition of PARP after injury is sufficient to improve regeneration . 10 . 7554/eLife . 12734 . 007Figure 4 . Chemical PARP inhibition enhances axon regeneration post-injury . ( A ) Micrographs of regenerating axons placed on plates containing DMS0 or PARP inhibitor ( A966492 , Selleckchem , 100 µM ) immediately after surgery . ( B ) Acute chemical inhibition of PARP function enhances regeneration ( *p<0 . 05 , Fisher’s exact test , n = 34 and 19 axons severed in animals exposed to DMSO or PARP inhibitor , respectively ) . ( C ) Axon regeneration is increased in murine cortical neurons exposed to chemical PARP inhibitor A966492 ( *p=0 . 0149 , Student’s t-test , n = 10 ) . ( D ) To assess functional regeneration , all GABA neurons were severed and animals were assessed for their ability to reverse in response to a touch on the nose from a platinum wire . ( E ) One hour after all GABA neurons are severed , animals are incapable of reversing in response to a touch on the nose ( shrinker ) . As functional connections are regenerated , animals recovered on PARP inhibitor displayed more backward movement than those recovered on DMSO ( measured as number of body bends ) . Significantly more animals on PARP inhibitors recovered wild type function ( 4+ body bends , *p<0 . 05 , Fisher’s exact test , n = 11 and 9 animals exposed to DMSO or PARP inhibitor , respectively ) . ( F ) The balance between PARP and PARG regulates axon regeneration and is altered by chemical PARP inhibitors . DOI: http://dx . doi . org/10 . 7554/eLife . 12734 . 00710 . 7554/eLife . 12734 . 008Figure 4—figure supplement 1 . Chemical PARP inhibitors have different effects on axon regeneration . Axon regeneration frequencies of wildtype animals placed on DMSO , Olaparib , Velaparib , or PJ34 , post-axotomy . Regeneration was scored as ( A ) proportion of cut axons that regenerated , ( B ) proportion of regenerating axons that fully regenerated to the dorsal nerve cord ( see Figure 2C ) , and ( C ) proportion of severed neurons that extended a secondary process rather than regenerate the severed axon . *p<0 . 05 , Fisher’s exact test , relative to wild type; error bars represent 95% confidence intervals; n = 32 , 22 , 22 , 26 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12734 . 008 Together , our findings suggest that regulation of PARylation is an important component of the DLK pathway role in the axon regeneration mechanism . Multiple lines of evidence suggest the parg genes are transcriptionally regulated downstream of DLK signaling to promote regeneration . First , we find both of the parg genes are upregulated in animals with activated DLK signaling . Second , we find that endogenous dlk-1 signaling drives parg-2 expression in an injury-dependent manner . Third , we find that loss of the parg genes reduces regeneration , both in animals with endogenous levels of dlk-1 activity and in animals with elevated DLK-1 signaling . These findings suggest a linear model in which DLK signaling induces parg expression , which in turn facilitates regeneration by removing PAR . In addition to regulating PARylation , our data indicate that DLK signaling regulates regeneration upstream of multiple effectors . Although PARylation has a strong effect on regeneration in both animals with wild-type levels of dlk-1 and in animals that overexpress dlk-1 , some GABA axons in parg-1 ( lf ) parg-2 ( lf ) mutants are still able to regenerate ( Figure 1F ) , indicating high levels of PAR do not completely prevent DLK-mediated axon regeneration . By contrast , GABA axons do not regenerate in animals lacking DLK ( Hammarlund et al . , 2009; Yan et al . , 2009 ) . These data suggest that DLK activity has other functions besides regulating PARylation . Some of these functions may be mediated by other transcriptional outputs of DLK signaling ( Watkins et al . , 2013 ) . Understanding these factors , as well as understanding the cellular effects of PAR on regeneration , await further study . Besides shedding light on functional outputs of DLK signaling , our findings identify a novel pathway , involving control of poly ( ADP-ribose ) levels , that regulates axon regeneration ( Figure 4F ) . Specifically , we find that PARG and PARP activity regulate the acute response of neurons to axon injury , and that chemical PARP inhibition after injury is sufficient to improve regeneration . The lack of additive phenotype in the double loss of function parg-1 ( - ) ; parg-2 ( - ) mutant suggests the two parg genes are not partially redundant . Rather , parg-1 and parg-2 may function together , for example as part of a complex . In Arabidopsis thaliana , the two PARG homologs physically interact ( Song et al . , 2015 ) , suggesting the PARG homologs may function coordinately . Alternatively , concerted action of both PARGs may be required to maintain PAR levels below a threshold . In this model , loss of either single PARG results in a sufficient PAR increase to block regeneration , but increasing PAR beyond this threshold does not further reduce regeneration . Previous investigations of the relationship between the two C . elegans parg homologs have been complicated by the physical proximity of the two genes in the genome . As a result , double loss of function mutants have been generated using RNAi . Since RNAi can result in incomplete knockdown of target genes , it has been difficult to determine the functional redundancy of the two genes using this approach . The parg-1; parg-2 mutant described here may be useful for further characterization of animals that completely lack PARG function . In vivo , injured mammalian axons must overcome extrinsic growth inhibition to regenerate . PARP1 is upregulated in murine cortical neurons exposed to inhibitory growth molecules ( myelin-associated glycoprotein , Nogo-A , Chondroitin sulfate proteoglycans ) in vitro and in crushed optic nerves in vivo . Moreover , inhibiting PARP1 promotes neurite outgrowth on inhibitory substrates in vitro ( Brochier et al . , 2015 ) . Since PARPs and PARGs have contrasting effects on PAR levels , NAD+ levels , which are substrates of PAR ( Bai , 2015 ) , and on axon regeneration , we conclude the balance between PARP and PARG function regulates axon regeneration , and present the hypothesis the PARG-PARP balance may determine axon regeneration by regulating PAR levels or by regulating NAD+ levels . Finally , the conservation of the role of PARP in mammalian axon regeneration may have important implications for nerve repair following injury or disease .
Strains were maintained as previously described at 20°C ( Brenner , 1974 ) . Some strains were provided by the CGC , which is funded by NIH Office of Research Infrastructure Programs ( P40 OD010440 ) . Specific mutations analyzed: parg-1 ( gk120 ) , parg-2 ( ok980 ) , parp-1 ( ok988 ) , parp-2 ( ok344 ) , wpIs9[Punc-47:DLK-1mini-GFP , ccGFP] , pmk-3 ( ok169 ) , hpIs3[punc-25::SYD-2::GFP; lin-15+] . To visualize GABA neurons in regeneration assays , mutants were crossed into the oxIs12 [Punc-47:GFP , lin-15+] background . XE1347 wpIs39[Punc-47:mCherry] , XE1551 wpIs9[Punc-47:DLK-1mini-GFP , ccGFP]; wpIs39[Punc-47:mCherry] , and XE1552 wpIs9[Punc-47:DLK-1mini-GFP , ccGFP]; pmk-3 ( ok169 ) ; wpIs39[Punc-47:mCherry] were analysed by RNA Seq . The SeqCel method was used to generate RNA Seq profiles of larval GABA neurons ( Spencer et al . , 2014 ) . Briefly , L4 stage larvae were dissociated and punc-47::mCherry-labeled GABA neurons were isolated by FACS ( BD FACSaria ) from wild-type ( XE1347 ) , dlk-1 ( OE ) ( XE1551 ) and dlk-1 ( OE ) ; pmk-3 ( ok169 ) ( XE1552 ) strains; dead and damaged cells were excluded by DAPI staining . Experiments were performed in triplicate for each genotype . For RNA-Seq analysis , total RNA ( 5–10 ng ) was amplified by SMARTer cDNA synthesis ( Clonetech ) and libraries sequenced ( PE-100 ) using the HiSeq 2500 system ( Illumina ) . RNA-Seq data were analyzed with CLC Genomics Workbench software ( Qiagen ) . A global comparison ( EDGE test ) ( Robinson et al . , 2010 ) of wild-type vs dlk-1 ( OE ) GABA neuron RNA-Seq data sets detected 386 transcripts that are significantly up-regulated ( >2 fold , p<0 . 05 ) in the dlk-1 ( OE ) GABA neuron profile . A comprehensive analysis of these data sets will be presented elsewhere . The parg-2 transcript was significantly enriched ( 187x , p=1 . 6 e−14 ) in the dlk-1 ( OE ) profile . The parg-1 transcript was 2 . 5 fold elevated in dlk-1 ( OE ) but was excluded from this initial analysis due to a conservative p-value correction for multiple testing . We identified parg-1 as a likely false negative in this global analysis due to statistically significant elevation of the parg-1 signal in a direct comparison with wild-type and pmk-3 ( ok169 ) ( XE1552 ) ( see Figure 1A ) . Pparg-2::NLS::mCherry::NLS was constructed by combining Gateway plasmids encoding the parg-2 promoter sequence ( obtained from GE Dharmacon Promoterome collection ) , NLS::mCherry::NLS coding sequence , the unc-54 UTR sequence , and pCFJ150 . Pparg-2::NLS::mCherry::NLS was injected along with the Pmyo-2::GFP co-injection marker ( expressed in the pharynx ) into wpIs9[Punc-47:DLK-1mini-GFP , ccGFP]; oxIs12[Punc-47::GFP] animals . wpIs9 was outcrossed from transgenic lines with wild type N2 males . mCherry expression was compared between animals carrying the same extra-chromosomal array in the presence or absence of wpIs9 . The posterior 7 VD/DD GABA neurons in 10 worms were analyzed for each genotype . Expression was analysed with an Olympus DSU mounted on an Olympus BX61 microscope , Andor Neo sCMOS camera , and Lumen light source . Error bars represent 95% confidence intervals . Significance is indicated with an asterisk ( p<0 . 0001 , Fisher exact test ) . Axotomy experiments were carried out as previously described ( Byrne et al . , 2011 ) . Post-axotomy images were acquired with an Olympus DSU mounted on an Olympus BX61 microscope , Andor Neo sCMOS camera , and Lumen light source . Error bars represent 95% confidence intervals . Significance is indicated with an asterisk ( p<0 . 01 , Fisher exact test ) . The double parg-1 ( wp20 ) parg-2 ( ok980 ) mutant was created by injecting sgRNA targeting parg-1 sequence: aaagactacgaagactatgt and Cas9 into parg-2 ( ok980 ) animals . The resulting deletion of the 20th and 21st base pairs of the second parg-1 exon is a frameshift mutation that creates a truncated protein . Punc-47::parg-2 expressing animals were obtained by injecting parg-2 ( ok980 ) ; oxIs12 worms with pAB1019 DNA at 50 ng/μl along with Punc-25::mCherry at 10 ng/ul and Pmyo-2:mCherry at 2 ng/μl as a co-injection marker . 1 kb ladder was added at 50 ng/μl as carrier . The pAB1019 plasmid was constructed by combining Gateway plasmids encoding the unc-47 promoter sequence , parg-2 coding sequence ( obtained from GE Dharmacon ORFeome collection ) , the unc-54 UTR sequence , and pCFJ150 . The mouse cortical neuron axon regeneration assay was performed by scrape injury of confluent cultures , as described previously ( Huebner et al . , 2011 ) . Primary cortical cultures were established from E17 C57BL/6 mice . Digested cells were plated on 96-well poly-D-lysine coated plates at a density of 25 , 000 cells per well in 200 µL of plating medium . Lentiviral particles encoding control non-targeting or PARP1 shRNA clones ( Sigma ) were added on DIV3 ( Day In Vitro 3 ) as described for other shRNAs ( Zou et al . , 2015 ) . On DIV8 , 96-well cultures were scraped using a custom-fabricated 96-pin array and allowed to regenerate for another 72 hr before fixing with 4% paraformaldehyde . Regenerating axons in the scrape zone were visualized using an antibody against β3 tubulin ( 1:2000 , mouse monoclonal; catalog #G712A; Promega ) . Growth cones were visualized by staining for F-actin using rhodamine-conjugated phalloidin ( 1:2000 , catalog #R415 , Life Technologies ) . Cell density was visualized using nuclear marker DAPI ( 0 . 1 μg/mL , catalog #4083 , Cell Signaling Technology ) . Images were taken on a 10X objective in an automated high-throughput imager ( ImageXpress Micro XLS , Molecular Devices ) under identical conditions . Regeneration zone identification , image thresholding and quantitation were performed using an automated MATLAB script in a fully automated fashion . The synapse marker hpIs3[punc-25::SYD-2::GFP; lin-15+] was crossed onto indicated combinations of parg ( lf ) and dlk-1 ( OE ) backgrounds . The dorsal cords of the resulting animals were imaged with a 40X oil objective on an UltraVIEW Vox ( PerkinElmer ) spinning disc confocal microscope ( Nikon Ti-E Eclipse inverted scope; Hamamatsu C9100-50 camera ) with Volocity software ( Improvision ) . Images were analyzed with ImageJ . All PARP inhibitors were acquired from Selleckchem . To assess the effects of PARP inhibitors on regeneration , GABA axons in L4 animals were axotomized and the worms placed on NGM plates ( Brenner , 1974 ) containing 100 µM of the respective PARP inhibitor . Control plates were made with the same amount of DMSO as the plates containing inhibitor . Axon regeneration was scored 24 hr post-axotomy . Functional recovery was assessed by counting the number of body bends an animal made after being tapped on the nose with a platinum wire . Zero body bends is referred to as ‘shrinker’ . | Neurons carry information around the body along slender projections known as axons . An injury that crushes or cuts an axon can lead to permanent disability if the axon fails to regenerate . While some damaged neurons in the body can repair themselves , typically those present in the brain and spinal cord cannot regenerate successfully after injury . The ability of a neuron to regenerate its axon depends in part on factors present inside the neuron itself . By understanding how these internal mechanisms regulate axon regeneration , researchers hope to develop new ways to boost the repair of damaged neurons . A protein called DLK acts inside neurons to promote regeneration of injured axons across a range of species including worms and mammals . In the absence of DLK , regeneration is impaired . The DLK signaling pathway is activated in damaged neurons and is thought to promote repair by altering the activity of genes and proteins that control the regeneration process . Byrne et al . have now identified genes that are activated by the DLK signaling pathway in the roundworm , Caenorhabditis elegans . The experiments show that DLK signaling increases the activity of genes encoding enzymes known as PARGs , which in turn enhance axon regeneration . PARG enzymes remove chain-like molecules called poly ( ADP-ribose ) that are attached to target proteins . Further experiments showed that other enzymes known as PARPs , which add the poly ( ADP-ribose ) markers to proteins , act to inhibit axon regeneration in both Caenorhabditis elegans and in injured neurons from mice . Consistent with this , Byrne et al . found that drugs that inhibit PARP enzymes improved axon regeneration when they were given to C . elegans with injured neurons . These results suggest that a critical role of the DLK signaling pathway is to regulate poly ( ADP-ribose ) levels and that reducing the amount of poly ( ADP-ribose ) added to proteins can promote axon regeneration . The next step is to understand exactly how poly ( ADP-ribose ) regulates axon regeneration and to identify the other factors – besides poly ( ADP-ribose ) , PARGs and PARPs – that act downstream of DLK signalling to regulate regeneration . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2016 | Inhibiting poly(ADP-ribosylation) improves axon regeneration |
The nonsense-mediated mRNA decay ( NMD ) pathway detects aberrant transcripts containing premature termination codons ( PTCs ) and regulates expression of 5–10% of non-aberrant human mRNAs . To date , most proteins involved in NMD have been identified by genetic screens in model organisms; however , the increased complexity of gene expression regulation in human cells suggests that additional proteins may participate in the human NMD pathway . To identify proteins required for NMD , we performed a genome-wide RNAi screen against >21 , 000 genes . Canonical members of the NMD pathway were highly enriched as top hits in the siRNA screen , along with numerous candidate NMD factors , including the conserved ICE1/KIAA0947 protein . RNAseq studies reveal that depletion of ICE1 globally enhances accumulation and stability of NMD-target mRNAs . Further , our data suggest that ICE1 uses a putative MIF4G domain to interact with exon junction complex ( EJC ) proteins and promotes the association of the NMD protein UPF3B with the EJC .
mRNA decay pathways have evolved to both maintain mRNA quality control and to regulate gene expression . For example , the nonsense-mediated mRNA decay ( NMD ) pathway was initially described for its role in targeting the destruction of messages harboring premature termination codons ( PTCs; [Maquat et al . , 1981; Peltz et al . , 1993] ) . The introduction of a PTC can result from genetic mutations or errors in gene expression , ultimately generating a message encoding a potentially deleterious truncated protein . As such , the NMD pathway protects the integrity of the proteome by detecting and destroying PTC-containing messages . In addition to this well-characterized role in quality control surveillance , NMD regulates the expression of ~5–10% of the mammalian transcriptome , highlighting its importance in determining levels of non-aberrant mRNAs ( Mendell et al . , 2004; Rehwinkel et al . , 2005; Tani et al . , 2012; Weischenfeldt et al . , 2008 ) . These apparently normal mRNAs often contain NMD-triggering features such as upstream open-reading frames ( uORFs ) in the 5’-UTR , the presence of a long 3’-UTR , or a 3’-UTR that harbors an intron ( Karousis et al . , 2016; Rebbapragada and Lykke-Andersen , 2009 ) . NMD is conducted by a conserved set of factors primarily identified via genetic screens in model organisms . The UPF1 , 2 , and 3 proteins ( up-frameshift ) , discovered in budding yeast , are required for NMD throughout eukaryotes , while the SMG ( suppressor with morphogenetic effects on genitalia ) proteins were originally identified in C . elegans and provide additional enzymatic and regulatory activities required for NMD in metazoans ( Leeds et al . , 1991; Pulak and Anderson , 1993 ) ; for a recent comprehensive review , see Karousis et al . ( 2016 ) . UPF1 , the central hub of the NMD pathway , is an ATP-dependent RNA helicase that acts at multiple steps in target discrimination and decay . UPF1’s ATPase activity and phosphorylation is enhanced by UPF2 binding , promoting decay ( Chakrabarti et al . , 2011; Chamieh et al . , 2008; Kashima et al . , 2006 ) . UPF1 phosphorylation by the PI3K-related kinase SMG1 in the context of translation termination promotes the recruitment of the NMD-specific endonuclease SMG6 , as well as the recruitment of the CCR4-NOT deadenylase complex via the SMG5-SMG7 heterodimer ( Eberle et al . , 2009; Gatfield and Izaurralde , 2004; Huntzinger et al . , 2008; Kashima et al . , 2006; Loh et al . , 2013 ) . UPF1-dependent RNA degradation can thus proceed through a combination of exo- and endonucleolytic pathways . A second major elaboration of the NMD pathway in vertebrates is the use of the exon junction complex ( EJC ) to identify potential targets of NMD . The EJC is a tetrameric complex comprising the RNA helicase eIF4AIII , the MAGOH-Y14 heterodimer , and CASC3 ( also known as MLN51 and Barentz ) and is deposited 20–24 nt upstream of exon-exon junctions during splicing ( Le Hir et al . , 2000 ) . The stable core participates in multiple stages of mRNA function by engaging in dynamic interactions with ‘peripheral’ EJC factors ( Le Hir et al . , 2016; Singh et al . , 2012 ) , including UPF3B , which binds EJCs consisting of at least eIF4AIII and MAGOH-Y14 through a conserved motif in its C-terminus ( Buchwald et al . , 2010; Gehring et al . , 2003 ) . Once bound in the nucleus , the EJC is believed to escort UPF3B into the cytoplasm , where the complex marks most splice sites prior to translation ( Hauer et al . , 2016; Saulière et al . , 2012; Singh et al . , 2012 ) . During translation , scanning and elongating ribosomes displace EJCs in the 5’-leader and coding sequence , respectively ( Gehring et al . , 2009a ) , leaving only EJC-UPF3B complexes > 50 nts downstream of the termination codon bound to the mRNA . UPF3B thus provides a link between nuclear mRNA biogenesis and the cytoplasmic NMD machinery , as the N-terminal region of UPF3B binds UPF2 , which in turn promotes UPF1 activity and phosphorylation ( Chamieh et al . , 2008; Kashima et al . , 2006 ) . Such mRNAs harboring an EJC sufficiently downstream of a stop codon are subject to ‘EJC-enhanced’ NMD , the branch of the NMD pathway responsible for the most prominent UPF1-dependent decay activities observed in mammalian cells ( Bühler et al . , 2006; Cheng et al . , 1994; Metze et al . , 2013; Singh et al . , 2008; Wang et al . , 2002; Zhang et al . , 1998 ) . Notably , the primary genetic screens responsible for NMD factor identification were conducted in organisms that do not use the EJC to demarcate targets of decay . Further indicating a potential need for an expanded machinery , the human NMD pathway surveils a transcriptome more complex than that found in yeast and worms and has adopted regulatory roles in development and stress responses ( Feng et al . , 2017; Gong et al . , 2009; Karam et al . , 2015; Medghalchi et al . , 2001; Wittkopp et al . , 2009 ) . To search for potential novel human NMD proteins , including those involved in ‘EJC-enhanced’ decay , we used a luciferase reporter to develop a gain-of-signal assay designed based on a well-characterized mRNA β-globin nonsense allele and performed a whole genome RNAi screen ( Maquat et al . , 1981; Thorne et al . , 2010 ) . By depleting >21 , 000 genes with three siRNAs/gene , we recovered core NMD and EJC factors as top hits , as well as a large cohort of potential NMD factors . From the latter , we used biochemical , genetic , and genomic approaches to validate ICE1 ( interactor of little elongation complex ELL subunit 1 , also known as KIAA0947 ) as a novel peripheral EJC factor essential for EJC-enhanced NMD . We show that ICE1 depletion leads to enhanced expression of many NMD targets , including those containing PTCs . Further , our data suggest that ICE1 uses a putative MIF4G domain to interact with mature EJCs and is required for proper association of UPF3B with the EJC core proteins . Importantly , overexpressing UPF3B to restore EJC-UPF3B assembly rescues the effects of ICE1 depletion on NMD , consistent with a model in which ICE1 serves an unanticipated role in linking nuclear EJC assembly to cytoplasmic detection of NMD substrates .
Genome-wide siRNA screening was conducted using a library of siRNAs targeting ~21 , 000 human genes with three independent siRNAs per gene ( Figure 1C and Supplementary File 1 ) . A major confounding factor in analysis of siRNA screen data is the prevalence of off-target effects mediated by siRNA seed sequences . The large number of siRNAs used in this screen enables application of Common Seed Analysis ( CSA ) , a computational approach to adjust for seed sequence-based off-target effects ( Marine et al . , 2012 ) . Following CSA , we ranked genes by median seed-corrected Z-score of the corresponding siRNAs and subjected the top ranked hits ( seed corrected Z-score >1 . 5 , 76 genes ) to pathway analysis using STRING ( Figure 1D , Table 1 , and Figure 1—figure supplement 1; [Szklarczyk et al . , 2017] ) . Several canonical members of the NMD pathway were identified with seed-corrected Z-scores > 1 . 5 , including UPF1 , SMG1 , SMG5 , SMG6 , and SMG7 . All four members of the core EJC also met this stringent cutoff ( RBM8A , EIF4A3 , MAGOH , and CASC3 ) , along with DHX8/PRP22 and Aquarius ( AQR ) , proteins implicated in EJC deposition ( Ideue et al . , 2007 ) . NMD pathway components UPF2 and UPF3B also exhibited a positive response in the screen , with seed corrected Z-scores of 0 . 98 and 0 . 72 , respectively . In addition to proteins known to directly participate in NMD , spliceosomal proteins were highly enriched among genes whose depletion caused elevated luciferase expression , with particular over-representation of proteins from the PRP19-related , Sm , and U2 snRNP sub-complexes ( Supplementary file 2 ) . Overall , the strong enrichment of genes involved in mRNA processing and NMD indicates that the expected biology was identified by the screen and that other high-scoring genes are likely to be members of the same processes . Moreover , the performance of the screen compares favorably to a recent CRISPR-based screen for NMD factors ( Figure 1—figure supplement 2 ) , particularly with respect to identification of EJC and related proteins . Despite the fact that both screens successfully enriched for proteins known to be involved in NMD , only 13 genes were shared as significant hits between the two screens , five of which were CASC3 , SMG5 , SMG6 , SMG7 , and UPF1 . Following the whole-genome screen , 135 genes were further interrogated with additional siRNAs ( Supplementary file 3 ) . Of these , 27 exhibited seed-corrected Z-scores greater than 1 , including known pathway components SMG1 , SMG6 , SMG7 , EIF4A3 , RBM8A , MAGOH , CASC3 , CWC22 , and AQR . Among the genes whose depletion caused reproducible effects on luciferase expression using multiple siRNAs , we manually curated candidates for downstream analysis , focusing on those with features suggestive of a possible role in NMD . Two promising examples were known interacting partners ICE1/KIAA0947 and the SUMO isopeptidase USPL1 ( primary screen seed-corrected Z-scores of 2 . 91 and 3 . 01 , respectively ) . These proteins are known to associate with the little elongation complex , a protein assembly that interacts with and promotes RNA polymerase II association and elongation at snRNA promoters ( Hu et al . , 2013 ) . ICE1 is a ~ 250 kDa nuclear protein proposed to have a scaffolding role in little elongation complex assembly but has not been structurally characterized or previously implicated in NMD ( Hu et al . , 2013; Hutten et al . , 2014 ) . To gain insight into possible functions of ICE1 in NMD , we performed protein structure prediction using the Phyre2 web portal ( Kelley et al . , 2015 ) . This analysis identified a putative MIF4G domain with high confidence at the C-terminus of the protein , a region highly conserved among vertebrates ( Figure 1—figure supplement 3 and Figure 1—figure supplement 4 ) . MIF4G domains are prevalent in proteins involved in translation and NMD ( Barbosa et al . , 2012; Buchwald et al . , 2013; Kadlec et al . , 2004 ) , leading us to hypothesize that ICE1 may use its putative MIF4G domain to interact with components of the NMD pathway . To assess a possible role for ICE1 in regulating NMD , we performed genome-wide RNAseq analyses of cells transfected with control ( siNT ) , ICE1 , and UPF1 siRNAs ( Supplementary file 4 ) . We began by evaluating the effect of ICE1 depletion on classes of transcripts with known NMD-inducing features . As the luciferase-β-globin reporter used to identify ICE1 undergoes EJC-stimulated decay , we first examined the impact of ICE1 depletion on transcript isoforms containing stop codons at least 50 nt upstream of the final exon-exon junction ( PTCs ) . For these analyses , we focused on genes for which at least one alternative PTC-containing isoform and one normal isoform were represented in our RNAseq data , finding that ICE1 depletion specifically increased the abundance of the PTC-containing isoforms relative to the normal isoforms ( Figure 2A ) . These data provide initial evidence that ICE1 is important for cellular elimination of many canonical NMD targets . To further investigate a role for ICE1 in EJC-dependent NMD , we tested its involvement in regulating the SR protein SRSF2 ( also known as SC35 ) . SRSF2 represents a classic example of NMD-based autoregulation of proteins involved in alternative splicing . Accumulation of SRSF2 protein leads to excision of two introns from its own 3’UTR , generating an NMD target mRNA referred to here as SRSF2 ( NMD+ ) ( [Sureau et al . , 2001]; Figure 2B , top schematic ) . Using previously characterized primers in the SRSF2 3’-UTR to distinguish SRSF2 transcript isoforms ( [Hauer et al . , 2016]; arrows denote region of amplification ) , our results recapitulate previous findings that SRSF2 ( NMD+ ) is a potent NMD substrate , as depletion of UPF1 resulted in a seven-fold increase in the NMD+ isoform compared to non-targeting controls ( Figure 2B , bottom right bar graph ) . Consistent with our RNAseq studies , depletion of ICE1 resulted in a four-fold increase in the abundance of this NMD target ( Figure 2B , bottom right graph ) , while levels of the unspliced 3’UTR isoform SRSF2 ( NMD- ) were not increased . These findings indicate that increases in the SRSF2 ( NMD+ ) isoform are due to disruption of EJC-stimulated NMD rather than increased transcription from the SRSF2 locus . Transcripts containing uORFs represent a second prominent class of NMD targets ( Gardner , 2008; Hurt et al . , 2013 ) . If uORFs are efficiently used relative to the major transcript ORF , the uORF TC can be recognized as a premature termination codon and degraded by NMD . We examined genes represented in a translation initiation site ( TIS ) database based on ribosome profiling data from HEK-293 cells ( Wan and Qian , 2014 ) . Of the 4310 genes expressed in our RNAseq data that also met ribosome profiling read coverage cutoffs for inclusion in the TIS database , 1578 showed evidence of uORF translation . As expected of a putative NMD factor , depletion of ICE1 systematically led to increased abundance of mRNAs with empirically identified uORFs , compared to those lacking evidence of uORF translation ( Figure 2C ) . The extent of up-regulation among uORF-containing mRNAs was less than that observed for PTC-containing mRNAs ( Figure 2A ) , possibly because leaky scanning or efficient re-initiation at downstream ORFs is sufficient to stabilize many mRNAs undergoing some degree of uORF translation ( Neu-Yilik et al . , 2011; Zhang and Maquat , 1997 ) . Because ICE1 participates in snRNA biogenesis as part of the little elongation complex , we asked whether the observed increases in uORF-containing mRNAs could be attributed to alterations in splicing . We detected alternative splicing events using two independent software packages , Majiq and Leafcutter , and excluded any genes found to undergo of splicing changes upon ICE1 depletion by either algorithm , using permissive cutoffs ( Supplementary file 5; see Materials and methods for details; [Li et al . , 2018; Vaquero-Garcia et al . , 2016] ) . Removing these genes from the analysis of uORF-containing mRNAs had no apparent impact ( Figure 2—figure supplement 1A ) . Likewise , removing the genes for which evidence of a PTC-containing isoform was detected ( Figure 2A ) had no effect on the relationship between ICE1 depletion and uORF mRNA abundance ( Figure 2—figure supplement 1B ) . Despite advances in uORF identification , there remain few human transcripts in which relative translation efficiencies of uORFs and primary ORFs are well characterized . For this reason , we chose to validate ICE1’s role in uORF-directed decay by studying key regulators of the Integrated Stress Response ( ISR ) known to be subject to translational regulation through uORFs that are inhibitory to translation of the downstream codon sequence ( CDS ) . Under normal conditions , expression of both the transcription factor CHOP/DDIT3 and the phosphatase regulatory subunit GADD34/PPP1R15A are impaired by a mechanism involving translation of an inhibitory uORF that prevents downstream reinitiation at the CDS ( Young and Wek , 2016 ) . The inability of the ribosome to translate the CDS and displace UPF1 and/or EJCs predisposes these transcripts for NMD ( Karam et al . , 2015; Schmidt et al . , 2015; Weischenfeldt et al . , 2008 ) . To investigate the effect of ICE1 depletion on transcript abundance of the ISR members CHOP/DDIT3 and GADD34/PPP1R15A , we performed RT-qPCR on RNA from cells treated with siICE1 or a non-targeting control siRNA . During ICE1 depletion , there was a 2–2 . 5 fold increase in the abundance of these uORF-containing transcripts ( Figure 2D ) . To ensure that this was not an artifact of ICE1 depletion potentially inducing the ISR , we also measured levels of the downstream ISR member ASNS , which is transcriptionally induced during the ISR but not under translational control ( Baird et al . , 2014; Barbosa-Tessmann et al . , 2000 ) . ICE1 depletion did not increase ASNS levels ( Figure 2D ) , and we did not observe alterations in mRNA processing of these and other genes induced by ICE1 depletion ( see below; Figure 2—figure supplement 2 ) , suggesting that the increase in CHOP/DDIT3 and GADD34/PPP1R15A levels was the product of reduced NMD activity . Due to the large number of human NMD targets made susceptible to decay via long 3’UTRs , we also examined the impact of ICE1 depletion on abundance of transcripts with varying 3’UTR lengths . We stratified genes into quartiles on the basis of the 3’UTR length of their most highly expressed isoforms in siNT-treated cells . The responses of genes with 3’UTRs in the shortest two quartiles ( <354 nt and 355–961 nt ) were indistinguishable , but ICE1 depletion resulted in significantly increased abundance of genes with 3’UTRs in the longest two quartiles ( 962–2158 nt and >2158 nt; Figure 2E ) . The failure of ICE1 depletion to enhance abundance of mRNAs with 3’UTRs of moderate length ( 355–961 nt ) differs from the significantly enhanced accumulation of such mRNAs in UPF1-depleted cells , but further work will be required to determine whether this is due to biological or technical differences ( Figure 2—figure supplement 1C ) . As with the uORF-containing mRNAs , exclusion of mRNAs exhibiting ICE1-dependent splicing changes or expression of PTC-containing isoforms had no effect on the analysis of ICE1 depletion and 3’UTR-dependent changes in mRNA abundance ( Figure 2—figure supplement 1D and E and Supplementary file 5 ) . In subsequent qRT-PCR studies , ICE1 knockdown led to increased abundance of some long 3’UTR-containing mRNAs sensitive to UPF1 depletion , including ZNF667 ( 1855 nt 3’UTR ) , CDK6 ( 10 , 235 nt ) , and SETD7 ( 5461 nt ) ( Figure 2F and Figure 2—figure supplement 2 ) . Together , our RNAseq analyses suggest that ICE1 depletion can disrupt NMD induced by uORFs , PTCs , and long 3’UTRs . To investigate whether the observed changes in mRNA abundance were due to increases in mRNA stability consistent with NMD inhibition , we used REMBRANDTS software to infer changes in mRNA stability from the relative abundance of exonic and intronic reads from each gene ( Alkallas et al . , 2017 ) . The REMBRANDTS algorithm is a refinement of earlier methods based on the idea that a change in exonic reads without a corresponding change in intronic reads is diagnostic of differential RNA stability , while concurrent changes in both exonic and intronic reads suggest altered transcription ( Ameur et al . , 2011; Gaidatzis et al . , 2015; Zeisel et al . , 2011 ) . The effects of ICE1 depletion on mRNA abundance and relative stability were highly correlated ( Spearman’s ρ=0 . 69; p<10−15; Figure 3—figure supplement 1A and Supplementary file 6 ) , and transcripts that significantly increased in steady-state abundance with ICE1 knockdown were preferentially stabilized upon ICE1 or UPF1 depletion ( Figure 3A and Figure 3—figure supplement 1B ) . Likewise , mRNAs that were increased in abundance upon UPF1 knockdown tended to be stabilized by ICE1 knockdown or UPF1 knockdown ( Figure 3—figure supplement 1C and D ) . Comparison of relative mRNA stability upon ICE1 depletion with responses to UPF1 or UPF3B siRNAs revealed that the populations of mRNAs stabilized by depletion of these two core NMD factors were stabilized to a similar extent by ICE1 depletion ( Figure 3B ) . Moreover , the mRNAs stabilized in both UPF1 and UPF3B knockdown cells exhibited an enhanced response to ICE1 depletion . Finally , we asked whether ICE1 depletion preferentially stabilizes mRNAs with uORFs or long 3’UTRs , with results that closely mirrored our findings based on steady-state mRNA abundance measurements ( compare Figure 2C and E with Figure 3—figure supplement 1E and F ) . To independently assess whether ICE1 affects NMD target mRNA stability , we used a metabolic labeling approach to quantify mRNA half-lives in cells transfected with non-silencing control , UPF1 , or ICE1 siRNAs ( Dölken , 2013; Russo et al . , 2017 ) . Following gene depletion , cells were pulse-labeled with 5-ethynyluridine , a modified nucleoside that allows covalent biotinylation and quantitative recovery of newly transcribed mRNA . Transcript half-lives can then be calculated based on the relative abundance of nascent and total mRNA . As observed in our transcriptome-wide analyses , these metabolic labeling studies indicated that either ICE1 or UPF1 depletion increased the stability of several well-characterized NMD target mRNAs , while failing to stabilize the non-NMD target control ASNS ( Figure 3C; Figure 3—figure supplement 2 ) . Together , these results indicate that the effect of ICE1 depletion on steady-state levels of mRNAs with NMD-inducing features is due to increased mRNA stability , consistent with a role for the protein in promoting NMD through UPF1 and UPF3B . NMD execution depends on mRNA nuclear export and translation in the cytoplasm , meaning that interference with multiple aspects of mRNA biogenesis , transport , and function can cause indirect inhibition of decay . The gain of luciferase signal observed with ICE1 siRNAs in the whole-genome screen suggests that ICE1 depletion did not prevent efficient NMD reporter export and translation ( Figure 1C and Supplementary file 1 ) . To corroborate this finding using endogenous NMD substrates , we detected several proteins produced from UPF1- and ICE1-sensitive mRNAs by immunoblotting ( Figure 3D ) . In all cases , we observed substantial increases in protein production from the target mRNAs , indicating that ICE1 depletion increases stability of NMD target mRNAs without interfering with their translation in the cytoplasm . Based on our observations implicating ICE1 in NMD , we next tested whether ICE1 associates with NMD or EJC proteins . Whole cell lysates from HEK-293 cells were subjected to immunoprecipitation assays using antibodies against ICE1 , with UPF3B antibodies as a positive control for recovery of both the trimeric UPF complex and EJC components . Immunoblot analysis of samples purified with these specific antibodies or control IgGs revealed ICE1 co-purification with the core EJC component eIF4AIII , ( Figure 4A ) , but not NMD proteins UPF1 , UPF2 , or UPF3B . Conversely , UPF3B co-purified with eIF4AIII , UPF1 , and UPF2 , but not ICE1 . The immunoprecipitation protocol used involves hypotonic lysis and gentle buffer conditions to retain RNA-binding protein interactions with RNAs ( see Materials and methods ) . To further investigate the nature of the interaction between ICE1 and eIF4AIII , we determined whether coprecipitation was disrupted by thorough digestion with an RNase A/T1 cocktail . Indicative of RNA-independent complex formation , ICE1 coprecipitation with eIF4AIII was not perturbed by RNase digestion . Importantly , in the UPF3B-positive control immunopurifications , RNase treatment abolished recovery of ribosomal protein S6 and PABPC1 , suggesting that the RNase conditions used efficiently disrupted RNA-mediated interactions ( Figure 4A ) . We also observed recovery of CASC3 and MAGOH in samples immunopurified with the antibody against ICE1 , indicating that ICE1 interacts with fully assembled EJCs ( Figure 4B ) . Reciprocally , we used monoclonal antibodies against eIF4AIII to coprecipitate ICE1 , along with the expected partners UPF3B , MAGOH , and CASC3 ( Figure 4C ) . Of note , we were unable to recover UPF3B or ICE1 from FLAG co-immunoprecipitations when the N- or C-termini of eIF4AIII were tagged , consistent with a previously reported mass spectrometry dataset ( data not shown; [Singh et al . , 2012] ) . Furthermore , while we did not observe co-purification of ICE1 and UPF3B using antibodies against endogenous proteins , stably overexpressed 3XMYC-UPF3B recovered ICE1 , in a manner dependent on the UPF3B EJC-binding domain ( Figure 4—figure supplement 1 ) . These findings indicate that ICE1 can participate in a EJC-UPF3B complex but preferentially associates with fully assembled EJCs not bound to UPF3B . As described above , ICE1 encodes a putative C-terminal MIF4G domain ( Figure 1—figure supplement 4 and Figure 5—figure supplement 1A ) . To determine if the C-terminus of ICE1 is important for EJC interaction , we transiently expressed 3XFLAG-tagged full-length ICE1 ( Figure 5—figure supplement 1A; 3XF-ICE1 ) , a variant lacking the C-terminus ( 3XF-ICE1 N-term ) , or the putative C-terminal MIF4G domain ( 3XF-MIF4GICE1 ) . The full-length ICE1 protein and the isolated C-terminus recovered endogenous eIF4AIII and CASC3 above background levels , but the protein lacking the C-terminus failed to enrich for the EJC proteins ( Figure 5—figure supplement 1B ) . To further probe the interaction between the putative MIF4G domain and the EJC , we employed stable Flp-In 293-TREx cell lines that allow for the integration of a single , inducible copy of a transgene of interest . Using stable expression of 3XF-MIF4GICE1 , we again observed that putative ICE1 MIF4G domain was sufficient for association with eIF4AIII ( Figure 5A ) . This interaction was retained during extensive RNase digestion , consistent with co-immunopurification studies using antibodies against endogenous proteins ( Figure 4A ) . As the C-terminal ~30 kDa of ICE1 is sufficient for its interaction with EJC proteins , we asked whether overexpression of the putative MIF4G domain could have a dominant negative effect . Consistent with this hypothesis , stable overexpression of the ( 3XFLAG-MIF4GICE1 ) protein used for immunoprecipitation experiments caused a reproducible several-fold increase in levels of two mRNAs that responded strongly to both ICE1 and UPF1 knockdown in RNAseq and qRT-PCR experiments , ANXA1 ( which contains a putative uORF; [Thierry-Mieg and Thierry-Mieg , 2006] ) and CGA ( previously identified as a strong target of UPF3B-dependent NMD; ( Chan et al . , 2007 ) ; Figure 5B ) . We note that this effect is not observed on all NMD targets tested , as the NMD substrate ATF3 was not stabilized during MIF4G overexpression , but is likely only apparent on highly NMD-sensitive transcripts such as ANXA1 and CGA ( Figure 5B ) . Together , these data indicate that ICE1’s interaction with eIF4AIII occurs through its putative MIF4G domain and that its pro-NMD function is disrupted when this domain is expressed in trans . Having observed ICE1 co-purification with endogenous core EJC proteins but not UPF3B , we next investigated whether NMD inhibition upon ICE1 depletion was due to disrupted EJC function . To determine whether ICE1 modulates EJC interactions with peripheral factors , we immunopurified fully assembled EJCs and associated proteins using an antibody against CASC3 . In lysates from HEK-293 cells depleted of ICE1 ( siICE1 ) , CASC3 co-immunoprecipitated with EJC core members eIF4AIII and MAGOH at comparable levels as in the non-targeting ( siNT ) control ( Figure 6A , left immunoblot , right bar graph illustrating eIF4AIII/CASC3 recovery efficiency quantified from three independent replicates ) , suggesting that ICE1 is not required for core EJC assembly . Interestingly , further immunoblotting of proteins co-purified with CASC3 showed a drastic reduction in the ability of the mature EJC to associate with NMD factors UPF3B and UPF2 when ICE1 was depleted . Following ICE1 depletion , CASC3 coprecipitated ~70% less UPF3B and , by association , nearly undetectable levels of UPF2 as compared to the non-targeting control ( Figure 6A , left immunoblot , right bar graph illustrating UPF3B/CASC3 recovery efficiency quantified from three independent replicates ) . We observed similar disruption of UPF3B-EJC interactions upon overexpression of the putative ICE1 MIF4G domain , suggesting that the phenotypes arising from these interventions share a mechanistic basis ( Figure 6—figure supplement 1 ) . The defect in UPF3B-EJC association in the absence of ICE1 could be due to impaired RNP assembly in the nucleus or failure to retain interactions in the cytoplasm . To begin to explore the role of ICE1 in regulating EJC-UPF3B interactions , we examined the localization of stably expressed GFP-tagged UPF3B in cells treated with control or ICE1 siRNAs . Following siRNA transfection , we marked nuclei with Hoechst stain and subjected cells to imaging flow cytometry ( Figure 6B ) . This approach allowed quantitative determination of the nuclear-cytoplasmic distribution of GFP-UPF3B in thousands of cells . In control cells , GFP-UPF3B was predominantly nuclear , but prominent diffuse cytoplasmic signal could also be observed , resulting in a median nuclear:cytoplasmic ratio of 2 . 89 ± 0 . 85 S . E . M . In contrast , ICE1 depletion caused UPF3B to accumulate in nuclei and be depleted from the cytoplasm . In knockdown cells , the median nuclear:cytoplasmic ratio shifted significantly to 5 . 59 ± 2 . 65 S . E . M . ( Mann Whitney test , p<10−15 ) , presumably decreasing UPF3B’s capacity to interact with mRNA targets of decay . It is thought that UPF3B is exported from the nucleus in association with mRNA-bound EJCs ( Gehring et al . , 2009b ) . To test whether reduced EJC binding contributes to increased nuclear accumulation of UPF3B , we analyzed the subcellular localization of the wild-type GFP-UPF3B and a mutant protein partially defective for EJC binding ( GFP-UPF3BΔ412-434 [Chamieh et al . , 2008; Gehring et al . , 2003] ) . The EJC-binding impaired UPF3B exhibited a modest defect in co-immunoprecipitation with eIF4AIII and significantly enhanced nuclear localization , consistent with the results from ICE1 knockdown cells ( Figure 6—figure supplement 2 ) . These findings suggest that inability of UPF3B to interact with EJCs may in part account for lack of export to the cytoplasm in the absence of ICE1 . Previous studies have shown that a UPF3B peptide can interact directly with EJCs assembled in vitro , albeit with modest affinity ( ~10 μM; [Buchwald et al . , 2010] ) . Our data suggest that ICE1 promotes association of UPF3B with the EJC , raising the possibility that ICE1 depletion may be overcome by boosting UPF3B levels . To test this hypothesis , we depleted ICE1 from parental HEK-293 cells or a cell line stably over-expressing 3XFLAG-UPF3B and assessed the ability of CASC3 to co-purify tagged and endogenous UPF3B . As shown in ( Figure 6A ) , ICE1 depletion from parental cells caused a decrease in UPF3B recovery with CASC3 ( Figure 7A , lanes 2 and 4 ) . However , this defect in association could be rescued by overexpressing 3XFLAG-UPF3B ( Figure 7A , lanes 6 and 8 ) . To determine whether restoration of the EJC-UPF3B interaction could also enhance NMD , we assayed the effect of ICE1 depletion on several NMD target mRNAs in parental and UPF3B-overexpression lines . As above , ICE1 depletion caused an increase in levels of ATF3 , GADD45B , GAS5 , and ANXA1 mRNAs . Importantly , over-expressing exogenous UPF3B partially rescued the NMD phenotype , causing significant reductions in the abundance of the EJC-mediated substrates ( Figure 7B ) . Consistent with previous findings , overexpression of UPF3B in the presence of ICE1 did not affect NMD target levels ( Huang et al . , 2011 ) , suggesting that the rescue of ICE1 depletion was not simply due to enhanced NMD efficiency . Together , the biochemical and functional rescue of ICE1 depletion by UPF3B overexpression indicates that the loss of NMD in the absence of ICE1 is due to a failure in UPF3B-EJC assembly or stability .
Beginning from whole-genome RNAi screening for proteins involved in the human NMD pathway , we present evidence that ICE1 is a EJC-associated protein that promotes UPF3B-EJC association and regulation of a large swath of NMD targets . This discovery was enabled by the CSA approach , which minimizes the impact of off-target effects mediated by siRNA seed sequences ( Marine et al . , 2012 ) . Indicative of the potential to identify novel NMD factors , this strategy resulted in high levels of enrichment of proteins known to be involved in NMD . The screen was particularly well suited to identification of factors involved in EJC-enhanced NMD , with all four core EJC components and two previously known assembly factors exhibiting high median seed-corrected Z-scores , in addition to the newly identified EJC-interacting factor ICE1 . We focus on the role of ICE1 in this study , but we expect that the screen data , particularly when combined with a recent CRISPR-based screen ( Alexandrov et al . , 2017 ) , will provide a valuable resource for further investigation of novel NMD factors . ICE1 depletion results in increased abundance of mRNAs selected for NMD on the basis of stop codons that violate the 50–55 nt rule , uORFs , and 3’UTR length . While decay of the former class is clearly stimulated by the presence of EJCs , the uORF class likely comprises a mix of transcripts that undergo EJC-dependent and EJC-independent decay , depending on transcript architecture , uORF translation frequency , and other factors . In our RNAseq studies , the largest effects of ICE1 loss appear to be on EJC-enhanced NMD , but our data also suggest a role for the protein in promoting decay of transcripts with longer than normal 3’UTRs . Even among the core NMD factors UPF2 and UPF3B , differential substrate preferences have been observed , leading to the proposal that there are multiple ‘branches’ of the NMD pathway , each with distinct cofactor requirements ( Chan et al . , 2007; Gehring et al . , 2005; Huang et al . , 2011; Metze et al . , 2013 ) . As yet , however , the RNA features underlying susceptibility to the various proposed NMD sub-pathways are not understood . Biochemical and functional assays suggest that ICE1 functions in NMD by promoting the proper association of UPF3B with EJCs ( Figure 7C ) . UPF2 association with EJCs is also decreased in ICE1 knockdown cells , presumably due to the inability of UPF3B to act as a bridging factor ( Chamieh et al . , 2008 ) . In addition to the defect in UPF3B-EJC association , we also observed reduced accumulation of UPF3B in the cytoplasm of cells depleted of ICE1 ( Figure 6B ) . Since UPF3B assembly with EJCs is thought to occur in the nucleus , enhanced UPF3B nuclear localization is unlikely to account for the observed defect in EJC binding . Instead , our data indicate that reduced export of UPF3B may be in part a consequence of decreased association with EJCs in the nucleus , preventing export of UPF3B with mRNPs . Alternatively , decreased stability of the UPF3B-EJC association could result in more rapid re-import of UPF3B . In turn , the reduced abundance of UPF3B in the cytoplasm may have the side-effect of impairing long 3’UTR-mediated NMD , explaining the apparent protection of such targets upon ICE1 depletion . Promotion of efficient UPF3B nuclear export by the EJC could also be necessary for its newly identified function in translation termination and help to explain why depletion of EJC factors has been observed to affect the decay of well-characterized 3’UTR length-dependent NMD targets ( Huang et al . , 2011; Neu-Yilik et al . , 2017 ) . Together , our data are consistent with a model in which ICE1 helps to prepare newly assembled EJCs for association with UPF3B ( Figure 7C ) . We find that ICE1 interacts with EJC components at endogenous levels , but have not observed an interaction between the endogenous ICE1 and UPF3B proteins . As these experiments were carried out under native conditions , we cannot exclude the possibility that this is in part due to dissociation during isolation . Arguing against this scenario , we readily observe UPF3B co-immunoprecipitation with the EJC under the same purification conditions and can detect co-purification of UPF3B with ICE1 when either UPF3B or the ICE1 C-terminus is overexpressed . Overexpressed UPF3B lacking the ability to bind EJCs also fails to associate with ICE1 , leading us to propose that ICE1 and UPF3B can concurrently interact with the EJC but that this complex is scarce or absent under normal cellular conditions . Together , our data suggest that ICE1 transiently interacts with fully assembled EJCs to promote UPF3B-EJC association . Possible effects of ICE1 binding to EJCs could be increased efficiency of EJC maturation or altered post-translational modification of EJC proteins , each of which could have the effect of increasing UPF3B recruitment or the stability of the complex . Alternatively , transient interactions between ICE1 and UPF3B could enhance UPF3B recruitment to EJCs . With increasing organismal complexity , the NMD pathway has evolved to use the exon junction complex to more efficiently discriminate between aberrant and normal mRNAs . The addition of components to the pathway in turn presents new opportunities for regulatory fine-tuning of decay . Interestingly , duplication of the ancestral UPF3 gene to yield UPF3A and UPF3B proteins has recently been shown to enable cell-type-specific control of NMD ( Shum et al . , 2016 ) . UPF3A has a reduced capacity to interact with EJCs but retains important UPF2-interacting residues , allowing it to disrupt UPF3B activity by competing for UPF2 binding . Our observations suggest that the association of UPF3B with the EJC is also a potential target for regulatory control of NMD . In this case , interference with ICE1 function appears to leave the UPF2-UPF3B interaction intact ( Figure 6—figure supplement 1 ) , while reducing the ability of UPF3B to associate with EJCs ( Figure 6 and Figure 6—figure supplement 1 ) . Since UPF3A and ICE1 affect distinct UPF3B interactions , they could be independently or concurrently used for cell-type- or condition-specific regulation of NMD . Further work will be required to understand whether ICE1 is subject to regulation , but our findings clearly point to an opportunity for cells and/or therapeutic interventions to manipulate ICE1 to decouple the EJC’s role in NMD from its contributions to pre-mRNA splicing , export , localization , and translation .
To generate an NMD-sensitive luciferase reporter , we constructed pCMV-3XFLAG-FLuc-β-globin ( 39PTC ) using PCR , traditional cloning methods , and β-globin reporters kindly provided by Professor Lynne Maquat ( University of Rochester ) . Pilot screens were performed by transfecting HEK-293 cells with pCMV-3XFLAG-FLuc-β-globin ( 39PTC ) using Lipofectamine 2000 according to the manufacturer’s protocol ( Thermo Fisher Scientific , Waltham , MA ) . 24 hr after transfection , selection for stable genomic integration was performed with 800 µg/mL Geneticin ( Thermo Fisher Scientific ) in 96-well plates at three cells/well density . Two weeks following the geneticin selection , 33 monoclonal cell colonies were isolated and expanded . The colony that showed the best response to the UPF1 siRNAs ( denoted pKC-4 . 06 ) was selected and expanded for use in the global assay . Whole-genome RNAi screening was conducted using the facilities of the Division of Pre-Clinical Innovation at NCATS ( Rockville , MD ) . Considerations for design , optimization , analysis and hit selection criteria of the RNAi assay were taken according to the extensive guidelines previously outlined ( Auld and Inglese , 2016; Martin et al . , 2012 ) . Briefly , siRNA screening was conducted using a genome-wide library of Silencer Select siRNAs ( Thermo Fisher Scientific ) comprising three siRNAs per gene for ~21 , 000 genes . 2 pmol of siRNA ( 20 nM final concentration ) was pre-spotted to 384-well plates , and 0 . 07 µL Lipofectamine RNAiMax was added in 20 µL of serum-free DMEM media . This complex was incubated at ambient temperature for 30 min prior to adding 1000 pKC-4 . 06 cells in 20 µL of 20% serum DMEM media . Cells were incubated for 72 hr prior to the addition of OneGlo ( Promega , Madison , WI ) luciferase assay reagent . For data analysis , screen data were filtered for off-target effects by applying the Common Seed Analysis ( CSA ) approach ( Marine et al . , 2012 ) . Hit selection was performed by converting normalized values into ranked Z-scores , and statistical significance determined with non-parametric Wilcoxon rank sum tests . STRING was used to identify any known interactions among the top screen hits ( median seed corrected Z > 1 . 5; [Szklarczyk et al . , 2015] ) . For comparison to the siRNA screen , high-throughput sequencing data ( NCBI SRA BioProject accession PRJNA353310 ) from a previously reported pooled CRISPR screen for NMD components were analyzed with MAGeCK software ( Alexandrov et al . , 2017; Li et al . , 2014 ) . Human Flp-In T-REx-293 cells ( Invitrogen , Carlsbad , CA; Cat . No . R78007 ) were maintained at 37°C and 5% CO2 in DMEM with 10% FBS and 1% pen/strep . Cells were obtained directly from the manufacturer and periodically assayed for mycoplasma contamination . Gene depletion studies were carried out by reverse transfection with siRNA non-targeting control ( siNT; Thermo Fisher Scientific , Silencer Select Negative Control #2 ) , siUPF1 ( 5’- AAGATGCAGTTCCGCTCCATTTT-3’; ( Mendell et al . , 2004 ) , siICE1 ( 5’-GGAAGATGATTATTCGTTATT-3’ ) , and siUPF3B ( 5’-GGAGAAGCGAGTAACCCTG-3’; ( Kim et al . , 2005 ) as previously described ( Ge et al . , 2016 ) . Briefly , 25 pmols of gene-specific or non-targeting siRNA was directly pipetted into each well of a Falcon six-well flat bottom multiwell cell culture plate ( Corning , Corning , NY; Cat . No . 353046 ) . Reverse transfections were then conducted using Lipofectamine RNAiMAX reagent according to the manufacturer’s protocol ( Thermo Fisher Scientific , Cat . No . 13778150 ) . As the transfection master mix is comprised of Opti-MEM reduced serum medium ( Thermo Fisher Scientific Cat . No . 31985062 ) , an equal volume of DMEM with 20% FBS with pen/strep was used to plate cells and attain a final 10% FBS concentration . 72 hr following the transfection , lysates were collected for protein or RNA isolation . Total RNA was extracted and purified from whole-cell lysates using the RNeasy Mini Kit with on-column DNase digestion ( Qiagen , Hilden , Germany; Cat . No . 74106 ) . RNA concentration was determined on a NanoDrop 1000 ( Thermo Fisher Scientific ) , and 1 µg of RNA used as a template for cDNA library preparation using the Maxima First Strand cDNA synthesis Kit for RT-qPCR ( Thermo Fisher Scientific , Cat . No . K1641 ) . The resulting cDNA libraries were further diluted 20-fold with ultrapure H2O and subsequently analyzed by RT-qPCR using iTaq Universal SYBR Green Supermix ( BioRad Laboratories , Hercules , CA; , Cat . No . 1725124 ) on a Roche LightCycler 96 instrument ( Roche Diagnostics Corporation , Indianapolis , IN ) . Sequences for gene-specific primers used for amplification are listed in Supplementary file 7 . 2-∆∆CT values were calculated using GAPDH or b-ACTIN for normalization , and all reported values are representative of three independent biological replicates ( Ge et al . , 2016 ) . pcDNA5-Flag-KIAA0947 ( ICE1 ) was a gift from Joan Conaway and Ronald Conaway ( Addgene plasmid # 49428; [Takahashi et al . , 2011] ) . Phusion High-Fidelity DNA Polymerase ( New England Biolabs , Cat . No . M0530 ) was used to PCR amplify ICE1 and UPF3B sequences from plasmid # 49428 and human cDNA , respectively , and cloned into pcDNA5-FRT-TO-3XF using traditional methods . Deletion constructs were generated using diverging 5’-phosphorylated primers followed by ligation with T4 DNA Ligase ( New England Biolabs , Cat . No . M0202S ) , and all constructs were sequence validated . To determine specific mRNA half-life measurements during ICE1 and UPF1 depletion , HEK-293 cells were reverse transfected with a gene-specific or non-targeting control siRNA for 72 hr . At the end of the depletion , cells were treated with 0 . 2 mM 5-ethynyl Uridine ( 5-EU ) for 60 mins . Cells were then immediately harvested and RNA was isolated and column purified using the RNeasy Mini Kit with on-column DNase digestion ( Qiagen , Hilden , Germany; Cat . No . 74106 ) . In addition , Drosophila S2 cells ( ThermoFisher Scientific , Cat . No . R690-07 ) were cultured at ambient temperature for 24 hr in Sf-900 media ( ThermoFisher Scientific , Cat . No . 10967032 ) containing 0 . 1 mM 5-EU , and RNA was isolated and processed as with the human cells for use as a spike-in control . The Drosophila spike in control was then quantified and later used to account for variations in biotinylation efficiency and recovery on the streptavidin beads . To differentiate total and nascent RNA levels in each sample , 2 µg of sample RNA was combined with 200 ng ( 10% ) of the spike-in control and partitioned using the Click-iT Nascent RNA Capture Kit ( ThermoFisher Scientific , Cat . No . C10365 ) following the manufacturer’s protocol . 1 µg of biotinylated RNA from each sample was used for ‘total’ cDNA library preparation , with the remaining 1 µg of RNA from that sample applied to the Streptavidin T1 magnetic beads for labeled RNA capture and cDNA synthesis . The resulting 50 µL cDNA libraries were diluted , and mRNA abundance was determined using qRT-PCR as previously described . Nascent RNA recovery was normalized to Drosophila RP49 mRNA levels , and individual half-lives were determined using the equation: t1/2 = -tL * ln ( 2 ) /ln ( 1/R ) , where tL is the EU labeling time in minutes and R is the abundance in nascent RNA fraction/abundance in total RNA fraction ( Dölken , 2013; Haque et al . , 2018 ; Russo et al . , 2017 ) . Total RNA from HEK-293 cells transfected with siICE1 , siUPF1 , siUPF3B , or siNT as above was assessed using an Agilent Bioanalyzer 2100 , subjected to ribosomal RNA removal using the Ribo-Zero rRNA Removal Kit ( Illumina , San Diego , CA ) , and used for library preparation with the Illumina TruSeq Stranded Total RNA Sample Preparation Kit ( Illumina ) . Paired-end 50 nt reads were generated on the Illumina HiSeq 2000 platform , and the resulting reads were mapped using HISAT2 software to the GRCh37/hg19 combined genome and transcriptome indexes provided by the authors ( Kim et al . , 2015 ) . Mapped reads and Ensembl gene models were used for transcriptome assembly by StringTie ( Pertea et al . , 2015 ) , after which individual assemblies were merged using TACO ( Niknafs et al . , 2017 ) to obtain a higher-confidence annotation better reflecting HEK-293 expression patterns . The resulting TACO-curated transcriptome was used for pseudoalignment-based transcript quantification and analysis of differential gene and isoform usage with Kallisto/Sleuth ( Bray et al . , 2016; Pimentel et al . , 2017 ) . To identify transcript isoforms violating the 50–55 nt rule , the longest ORF in each TACO-derived transcript was annotated with IsoformSwitchAnalyzeR ( Vitting-Seerup and Sandelin , 2017 ) . The most highly expressed isoform from each gene , as judged by the transcripts per million ( TPM ) Kallisto calculation , was used to compute 3’UTR lengths for downstream gene-level analyses . Changes in RNA stability were inferred from RNAseq data using REMBRANDTS ( Alkallas et al . , 2017 ) , following HISAT2 mapping to the GRCh38 combined genome and transcriptome indexes provided by the authors ( Kim et al . , 2015 ) . Reads mapping to constitutive exons and introns of Ensembl GRCh37 . 87 annotations were quantified with HTSeq ( Anders et al . , 2015 ) , and a read cutoff stringency of 0 . 99 was used for REMBRANDTS analysis . Majiq and Leafcutter were used to analyze splicing changes following siICE1 and siNT treatment ( Li et al . , 2018; Vaquero-Garcia et al . , 2016 ) . For Majiq , a change in isoform usage of 10% or greater at the 95% confidence level was used to identify genes potentially undergoing alternative splicing . For Leafcutter , genes with a change in isoform usage of 10% or greater and adjusted p<0 . 05 were selected . Raw and processed RNAseq data are available in the NCBI GEO database , accession GSE105436 . A freeze/thaw lysis protocol was implemented to harvest lysates for protein assays , as described ( Hogg and Collins , 2007 ) . Briefly , cells were collected in 15-mL conical tubes following trypsinization and resuspension in media , centrifuged at 500 xg for 5 min at 4°C , and rinsed once with 1 mL ice-cold PBS . Lysates were then transferred to a 1 . 5 mL Eppendorf tube and centrifuged at 2000 xg for 1 min at 4°C . Cell pellet volume was estimated with an analytical balance , and the pellet resuspended in 5X packed cell volume of ice-cold hypotonic lysis buffer ( e . g . , 500 µL of lysis buffer for a 100 µL cell pellet; buffer contains 20 mM HEPES , pH 7 . 4 , 2 mM MgCl2 , 10% glycerol , 1 mM DTT , 1X Protease and Phosphatase Inhibitor Cocktail , Thermo-Fisher Cat . No . 78440 ) . Following a 5 min incubation on ice , cell extracts were snap frozen in liquid nitrogen , briefly thawed in a 37°C H2O bath , and frozen and thawed again . With the second and final thaw , 5M NaCl was added to achieve a final 150 mM concentration , and extracts were allowed to incubate on ice for 5 min . To isolate the soluble fraction , extracts were centrifuged for 15 min at 20 , 000 x g at 4°C , and the supernatant was aliquoted to fresh 1 . 5 mL Eppendorf tubes for a final snap freeze in liquid nitrogen prior to storage at −80°C . Western blots were performed using the NuPAGE electrophoresis and transfer systems ( Invitrogen ) . Denatured proteins were resolved on 3–8% Tris-acetate or 4–12% bis-Tris gels depending on molecular weight , and transferred to nitrocellulose membranes according to the NuPAGE manufacturer’s protocol ( Invitrogen ) . Membranes were incubated on an orbital shaker with blocking buffer for fluorescent western blotting ( Rockland Immunochemicals , Limerick , PA ) for 1 hr at room temperature . Incubations with the primary antibody were performed overnight at 4°C on an orbital shaker . Following three 10 min washes with 1XTBS ( 0 . 1% Tween-20 ) , membranes were incubated with the appropriate secondary antibody for one hour at room temperature , followed by three additional 10 min washes with 1XTBS . Final quantitative western blot images were obtained on a Licor Odyssey imaging system ( LI-COR Biosciences , Lincoln , NE ) . All antibodies used for immunoprecipitation and/or western blotting are provided in the Key Resources table . The eIF4AIII hybridoma cell line was developed by the Protein Capture Reagents Program and obtained from the Developmental Studies Hybridoma Bank ( DHSB ) , created by the NICHD of the NIH and maintained at The University of Iowa , Department of Biology , Iowa City , IA 52242 . A murine hybridoma cell line ( PCRP-EIF4A3-3A2-f ) that expresses secreted monoclonal antibody against the full-length human eIF4AIII epitope was obtained from the Developmental Studies Hybridoma Bank at the University of Iowa . Hybridoma cultures were maintained at 37°C and 5% CO2 in DMEM ( Gibco ) supplemented with 1% pen/strep and 10% Ultra Low IgG FBS ( ThermoFisher Scientific , Catalog # 16250078 ) . Cultures were visually inspected for overall health and divided while the cells were still in log phase growth , or approximately 5 × 105 - 1 × 106 cells/mL . To purify monoclonal eIF4AIII antibody from the culture supernatant , 50 mL of culture media from cells in log phase growth was centrifuged at 4°C for 5 min at 1000 RPM . The supernatant was then sterile filtered using a 0 . 4 mm filter and stored at 4°C overnight . Prior to affinity purification , a protein G column ( HiTrap Protein G HP 1 mL , Catalog # 339-0485-81 ) was equilibrated with 10 column volumes of 100 mM Tris-HCl pH 8 . 0 . Meanwhile , the pH of the cell culture supernatant was adjusted by adding 5 mL ( ~10% total volume ) of 1 . 0M Tris-HCl . The total supernatant volume was passed through the protein G column attached to an AKTApurifier 10 ( G . E . ) at a speed of 1 mL/min . After loading the sample , the column was washed with 10 column volumes of 100 mM Tris-HCl , followed by another 10 column volumes of 10 mM Tris-HCl . The antibody was then eluted from the column with 50 mM glycine pH 3 . 0 , and collected into Falcon tubes containing a neutralization buffer of 100 mL 1M Tris-HCl . Fractions containing the antibody were determined by monitoring UV280 values during the elution process , and purity and concentration determined by Coomassie staining and Pierce 660 calorimetric protein assay reagent ( Catalog # 22660 ) , respectively . eGFP-UPF3B and eGFP-UPF3BΔ412–434 expression constructs were cloned into a tetracycline-inducible pcDNA5 vector and integrated into human Flp-In T-REx-293 cells . Gene-specific depletions for ICE1 and a non-targeting control siRNA were performed as described above . At the time of transfection , eGFP-fusion protein expression was induced by the addition of doxycycline to a final concentration of 200 ng/mL . Prior to harvesting the samples , Hoechst dye was added directly to the media to a final concentration of 10 µg/mL for 45 min at 37°C . At the end of the incubation period , cells were trypsinized , rinsed with 1X PBS , and resuspended in 50 μL of 1X PBS with 2% FBS in low protein binding tubes . Spectral compensation controls including non-fluorescent parentals , GFP-only , and Hoechst-only labeled cells were processed in parallel to allow for subsequent data acquisition and analysis . Data were collected and analyzed on an ImageStream imaging flow cytometer ( Amnis , Millipore Sigma , Seattle , WA ) , according to the manufacturer’s protocol . Briefly , data were collected using INSPIRE software ( Amnis ) and 405 nm and 488 nm lasers were used to excite Hoechst and eGFP , respectively . Laser powers were chosen in order to prevent pixel saturation , and 10 , 000 single and focused events were captured per experimental condition . Single color compensation controls were merged to generate a compensation matrix and all sample files were analyzed with this matrix applied . Data analysis was performed with IDEAS software ( version 6 . 2 , Amnis ) , with eGFP expression levels or ‘intensity’ calculated by the software as the sum of pixel values minus the background pixel values . To determine nuclear localization of the eGFP signal , a morphology mask was created to conform to the shape of the nuclear Hoechst imagery . For determining the cytoplasmic area , a combined mask that subtracted the nuclear Hoechst imagery from an erode mask of the brightfield image of the total cell area was made . The intensity of GFP was then calculated by the ratio of GFP expression in the nucleus to the cytoplasmic region . | The DNA in our cells contains the hereditary information that makes each of us unique . Molecules called mRNAs are copies of this information and are used as templates for making proteins . When a strand of incorrectly copied mRNA , or one including errors from the original DNA template , is recognized , our cells destroy the mRNA to prevent it from producing a damaged protein . Organisms from yeast to humans have evolved a mechanism for finding and destroying faulty mRNAs , called mRNA surveillance . Animals are particularly reliant on mRNA surveillance , as their proteins are often made from cutting and pasting together mRNA from different portions of DNA , in a process known as splicing . Despite being a vital process , we still lack a good understanding of how mRNA surveillance works . Now , Baird et al . used human kidney cells that produced an error-containing mRNA that could be tracked . To investigate how efficient RNA surveillance is under different conditions , the levels of individual proteins were reduced one at a time . By tracking the amount of faulty mRNA , it was possible to find out if a single protein plays a role in human mRNA surveillance . If the number of faulty mRNAs is high when a protein is reduced , it suggests that this protein may be involved in mRNA surveillance . Baird et al . screened more than 21 , 000 proteins , the majority of proteins made in human cells . Many of the proteins that stood out as important in mRNA surveillance were the ones already known to be relevant in yeast and worm cells . But the experiments also identified new proteins that appear to play a role specifically in human RNA surveillance . One of the proteins , ICE1 , is essential for the relationship between mRNA splicing and mRNA surveillance . Without ICE1 , the mRNA surveillance machinery can no longer find and destroy faulty mRNAs . Nearly one-third of genetic diseases are caused by mutations that result in faulty mRNAs , which can be detected by mRNA surveillance pathways . Depending on the disease , destroying these error-containing mRNAs can either improve or worsen disease symptoms . A better understanding of the factors that control human RNA surveillance could one day help to develop treatments that affect mRNA surveillance to improve disease outcomes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"biochemistry",
"and",
"chemical",
"biology"
] | 2018 | ICE1 promotes the link between splicing and nonsense-mediated mRNA decay |
Retinal ganglion cell ( RGC ) axons of binocular animals cross the midline at the optic chiasm ( OC ) to grow toward their synaptic targets in the contralateral brain . Ventral anterior homeobox 1 ( Vax1 ) plays an essential role in the development of the OC by regulating RGC axon growth in a non-cell autonomous manner . In this study , we identify an unexpected function of Vax1 that is secreted from ventral hypothalamic cells and diffuses to RGC axons , where it promotes axonal growth independent of its transcription factor activity . We demonstrate that Vax1 binds to extracellular sugar groups of the heparan sulfate proteoglycans ( HSPGs ) located in RGC axons . Both Vax1 binding to HSPGs and subsequent penetration into the axoplasm , where Vax1 activates local protein synthesis , are required for RGC axonal growth . Together , our findings demonstrate that Vax1 possesses a novel RGC axon growth factor activity that is critical for the development of the mammalian binocular visual system .
Development of the mammalian binocular visual system requires topographic synaptic connections of retinal ganglion cell ( RGC ) axons to neurons on the lateral geniculate nucleus and superior colliculus of the brain ( Lemke and Reber , 2005 ) . To access their synaptic targets , RGC axons exit from the retina and grow in selective directions by recognizing guidance cues expressed in optic pathway structures , including the optic disc ( OD ) , optic stalk ( OS ) , optic chiasm ( OC ) , and optic tract ( OT ) ( Petros et al . , 2008 ) . RGC axon guidance cues include cell surface ligands such as semaphorins in the OS and ephrinB2 in the OC , and soluble factors such as netrin-1 in the OD and Slit1 in areas surrounding the OC ( Erskine and Herrera , 2007 ) . Ultimately , only about 3% of mouse RGC axons , which originate from the ventral and temporal part of the retina , are linked to targets on the same side of the brain , whereas a majority of RGC axons are connected to those on the opposite side after crossing the midline at the OC , located at the ventral-medial hypothalamic ( vHT ) area . Subsets of vHT cells therefore express molecules that determine the directionalities of RGC axons at the OC . It has been shown that vHT radial glial cells express ephrinB2 , which binds to EphB1 receptors expressed in ventral-temporal RGC axons and repels the axons toward the ipsilateral optic tract ( Nakagawa et al . , 2000; Williams et al . , 2003 ) . In addition , vascular endothelial growth factor 164 ( VEGF164 ) , an isoform of the vascular endothelial growth factor VEGF-A ( Soker et al . , 1996 ) , and neuronal cell adhesion molecule ( NrCAM ) expressed in the vHT have been suggested to support the growth of RGC axons across the vHT midline by binding to neuropilin-1 and plexin-A1 , respectively ( Williams et al . , 2006; Erskine et al . , 2011; Kuwajima et al . , 2012 ) . To receive the directional guidance of these molecules at the vHT , RGC axons must pass through the ventral-lateral diencephalic area , where repulsive guidance cues , such as Slit and semaphorins , are expressed at high levels ( Erskine et al . , 2000; Plump et al . , 2002 ) . However , the molecules that support RGC axon growth toward the vHT midline are still unknown . Ventral anterior homeobox 1 ( Vax1 ) is a homeodomain transcription factor expressed in various ventral-medial forebrain-derived structures , including the medial and lateral geniculate eminences , the ventral septum , the anterior entopeduncular area , the preoptic area , the vHT , and the OS ( Hallonet et al . , 1998; Bertuzzi et al . , 1999 ) . Genetic inactivation of Vax1 in humans and mice causes agenesis of multiple midline structures of the brain , including the anterior commissure , the corpus callosum , and the OC , in addition to the coloboma of the eye ( Bertuzzi et al . , 1999; Hallonet et al . , 1999; Slavotinek et al . , 2012 ) . RGC axons in Vax1-deficient mice can grow through the OS but cannot access the vHT area and fail to form an OC . Vax1 is not expressed in RGCs despite its critical roles in growth and fasciculation of RGC axons ( Bertuzzi et al . , 1999; Mui et al . , 2005 ) . Therefore , it has been suggested that defects in OC formation in Vax1-deficient mouse RGCs might be caused by the loss or gain of axon guidance cues that are potential transcription targets of Vax1 in vHT cells . Contrary to expectation , we here found that Vax1 promoted RGC axon growth in a transcription-independent manner . Moreover , Vax1 is secreted from vHT cells and binds and enters RGC axons to stimulate axonal growth . This unexpected trafficking of Vax1 to RGC axons was mediated by heparan sulfate proteoglycans ( HSPGs ) , such as syndecan and glypican , expressed in RGC axons . However , Vax1 binding to HSPGs was not sufficient to trigger RGC axon growth; penetration into the RGC axoplasm and subsequent stimulation of local protein synthesis were also necessary . Taken together , our findings reveal an unconventional function of Vax1 as an RGC axon growth factor that enables RGC axons to grow toward the midline during development .
Vax1 is expressed in cells located in optic pathway structures , such as the OS and vHT , and plays an essential role in fasciculation of RGC axons and formation of the OC ( Bertuzzi et al . , 1999; Hallonet et al . , 1999 ) . At the vHT of days post coitum 14 . 5 ( E14 . 5 ) mouse embryo , Vax1 is expressed in Sox2 ( SRY box 2 ) -positive neural progenitor cells ( NPCs ) and RC2-detectable nestin-positive radial glia ( Figure 1A , B; top rows ) , which is known to provide RGC axon guidance cues ( Petros et al . , 2008 ) . Although the morphology of the chiasm is abnormal in Vax1-deficient ( Vax1−/− ) mice ( Bertuzzi et al . , 1999 ) , these OC-forming cells and several chiasm-localized cues ( NrCAM and Vegfa ) are still present ( Figure 1A , B , bottom rows; Figure 1—figure supplement 1 ) . However , vHT explants from Vax1−/− mice were unable to attract RGC axons regardless of the Vax1 gene status of co-cultured retinal explants , whereas wild-type ( WT; Vax1+/+ ) vHT explants were able to attract RGC axons projected from Vax1−/− explants as well as WT explants ( Figure 1C , D ) . We therefore concluded that Vax1 controls the RGC axonal growth in a non-cell autonomous manner , potentially by regulating the expression of unidentified secreted axon-guidance molecules . 10 . 7554/eLife . 02671 . 003Figure 1 . Vax1 regulates RGC axonal growth in a non-cell autonomous manner . ( A ) Cells expressing Vax1 ( green ) in brain sections ( coronal; 16 μm ) from E14 . 5 Vax1+/+ ( top ) and Vax1−/− ( bottom ) embryos were detected by co-immunostaining for the NPC marker Sox2 ( red ) and post-mitotic neuronal marker tubulin-βIII ( blue ) , detected with the Tuj1 antibody . The right-most three columns are the magnified images of dotted boxes in the left column image . The results indicate that Vax1 is expressed in a subpopulation of Sox2-positive NPCs ( arrowheads ) but is not detectable in Tuj1-positive neurons . ( B ) Vax1-expressing cells in the vHT were also compared with RC2-positive radial glia . Arrowheads indicate RC2-positive radial glial cells expressing Vax1 . Scale bars: 50 μm . ( C ) The vHT and dorsal neural retina ( NR ) were isolated from WT ( Vax1+/+ ) and Vax1-knockout ( Vax1−/− ) E13 . 5 mouse embryos and co-cultured in a combinatorial manner for 48 hr . The explants were fixed and immunostained with an anti-NF160 antibody ( α-NF160; red ) ; nuclei were counterstained with DAPI ( blue ) . Dotted boxes indicate the area magnified in each inset . Red dotted lines link centers of retinal explants and vHT explants . Scale bars: 500 μm . ( D ) The angular distribution of RGC axons in images was measured by counting pixels containing immunostaining for the axon marker NF160 ( axon counts ) , as described in ‘Materials and methods’ , and presented graphically . + , forward direction angle segment; 0 , neutral direction angle segments; − , reverse direction angle segment . The values in the bar are averages , error bars denote standard deviations ( SDs ) , and numbers under y-axis labels are the numbers ( n ) of explants analyzed from three independent experiments . p-values determined by the analysis of variance ( ANOVA ) are between 0 . 01 and 0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 00310 . 7554/eLife . 02671 . 004Figure 1—figure supplement 1 . Expression of RGC axon attractive cues in Vax1−/− mouse vHT . To determine whether defective projection of RGC axons to the vHT midline of Vax1−/− mice resulted from defective expression of RGC axon growth factors , such as VEGF164 or NrCAM ( Williams et al . , 2006; Erskine et al . , 2011 ) , in the vHT , we compared expression of these cues between WT ( Vax1+/+ ) and Vax1-knockout ( Vax1−/− ) mouse embryos . Expression of Vegfa mRNA ( A ) and NrCAM protein in E14 . 5 Vax1+/+ and Vax1−/− mouse embryonic sections ( coronal; 16 μm ) was examined by in situ RNA hybridization ( ISH; A ) and immunofluorescence staining ( IF; B ) , respectively . The vHT of the Vax1−/− mouse brain retained , but expanded , expression of both Vegfa and NrCAM . Scale bars: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 004 To identify Vax1-regulated secreted factors that control RGC axonal growth from co-cultured retinal explants , we overexpressed mouse Vax1 in COS7 cells . RGC axons from retinal explants grew preferentially toward co-cultured Vax1-expressing COS7 cell aggregates , whereas RGC axons projected in random directions upon co-incubation with untransfected or Vax2-overexpressing COS7 cell aggregates ( Figure 2A , B ) . Because Vax2 shares an identical homeodomain with Vax1 ( Barbieri et al . , 1999 ) , these results indicate that the RGC axon growth stimulatory activity is specific for Vax1 . 10 . 7554/eLife . 02671 . 005Figure 2 . Vax1 homeodomain protein is a secreted protein . ( A ) COS7 cells overexpressing Myc-tagged mouse Vax1 , Vax1 ( R152S ) , or Vax2 were co-cultured with E13 . 5 mouse retinal explants ( NR ) for 48 hr . The explants were then stained with a rabbit anti-Myc antibody ( green ) and a mouse anti-NF160 antibody ( red ) . Nuclei of explant cells were counterstained with DAPI ( blue ) . Dotted red lines indicate the connections between the centers of two explants . Scale bars: 500 μm ( left column ) and 100 μm ( magnified immunostained images in two right-hand columns ) . ( B ) The angular distribution of RGC axons was measured as described in Figure 1D . The values in the bar are averages and error bars denote SDs . Numbers under y-axis labels are the number of explants analyzed from three independent experiments . p-values are between 0 . 01 and 0 . 005 ( ANOVA ) . ( C ) Growth medium from COS7 cells overexpressing Myc-Vax1 , Myc-Vax1 ( R152S ) , or Myc-Vax2 was collected , and the presence of Vax protein in the growth medium ( GM ) was detected by Western blotting ( WB ) with an anti-Myc antibody . The relative amounts of secreted protein were also measured by analyzing the level of proteins in the COS7 cell lysates ( CL; 5% of total ) . ( D ) vHTs isolated from WT and Vax1−/− E13 . 5 mouse embryos were cultured for 24 hr , after which GM was collected for detection of secreted Vax1 protein by Western blotting . CL , cell lysates of vHT explants ( 5% of total ) . ( E ) Cerebrospinal fluid ( CSF ) from E14 . 5 mouse embryos ( n = 20 ) was collected , a supernatant fraction ( S3 ) was separated from cell debris ( P2 ) by step-wise centrifugation ( see ‘Materials and methods’ for details ) , and the presence of Vax1 protein was examined by Western blotting . The presence of β-tubulin , a cytoplasmic protein , was also examined in GM and CSF fractions to check for the possible non-specific release of intracellular proteins from dead cells . CL , E14 . 5 vHT cell lysates ( 2% of total lysates from one embryonic vHT ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 00510 . 7554/eLife . 02671 . 006Figure 2—figure supplement 1 . Relative transcriptional activities of Vax1 mutants used in this study . HEK293T cells ( 105 ) were transfected with pcDNA6-V5 expression vectors ( 1 μg ) for Vax1 , Vax1 ( R152S ) , or Vax1 ( WF/SR ) together with pGL3-Tcf7l1-luciferase ( 0 . 2 μg ) ( Vacik et al . , 2011 ) and pCMV-β-gal ( 0 . 2 μg ) plasmids . Luciferase activity in the transfected cells was measured 24 hr post-transfection . The values were then normalized to β-galactosidase activity in the same cells to obtain the relative luciferase activity of the cells . The values are averages obtained from three independent experiments , and error bars denote SDs ( **p < 0 . 001; ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 00610 . 7554/eLife . 02671 . 007Figure 2—figure supplement 2 . Interference of Vax1 intercellular transfer by sequestering extracellular Vax1 . E13 . 5 retinal explants were co-incubated with Myc-Vax1 transfected COS7 cell aggregates for 24 hr in the presence of rb-IgG ( 1 μg/ml; top ) or anti-Vax1 antibody ( α-Vax1; 1 μg/ml; bottom ) . The explants were then fixed for immunostaining with mouse anti-NF160 antibody ( red ) and rabbit anti-Myc antibody ( green ) . Immunostaining images of dotted boxed area in the left column are shown in the right column . Arrows indicate Vax1-Myc proteins internalized to RGC axons . Scale bars: left , 500 μm; right , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 007 We next examined whether Vax1-induced RGC axonal growth is dependent on Vax1 transcription activity by co-incubating retinal explants with COS7 cells expressing a transcriptionally inactive Vax1 ( R152S ) mutant ( Figure 2 , Figure 2—figure supplement 1 ) . This mutation was reported in a human patient who exhibited coloboma , cleft palate , and agenesis of corpus callosum ( ACC ) , phenotypic manifestations similar to those of Vax1−/− mice ( Slavotinek et al . , 2012 ) . Unexpectedly , we found that Vax1 ( R152S ) -expressing COS7 cells were also able to induce RGC axonal growth as efficiently as WT Vax1-expressing COS7 cells ( Figure 2A , third row , B ) , suggesting that Vax1 induces RGC axonal growth in a transcription-independent manner . More strikingly , Vax1 and Vax1 ( R152S ) proteins were not only expressed in transfected COS7 cells , they were also detectable in neurofilament 160 kDa ( NF160 ) -positive RGC axons projecting from co-cultured retinal explants ( Figure 2A , right two columns ) . These axonal Vax1-immunostaining signals were remarkably decreased in the presence of a rabbit anti-Vax1 polyclonal antibody ( α-Vax1 ) that sequesters Vax1 in the growth medium ( Figure 2—figure supplement 2 ) . Furthermore , Vax1 and Vax1 ( R152S ) proteins were found in the growth medium of transfected COS7 cells , whereas Vax2 protein was not ( Figure 2C ) . Since the viability of the transfected COS7 cells were not different from each other ( data not shown ) , these results suggest that Vax1 proteins in the growth medium and co-cultured RGC axons did not originate from dead cells . Similar to overexpressed Vax1 in COS7 cells , endogenous Vax1 expressed in vHT explants was detectable in the growth medium ( Figure 2D ) . Furthermore , Vax1 protein was also identified in the cerebral spinal fluid ( CSF ) of E14 . 5 mouse embryos ( Figure 2E ) , suggesting that Vax1 is secreted in vivo as well as in vitro . We further tested whether secreted Vax1 is capable of directly binding to RGC axons and regulating axonal growth using purified recombinant Vax1 protein . Time-lapse recordings of RGC axons revealed that fluorescein isothiocyanate ( FITC ) -labeled , His-tagged Vax1 ( Vax1-His ) protein added to the growth medium of retinal explants accumulated in RGC axons and exerted strong growth stimulatory effects on them ( Figure 3A , B; Videos 1–3; Figure 3—figure supplement 1 ) . The axon growth stimulating effects of Vax1-His were applied equally to retinal quadrants ( Figure 3—figure supplement 2 ) , implicating Vax1 is not a region-specific axon growth factor . The transcriptionally inactive Vax1 ( R152S ) -His mutant protein was also detectable in RGC axons and stimulated axonal growth as efficiently as WT Vax1-His ( Figure 3C , D ) , suggesting that extracellular Vax1 moves to RGC axons and induces axonal growth in transcription-independent manner . Despite that Vax2 is not secreted ( Figure 2C ) , the ability of recombinant Vax2-His to be internalized and induce RGC axonal growth is almost equivalent to that of Vax1 ( Figure 3—figure supplement 3 ) , implicating that internalization but not secretion is a conserved characteristic of VAX transcription factors . 10 . 7554/eLife . 02671 . 008Figure 3 . Vax1 protein is a retinal axon growth factor . ( A ) E13 . 5 mouse retinal explants were cultured for 24 hr and then treated with 6X-His-FITC peptide ( 100 ng/ml ) or recombinant Vax1-His-FITC protein ( 500 ng/ml ) for an additional 24 hr . Images of RGC axons were taken every 15 min for 16 hr before immunostaining with anti-Vax1 and anti-His antibodies ( Videos 1 and 2; Figure 3—figure supplement 1 ) . The accumulation of 6X-His-FITC and Vax1-His-FITC in growing RGC axons was also visualized by detecting FITC fluorescence signals ( inset images ) . Red arrowheads indicate the area magnified in each inset . ( B ) The changes in RGC axonal length during the recording were plotted after adjusting the initial length to 100% . ( C ) Retinal explants treated with 6X-His ( 25 ng/ml ) , Vax1-His ( 100 ng/ml ) , or Vax1 ( R152S ) -His ( 100 ng/ml ) for 24 hr were stained with rabbit anti-Vax1 ( green ) and mouse anti-NF160 ( red ) antibodies to visualize Vax1 protein in RGC axons . Arrowheads indicate the area magnified in each inset . Scale bars: 500 μm . ( D ) Relative numbers of axon bundles projecting from retinal explants were indirectly measured by counting the pixels containing NF160 immunofluorescence in RGC axons between 20 and 40 μm from the rim of the explants ( total axon bundle ) . The relative thickness of individual axon bundles was also measured by comparing the total pixel counts of NF160 in the 20–40-μm area ( individual axon bundle ) . The values in the graph are averages expressed relative to those of 6X-His peptide-treated samples , presented as 1; error bars denote SDs ( **p < 0 . 001; ANOVA ) . The scores on top of the graph columns are the number of axons ( individual axon bundle ) and the number of explants ( total axon bundle ) analyzed , respectively . Results were obtained from three independent experiments . The number of explants analyzed: for 6X-His , n = 10; Vax1-His , n = 11; Vax1 ( R152S ) -His , n = 6 . ( already shown in total axon bundle ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 00810 . 7554/eLife . 02671 . 009Figure 3—figure supplement 1 . Penetration of exogenous Vax1 protein into RGC axons . Retinal explants isolated from E13 . 5 mouse embryos were cultured for 24 hr before recording phase contrast and fluorescence images every 15 min on a Zeiss Axio Observer Z1 inverted microscope . After 180 min of recording , 6X-His-FITC peptide ( 100 ng/ml ) or Vax1-His-FITC protein ( 500 ng/ml ) was added to the growth medium . The images were then combined into video clips , provided in Videos 1 and 2 . After recording for 16 hr , the explants were washed with PBS , and Vax1 protein present at the surface of axons was detected by incubating with rabbit α-Vax1 ( green ) in growth medium for 20 min . The explants were then washed three times with PBS and fixed in 4% PFS/PBS for 1 hr prior to the detection of 6X-His peptide or Vax1-His protein inside and on the surface of axons using mouse anti-His antibody ( red ) . The explants were then further incubated with Alexa 488-labeled α-rabbit IgG and Cy3-labeled α-mouse IgG , and the distribution of Vax1 on the surface and within the intracellular space of RGC axons was analyzed by confocal microscopy . Phase contrast images in the left column show representative snapshot images from Videos 1 and 2 . Immunostained images in the right columns correspond to dotted boxes in the phase contrast images in the left column . Arrows indicate Vax1-His proteins internalized to RGC axons . Scale bars: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 00910 . 7554/eLife . 02671 . 010Figure 3—figure supplement 2 . Region non-selective stimulation of retinal axonal growth by recombinant Vax1 . Dark field images of retinal quadrant explants were taken before ( 0 hr ) and after ( 24 hr ) treating them with 6X-His peptide ( 25 ng/ml; white column ) or Vax1-His protein ( 100 ng/ml; black column ) . The changes in axonal length during the 24-hr incubation period were shown in a graph . The values in the graph are averages and error bars denote SDs . The scores in the graph columns are the numbers of axons analyzed . Numbers of explants analyzed are shown on the top of the columns . p-values were determined by Student t-test ( **p < 0 . 001 ) . The results were obtained from two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 01010 . 7554/eLife . 02671 . 011Figure 3—figure supplement 3 . Recombinant Vax2 is capable for inducing RGC axon growth in vitro . ( A ) E13 . 5 WT retinal explants were incubated in the presence and absence of Vax2-His ( 100 ng/ml ) for 24 hr . The explants were then fixed for immunostaining with rabbit anti-Vax2 antibody ( green ) and mouse anti-NF160 antibody ( red ) . DAPI ( blue ) , nuclear counter staining . Immunostaining images in the bottom panel are magnified versions of dotted boxed area in the top panel . Arrow indicates Vax2-His proteins internalized to RGC axons . Scale bars are 500 μm ( top ) and 100 μm ( bottom ) , respectively . ( B ) Dark field images of the explants were taken before ( 0 hr ) and after ( 24 hr ) incubation period . The changes in axonal length during the 24-hr incubation period were measured and compared with untreated control samples . The values in the graph are averages and error bars denote SDs . The scores on top of the graph columns are the number of axons analyzed . The number of explants analyzed are: 6X-His , n = 5; Vax1 , n = 4 . p-values were determined by Student t test ( **p < 0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 01110 . 7554/eLife . 02671 . 012Video 1 . Time-lapse video of E13 . 5 mouse retinal explants cultured in the presence of 6X-His-FITC peptides ( 100 ng/ml ) as described in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 01210 . 7554/eLife . 02671 . 013Video 2 . Time-lapse video of E13 . 5 mouse retinal explants cultured in the presence of Vax1-His-FITC proteins ( 500 ng/ml ) as described in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 01310 . 7554/eLife . 02671 . 028Video 3 . Time-lapse video of E13 . 5 mouse retinal explants cultured in the presence of 6X-His-FITC peptides ( 500 ng/ml ) as described in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 028 We next sought evidence for the transfer of Vax1 to RGC axons in vivo . As reported previously ( Hallonet et al . , 1998; Bertuzzi et al . , 1999 ) , Vax1 mRNA is expressed in RGC axon-associated structures , including the OD , the OS , and the vHT , but not the retina , of E14 . 5 mice ( Figure 4A , Figure 4—figure supplement 1A ) . However , an examination of Vax1 protein distribution in E14 . 5 Vax1lacZ/+ heterozygous mouse embryos showed that Vax1 was detectable in the retina as well as the OS and OD ( Figure 4B , top row ) . In these mice , β-galactosidase ( β-gal ) is expressed from a lacZ gene replacing one Vax1 gene locus while Vax1 is expressed from the other intact Vax1 gene locus ( Hallonet et al . , 1999 ) ; therefore , β-gal should be expressed in cells expressing Vax1 . However , we found that RGCs in Vax1lacZ/+ mice did not express β-gal but did express Vax1 , this contrasts with OS APCs which co-expressed Vax1 and β-gal ( Figure 4B , top row ) . Vax1-immunostaining signals were completely absent in RGCs as well as β-gal-positive OS APCs from homozygous lacZ knock-in ( Vax1lacZ/lacZ ) Vax1lacZ/lacZ littermates , suggesting that Vax1 immunostaining signals in the Vax1lacZ/+ mouse RGC were specific ( Figure 4B , bottom row ) . Collectively , these data demonstrate that Vax1 protein in RGCs might originate from the neighboring Vax1/β-gal double-positive APCs in the OS or NPCs and radial glia in the vHT ( Figure 4B , Figure 4—figure supplement 1B ) . 10 . 7554/eLife . 02671 . 014Figure 4 . Mouse RGC axons import Vax1 protein . ( A ) Vax1 mRNA expression in E14 . 5 WT ( Vax1+/+ ) mouse retinas was examined by in situ RNA hybridization using a [33P]-CTP-labeled antisense Vax1 probe , as described elsewhere ( Mui et al . , 2005 ) . Vax1 transcripts were detected in the OS and OD , but not in the neural retina ( NR ) . This in situ hybridization signal was absent in Vax1lacZ/lacZ homozygous knock-in mouse eyes ( bottom ) . ( B ) The distribution of Vax1 protein in the NR ( i and iii ) and OS ( ii and iv ) of E14 . 5 Vax1lacZ/+ and Vax1lacZ/lacZ embryos was compared with that of β-gal expressed from the lacZ gene at the Vax1 locus by co-staining with rabbit anti-Vax1 ( green ) and mouse anti-β-gal ( red ) antibodies . Vax1 protein detected in Vax1lacZ/+ mouse retinal cells , where β-gal signals were absent , is presumed to originate from external sources that co-express Vax1 and β-gal . Red dots in ( iii ) are non-specific background β-gal immunostaining . ( C ) Distribution of Vax1 in RGC axons and cell bodies was examined by co-immunostaining for Vax1 ( green ) and the RGC axonal marker NF160 ( red ) . Nuclei were counterstained with DAPI ( blue ) . Arrowheads in ( ii ) indicate Vax1 protein that co-localizes with NF160 , whereas arrows point to Vax1 in APC nuclei . Vax1 immunostaining signals were completely absent in the OS and NR of Vax1lacZ/lacZ mice , whereas NF160 immunostaining was still detectable in defasciculated RGC axons ( iii and iv ) . ( D ) Sections of E18 . 5 WT and Vax1lacZ/lacZ mouse retinas ( top ) and optic nerves ( ON; bottom ) were immunostained with rabbit anti-Vax1 antibody and gold ( 25 nm ) -labeled anti-rabbit IgG . Subcellular localization of Vax1 proteins was then examined by electron microscopy . Arrowheads in RGC images point to Vax1 proteins in the intracellular vesicle , whereas arrows in the images indicate Vax1 proteins bound to the extracellular surface of the RGC plasma membrane ( top ) . Arrowheads in APC images indicate Vax1 proteins in trafficking vesicles , whereas arrows mark Vax1 proteins associated with chromatin in the nucleus ( bottom ) . Scale bars in ( A ) to ( C ) : 200 μm ( left column ) and 20 μm ( right two columns ) . Scale bars in ( D ) : 0 . 5 μm ( left column ) and 0 . 2 μm ( right two columns ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 01410 . 7554/eLife . 02671 . 015Figure 4—figure supplement 1 . Expression of Vax1 mRNA and protein in the vHT area . ( A ) Vax1 mRNA was detected in the HT , hypothalamic cell cord ( HCC ) , and optic nerve ( ON ) approaching the OC of E14 . 5 WT ( Vax1+/+ ) embryonic brains ( top left ) . Vax1 proteins , assessed by immunostaining with an anti-Vax1 antibody , were also detectable in these structures ( top right ) . However , Vax1 immunostaining signals as well as Vax1 mRNA were absent in these structures in the brains of Vax1−/− littermates ( bottom ) . ( B ) Vax1 localization in E14 . 5 Vax1lacZ/+ mouse vHT cells , which also express β-gal from one Vax1 locus , was examined by co-immunostaining for Vax1 and β-gal . All cells that expressed β-gal also expressed Vax1 ( arrowhead ) , but some Vax1-positive cells lacked β-gal ( arrow ) . Vax1 was not detected in the brains of Vax1lacZ/lacZ homozygous knock-in mice , which express β-gal from both Vax1 loci , confirming the specificity of Vax1 immunostaining . Scale bars: 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 01510 . 7554/eLife . 02671 . 016Figure 4—figure supplement 2 . Cytoplasmic localization of Vax1 in RGCs . Sections of E18 . 5 WT ( top ) or Vax1lacZ/lacZ mouse retinas ( bottom ) were immunostained with rabbit α-Vax1 and gold ( 25 nm ) -labeled anti-rabbit IgG and analyzed by electron microscopy , as described in Figure 4D . Arrowheads indicate Vax1 proteins in trafficking vesicles , whereas arrows mark soluble Vax1 in the cytoplasm ( top ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 016 Vax1 protein in OS APCs was present mainly in nuclei , whereas a majority of Vax1 protein in β-gal-negative RGCs was non-nuclear ( Figure 4B , i and ii ) . Furthermore , Vax1 co-localized with NF160 in E14 . 5 WT mouse RGC axons but was lost in Vax1lacZ/lacZ mouse RGC axons ( Figure 4C ) . These Vax1 localization patterns in the OS APCs and RGCs were further confirmed by immuno-transmission electron microscopy . Vax1 protein was highly enriched at the extracellular part of the RGC membrane and was also detected in the cytoplasm and intracellular vesicles ( Figure 4D , top row; Figure 4—figure supplement 2 ) . Vax1 protein in OS APCs was not only enriched in nuclei but was also detectable in cytoplasmic membrane structures ( Figure 4D , bottom row ) . These results therefore suggest that OS- and/or vHT-secreted Vax1 might enter RGCs after docking with the RGC axon membrane . To investigate whether the secreted Vax1 is necessary for the growth of RGC axons toward vHT explants , we sequestered extracellular Vax1 using α-Vax1 . α-Vax1 not only interfered with the transfer of Vax1 from vHT cells to RGC axons , it also antagonized RGC axonal growth toward vHT explants ( Figure 5A , center ) . In contrast , neither rb-IgG ( pre-immune rabbit IgG ) nor α-Vax2 influenced Vax1 transfer into RGC axons or RGC axonal growth toward vHT explants ( Figure 5A , left and right ) . α-Vax1 treatment not only reduced the population of retinal axons growing toward the vHT ( Figure 5B ) , it also decreased the total number of retinal axons growing from the explants ( Figure 5C ) , suggesting an axogenic activity as well as an axon growth-stimulating activity of extracellular Vax1 . 10 . 7554/eLife . 02671 . 017Figure 5 . Secreted Vax1 protein is necessary for RGC axon growth . ( A ) vHTs and retinas isolated from E13 . 5 mouse embryos were co-cultured for 48 hr in the presence of pre-immune rabbit IgG ( rb-IgG; 1 μg/ml ) , anti-Vax1 ( α-Vax1; 1 μg/ml ) , or anti-Vax2 ( α-Vax2; 1 μg/ml ) antibodies . Vax1 localization in RGC axons was then determined by staining explants with rabbit α-Vax1 ( green ) and mouse α-NF160 ( red ) . Arrowheads indicate the area magnified in each inset . Scale bars: 500 μm . ( B ) The distribution of RGC axons in each angle segment was determined as described in ‘Materials and methods’ . The values in the bar are averages , and error bars denote SDs . p-values are between 0 . 01 and 0 . 005 ( ANOVA ) . ( C ) Total image pixel counts of NF160 immunofluorescence in a 20–40-μm area were compared to obtain the relative number of axons projected from each explant . Scores under y-axis labels of ( B ) and ( C ) are the numbers of explants analyzed in three independent experiments ( **p < 0 . 001; ANOVA ) . ( D ) Slabs of mouse heads including eyes , forebrain , and midbrain structures were prepared from E13 . 5 WT mouse embryos . The third brain ventricles of mouse-head slabs were then implanted with collagen gels containing rb-IgG ( 1 μg/ml ) or α-Vax1 ( 1 μg/ml ) and subsequently incubated at 37°C in a CO2 incubator for 12 hr ( top row; see ‘Materials and methods’ for details ) . The slabs were then fixed and frozen to obtain horizontal sections ( 18 μm thick ) . The slides containing optic nerves ( ON ) were then further co-stained with α-Vax1 ( green ) and α-NF160 ( red ) and analyzed using an Olympus FV1000 confocal microscope . Images in the bottom row are magnifications of dotted boxes in the top row . Scale bars: 200 μm . Relative fluorescence intensities of Vax1- and/or NF160-positive immunostaining intensities in the midline area ( dotted box ) were measured using ImageJ software and presented graphically . White column , rb-IgG; black column , α-Vax1 . The values are relative intensities compared with rb-IgG-treated samples; error bars denote SDs and values on the top of graph columns are number of slabs analyzed ( *p < 0 . 01; Student t test ) . A , anterior; P , posterior; M , medial; L , lateral; * , optic chiasm; V3 , third ventricle . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 017 The roles of extracellular Vax1 in RGC axon growth were also investigated in vivo . Collagen gels releasing rb-IgG or α-Vax1 were implanted in the third ventricle of E13 . 5 mouse embryonic brain slabs to sequester extracellular Vax1 in the vHT area ( Figure 5D , diagram on top panel ) . Mouse embryos implanted with α-Vax1-releasing gels showed a remarkable reduction in RGC axons accessing the midline compared with embryos implanted with rb-IgG-releasing gels ( Figure 5D ) . In α-Vax1-implanted mouse embryos , a significant number of RGC axons showed reduced Vax1 immunoreactivity and stopped at the lateral wall of the ventral diencephalon , properties similar to those observed in Vax1−/− mice ( Bertuzzi et al . , 1999 ) ( Figure 5D , bottom row ) . Taken together , these results suggest that extracellular Vax1 is necessary for RGC axonal growth to the ventral midline . Intercellular transfer has also been reported for other homeodomain transcription factors , such as engrailed-2 ( En2 ) and orthodenticle homeobox 2 ( Otx2 ) ( Joliot et al . , 1998; Sugiyama et al . , 2008; Spatazza et al . , 2013 ) . However , little is known about the regulatory mechanisms underlying the trafficking of homeodomain transcription factors . We therefore searched for the gene encoding proteins capable of modifying the intercellular transfer of Vax1 in Drosophila ( Figure 6; screening results are unpublished ) . One of the genes isolated in this screen encodes the transmembrane heparan sulfate proteoglycan ( HSPG ) protein , syndecan ( Sdc ) ( Spring et al . , 1994 ) . HSPGs , including Sdc2 , Sdc3 , and glypican 1 ( Glp1 ) , are highly expressed in mouse RGC axons and have been proposed to play critical roles in RGC axon guidance in various vertebrates ( Chung et al . , 2001; Inatani et al . , 2003; Lee and Chien , 2004; Pratt et al . , 2006 ) . We thus focused on the potential role of HSPGs as receptors for Vax1 in RGC axons . 10 . 7554/eLife . 02671 . 018Figure 6 . Regulation of intercellular Vax1 transfer by HSPGs in Drosophila wing imaginal discs . Vax1-EGFP ( green ) and DsRed ( red ) were co-expressed under the control of Ptc-Gal4 in the A/P boundary cells of Drosophila wing imaginal discs in wt ( top row ) and Sdc mutant ( Sdc23; second row ) flies . The number of cells positive for Vax1-EGFP but negative for DsRed at the posterior part of the wing disc was increased in Sdc23 flies . In contrast , the number of Vax1-EGFP-positive cells in the posterior wing disc was significantly decreased in fly embryos that co-expressed Sdc ( third row ) or Dlp ( bottom row ) together with Vax1-EGFP and DsRed . Diagrams in the right column show the distribution of Vax1-EGFP-positive cells ( green ) and DsRed ( red ) in the corresponding fly wing discs . Scale bars: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 018 Vax1-GFP protein was co-expressed with DsRed protein in the A/P ( anterior/posterior ) boundary cells of Drosophila wing imaginal disc under the control of a Ptc-Gal4 driver . Vax1-EGFP , but not DsRed , was transferred to neighboring cells , results similar to those observed in the mammalian systems ( Figure 6 , top row ) . However , the transfer of Vax1-EGFP to neighboring wing disc cells was suppressed upon co-expression of Sdc in the A/P boundary cells ( Figure 6 , third row ) . Overexpressed dally-like protein ( Dlp ) , a Drosophila homolog of Glp , also suppressed Vax1 transfer in the same manner as Sdc ( Figure 6 , bottom row ) , suggesting that the intercellular transfer of Vax1 in Drosophila wing imaginal disc cells is mediated by HSPGs but not specifically by Sdc . These results also imply that these overexpressed HSPGs in the A/P boundary cells captured co-expressed Vax1-EGFP protein , thereby interfering with the transfer of Vax1-EGFP to neighboring cells ( Figure 6 , diagram in the right column ) . Conversely , Vax1-EGFP proteins were transferred farther in the Sdc mutant ( Sdc23 ) wing imaginal disc , where total HSPG levels were reduced owing to the loss of Sdc ( Figure 6 , second row ) . HSPG-regulated Vax1 transfer was further investigated in the mammalian systems . We found that Vax1 , but not Vax2 , interacted with Sdc2 in E14 . 5 mouse optic nerves as well as with overexpressed Sdc1 and Sdc2 in human embryonic kidney ( HEK ) 293T cells ( Figure 7A , Figure 7—figure supplement 1A , B ) . Sdc2-N , lacking the C-terminal cytoplasmic domain , was able to interact with Vax1 , whereas Sdc2-C , which lacks the N-terminal extracellular domain , failed to interact with Vax1 ( Figure 7—figure supplement 1C ) , suggesting that Vax1 binds to the extracellular domain of Sdc . 10 . 7554/eLife . 02671 . 019Figure 7 . Vax1 binding to HSPGs is necessary for RGC axonal growth . ( A ) Interaction between Vax1 and Sdc2 in the E14 . 5 mouse optic nerve was investigated by immunoprecipitation with a rabbit anti-Vax1 ( top ) or goat anti-Sdc2 ( bottom ) antibody and subsequent Western blotting with reciprocal antibodies . The specificity of anti-Vax1 and anti-Sdc2 antibodies was confirmed by immunoprecipitation with pre-immune rabbit IgG ( rb-IgG ) and goat IgG ( gt-IgG ) , respectively . ( B ) Immunoprecipitation of Vax1 in the E14 . 5 mouse OS was also performed in the presence or absence of heparin ( 1 mg/ml ) to determine whether the Vax1 protein bound to HS sugar groups of Sdc2 , Sdc3 , and/or Glp1 HSPGs expressed in RGC axons . ( C ) Vax1-His protein ( final concentration , 2 μg/ml ) was incubated at 4°C for 1 hr with Sepharose 4B resin ( Sigma , St . Louis , MO , USA ) coated with HS or CS . The resins were washed three times with PBS , and Vax1 protein bound to the resins was eluted in SDS sample buffer for subsequent SDS-PAGE on 10% gels and Western blotting with α-Vax1 . Relative intensities of Vax1 bands in Western blot images were analyzed using ImageJ software . ( D ) Retinal explants were treated with heparinase I ( 2 . 5 U/ml ) or ChnaseABC ( 2 . 5 U/ml ) for 3 hr and then incubated with 6X-His peptide ( 25 ng/ml ) or Vax1-His recombinant protein ( 100 ng/ml ) for an additional 24 hr . The presence of Vax1-His in RGC axons was then examined by co-immunostaining with rabbit anti-Vax1 ( green ) , mouse anti-NF160 ( red ) , and goat anti-Sdc3 ( blue ) antibodies . Dotted boxes indicate the area magnified at right . Scale bars: 500 μm . ( E ) The graph shows relative distances that RGC axons grew during the 24-hr incubation period . The values in the graph are averages of fold ratios compared with those of 6X-His-treated samples , error bars denote SDs , and the scores on top of graph columns are the number of axons analyzed ( *p < 0 . 01 , **p < 0 . 001; ANOVA ) . Results were obtained from three independent experiments . Numbers of explants analyzed: 6X-His , n = 6; Vax1 , n = 7; heparinase , n = 5; chondroitinase , n = 5; Vax1+heparinase I , n = 6; Vax1+ChnaseABC , n = 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 01910 . 7554/eLife . 02671 . 020Figure 7—figure supplement 1 . Molecular determination of the interaction between Vax1 and Sdc . Vax1 ( A ) or Vax2 ( B ) was co-expressed with GFP-tagged Sdc1 and Sdc2 in HEK293T cells . GFP-fused Sdc1 and Sdc2 protein complexes were isolated by immunoprecipitation ( IP ) , and Vax1 and Vax2 proteins in GFP-Sdc1 and GFP-Sdc2 immune complexes were detected by Western blotting using α-Vax1 ( A ) or α-Vax2 ( B ) . ( C ) The Vax1-interacting domains of Sdc2 were examined by co-expressing Vax1 with Sdc2-N , which lacks the cytoplasmic domain , or with Sdc2-C , which lacks the extracellular domain . GFP-Sdc-N or GFP-Sdc-C protein that co-immunoprecipitated with Vax1 was detected by Western blotting with α-GFP . Relative levels of Vax1 , Vax2 , and GFP-Sdc2 proteins in cell lysates ( CL ) used for IP were also examined by Western blotting ( bottom two panels in A–C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 02010 . 7554/eLife . 02671 . 021Figure 7—figure supplement 2 . Co-localization of Vax1 and Sdc3 in RGC axons . ( A ) Distribution of Vax1 and Sdc3 in E14 . 5 mouse embryonic eyes was examined by co-immunostaining with rabbit anti-Vax1 ( green ) and goat anti-Sdc3 ( red ) antibodies . Image in the third is a merged magnification of area marked by dotted boxes in the left two images . Plots in the right column indicate fluorescence intensities of areas marked by dotted lines in the third . Arrowheads on plots indicate fluorescence intensities of the corresponding points in the images . N , nasal; T , temporal . Scale bars: 200 μm ( left two panel ) and 20 μm ( third panel ) . ( B ) Retinal explants isolated from dorsal-temporal ( DT ) and ventral-temporal ( VT ) parts of E13 . 5 mouse retinas were treated with Vax1-His protein ( 100 ng/ml ) for 24 hr . The explants were then fixed for immunostaining with rabbit anti-Vax1 ( green ) , mouse anti-NF160 ( red ) , and goat anti-Sdc3 ( blue ) antibodies . The localization of Vax1 and Sdc3 in NF160-positive RGC axons was examined by confocal microscopy ( see details in ‘Materials and methods’ ) . Scale bars: 500 μm ( left column ) and 50 μm ( magnified images at right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 021 The extracellular domain of Sdc is attached by heparan sulfate ( HS ) sugars ( Bishop et al . , 2007 ) ; thus , Vax1 could interact with the sugar groups as well as the protein backbone of Sdc . To determine the potential binding of Vax1 to HS side chains of Sdc2 in RGC axons , we used co-immunoprecipitation assays to test whether excess free heparin competed with HS side chains of HSPGs , including Sdc2 , Sdc3 , and Glp1 , for binding to Vax1 . Vax1 interactions with each of these HSPGs expressed in RGC axons in E14 . 5 optic nerves were disrupted in the presence of free heparin , whereas interactions with Pax2 ( paired homeobox 2 ) , which complexes with Vax1 in OS APCs , were not ( Figure 7B ) . Moreover , recombinant Vax1-His specifically bound HS-sepharose resin with high affinity ( Figure 7C ) . Collectively , these results suggest that Vax1 preferentially binds to HS side chains of HSPG proteins expressed in RGC axons . We also examined the influence of Vax1 binding to HSPGs on RGC axon growth . Sdc3 was expressed in E14 . 5 mouse RGCs in a non-polarized manner and co-localized with Vax1 in RGC axons ( Figure 7—figure supplement 2 ) . The growth stimulatory effects of Vax1 on RGC axons were abolished by the treatment of retinal explants with heparinase I , which cleaves heparin and HS sugar chains , but not by incubation with chondroitinase ABC ( ChnaseABC ) , which digests chondroitin sugar chains ( Bishop et al . , 2007; Figure 7D , E ) . Heparinase I treatment also decreased the immunostaining intensity of exogenously provided Vax1-His in RGC axons ( Figure 7D ) . Neither heparinase I nor ChnaseABC influenced RGC axon growth in the absence of Vax1 ( Figure 7D , E ) . Collectively , these results suggest that the binding of Vax1 to HSPGs is necessary for the induction of RGC axonal growth . Secreted Vax1 not only bound to HSPGs at the RGC cell surface , it also moved into the RGC axoplasm by exploiting the cell-penetrating property of its homeodomain ( Joliot and Prochiantz , 2004; Figure 4D; Figure 4—figure supplement 2 ) . It is therefore possible that Vax1 stimulates RGC axonal growth either by acting as a ligand for HSPGs or by regulating cytoplasmic events after penetration . To answer this question , we tested the function of a Vax1 ( WF/SR ) mutant , in which conserved Trp147 and Phe148 amino acids responsible for cell penetration were replaced with Ser167 and Arg148 ( Joliot et al . , 1998 ) ; this mutant lacks the ability to cross the cell membrane but remains capable of binding to Sdc2 ( Figure 8—figure supplement 1A , B ) . We found that Vax1 ( WF/SR ) barely penetrated RGC axons and induced RGC axonal growth less efficiently than WT Vax1 ( Figure 8—figure supplement 1C , D ) . These results suggest that Vax1-induced RGC axonal growth requires cell penetration . To determine which cytoplasmic events Vax1 might affect , we identified cytoplasmic Vax1-interacting proteins by MALDI-TOF ( matrix-assisted laser desorption/ionization-time of flight ) mass spectrometry ( Figure 8A ) . Interestingly , a majority of proteins isolated by Vax1-affinity purification were related to protein synthesis , including ribosomal proteins ( RPs ) L11 , L23A , L26 , S14 , and S16; translation regulators , such as eukaryotic translation initiation factor ( eIF ) 3B and 3C; and the chaperone HSPA1A ( heat shock 70-kDa protein 1A ) . These data suggest that Vax1 might act in RGC axons by modulating protein synthesis , a mechanism similar to that by which cytoplasmic En2 controls RGC axonal growth ( Brunet et al . , 2005; Yoon et al . , 2012 ) . To confirm this , we tested the effects of Vax1 on protein synthesis in RGC axons by quantifying the newly synthesized proteins , measuring the fluorescence intensity of incorporated bioorthogonal noncanonical amino acid azidohomoalanine ( AHA ) , labeled with Alexa Fluor 488 by click chemistry ( Dieterich et al . , 2007 ) . Treatment with Vax1-His induced a remarkable increase in the fluorescence intensities of AHA-Fluor 488-labeled proteins in RGC axons ( Figure 8B , middle column ) , whereas the Vax1 ( WF/SR ) -His had no effect on protein synthesis ( Figure 8B , right column ) , indicating that intracellular Vax1 stimulated protein synthesis . 10 . 7554/eLife . 02671 . 022Figure 8 . Imported Vax1 induces RGC axonal growth by stimulating local protein synthesis . ( A ) GST and GST-Vax1 protein complexes were affinity purified from cytoplasmic fractions of HEK293T cells overexpressing GST and GST-Vax1 , respectively ( see ‘Materials and methods’ for details ) . Complexes were then analyzed by SDS-PAGE on 10% gels and subsequent silver staining to detect proteins specifically enriched in GST-Vax1 complexes . The identities of protein bands , shown to the right of the gel photograph , were determined by MALDI-TOF mass spectrometry . Vax1-FL , full-length Vax1; Vax1-N* , Vax1 N-terminal fragment . ( B ) E13 . 5 mouse retinal explants were cultured for 24 hr before changing to medium containing Vax1-His ( 100 ng/ml ) or Vax1 ( WF/SR ) -His ( 100 ng/ml ) for an additional 16 hr-incubation . The explants were further incubated for 6 hr after addition of the bioorthogonal noncanonical amino acid AHA ( L-azidohomoalanine ) . Newly synthesized proteins incorporating these noncanonical amino acids were labeled with Alexa Fluor 488-alkyne by click chemistry ( Dieterich et al . , 2007 ) , and the rates of protein synthesis in RGC axons ( middle row ) and explant cell body ( bottom ) were assessed by measuring the fluorescence intensities of AHA-Alexa Fluor 488-labeled proteins ( see ‘Materials and methods’ for details ) . Scale bars: 500 μm ( top ) and 100 μm ( bottom ) . ( C ) The influence of nuclear events in Vax1-induced RGC axon growth was excluded by isolating axons from the cell body before treatment with Vax1 proteins ( 100 ng/ml ) for 6 hr . Arrowheads indicate the area magnified in each inset . Scale bars: 500 μm . ( D ) Relative AHA-Alexa488 fluorescence intensities in cell body-free axons were measured using ImageJ software and are shown graphically . Error bars denote SDs . ( E ) Relative distances that RGC axons grew during this 6-hr incubation period are presented graphically . The values in the graph are averages of fold ratios compared with those of untreated samples . Scores on top of the graph columns in ( D ) and ( E ) are number of axons analyzed ( **p < 0 . 001; ANOVA ) . Results were obtained from two independent experiments . Numbers of explants analyzed: untreated , n = 4; Vax1 , n = 5; Vax1 ( WF/SR ) , n = 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 02210 . 7554/eLife . 02671 . 023Figure 8—figure supplement 1 . Biophysical properties of Vax1 ( WF/SR ) mutant protein . ( A ) Trp and Phe ( WF ) at amino acid residues 147 and 148 of mouse Vax1 , which are homologous to the critical residues of the cell-penetrating region of the Antp ( antennapedia ) homeodomain ( Joliot et al . , 1998 ) , were mutated to Ser–Arg ( SR ) in Vax1 ( WF/SR ) . Vax1 and Vax1 ( WF/SR ) interactions with GFP-Sdc2 were assessed by immunoprecipitating cell lysates with α-Vax1 and subsequent Western blotting with α-GFP . Successful expression of Vax1 and GFP-Sdc2 was also assessed by Western blotting of cell lysates . ( B ) COS7 cells were incubated for 3 hr with growth medium ( S3 fraction ) from HEK293T cells overexpressing Vax1-V5 or Vax1 ( WF/SR ) -V5 . Vax1 protein at the intact cell surface was detected with rabbit α-Vax1 ( green ) , whereas Vax1 protein inside cells and at the cell surface of fixed cells was labeled with mouse α-V5 ( red ) . After incubating with Alexa 488-conjugated α-mouse IgG and Cy3-conjugated α-rabbit IgG , the distribution of cell surface and intracellular Vax1 was analyzed by confocal microscopy . Scale bars: 20 μm . ( C ) The importance of cell penetration of Vax1 for RGC axonal growth was investigated by co-culturing mouse retinal explants ( NR ) with COS7 cells overexpressing WT Vax1 or Vax1 ( WF/SR ) mutant for 48 hr . The explants were then stained with rabbit α-Vax1 ( green ) and mouse α-NF160 ( red ) . Scale bars: 500 μm ( top ) and 200 μm ( bottom ) . ( D ) Relative axon content in each angle segment was analyzed as described in ‘Materials and methods’ and is shown graphically . The values in the graph are averages , and error bars denote standard deviations ( SD ) . Results were obtained from three independent experiments . The scores under y-axis labels are the number of explants analyzed . p-values are between 0 . 001 and 0 . 005 ( ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 023 In contrast to its significant induction of protein synthesis in axons , Vax1-His-induced effects on AHA-Fluor 488 fluorescence intensities in retinal cell bodies were not notably different from those of Vax1 ( WF/SR ) -His or the 6X-His ( Figure 8B , bottom row ) . Similarly , Vax1 failed to stimulate general translation in the cultured cell-lines and purified polysomes in vitro ( data not shown ) . In contrast , isolated retinal axons from the cell body still responded to Vax1 as efficiently as those that projected from intact retinal explants ( Figure 8C , D ) . Taken together , these results suggest that Vax1 stimulates RGC axon growth by stimulating local translation of specific mRNA in the axon rather than by modulating gene expression in the cell body . We further investigated whether extracellular Vax1 could restore RGC axon growth towards the midline in Vax1−/− mice by re-supplying Vax1 protein to the vHT extracellular space . In contrast to the lack of RGC axon access to the vHT observed in Vax1−/− mouse embryos , remarkable numbers of RGC axons were detectable in the vHT of Vax1−/− mouse embryos implanted with collagen gels that released recombinant Vax1-His ( Figure 9A , top rows of left two columns , B ) . Remarkable number of RGC axons were able to grow to the source of extracellular Vax1 protein ( i . e . , the third ventricle ) , although they failed to restore the OC . In contrast , implantation of collagen gels that released cell-penetration–defective Vax1 ( WF/SR ) -His did not induce the regrowth of RGC axons ( Figure 9A , [third column] , B ) . Implanted recombinant Vax1-His was detectable in RGC axons , whereas Vax1 ( WF/SR ) -His was not , suggesting that the implanted Vax1 stimulated RGC axon growth by penetrating into the axoplasm . 10 . 7554/eLife . 02671 . 024Figure 9 . RGC axons re-grow in Vax1-implanted , Vax1−/− mouse brains . ( A ) The third ventricles of E13 . 5 Vax1−/− mouse-head slabs were implanted with collagen gels mixed with 6X-His peptide ( 4 . 78 μg/ml; 5 . 69 nmol ) , Vax1-His ( 200 μg/ml; 5 . 69 nmol ) , or Vax1 ( WF/SR ) -His ( 200 μg/ml; 5 . 69 nmol ) and incubated for 12 hr ( see diagram in Figure 5D and ‘Materials and methods’ for details ) . Vax1−/− mouse-head slabs were also implanted with collagen gels mixed with Robo1-Fc fragment ( 1 μg/ml; 12 . 35 pmol; R&D Systems , Minneapolis , MN , USA ) in the presence of 6X-His peptide ( 4 . 78 μg/ml; 5 . 69 nmol ) or Vax1-His ( 200 μg/ml; 5 . 69 nmol ) and incubated for 12 hr . The fluorescence images of horizontal sections of head slabs were obtained using an Olympus FV1000 confocal microscopy equipped with a transmitted light detector ( top row ) . The same embryonic sections were further stained with rabbit anti-Vax1 ( green ) and mouse anti-NF160 ( red ) antibodies ( bottom row ) . ( B ) Fluorescence intensities of NF160 immunostains in the boxed areas in ( A ) were measured using ImageJ software and are presented graphically . The values are intensities expressed relative to rb-IgG-treated samples , and error bars denote SDs ( **p < 0 . 001; ANOVA ) . Numbers on top of the graph columns are head-slab preparations . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 02410 . 7554/eLife . 02671 . 025Figure 9—figure supplement 1 . vHT-secreted Silt inhibits RGC axon growth . ( A ) E13 . 5 WT retinal explants were co-incubated with E13 . 5 Vax1-ko vHT explants for 24 hr in the presence ( bottom ) or absence ( top ) of Robo1-Fc ( 100 ng/ml ) . Dark field images of the explants were taken before ( 0 hr; left column ) and after ( 24 hr; center column ) the incubation . The explants were then fixed for immunostaining with mouse anti-NF160 antibody ( red ) . Immunostaining images of dotted boxed areas in the center column are shown in the right column . DAPI , nuclear counter staining . Scale bars denote 500 μm . ( B ) The angular distribution of RGC axons was measured by counting axon marker NF160 immunostaining image pixels ( axon counts ) as described in ‘Materials and methods’ and presented graphically . + , forward direction angle segment; 0 , neutral direction angle segments; − , reverse direction angle segment . The values in the bar are averages , and error bars denote SDs . Numbers under y-axis labels are the number of explants analyzed . p-values determined by ANOVA test are <0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 02510 . 7554/eLife . 02671 . 026Figure 9—figure supplement 2 . Reciprocal antagonism between Vax1 and Slit2 in vitro . ( A ) E13 . 5 mouse retinal explants were cultured for 24 hr prior to treatment with Slit2-His ( 10 ng/ml; R&D Systems ) for 24 hr in the absence ( left ) of Vax1-His or in the presence of 10 ng/ml ( middle ) or 100 ng/ml ( right ) of Vax1-His ( top row ) . In reciprocal experiments , explants were treated with Vax1-His ( 10 ng/ml ) for 24 hr in the absence ( left ) of Slit2-His or the presence of 10 ng/ml ( middle ) or 100 ng/ml ( right ) Slit2-His ( bottom row ) . ( B ) The changes in RGC axon length during the last 24 hr are shown graphically . Error bars denote SD and the scores on top of the graph columns are the number of axons analyzed . Number of explants analyzed: untreated , n = 22; Vax1 ( 10 ng/ml ) , n = 10; Vax1 ( 100 ng/ml ) , n = 9; Slit2 ( 10 ng/ml ) , n = 5; Slit2 ( 100 ng/ml ) , n = 5; Vax1 ( 10 ng/ml ) + Slit2 ( 10 ng/ml ) , n = 7; Vax1 ( 10 ng/ml ) + Slit2 ( 100 ng/ml ) , n = 6; Vax1 ( 10 ng/ml ) + Slit2 ( 10 ng/ml ) , n = 13; and Vax1 ( 100 ng/ml ) + Slit2 ( 10 ng/ml ) , n = 7 . Results were obtained from four ( for the first two ) or three ( for the rest ) independent experiments . ( C ) Vax1-His and Slit2-His added to the retinal explants were detected by immunostaining with rabbit α-Vax1 ( green ) and mouse α-Slit2 ( red ) . Scale bars: 20 μm . Fluorescent intensities Vax1 and Slit2 immunostaining images were measured by ImageJ software and shown graphically . Scores on top of the columns are number of areas analyzed . Number of explants analyzed for immunostainings are same as ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 026 It has been suggested that the avoidance of RGC axons in Vax1−/− mouse vHT might result from high concentrations of Slit protein in the ventral-lateral diencephalon ( Bertuzzi et al . , 1999 ) . In support of this , the growth of RGC axons towards Vax1−/− vHT explants were partly recovered by sequestering extracellular Slit protein using an Fc-fused extracellular fragment of the Slit receptor Robo1 ( Robo1-Fc ) ( Figure 9—figure supplement 1 ) . Thus , we further investigated the roles of Slit in RGC axon avoidance in Vax1−/− mouse embryos using Robo1-Fc . The growth of RGC axons into the vHT was partially rescued by implanting a Robo1-Fc–releasing collagen gel into the third ventricle , an effect that was further enhanced by implantation of gels co-releasing Vax1-His and Robo1-Fc ( Figure 9A , two right columns , B ) .
RGC axonal projection to the OC is often compared to the spinal commissural axonal projection toward the floor plate ( FP ) . Spinal commissural axons are prevented from prematurely entering the midline by Slit1 expressed in the medial spinal cord and grow in the ventral direction ( Stein and Tessier-Lavigne , 2001 ) . In much the same way , Slit1 in the preoptic area and ventral-lateral diencephalon prevents RGC axons from accessing the brain anywhere but at the vHT to form the OC ( Erskine et al . , 2000; Plump et al . , 2002 ) . The spinal commissural axons sense attractive cues , such as netrin and Shh , secreted from the FP ( Stein and Tessier-Lavigne , 2001; Charron et al . , 2003 ) . The attractive netrin and Shh signals are expected to compete with the co-existing repulsive signal of Slit2 , which can be accumulated locally by HSPGs , including α-dystroglycan , at the ventral FP ( vFP ) , to determine the directionality of spinal commissural axon growth cones ( Matsumoto et al . , 2007; Wright et al . , 2012 ) . RGC-expressed HSPGs were also reported to play roles as co-receptors for netrin and Slit ( Johnson et al . , 2004; Hussain et al . , 2006; Piper et al . , 2006; Ogata-Iwao et al . , 2011 ) , suggesting a similar HSPG-based antagonistic regulation of RGC axon growth . However , both netrin and Shh are dispensable with respect to attracting RGC axons toward the vHT and instead function as repulsive cues for RGC axons ( Deiner and Sretavan , 1999; Sanchez-Camacho and Bovolenta , 2008 ) . In this study , we propose vHT-secreted Vax1 as a RGC axon growth factor that is analogous to vFP-secreted netrin and Shh . This unconventional axon growth factor also utilizes HSPGs for anchoring to RGC axons ( Figure 7 ) and could compete with Slit for HSPGs binding in vitro ( Figure 9—figure supplement 2 ) . However , it is unclear whether this competition is valid in physiological conditions because Slit does not inhibit RGC axon growth toward the vHT midline ( Erskine et al . , 2000; Plump et al . , 2002 ) . Moreover , cell penetration-defective Vax1 ( WF/SR ) mutant could not induce RGC axon growth in vivo as well as in vitro , despite of its capability of HSPG binding ( Figures 8 and 9 , Figure 8—figure supplement 1 ) . It suggests that HSPG binding of Vax1 is not sufficient to induce RGC axon growth but local protein synthesis induced by internalized Vax1 is necessary . Collectively , we propose a hypothetical model that Vax1 promotes RGC axon growth towards the vHT midline by directly targeting mRNA in the axons rather than by serving for a conventional axon guidance molecule that binds specific receptors and triggers on intracellular signaling cascades ( Figure 10 ) . Identification of axonal target mRNA awaits future investigations . 10 . 7554/eLife . 02671 . 027Figure 10 . Model depicting Vax1 functions as a secreted retinal axon growth factor . Vax1 is expressed in radial glia and NPCs of the vHT as well as the OS APCs ( A ) and secreted to the extracellular space ( B ) . RGC axons that grow in the OS capture APC-secreted Vax1 by HSPGs , resulting in an increase in Vax1 concentration at RGC axons for subsequent penetration and local activation of translation in the axon ( B ) . The imported Vax1 in RGC axoplasm not only promotes axonal growth towards the vHT , but it also enhances fasciculation of RGC axons . This axon growth stimulatory activity of Vax1 supports sustained RGC axon growth to the vHT midline after the axons were avoided from progressing to dorsal diencephalon , which expresses high concentration of Slit repulsive axon guidance cue ( C , top ) . Therefore , RGC axons stop at the lateral wall of Vax1-ko mouse diencephalon and fail to access the midline ( C , bottom ) . At the vHT midline , RGC axon guidance cues , including VEGF164 , NrCAM , and ephrinB2 , determine the directionalities of RGC axon growth cones by acting their specific receptors ( D ) . Vax1 does not likely determine the directionalities of RGC axon growth cone at the midline but does promote the growth of the RGC axon shaft as well as the growth cone regardless of their original positions in the retina . DOI: http://dx . doi . org/10 . 7554/eLife . 02671 . 027 Translation-dependent , but transcription-independent , RGC axon guidance by a secreted transcription factor has also been reported for En2 , which regulates RGC growth cone turning by increasing the expression of mitochondrial proteins involved in elevating the local ATP , which potentiates ephrin A5 signaling ( Brunet et al . , 2005; Stettler et al . , 2012; Yoon et al . , 2012 ) . Since Vax1 and En2 share a homologous homeodomain , Vax1 could function in a similar manner by cooperating with attractive RGC axon guidance cues , such as VEGF164 and NrCAM ( Williams et al . , 2006; Erskine et al . , 2011; Kuwajima et al . , 2012 ) , and by modulating mitochondrial activity . Conversely , En2 could also use HSPGs to bind target axons and fine-tune their selective growth . In their use of HSPGs as docking sites for RGC axons , Vax1 can be compared to Otx2 that binds specifically to CSPGs of the perineuronal net surrounding parvalbumin ( PV ) neurons in the visual cortex ( Beurdeley et al . , 2012; Miyata et al . , 2012 ) . The differential affinities of Vax1 and Otx2 for HS and CS might be related to their different homeodomains . Among homeodomain proteins proven to exhibit transfer , Vax1 possesses an antennapedia class homeodomain homologous to that of Emx2 and En2 , whereas Otx2 shares a paired class homeodomain similar to that of Pax6 ( paired box 6 ) ( Bürglin , 2011; Spatazza et al . , 2013 ) . One intriguing possibility that has not yet been explored is that these secreted homeodomain proteins share the property of preferential binding to HS and CS; however , it is at least as likely that intercellular transfer of homeodomain proteins are target-selective events . Vax2 was not as effectively secreted from COS7 cells as Vax1 ( Figure 2 ) , despite sharing an identical homeodomain with Vax1 . This suggests that secretion of Vax1 is not solely mediated by the homeodomain but is also dependent on a three-dimensional structure that supports the secretion property of the homeodomain . Unlike Vax1 , Vax2 undergoes a specific phosphorylation that results in its cytoplasmic retention ( Kim and Lemke , 2006 ) . Phosphorylation of En2 inhibits its secretion ( Maizel et al . , 2002 ) , suggesting that the phosphorylation might change the structure of these proteins in such a way that it interferes with homeodomain recognition by secretion regulators . However , phosphorylation-defective Vax1 ( S170A ) was still unable to be secreted from COS7 cells ( data not shown ) . Instead , Vax2 could be secreted from different types of cells , where Vax2 might form three-dimensional structures that can be recognizable by secretion machineries ( Lee et al . , unpublished data ) . The results suggest that the secretion of homeodomain proteins is a cell context-dependent event . Multiple midline-crossing defects , including agenesis of the corpus callosum , anterior commissure , and hippocampal commissure , are observed in Vax1-deficient mice and homozygous VAX1 mutant human patients ( Bertuzzi et al . , 1999; Slavotinek et al . , 2012 ) . However , the molecular functions of Vax1 in the development of these structures remain unknown . In this study , we show that Vax1-induced RGC axonal growth is independent of its transcription factor activity ( Figures 2A and 3C ) . Instead , Vax1 acts as a regulator of translation after penetrating into RGC axons ( Figure 8 ) . Although we cannot rule out Vax1 functions as a transcription factor in the development of those commissures , our results suggest a potential role of secreted Vax1 in the growth of cortical axons ( Min et al . , unpublished data ) . Whether axonal growth of these commissural axons in the mammalian forebrain also requires local protein synthesis triggered by Vax1 protein secreted from cells located in other midline structures , such as the ventral-medial telencephalon and the septum , is a question that warrants further investigation .
Vax1 knock-out ( Vax1−/− ) and Vax1 knock-in ( Vax1lacZ/lacZ ) mice were reported previously ( Bertuzzi et al . , 1999; Hallonet et al . , 1999 ) . Retinal and vHT explants were cultured as described previously ( Sato et al . , 1994 ) . Briefly , retinal or vHT explants were added to a collagen mixture and positioned on plates coated with poly-L-lysine ( 10 μg/ml ) and laminin ( 10 μg/ml ) . The explants were then incubated at 37°C for 1 hr to allow gelling before adding Neurobasal medium containing B27 supplement ( Invitrogen , Carlsbad , CA , USA ) . COS7 cell droplets ( 105 cells/droplet ) were also prepared using the same procedures . The explants were cultured alone or co-cultured with vHT or COS7 explants for 48 hr before treating with proteins and antibodies . For time-lapse recording of retinal axon growth , mouse retinal explants were treated with FITC-labeled 6X-His peptide ( 100 ng/ml , 500 ng/ml ) or Vax1-His protein ( 500 ng/ml ) for 24 hr and photographed every 15 min using a Zeiss Axio Observer Z1 inverted microscope . The explants were washed twice with phosphate-buffered saline ( PBS ) prior to incubation with rabbit anti-Vax1 polyclonal antibody ( α-Vax1; 1:100 ) for 30 min to detect Vax1-His located on the cell surface . The explants were then washed with PBS and fixed in 4% paraformaldehyde ( PFA ) /PBS for subsequent immunostaining procedures to detect total Vax1-His protein inside and outside of cells . For slab embryo culture with collagen gel implantation , E13 . 5 mouse embryos were moved onto culture slide chambers containing collagen mixture , positioning the dorsal part on top , after dissecting out the dorsal half of the brain and lower part of the mouth . Droplets of collagen solution mixed with pre-immune IgG ( 1 μg/ml ) , α-Vax1 ( 1 μg/ml ) , 6X-His peptide ( 4 . 78 μg/ml ) , or Vax1-His ( 200 μg/ml ) protein were then delivered into the third ventricle of the slab embryos . The embryos were then filled with culture medium and incubated for 12 hr at 37°C in a humidified atmosphere supplemented with 7% CO2 . The embryos were fixed in 4% PFA/PBS for subsequent freezing in OCT ( optimal cutting temperature ) medium . RGC axon growth at the vHT was monitored along the horizontal axis of slide-mounted embryonic brain sections under an Olympus BX-71 microscope . The slides were then stained with appropriate antibodies and examined under an Olympus FV1000 confocal microscope to detect the penetration of implanted Vax1 protein into RGC axons . Relative axon counts ( combined values of numbers and lengths of axons ) of retinal explants were obtained by measuring NF160-fluorescent pixels in images of retinal axons using the ImageJ software . Relative axon counts of retinal explants co-cultured with the vHT explants or COS7 cell aggregates were obtained along three angle segments: forward ( + ) , neutral ( 0 ) , and reverse ( − ) . A clockwise angle from a line connecting two centers of explants was obtained and classified as forward direction ( + ) if it was between 0° and 60° or between 301° and 360°; neutral direction ( 0 ) if it was between 61° and 120° or between 241° and 300°; and reverse direction ( − ) if it was between 121° and 240° ( Figures 2B and 5B ) . Relative axon counts in each angle segment were then obtained by comparing the pixel counts of NF160 immunofluorescence in RGC axons of the explants in each angle segment . α-Vax1 and α-Vax2 were produced as reported previously ( Mui et al . , 2005 ) . Commercially available antibodies against the following proteins were used: mouse anti-Myc ( Santa Cruz Biotechnology , Dallas , TX , USA ) , mouse anti-GFP ( Santa Cruz Biotechnology , Dallas , TX , USA ) , mouse anti-tubulin β-III ( Tuj1; Covance , Princeton , NJ , USA ) , goat anti-Sox2 ( Santa Cruz Biotechnology , Dallas , TX , USA ) , mouse anti-Nestin ( RC2; Millipore , Billerica , MA , USA ) , mouse anti-β-galactosidase ( Developmental Studies Hybridoma Bank , DSHB ) , mouse anti-NF160 ( Developmental Studies Hybridoma Bank [DSHB] , Iowa City , IA , USA ) , goat anti-Sdc2 ( Santa Cruz Biotechnology , Dallas , TX , USA ) , goat anti-Sdc3 ( Santa Cruz Biotechnology ( Dallas , TX , USA ) , for immunohistochemistry ) , rabbit anti-Sdc3 ( Abcam ( UK ) , for Western blot ) , rabbit anti-Glp1 ( Santa Cruz Biotechnology , Dallas , TX , USA ) , and rabbit anti-Pax2 ( Invitrogen , Carlsbad , CA , USA ) antibodies . The heads of embryonic mice were fixed in 4% PFA/PBS at 4°C for 2–16 hr , depending on the protein to be detected , and then incubated in a 20% sucrose/PBS solution at 4°C for 16 hr before embedding in OCT medium for freezing . Sections of frozen tissue were incubated for 1 hr in a blocking solution containing 0 . 2% Triton X-100 , 5% normal donkey serum , and 2% bovine serum albumen ( BSA ) in PBS . Sections were first incubated with the indicated primary antibodies in blocking solution without 0 . 2% Triton X-100 at 4°C for 16 hr and then with the appropriate Alexa488- , Cy3- , or Cy5-conjugated secondary antibody . Immunofluorescence was subsequently analyzed using Olympus FV1000 and Zeiss LSM710 confocal microscopes . Growth medium or CSF from the lateral ventricle of E14 . 5 mice was centrifuged twice at 500×g for 10 min and then twice at 2000×g for 15 min to obtain supernatant ( S3 ) fractions . The S3 fractions were then mixed with an equal volume of 3 M trichloroacetic acid ( TCA ) solution to precipitate the macromolecules . The TCA precipitates were washed twice with 100% acetone , air-dried pellets , and dissolved in 2×-sodium dodecyl sulfate ( SDS ) sample buffer for SDS-PAGE ( polyacrylamide gel electrophoresis ) analysis . Ptc-Gal4 , UAS-DsRed , UAS-Sdc , UAS-Dlp , and sdc23 flies were obtained from the Bloomington stock center , IN , USA . The UAS-Vax1-EGFP fly was generated by injection of pUAS-Vax1-EGFP constructed by cloning mouse Vax1 cDNA into the pUAST-EGFP vector . Third-instar larvae of Ptc-Gal4>UAS-Vax1-EGFP , UAS-DsRed;+ were obtained from a cross of the Ptc-Gal4>UAS-DsRed;TM6B fly with a UAS-Vax1-EGFP fly . After fixing larval wing imaginal discs with 4% PFA/PBS for 30 min , the Ptc-Gal4-induced green fluorescence signals from EGFP and Vax1-EGFP proteins were compared with red fluorescence signals from DsRed proteins by confocal microscopy ( Olympus FV1000 ) . HEK293T cells and E14 . 5 mouse optic nerves were lysed in a buffer consisting of 10 mM Tris–HCl ( pH 7 . 4 ) , 200 mM NaCl , 1% Triton X-100 , and 1% NP-40 . Cell lysates were centrifuged at 12000×g for 10 min at 4°C . The supernatants were collected and incubated with the indicated antibodies at 4°C for 16 hr; then , protein A-agarose beads were added and incubation was continued at 4°C for 1 hr . After washing the immune complexes five times with lysis buffer , proteins were eluted with 2× SDS sample buffer . Samples were then analyzed by SDS-PAGE and Western blotting . HEK293T cells transfected with pEBG or pEBG-Vax1 were lysed with a buffer consisting of 10 mM Tris–HCl ( pH 7 . 4 ) , 200 mM NaCl , and 1% NP-40 . Cell lysates were centrifuged at 12000×g for 10 min at 4°C . The supernatants were collected and incubated with glutathione Sepharose 4B resin ( GE Healthcare ) at 4°C for 1 hr . After washing five times with lysis buffer , proteins were eluted with 2× SDS sample buffer and samples were analyzed by SDS-PAGE on 10% gels . Gels were stained using Silver Stain Kit for Mass Spectrometry ( Pierce ) to isolate bands for MALDI-TOF mass spectrometry analysis at the Korea Basic Science Institute ( KBSI ) , Daejeon , South Korea . Explant culture media were replaced with methionine-free media , 30 min prior to the addition of 50 μM L-azidohomoalanine ( AHA , Invitrogen ( Carlsbad , CA , USA ) ) . After 6 hr , retinal explants were washed twice with PBS containing 1% fetal bovine serum ( FBS ) and then 30 μM DIBO-Alexa Fluor 488 ( Invitrogen , Carlsbad , CA , USA ) in PBS containing 1% FBS was added . The explants were then incubated at room temperature in the dark for 1 hr . After washing four times in PBS containing 1% FBS , retinal explants were fixed with 4% PFA in PBS for 15 min at room temperature for subsequent detection of the fluorescence of proteins incorporating AHA-Alexa Fluor 488 . | We see the world around us when light bounces off of objects and hits the retina at the back of our eyes . This triggers electrical signals in neurons called retinal ganglion cells ( RGCs ) , which have long structures called axons that extend out from the retina and into the parts of the brain where the signals are interpreted . As the axons grow , various ‘guidance’ molecules direct the axons to the correct part of the brain . One molecule that is important for the growth of retinal ganglion cells' axons is a protein called Vax1 . This protein is a transcription factor and binds to DNA to control how and when the molecular templates used to make proteins are made—a process called transcription . Vax1 is not produced in retinal ganglion cells , but it does control the extension of these cells' axons into part of the brain called the ventral hypothalamus . In this study , the axons cross to the other side of the brain by forming a structure called optic chiasm . Humans and mice lacking Vax1 are unable to develop the optic chiasm , and the axons of their retinal ganglion cells do not reach their targets in the brain . These defects were thought to occur because the guidance molecules whose transcription is normally controlled by Vax1 were not produced in the correct amounts when Vax1 is absent . Kim et al . now challenge this view by creating a mutant version of Vax1 that cannot bind to DNA or regulate the transcription of other proteins . Retinal ganglion cell axons could still grow correctly when they were put close to cells expressing this version of the Vax1 protein . This contradicts a hypothesis that Vax1 supports axonal growth by transcribing guidance molecules . Kim et al . followed up these results by examining developing mice and reached the unexpected conclusion that Vax1 is secreted from cells in the ventral hypothalamus and binds to a type of sugar molecule found on the surface of the axons . Once bound , Vax1 can enter the axons where it appears to stimulate the production of proteins inside axons , which helps the axons to grow . These findings reveal unconventional functions for Vax1 that occur in addition to its role as a transcription factor . Vax1 is known to regulate the development of several structures in the brain , so the work of Kim et al . also raises new questions about how Vax1 controls these processes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2014 | Regulation of retinal axon growth by secreted Vax1 homeodomain protein |
Inflammation-induced release of prostaglandin E2 ( PGE2 ) changes breathing patterns and the response to CO2 levels . This may have fatal consequences in newborn babies and result in sudden infant death . To elucidate the underlying mechanisms , we present a novel breathing brainstem organotypic culture that generates rhythmic neural network and motor activity for 3 weeks . We show that increased CO2 elicits a gap junction-dependent release of PGE2 . This alters neural network activity in the preBötzinger rhythm-generating complex and in the chemosensitive brainstem respiratory regions , thereby increasing sigh frequency and the depth of inspiration . We used mice lacking eicosanoid prostanoid 3 receptors ( EP3R ) , breathing brainstem organotypic slices and optogenetic inhibition of EP3R+/+ cells to demonstrate that the EP3R is important for the ventilatory response to hypercapnia . Our study identifies a novel pathway linking the inflammatory and respiratory systems , with implications for inspiration and sighs throughout life , and the ability to autoresuscitate when breathing fails .
Breathing is essential for life , but the underlying mechanisms that control breathing movements and neuronal pattern generation are under debate ( Jasinski et al . , 2013 ) . Breathing maintains tissue homeostasis , and an adequate response to increased carbon dioxide ( CO2 ) levels is crucial ( Kaila and Ransom , 1998; Guyenet and Bayliss , 2015 ) . Failure to adequately respond to pCO2 alterations is linked to breathing disturbances; apnea of prematurity; centrally mediated sickness , such as noxious sensations and panic; and premature death , as in sudden infant death syndrome ( Guyenet and Bayliss , 2015 ) . Neuronal networks in the parafacial respiratory group/retrotrapezoid nucleus ( pFRG/RTN ) and the preBötzinger complex ( preBötC ) are important networks implicated in the central control of breathing . pFRG/RTN paired-like homeobox 2b ( Phox2b ) -expressing neurons are sensitive to changes in CO2 levels or their proxy , pH ( [H+] ) ( Mellen and Thoby-Brisson , 2012; Onimaru and Dutschmann , 2012 ) . This responsiveness to hypercapnia is independent of synaptic transmission , and the Phox2b+ neurons detect CO2/H+ via intrinsic proton receptors ( TASK-2 and GPR4 ) in parallel pathways ( Kumar et al . , 2015 ) . Moreover , medullary astrocytes contribute to central chemosensitivity . Slight acidification leads to an increased astrocytic intracellular concentration of calcium ions ( Ca2+ ) , resulting in vesicle-independent ATP release ( Gourine et al . , 2010 ) . In addition , a CO2 sensitivity of astrocytes also mediates a vesicular-independent ATP release ( Huckstepp and Dale , 2011 ) . Some connexins , which are expressed on astrocytes , e . g . , connexin 26 ( Cx26 ) and Cx30 , are indeed sensitive to CO2 ( Meigh et al . , 2013; Reyes et al . , 2014 ) . These cellular processes of chemosensitivity result in an altered respiratory pattern that lowers the blood CO2 levels . Inflammation reduces the CO2 response and , particularly in neonatal mammals , can induce sighs , an altered response to hypoxia and potentially life-threatening apnea episodes as shown in humans , sheep , piglets and rodents ( Guerra et al . , 1988; Long , 1988; Herlenius , 2011; Siljehav et al . , 2014; Koch et al . , 2015; Siljehav et al . , 2015 ) . In the inflammatory pathway , prostaglandin E2 ( PGE2 ) is an important molecular mediator , that together with its main receptor , the EP3R , play roles in the hypoxic and hypercapnic responses , e . g . seen in patients with bronchopulmonary dysplasia ( Kovesi et al . , 2006; Siljehav et al . , 2014; Koch et al . , 2015 ) . PGE2 also seems to induce a sigh oriented respiratory pattern ( Koch et al . , 2015 ) . Sighs are regularly occurring events of augmented breaths with a biphasic inspiratory pattern with the initial phase being comparable to eupnea and the second having larger amplitude ( Toporikova et al . , 2015 ) . Such breaths are necessary for life and have been linked to several pathological states ( Ramirez , 2014; Li et al . , 2016 ) . Here , we hypothesized that both PGE2 and EP3R constitute parts of the respiratory machinery and that they are involved in the induction of sighs and the hypercapnic response . We established a viable brainstem organotypic slice culture that maintains respiratory-related activity for several weeks in vitro and used this to investigate how PGE2 and EP3R alter breathing and control of chemosensitivity . Our novel data reveal an important role of the EP3R in the pFRG/RTN hypercapnic response and furthermore suggest that PGE2 is released during hypercapnia , possibly through CO2-sensitive connexin hemichannels . Inflammation , with its associated PGE2 release , exogenous PGE2 and a lack of EP3R , blunts the hypercapnic response . These data link the inflammatory and respiratory systems , with implications for sighs and inspiration throughout life as well as for the ability to autoresuscitate when breathing fails .
To investigate the role of PGE2 and EP3R in respiration and sigh activity , we performed whole body plethysmography on 9-day old mice . We found EP3R and its ligand PGE2 to be important modulators of breathing and the response to hypercapnia ( 5% CO2 in normoxia; Table 1 ) . The sigh frequency increased after the intracerebroventricular ( i . c . v . ) injection of PGE2 ( 1 µM in 2–4 µl artificial cerebrospinal fluid , aCSF ) in an EP3R-dependent manner ( Figure 1c–d , Table 2 ) , as did the tidal volume ( VT ) ( during eupnea , excluding sighs ) in wild-type mice ( Figure 1e ) . Furthermore , hypercapnic exposure also induced an increase in sigh frequency ( Figure 1f , Table 2 ) . This increase was larger in wild-type mice than in mice lacking the EP3R ( Ptger3-/- mice ) . This CO2-induced increase in sigh frequency was abolished in wild-type mice after i . c . v . injection of PGE2 ( Figure 1f , Table 2 ) . The mice also responded to hypercapnia with increases in respiratory frequency ( FR ) , VT and minute ventilation ( VE; Figure 1g ) . I . c . v . injection of PGE2 abolished the VT but not the FR response during hypercapnia ( Table 1 ) . This provides new information on how PGE2 induces sigh activity and how increased PGE2 levels , as during inflammation , may both induce sighs and attenuate responsiveness to CO2 . 10 . 7554/eLife . 14170 . 003Table 1 . Respiratory parameters under basal conditions . Ptger3-/- mice are heavier than wild-type mice of the same age . They do not , however , differ in respiratory frequency ( FR ) , tidal volume ( VT ) , or minute ventilation ( VE ) . I . c . v . injection of PGE2 increases VT and VE in wild-type mice but not Ptger3-/- mice . Respiratory frequency , tidal volume , and minute ventilation all increased during hypercapnic exposure . n: number of animals . Data are presented as mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 003Weight ( g ) FR ( breaths/min ) VT ( µl /g ) VE ( µl/g/min ) FR ( breaths/min ) HypercapniaVT ( µl /g ) HypercapniaVE ( µl /g ) HypercapniaWT - vehicle n=53 . 7 ± 0 , 5#206 ± 28*9 . 7 ± 2 . 9* †2 . 0 ± 0 . 8*†259 ± 20*#12 . 3 ± 3 . 1*#3 . 2 ± 1 . 0*#WT - PGE2 n=83 . 9 ± 0 . 4210 ± 15*15 . 1 ± 3 . 3†3 . 2 ± 0 . 8*†267 ± 31*#15 . 9 ± 2 . 64 . 2 ± 0 . 8*#Ptger3-/- - vehicle n=54 . 8 ± 0 . 4#215 ± 32*14 . 2 ± 2 . 4*3 . 0 ± 0 . 5*240 ± 37*#15 . 7 ± 3 . 1*#3 . 4 ± 1 . 1*#Ptger3-/- - PGE2 n=74 . 4 ± 0 . 3211 ± 18*13 . 8 ± 2 . 7*2 . 9 ± 0 . 6*241 ± 30*#14 . 9 ± 2 . 7*#3 . 5 ± 0 . 4*#*p<0 . 05 ( normocapnia vs . hypercapnia ) , #p<0 . 05 ( WT vs . Ptger3-/- ) , †p<0 . 05 ( vehicle vs . PGE2 ) . 10 . 7554/eLife . 14170 . 004Figure 1 . PGE2 and CO2 increase sigh activity via EP3R signaling . Respiratory activity was recorded in vivo in a two-chamber plethysmograph ( a ) . Sighs , defined by an increase in inspiratory volume and respiratory cycle period with a biphasic inspiration ( b ) , increase in frequency after intracerebroventricular injection ( i . c . v . ) of PGE2 . This effect is absent in mice lacking EP3R ( Ptger3-/- , c , arrows , d ) . I . c . v . injection of PGE2 also increases the tidal volume ( VT ) in wild-type C57BL/6J ( WT ) mice ( e ) . The sigh frequency is increased by hypercapnic ( 5% CO2 in normoxia ) conditions in wild-type and Ptger3-/- mice but less so in Ptger3-/- mice ( f ) . In wild-type mice , the increase is abolished after i . c . v . injection of PGE2 ( f ) . Hypercapnic exposure causes an increase in respiratory frequency ( FR ) , tidal volume ( VT ) , and minute ventilation ( VE ) ( g ) , but the increase is attenuated in Ptger3-/- mice . Data are presented as means ± SD . *p<0 . 05 Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 00410 . 7554/eLife . 14170 . 005Figure 1—source data 1 . In vivo plethysmography data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 00510 . 7554/eLife . 14170 . 006Table 2 . PGE2 and hypercapnia induce sighs . Sigh frequency does not differ between wild-type mice and Ptger3-/- mice . In wild-type mice , PGE2 increases sigh frequency . Hypercapnia also increases sigh frequency more in wild-type mice than in Ptger3-/- mice . PGE2 abolishes this increase in wild-type mice but not in Ptger3-/- mice ( *p<0 . 05 ) . n: number of animals . Data are presented as mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 006Sighs/min NormocapniaSighs/min HypercapniaWT - vehicle n=50 . 4 ± 0 . 11 . 0 ± 0 . 2*WT - PGE2 n=80 . 7 ± 0 . 1*0 . 9 ± 0 . 3Ptger3-/-- vehicle n=50 . 4 ± 0 . 10 . 7 ± 0 . 3*Ptger3-/—- PGE2 n=70 . 4 ± 0 . 10 . 6 ± 0 . 2* To unravel the mechanistic details of the PGE2-EP3R system in respiratory regulation and its connection to the hypercapnic response and sighs , we set out to create a model system that would allow long-term , detailed studies of the respiratory neural networks , i . e . , networks with neurons as well as glial cells . Brainstem organotypic slice cultures of the mouse brainstem from 3-day-old mice were prepared at the preBötC brainstem level ( Figure 2a ) . To validate this new model system , we first examined survival and expression of various neural markers in the brainstem slice cultures during cultivation . Neural marker staining showed intact neurons , and neurokinin 1 receptor ( NK1R ) -positive respiratory regions were cytoarchitectonically well preserved ( Figure 2b , e , g , Figure 2—figure supplement 1 ) . The expression pattern of vesicular glutamate transporter 2 ( VGlut2 ) , similar to that in vivo , indicates the functional potential of the brainstem slice culture because glutamatergic synapses are essential for the development of the breathing rhythm generator ( Wallén-Mackenzie et al . , 2006 ) ( Figure 2d ) . Neuronal markers MAP2 and KCC2 ( Kaila et al . , 2014 ) were expressed in the preBötC ( Figure 2c–f , Figure 2—figure supplement 2 ) . The protein expression in the preBötC remained stable for 3 weeks of cultivation ( Figure 2—figure supplement 1 ) . The brainstem slice cultures became thinner with longer cultivation as the tissue spread out ( Figure 2—figure supplement 2 ) . However , they remained viable and exhibited a low degree of necrosis and apoptosis , even after 3 weeks ( Figure 2—figure supplement 3 ) . 10 . 7554/eLife . 14170 . 007Figure 2 . Brainstem slice cultures have a preserved structure and neurons with functional potential . Brainstem slices containing the preBötC were used to create slice cultures . Anatomical landmarks , including the nucleus ambiguus ( NA ) , nucleus tractus solitarius ( NTS ) , and nucleus hypoglossus ( XII; a ) , as well as the distinct expression of NK1R ( b , c , g ) enabled the identification of the preBötC region . The brainstem slice displayed MAP2-/Tuj1-positive neurons expressing NK1R ( b , c ) , VGlut2 ( d ) , and/or KCC2 ( e ) . The abundant MAP2-/Tuj1-positive cells demonstrated a preserved neuronal network within the preBötC ( g ) . KCC2 expression was found in the NTS , NA , and preBötC ( e ) . DIV; days in vitro . Arrowheads: double-labeled cells . Scale bars: 100 µm in b–f , 500 µm in g . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 00710 . 7554/eLife . 14170 . 008Figure 2—figure supplement 1 . Protein expression pattern is preserved during cultivation . The expression pattern of neuronal the markers NK1R , MAP2 , Tuj1 and KCC2 and the astrocyte marker GFAP did not change during cultivation for 3 weeks . DIV: days in vitro . Scale bars: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 00810 . 7554/eLife . 14170 . 009Figure 2—figure supplement 2 . Slices flatten during cultivation . The gross morphology of the slices changed slightly during cultivation due to thinning and spreading . DIV: days in vitro . Scale bar: 500 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 00910 . 7554/eLife . 14170 . 010Figure 2—figure supplement 3 . Brainstem slice cultures are viable . An individual single necrotic cells ( 8 ± 3% , n=257 , propidium iodide-stained ) were found in the brainstem slice cultures ( N=12 ) that had been cultivated for 3 weeks , but no large clusters of necrotic cells were detected ( a , d ) . The few necrotic cells were observed in the thickest regions , indicating that diffusion-based oxygenation is critically dependent on slice thickness . Oxygen glucose deprivation ( OGD ) for 1 hr produced clear positive PI-staining throughout the slice ( b; N=5 ) . Cultures also showed low apoptotic activity ( 2 ± 1% , n=187 ) , as evaluated by caspase-3 staining ( c , d; N=20 ) . DIV: days in vitro . N: slices , n: cells . Scale bars: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 010 After evaluating morphology , we investigated the cellular activity within the brainstem slice culture . Neurons in the brainstem slice cultures retained their electrical properties at 7 days in vitro ( DIV ) , including a resting membrane potential of −55 ± 6 mV ( Figure 3b–c ) and overshooting action potentials ( Figure 3c ) . The resting membrane potential , action potential threshold , half-width and peak amplitudes of the action potential , and membrane time constant were within the ranges of acute respiratory slices ( Figure 3c , Figure 3—figure supplement 1 ) . Action potentials occurred in clusters of regular rhythmic bursting activity . Neuronal connections were also similar to those seen immediately ex vivo , e . g . , in acute slices , ( Ballanyi and Ruangkittisakul , 2009 ) as evidenced by the postsynaptic potentials and concurrent inputs to neighboring neurons , resulting in correlated activity ( Figure 3b , Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 14170 . 011Figure 3 . Neuronal electrical activity indicates preserved networks . Neurons in a preBötC slice ( 7 DIV ) , patched in the whole-cell configuration in current-clamp mode ( a ) , exhibit regular rhythmic bursting activity ( b ) . The neurons exhibited a hyperpolarized resting potential , action potentials , synaptic input , and spontaneous electrical activity , with epochs of action potential activity ( b , c ) . The different measured variables indicated healthy and normally functioning neurons ( d ) . Depicted here are two simultaneously patched neurons that also received common synaptic input ( e , arrows ) . Spiking epochs occurred simultaneously , suggesting synchronized network oscillations . Direct connectivity between the depicted neurons showed that they were neither chemically nor electrically synaptically connected to each other . This finding indicates that the observed correlation was induced by common input from a preserved network structure . AP: action potential . DIV: days in vitro . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 01110 . 7554/eLife . 14170 . 012Figure 3—source data 1 . Electrophysiology patch clamp data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 01210 . 7554/eLife . 14170 . 013Figure 3—figure supplement 1 . Cells of brainstem slice cultures retain neuronal electrical properties . Recorded neurons showed correlated epochs of spontaneous activity ( a , b ) and received synaptic input that was often synchronized ( a , arrow ) , suggesting common presynaptic neurons . Note the similarity in neuronal properties of the recorded neighboring neurons ( a , b ) . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 013 Thus , on an individual neuronal level , the cells behave as expected . However , breathing is generated through cellular interactions in respiration-related neural networks . To investigate how individual cells interact , we applied live time-lapse Ca2+ imaging to allow simultaneous recording of the activity of hundreds of cells . Tetramethyl rhodamine coupled Substance P ( TMR-SP ) , visualizing NK1R-expressing neurons , was used to identify the preBötC . In the brainstem slice cultures , the preBötC contained networks with correlated activity between cells ( Figure 4b–d ) , which was analyzed using a recently reported cross-correlation analysis method ( Smedler et al . , 2014 ) ( Figure 4—figure supplement 1 ) . We found clusters of cells with highly correlated activity . Such groups of cells in close proximity to each other were interconnected via a few cells that seem to function as hubs ( Watts and Strogatz , 1998 ) . The correlated network activity in the preBötC was preserved for 1 , 2 and 3 weeks ( Figure 4b–e ) . The number of active cells and the correlations per active cell remained similar over time ( Figure 4e ) . These data suggest that the brainstem slice culture approach can indeed be used to perform long-term studies of respiratory neural network activity . 10 . 7554/eLife . 14170 . 014Figure 4 . Neural activity in the preBötC is arranged in a functional respiratory network with respiratory-related motor output . In the preBötC slice ( a ) , a cross-correlation analysis of Ca2+ time-lapse imaging data ( Figure 4—figure supplement 1 ) revealed small-world network-structured correlated activity in the preBötC ( b–d ) . The number of correlating cell pairs did not change over time ( e ) , nor did the small-world network parameter or connectivity ( f ) . TMR-SP-positive regions contained more correlated cell pairs than TMR-SP-negative regions ( 621 ± 284 , N=14 and 56 ± 48 , N=9 , respectively; p<0 . 05 ) , although there was no difference in the number of active cells ( 112 ± 57 , N=14 and 144 ± 68 , N=9 , respectively , N . S . ; g ) . As in the preBötC , the nucleus hypoglossus maintained correlated neural network activity ( h ) . Ten percent of the cells ( n=8–12/slice ) in the hypoglossal nucleus exhibited a regular spiking frequency of ~50–100 mHz ( i ) . The multicolored bar indicates the correlation coefficient in b–h; warmer colors indicate more strongly correlated activity between two cells connected by the line . DIV: days in vitro . A . U . : arbitrary units . w: week . N: number of slices , n: number of cells . Scale bars: 500 µm in a , 100 µm in b–d and g–h . Multicolored bar: color-coded correlation coefficient values . Data are presented as means ± SD . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 01410 . 7554/eLife . 14170 . 015Figure 4—source data 1 . Correlation data preBötC . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 01510 . 7554/eLife . 14170 . 016Figure 4—source data 2 . Frequency data with DAMGO . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 01610 . 7554/eLife . 14170 . 017Figure 4—figure supplement 1 . Single cell events provide information about correlated activity . Ca2+ signals are measured in over 200 regions of interest in a single experiment . Events , here simplified as peak maxima with a minimum 20% increase above baseline , over time ( lines ) are identified . The locations of such events , or rather variations in the Ca2+ signal , in both time and space are used to calculate the correlation coefficient between the cells ( Smedler et al . , 2014 ) . Those correlation coefficients are then drawn as lines between their corresponding cells , providing a graphical image of the network structure . DIV: days in vitro . Scale bar: 100 µm . Multicolored bar: color-coded correlation coefficient values . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 01710 . 7554/eLife . 14170 . 018Figure 4—figure supplement 2 . Spontaneous Ca2+ activity is preserved for 3 weeks . NK1R-expressing neurons exhibit rhythmic Ca2+ activity after 1 , 2 , and 3 weeks of cultivation , even during treatment with TTX , which inhibits synapse signaling . There were no significant differences in the average frequency or regularity among slice cultures of different ages . n=840 at 7 DIV , n=621 at 14 DIV and n=456 at 21 DIV . DIV: days in vitro . n: number of cells . *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 018 Analysis of the network structure revealed stable connectivity values ( i . e . , the number of cell pairs with a correlation coefficient exceeding the cut-off value , divided by the total number of cell pairs ) during the cultivation of preBötC slices for up to 3 weeks ( Figure 4f , Table 3 ) . These values were slightly higher than those estimated in a previous study ( Hartelt et al . , 2008 ) , in which only neurons were accounted for . However , both neurons and glia are involved in respiratory control ( Erlichman et al . , 2010; Giaume et al . , 2010 ) , and our analysis provides information on both cell types . Moreover , other analyzed network parameters , i . e . , the normalized mean path-length ( λ ) and the normalized mean clustering-coefficient ( σ ) , also remained stable ( Figure 4f , Table 3 ) . Overall , the small-world parameter ( Watts and Strogatz , 1998 ) γ=σλ was unchanged after 3 weeks in culture . Inhibiting the firing of action potentials and consequent activation of synapses by tetrodotoxin ( TTX , 20 nM ) abolished the coordinated network activity and revealed a population of cells that retained rhythmic alterations of cytosolic Ca2+ levels ( 31 ± 4% of the total number of cells , N=14 slices ) . Most of these cells ( 76 ± 12% , N=14 ) were NK1R-positive neurons , indicating the presence of functioning pacemaker neurons ( Figure 4—figure supplement 2 ) . The Ca2+ signals from synapse-independent cells remained , however with a lower frequency and higher coefficient of variation ( Figure 4—figure supplement 2 ) . Regions outside the brainstem nuclei contained active cells , without intercellular coordination ( Figure 4g ) . This cellular activity ceased during TTX treatment . In conclusion , the brainstem slice cultures contain a preserved preBötC network with a small-world structure . 10 . 7554/eLife . 14170 . 019Table 3 . The preBötC network parameters remain unchanged for 21-DIV cultures . The results of correlation analysis for the preBötC are shown . N . S . : not significant . N: number of slices . Data are presented as mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 019preBötC7 DIV ( N=12 ) 14 DIV ( N=13 ) 21 DIV ( N=8 ) Correlating cell pairs560 ± 325501 ± 277517 ± 327N . S . Active cells110 ± 40100 ± 59110 ± 69N . S . Correlations per active cell6 ± 46 ± 57 ± 6N . S . Connectivity0 . 8 ± 0 . 10 . 8 ± 0 . 20 . 8 ± 0 . 2N . S . Mean shortest path length ( λ ) 0 . 7 ± 0 . 20 . 8 ± 0 . 20 . 8 ± 0 . 2N . S . Clustering coefficient ( σ ) 2 . 7 ± 1 . 52 . 8 ± 1 . 62 ± 0 . 6N . S . Small-world parameter ( γ ) 4 . 2 ± 3 . 03 . 4 ± 1 . 72 . 7 ± 1 . 7N . S . As the preBötC delivers part of its motor output through the hypoglossal nerve ( Smith et al . , 2009 ) , we also examined the hypoglossal motor nucleus . In this region of the hypoglossal motor nucleus , we found correlated cell activity organized similarly to that found in the preBötC network ( Figure 4h ) . Within this network , frequency analysis revealed regularly spiking cells with a frequency between 50 and 100 mHz , corresponding to a rhythmic motor neuron output of 3–6 bursts of respiration-related activity/min ( average 3 . 7 ± 0 . 9 bursts/min; Figure 4i ) . This suggests a preserved respiratory-related output in the brainstem slice cultures . Subsequent recordings of extracellular potentials from the 12th cranial nerve and hypoglossal nucleus revealed a corresponding rhythmic respiratory-related output at 7 ( N=16 ) , 14 ( N=3 ) , and 21 DIV ( N=6 ) . Respiratory output from acute slices varied between 1 and 8 bursts per min ( neonatal mice , 3 mM K+ ) , with frequencies in the lower range after a longer incubation time in vitro ( Ramirez et al . , 1997; Ruangkittisakul et al . , 2011 ) . In our model we observed a respiratory-related frequency of 3 . 7 ± 2 . 5 bursts per min ( average of frequencies at 7 , 14 and 21 DIV , no significant difference was observed between different DIV , Figure 5a ) , which is within the expected range for a slice . Among individual cultures , there was some variability in frequency ( Figure 5a ) . However , the intrinsic rhythm was stable , with an average coefficient of variation of 22 ± 8 ( no difference between the different DIV , Figure 5b ) . Rhythmic XII activity was observed for more than 2 hr during recordings ( Figure 5—figure supplement 1 ) . The activity could be inhibited by a µ-opioid receptor agonist , [D-Ala2 , N-Me-Phe4 , Gly5-ol]-enkephalin ( DAMGO , 0 . 5 µM; Figure 5c , Figure 5—figure supplement 1 ) and stimulated by NK1R agonist Substance P ( 1 µM; 19 ± 13% increase in frequency , p<0 . 05; N=7; Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 14170 . 020Figure 5 . Breathing brainstem in a dish: ongoing/persistent rhythmic XII motor activity . The connected preBötC neural networks generate respiratory-related motor neuronal output delivered through the 12th cranial nerve ( XII ) . The hypoglossal nucleus/nerve discharge frequency varied among the brainstem slice cultures but did not depend on brainstem slice culture age ( a , N=16 at 7 DIV , N=3 at 14 DIV , and N=6 at 21 DIV ) . The regularity of respiration-related motor activity , measured as CV ( coefficient of variation ) , remained stable during 3 weeks of culture ( b ) . The µ-opioid receptor agonist DAMGO ( 0 . 5 µM ) silenced the XII nerve activity in 5/5 brainstem slice cultures , as depicted here in ( c ) from a 7-DIV brainstem culture ( filtered trace , above , and rectified and smoothed trace , below ) . DAMGO lowered the Ca2+ . In the hypoglossal nucleus , DAMGO ( 0 . 5 μM ) lowered the frequency of regularly-spiking cells ( f , g ) . N: number of slices . Data are presented as means ± SD . *p<0 . 05 Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 02010 . 7554/eLife . 14170 . 021Figure 5—source data 1 . 12th cranial nerve electrophysiology recordings . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 02110 . 7554/eLife . 14170 . 022Figure 5—source data 2 . Frequency data with DAMGO . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 02210 . 7554/eLife . 14170 . 023Figure 5—source data 3 . High potassium frequency data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 02310 . 7554/eLife . 14170 . 024Figure 5—source data 4 . Network topology and frequency data with DAMGO . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 02410 . 7554/eLife . 14170 . 025Figure 5—figure supplement 1 . Rhythmic respiratory-related output is preserved . Here , the rhythmic respiratory-related output , recorded from the hypoglossal motor nucleus of a 3-week-old brainstem slice culture ( N=6 ) , is displayed . Panels show filtered ( above ) and rectified and smoothed ( below ) traces of the extracellular recording . Rhythmic activity was maintained for 137 min ( a ) . The frequency was regular ( b ) and increased after Substance P application ( 1 µM; N=7; c ) . Thirty minutes after Substance P application , the respiratory-related rhythm returned to the control frequency ( d ) . The activity was inhibited by DAMGO ( 0 . 5 µM; N=5 , e ) . N: number of slices . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 025 In the preBötC , DAMGO also inhibited the Ca2+ activity of individual NK1R+ neurons and lowered the network frequency significantly ( Figure 5d , Video 1 ) . This was accompanied by an increase in the coefficient of variation in this area ( 36 ± 4 vs . 47 ± 6 , N=7 slices , p<0 . 05 ) . The network structure was not affected . An increase in [K+] from 3 mM to 9 mM , with subsequent membrane potential depolarization , increased the frequency in the preBötC ( Figure 5e ) . In the hypoglossal nucleus , DAMGO caused a frequency reduction in the regularly spiking cells ( Figure 5f , g ) . Thus , the preBötC brainstem slice culture remained active and responsive and generated rhythmic respiration-related motor output activity . 10 . 7554/eLife . 14170 . 026Video 1 . NK1R+ respiratory neurons in the preBötC are identified using TMR-SP ( red dye ) , followed by Ca2+ oscillations visualized with Fluo-4 . After 25 s , the µ-opioid receptor agonist DAMGO ( 0 . 5 µM ) is added and reduces the signaling frequency of the network . fps: frames per second . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 026 Gap junction signaling plays an important role in the development of the respiratory system , the maintenance of respiratory output and likely the CO2/pH response ( Elsen et al . , 2008; Fortin and Thoby-Brisson , 2009; Gourine et al . , 2010; Huckstepp et al . , 2010a ) . Thus , we used the brainstem slice cultures to investigate the involvement of gap junctions in the neural networks and their response to CO2 . In the brainstem slice cultures , immunohistochemistry showed high Cx43 expression in neurons of the preBötC ( Figure 6a ) and lower and persistent Cx26 and Cx32 expression in the respiratory regions ( Figure 6b–d ) at 7 DIV . To assess the function of these intercellular gap junctions and hemichannels , we treated the brainstem slice cultures at 7 DIV with gap junction inhibitors carbenoxolone ( CBX ) or 18α-glycyrrhetinic acid ( 18-α-GA ) . Both inhibitors decreased the number of correlating cell pairs and active cells in the preBötC , whereas glycyrrhizic acid ( GZA ) , an analog to CBX that lacks the ability to block gap junctions , and the aCSF control did not ( Figure 6e–g , k–l ) . However , the individual activity of NK1R expressing neurons was not affected ( Figure 6h–j , m ) . These findings suggest a role for gap junctions in the maintenance of correlated network activity in the preBötC . 10 . 7554/eLife . 14170 . 027Figure 6 . Gap junctions are necessary to maintain part of the correlated respiratory network . In the respiratory regions , the gap junction proteins Cx43 ( a , N=9 ) , Cx32 ( b , c , N=8 ) , and Cx26 ( d , arrowheads; double-labeling with NK1R , N=5 ) are present . Gap junction inhibitors CBX ( e ) and 18-α-GA ( f ) reduced network synchronization in the preBötC . Notably , the Ca2+ activity of individual NK1R-positive cells was not affected ( h–j , m ) . Correlating cell pair numbers decreased to 21% ( N=8 ) and 20% ( N=6 ) of their respective controls after treatment with CBX and 18-α-GA , respectively ( k ) . Network properties were not affected by GZA , an analog to CBX that lacks the ability to block gap junctions , ( g , j–k , N=7 ) or aCSF ( N=8 ) . An initial increase in fluorescence intensity was noted after adding CBX and GZA but not after adding 18-α-GA , indicating an immediate excitatory effect of CBX and GZA ( l ) . 18-α-GA reduced the number of active cells in the network at 1 min after application ( 53% ) , but CBX did not ( 91% , N . S . ) . At the same time point , an increased number of active cells were observed with GZA treatment ( 139% ) . After 10 min , a reduction of the number of active cells was found after treatment with both 18-α-GA and CBX ( 54% and 43% ) . However , the number of active cells returned to normal after GZA application ( 89% , N . S . ; l ) . DIV: days in vitro . N: number of slices . Scale bars: 10 µm in a , c , and d , 100 µm in others . Multicolored bar: color-coded correlation coefficient values . Data are presented as means ± SD . *p<0 . 05 . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 02710 . 7554/eLife . 14170 . 028Figure 6—source data 1 . Gap junction inhibition data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 02810 . 7554/eLife . 14170 . 029Figure 6—figure supplement 1 . A gap junction-independent network is present within the preBötC . Gap junction inhibitors did not affect the general topology of the respiratory network ( a–c ) . Data are presented as means ± SD . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 029 Conversely the rhythmic activity of NK1R+ neurons does not depend on gap junctions . Moreover , gap junction inhibition did not affect the mean correlation values , connectivity , or small-world parameter of the remaining correlated cell pairs ( Figure 6—figure supplement 1 ) . This demonstrates that the cells connected in a gap junction-independent manner are organized as a small-world network . These results are in line with topological data showing that respiratory neurons are organized in small clusters in the preBötC ( Hartelt et al . , 2008 ) . Our in vivo data , as well as others’ , indicate that PGE2 and hypercapnia induce sigh activity ( Ramirez , 2014; Koch et al . , 2015 ) . We hypothesized that this is due to effects on the respiratory centers in the brainstem . We used our brainstem slice cultures of the preBötC to study the direct effects of PGE2 and hypercapnia in vitro . PGE2 levels in cerebrospinal fluid measured in experimental models and in human infants are in the pico- to nanomolar range ( Hofstetter et al . , 2007 ) . In the brainstem slice cultures at 7 DIV , the application of PGE2 ( 10 nM ) lowered the Ca2+ signaling frequency of respiratory neurons in the preBötC ( Figure 7a–b ) . PGE2 also induced longer Ca2+ transients , and the signal amplitudes increased compared to those of the controls ( Figure 7b ) . Koch and colleagues ( Koch et al . , 2015 ) suggested that the increase in sighs induced by PGE2 is mediated through persistent sodium channels ( INaP ) ( Koch et al . , 2015 ) . Indeed , in the preBötC , 10 µM Riluzole , a blocker of the persistent sodium current ( INaP ) , attenuated effect of PGE2 on Ca2+ signal amplitude and length as well as decreasing the signal frequency ( Figure 7b ) . As in previous studies ( Toporikova et al . , 2015 ) , Riluzole did not affect the Ca2+ signal compared to control periods . Riluzole is used as an INaP blocker , but may also affect other parts of neuronal signaling , such as glutamate release ( Wang et al . , 2004 ) . Therefore , we cannot completely determine whether the PGE2 effect is due to an effect on the persistent sodium current or interference with glutamate signaling , although an effect on INaP is likely ( Koch et al . , 2015 ) . 10 . 7554/eLife . 14170 . 030Figure 7 . PGE2 modulates preBötC network activity . PGE2 lowered the Ca2+ signaling frequency of the preBötC network in WT mice but not in Ptger3-/- mice ( a–b ) . The effect was attenuated but not abolished by Riluzole ( b ) . PGE2 also increased signal amplitude and length ( a–b ) , an effect that was abolished after Riluzole application ( b ) . Ptger3 is expressed in the preBötC ( c , d ) , and 20% of the EP3Rs were of the α ( Gi-protein coupled ) subtype and 77% of the γ ( Gs-protein coupled ) subtype ( e ) . Hypercapnic exposure ( pCO2 elevated from 4 . 6 to 6 . 6 kPa ) did not affect the signal frequency of the preBötC ( f–g ) . DIV: days in vitro . Scale bars: 50 µm in c and 10 µm in d . *p<0 . 05 Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 03010 . 7554/eLife . 14170 . 031Figure 7—source data 1 . PGE2 data preBötC . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 03110 . 7554/eLife . 14170 . 032Figure 7—source data 2 . Hypercapnia data preBötC . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 03210 . 7554/eLife . 14170 . 033Figure 7—source data 3 . Hypercapnia data preBötC 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 03310 . 7554/eLife . 14170 . 034Figure 7—figure supplement 1 . Hypercapnia had no effect on the preBötC . Hypercapnic exposure ( pCO2 elevated to 6 . 6 kPa ) did not affect the preBötC network structure . DIV: days in vitro . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 034 EP3Rs were present in the preBötC ( Figure 7c–d ) . qRT-PCR showed that 20% of the EP3Rs were of the α-subtype ( Figure 7e ) . EP3Rα inhibits adenylate cyclase via Gi-protein , and reduced cAMP levels inhibit FR ( Ballanyi et al . , 1997 ) . The EP3Rγ subtype , however , which couples to the GS-protein , was the most abundant ( Figure 7e ) . In vivo , hypercapnia increases sigh activity , VT , FR , and VE ( Figure 1 ) . Therefore , we exposed the preBötC brainstem slice culture to increased levels of CO2 by raising the pCO2 levels from 4 . 6 kPa to hypercapnic 6 . 6 kPa , while maintaining a constant pH of 7 . 5 in the aCSF by the addition of bicarbonate . This did not have any effect on the Ca2+ signaling frequency , the Ca2+ signaling pattern or the network structure in wild-type or Ptger3-/- mice ( Figure 7f–g , Figure 7—figure supplement 1 ) . However , the preBötC is not the main central chemosensitive region . Instead , the sensitivity to CO2 is more profound in the pFRG . Therefore , we generated organotypic slice cultures of the pFRG/RTN brainstem level . The analysis of network structure and function that we conducted on the preBötC was previously not possible to perform in the pFRG/RTN on acute transverse slices . Studies of the pFRG/RTN are particularly interesting because of its crucial role in central respiratory chemosensitivity ( Onimaru et al . , 2009 ) . We therefore created the same type of brainstem slice culture as with the preBötC slice using slices containing the pFRG/RTN instead ( Figure 8a ) . These brainstem slice cultures expressed neuronal markers as expected ( Figure 8b–d , Figure 8—figure supplement 1 ) and displayed retention of electrical properties , in a manner similar to the preBötC brainstem slice cultures ( Figure 8e–f ) . 10 . 7554/eLife . 14170 . 035Figure 8 . pFRG/RTN brainstem slice culture . pFRG/RTN slices were selected based on the location of the facial nucleus ( VII; a ) . In the brainstem slice culture , pFRG/RTN expressed the neuronal markers NK1R ( b ) , KCC2 ( c ) , Phox2b ( c ) , vGlut2 ( d ) , and MAP2 ( d ) . The pFRG/RTN neurons also retained adequate electrical properties and generated spontaneous action potentials individually or in clusters ( e–f ) . Data are presented as box plots with minimum and maximum values . DIV: days in vitro . Scale bars: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 03510 . 7554/eLife . 14170 . 036Figure 8—source data 1 . pFRG/RTN characterization . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 03610 . 7554/eLife . 14170 . 037Figure 8—figure supplement 1 . Cultivation of pFRG/RTN slices . NK1R expression was preserved in the pFRG/RTN during cultivation ( upper panel ) . Whole-cell recordings from neurons in 1-week-old pFRG/RTN brainstem slice cultures shows the existence of spontaneous synaptic input as well as excitable membrane properties ( middle panel ) . The overall morphology of the pFRG/RTN brainstem slice cultures changed slightly during cultivation , becoming thinner ( lower panel ) . DIV: days in vitro . Scale bars: 100 µm in upper panel , 500 µm in lower panel . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 037 Looking at multiple cells using time-lapse Ca2+ imaging , the activity of the pFRG/RTN was correlated in a scale-free small-world network , akin the one in the preBötC ( Figure 9b–d ) and was stable during cultivation ( Figure 9e–f ) . There was a slight difference in the number of active cells between 2 week and 3 week cultures ( Figure 7e ) . However , all network properties remained unchanged ( Figure 9f and Table 4 ) . The inhibition of neuronal spiking and synapses by TTX ( 20 nM ) disrupted the coordinated activity ( 21 ± 9% of correlated cell pairs remained , N=11 ) . However , rhythmic Ca2+ activity persisted in a subset of primarily ( 64 ± 9% , N=11 ) NK1R-positive cells ( Figure 9—figure supplement 1 ) . The pFRG/RTN cells did not exhibit any change in signaling frequency after DAMGO application ( Figure 9g , average levels from 7- , 14- , and 21-DIV cultures are displayed , as there were no significant differences among cultures of these ages ) , confirming the absence of preBötC µ-opioid-sensitive regions in these slices ( Ballanyi and Ruangkittisakul , 2009 ) . Similarly to the preBötC brainstem slice culture , the pFRG/RTN responded to higher [K+] with an increase in frequency ( Figure 9h; average levels from 7- , 14- , and 21-DIV cultures are displayed , as there were no significant differences among cultures of these ages ) . 10 . 7554/eLife . 14170 . 038Figure 9 . The pFRG/RTN respiration-related network generates correlated neural activity and responds to CO2 . The pFRG/RTN network is arranged in a small-world manner just ventral to the facial nucleus . The network structure was preserved during cultivation ( a–d ) . The number of correlating cell pairs did not change with longer culturing times , but the number of active cells was higher at 3 weeks than at 2 weeks ( 45 ± 27<76 ± 19 , p<0 , 05; e ) . The network parameters were stable during cultivation ( f ) . The pFRG/RTN network did not respond to the µ-opioid receptor agonist DAMGO ( 0 . 5 µM; n=420 , N=4; a ) , but the average network frequency increased with higher potassium concentrations ( 22 ± 5 mHz and 38 ± 7 mHz , N=12; b ) . Both the neural network and individual NK1R/TMR-SP-labeled cells responded to increases in CO2 pressure ( pCO2 elevated to 6 . 6 kPa ) , indicating that the chemosensitivity was preserved in the pFRG/RTN brainstem slice culture . Suramin , a P2 receptor antagonist , and TNP-APT , a P2X receptor antagonist , attenuated the CO2 response but did not abolish it ( g ) . DIV: days in vitro . Scale bars: 100 µm . Multicolored bar: color-coded correlation coefficient values . N: number of slices , n: number of cells . Data are presented as means ± SD . *p<0 . 05 . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 03810 . 7554/eLife . 14170 . 039Figure 9—source data 1 . Correlation data pFRG/RTN . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 03910 . 7554/eLife . 14170 . 040Figure 9—source data 2 . Hypercapnia data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 04010 . 7554/eLife . 14170 . 041Figure 9—source data 3 . High potassium frequency data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 04110 . 7554/eLife . 14170 . 042Figure 9—source data 4 . Riluzole and TTX data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 04210 . 7554/eLife . 14170 . 043Figure 9—figure supplement 1 . Spontaneous Ca2+ activity is preserved during cultivation . NK1R-expressing neurons have spontaneous Ca2+ activity after 1 , 2 , and 3 weeks of cultivation . This activity remains during synaptic inhibition via addition of TTX . Within the culture , the frequency varied between cells , but there were no significant differences in average frequency or regularity among cultures of different ages . n=315 at 7 DIV , n=429 at 14 DIV , and n=192 at 21 DIV . DIV: days in vitro . n: number of cells . *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 04310 . 7554/eLife . 14170 . 044Figure 9—figure supplement 2 . Hypercapnia reduces mean path lengths in the pFRG/RTN of wild-type mice . A decrease in mean path length during hypercapnia was seen in the wild-type pFRG/RTN ( N=18 ) . Other parameters remained unchanged . No parameters were affected in the Ptger3-/- pFRG/RTN ( N=7 ) . N: number of slices . DIV: days in vitro . Scale bars: 100 µm . Multicolored bar: color-coded correlation coefficient values . Data are presented as means ± SD . *p<0 . 05 . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 04410 . 7554/eLife . 14170 . 045Table 4 . The pFRG/RTN network parameters remain unchanged for 21-DIV cultures . The results of correlation analysis for the pFRG/RTN are shown . Among the analyzed network parameters , only the number of active cells differed at the analyzed time points , and only between 14 and 21 DIV . N . S . : not significant . N: number of slices . Data are presented as mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 045pFRG/RTN7 DIV ( N=12 ) 14 DIV ( N=11 ) 21 DIV ( N=6 ) Correlating cell pairs118 ± 6961 ± 3174 ± 42N . S . Active cells49 ± 2641 ± 21*76 ± 19**p<0 . 05Correlations per active cell3 . 1 ± 2 . 21 . 7 ± 1 . 11 . 0 ± 0 . 7N . S . Connectivity0 . 7 ± 0 . 20 . 6 ± 0 . 20 . 7 ± 0 . 2N . S . Mean shortest path length ( λ ) 0 . 7 ± 0 . 20 . 7 ± 0 . 30 . 8 ± 0 . 5N . S . Clustering coefficient ( σ ) 2 . 7 ± 1 . 63 . 0 ± 1 . 74 . 5 ± 3 . 6N . S . Small-world parameter ( γ ) 3 . 6 ± 2 . 54 . 2 ± 2 . 63 . 3 ± 1 . 6N . S Next we examined the CO2 sensitivity of the pFRG/RTN ( Onimaru et al . , 2008 ) . This resulted in increased signal frequency of the Ca2+ oscillations ( Figure 9i , Table 5 , Video 2; data from 7-DIV cultures are displayed , and no significant differences in the response among 7- , 14- , and 21-DIV cultures were observed ) and the activation of some previously dormant cells . During hypercapnic exposure , the pFRG/RTN network topology remained essentially unchanged ( Figure 9—figure supplement 2 ) . 10 . 7554/eLife . 14170 . 046Table 5 . pFRG/RTN slices respond to CO2 if the EP3R is present . The average mean frequency of all cells in the network and the average mean frequency of NK1R-positive cells during the control period or during exposure to hypercapnia are shown ( pCO2 = 55 mmHg , pH = 7 . 5 ) . N . S . : not significant . N: number of slices , n: number of cells . Data are presented as mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 046Mean frequency ( mHz ) NetworkNK1R+ cellsControlHypercapniaControlHypercapniapFRG/RTN - WT ( N=7 , n=343 ) 21 . 6 ± 3 . 254 . 1 ± 2 . 7*p<0 . 0514 . 4 ± 0 . 938 . 5 ± 5 . 5*p<0 . 05pFRG/RTN - Ptger3−/− ( N=5 , n=448 ) 25 . 0 ± 7 . 926 . 0 ± 1 . 9N . S . 11 . 4 ± 5 . 811 . 6 ± 3 . 8N . S . preBötC - WT ( N=5 , n=1737 ) 16 . 4 ± 2 . 516 . 5 ± 1 . 3N . S . 16 . 6 ± 4 . 615 . 7 ± 5 . 3N . S . preBötC - Ptger3−/− ( N=4 , n=822 ) 21 . 1 ± 8 . 617 . 3 ± 3 . 8N . S . 22 . 7 ± 5 . 917 . 8 ± 7 . 7N . S . 10 . 7554/eLife . 14170 . 047Video 2 . Ca2+ oscillations visualized with Fluo-4 in the chemosensitive region pFRG/RTN . Low network activity is increased by exposure to hypercapnia after 15 s . fps: frames per second . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 047 Response to hypercapnia involves pFRG/RTN astrocytes , which release ATP that acts on purinergic P2-receptors ( Erlichman et al . , 2010; Gourine et al . , 2010; Huckstepp et al . , 2010a ) . We sought to examine whether this kind of signaling pathway was active in the 7-DIV brainstem slice cultures , and we found that blocking purinergic receptors with Suramin or TNP-ATP application did not abolish the hypercapnic response , in agreement with previous data ( Sobrinho et al . , 2014 ) . However , both the unspecific P2 receptor and the more specific P2X receptor antagonist attenuated the CO2 response by approximately one third ( 30 ± 6%; Figure 9i ) , as observed in adult and neonatal rats ( Wenker et al . , 2012 ) and 9-day-old mice ( Gourine et al . , 2010 ) . Thus , the CO2-induced release of ATP acting on P2 receptors may contribute to the CO2 response . In conclusion , our brainstem organotypic slice culture contains an active pFRG/RTN network that retains its structural integrity over time and responds to CO2 exposure with increased activity . Gap junctions , both intercellular and hemichannels , are linked to respiratory chemosensitivity ( Huckstepp et al . , 2010a; Meigh et al . , 2013; Reyes et al . , 2014 ) . Recently , CO2 was shown to interact with the hemichannel Cx26 , inducing an open state through the formation of carbamate bridges , thus increasing the release of compounds such as ATP ( Meigh et al . , 2013 ) . Therefore , we hypothesized that gap junctions exert functions within the pFRG/RTN network . However , gap junction inhibitors did not affect signaling frequency or network topology of the pFRG/RTN ( Figure 10a , Figure 10—figure supplement 1 ) . Instead , the frequency response to hypercapnia was both inhibited and reversed by the application of the gap junction inhibitor 18-α-GA ( Figure 10b–c ) . GZA ( a structural analog of CBX without gap junction-inhibiting properties ) did not alter the CO2 response ( Figure 10b–c ) . 10 . 7554/eLife . 14170 . 048Figure 10 . Correlated pFRG/RTN network activity is not dependent on gap junctions , but hypercapnic responses are . Blocking gap junctions in the pFRG/RTN did not change the functional network structure of the respiratory center or alter its frequency ( a and c , N=7 ) . However , hypercapnic responses ( CO2↑ ) were abolished when gap junctions were inhibited by 18-α-GA ( b , top trace; c , left graph , N=7 ) . GZA ( a structural analog of CBX without gap junction-inhibiting properties ) increased the frequency , and hypercapnia increased it further ( b , middle trace; c , middle graph , N=7 ) . An initiated hypercapnic response was attenuated but not completely reversed by 18-α-GA ( b , bottom trace; c , lower graph , N=5 ) . This dynamic was not seen after application of GZA . DIV: days in vitro . Scale bars: 200 µm . N: number of slices . Multicolored bar: color-coded correlation coefficient values . Data are presented as means ± SD . *p<0 . 05 . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 04810 . 7554/eLife . 14170 . 049Figure 10—source data 1 . Hypercapnia and gap junction inhibition frequency data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 04910 . 7554/eLife . 14170 . 050Figure 10—source data 2 . Hypercapnia and gap junction inhibition network data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 05010 . 7554/eLife . 14170 . 051Figure 10—figure supplement 1 . Network structure in the pFRG/RTN is not dependent on gap junctions . Blocking gap junctions in the pFRG did not change the network parameters , correlating cell pairs , or number of active cells ( N=7 ) . Data are presented as means ± SD . N: slices . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 051 We conclude that 18-α-GA inhibits the hypercapnic response , while inhibition of purinergic signaling pathways attenuates it . Thus , we suggest that the CO2 response is not entirely explained by the connexin-mediated release of ATP . Furthermore , inflammation via PGE2 and EP3R alters the hypercapnic response in vivo and in brainstem spinal cord en bloc preparations ( Figure 1 and Siljehav and colleagues Figures 1 and 4 [Siljehav et al . , 2014] ) . Therefore , we hypothesized that hypercapnic responses involve PGE2 signaling and next analyzed the PGE2 content of the aCSF under control and hypercapnic conditions . In all examined slices ( N=12/12 , 7 DIV ) , a transient doubling of the PGE2 concentration after pCO2 elevation was evident ( Figure 11 ) . When gap junction blockers were applied , this peak was absent ( N=4/4 , 7 DIV; Figure 11 ) . This indicates a hypercapnia-induced , gap junction-mediated release of PGE2 . 10 . 7554/eLife . 14170 . 052Figure 11 . PGE2 is released during hypercapnia . The aCSF contents exhibited an increase in microenvironmental PGE2 levels during hypercapnia in 12 out of 12 slices . Here , the PGE2 concentration of a brainstem slice culture is displayed during control and hypercapnic periods ( a ) . When gap junctions were inhibited ( 18-α-GA , blue line ) , the PGE2 levels remained unaltered during hypercapnia ( N=4 ) . The average PGE2 level throughout the whole experiment was not affected by hypercapnia , but the peak value was higher during hypercapnia than under control conditions ( b ) . N: number of slices . Data are presented as means ± SD . *p<0 . 05 . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 05210 . 7554/eLife . 14170 . 053Figure 11—source data 1 . Hypercapnia PGE2 ELISA data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 05310 . 7554/eLife . 14170 . 054Figure 11—figure supplement 1 . mPGEs-1 is expressed in astrocytes in the proximity of the ventral border of the pFRG . Expression of mPGEs-1 , critical for PGE2 production , was found in GFAP-expressing astrocytes ( arrowheads ) proximal to the ventral medullary border in acute frozen brainstem tissue . This was evident in both wild-type mice and transgenic mice with GFAP-driven expression of GFP ( N=11/11 and 6/6 respectively ) . N: number of slices . Scale bars: 100 µm . * indicates the ventrolateral edge of the brainstem . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 054 Immunohistochemistry showed expression of microsomal prostaglandin E synthase 1 ( mPGEs-1 ) in GFAP positive astrocytes ( Figure 11—figure supplement 1 ) . mPGEs-1 , the main PGE2 producing enzyme , has previously been found mainly in endothelial cells of the blood brain barrier of adult rats ( Yamagata et al . , 2001 ) . Our findings suggest that astrocytes in the vicinity of the ventral brainstem border of neonates express mPGEs-1 and might therefore be candidates for modulation of breathing through CO2-induced release of PGE2 . PGE2 has a primarily inhibitory effect on respiration in neonatal mice and humans ( Hofstetter et al . , 2007 ) , which we confirmed to account for its effects on the preBötC ( Figure 7 ) . However , as hypercapnia seems to induce a release of PGE2 while stimulating breathing activity , we hypothesized that PGE2 has a direct stimulatory effect on the pFRG/RTN . Indeed , PGE2 increased the signaling frequency of pFRG/RTN neurons ( Figure 12a–b , Table 6 ) . This effect was EP3R dependent , and EP3Rs were present in the pFRG/RTN , expressed both on respiratory neurons and on astrocytes ( Figure 12c–e ) . We also observed a non-significant increase in amplitude ( 8 ± 3% and 11 ± 4% increase compared to control period , N . S . ) . Neither the PGE2 effect nor the hypercapnic response of the pFRG/RTN was affected by Riluzole ( 30 ± 5 mHz vs 25 ± 2 mHz , N . S . , N=6 , and 36 ± 2 mHz vs 35 ± 6 mHz , N . S . , N=6 ) . qRT-PCR showed abundant expression of the EP3Rγ subtype , which couples to the GS-protein ( Namba et al . , 1993 ) . This would lead to an increase in intracellular cAMP in the pFRG/RTN Ptger3-expressing cells in response to PGE2 ( Figure 12f ) . 10 . 7554/eLife . 14170 . 055Figure 12 . PGE2 alters respiratory network activity . In the pFRG/RTN , PGE2 increased the frequency of respiratory ( NK1R-expressing ) neurons . This PGE2 effect was absent in brainstem slice cultures lacking EP3R ( Ptger3-/-; a–b ) . EP3Rs were present in NK1R-expressing neurons in the pFRG/RTN ( c , arrowheads , f ) and co-localized with Phox2b ( d , arrowheads ) . EP3Rs were also found on S100B-expressing astrocytes ( e , arrowheads ) . Staining was performed on acutely fixed tissue ( c–e ) and brainstem slice cultures ( f ) . qRT-PCR showed an abundance of the EP3Rγ ( Gs-protein coupled ) in the pFRG/RTN ( N=7; f ) . N: number of slices . DIV: days in vitro . Scale bars: 100 µm . Data are presented as means ± SD . *p<0 . 05 Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 05510 . 7554/eLife . 14170 . 056Figure 12—source data 1 . PGE2 frequency data pFRG/RTN . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 05610 . 7554/eLife . 14170 . 057Table 6 . PGE2 increases the frequency of pFRG/RTN neurons and decreases the frequency of preBötC neurons . The mean frequencies of NK1R-positive cells during the control period or during exposure to 10 nM PGE2 are shown . N . S . : not significant . N: number of slices , n: number of cells . Data are presented as mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 057Mean frequency ( mHz ) ControlPGE2pFRG/RTN - WT ( N=5 , n=343 ) 13 . 7 ± 1 . 121 . 5 ± 2 . 9*p<0 . 05pFRG/RTN - Ptger3−/− ( N=4 , n=448 ) 12 . 1 ± 2 . 08 . 5 ± 2 . 9N . S . preBötC - WT ( N=7 , n=1737 ) 20 . 3 ± 2 . 28 . 7 ± 1 . 4*p<0 . 05preBötC - Ptger3−/− ( N=5 , n=822 ) 22 . 8 ± 2 . 316 . 4 ± 1 . 1N . S . To further characterize the PGE2 signaling during hypercapnia , we blocked its main receptor , EP3R . Notably , pharmacological blocking of EP receptors ( using AH6809 , 10 µM ) abolished the hypercapnic response ( Figure 13a–b , 7 DIV ) , in line with our in vivo data from Ptger3-/- mice . pFRG/RTN slices ( 7 DIV ) from Ptger3-/- mice did not respond to hypercapnia ( Figure 13c–d ) . Thus , EP3R is important for pFRG/RTN CO2 responsiveness . We next generated a lentiviral vector in which the mouse EP3R ( Ptger3 ) promoter controls the expression of the red light-activated halorhodopsin Halo57 fused to eGFP ( Figure 13e ) . After transduction , we detected eGFP expression in 90 ± 6% of Phox2b-positive neurons in the pFRG/RTN ( Figure 13—figure supplement 1 ) . Stimulation by red ( 625 nm ) light of the transduced brainstem slice cultures ( 7 DIV ) triggered hyperpolarization of Ptger3-halo57-expressing cells and immediately reduced the calcium signaling frequency of both the network and individual NK1R+ neurons ( Figure 13—figure supplement 1 ) . This finding indicates a fundamental role for Ptger3-expressing cells in the network . Additionally , the response to hypercapnia in the pFRG/RTN was abolished during the light-induced silencing of Ptger3-expressing cells . The CO2 response was also reversed by the light-induced halo57 hyperpolarization of Ptger- expressing cells ( Figure 13f–g , Table 7 ) . 10 . 7554/eLife . 14170 . 058Figure 13 . PGE2 , acting through EP3R , is crucial for the hypercapnic response . Pharmacological inhibition of EP3R by the EP receptor antagonist AH6809 inhibited the response to hypercapnia ( increased pCO2[CO2↑] ) in the pFRG/RTN ( N=6 , n=472 , N . S . ; a–b ) . The hypercapnic response was also absent in pFRG/RTN slices lacking EP3R ( Ptger3−/−; N=5 , n=348 , N . S . ; c–d ) . Layout of the lentivirus containing Halo57 ( ER2 ) and eGFP genes under the control of the EP3R promoter ( Ptger3 ) used for optogenetics ( WPRE=gene enhancing element; e ) . During optogenetic silencing of Ptger3-expressing cells , no frequency changes were observed in response to hypercapnia ( f , top trace; g , left graph ) . The hypercapnic response was also reversed by activating Ptger3-Halo57 ( f , middle and bottom trace; g , middle and right graph ) . Red line: Halo57 activation in response to 625 nm light . N: slices , n: cells . Data are presented as means ± SD . *p<0 . 05 . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 05810 . 7554/eLife . 14170 . 059Figure 13—source data 1 . EP antagonist data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 05910 . 7554/eLife . 14170 . 060Figure 13—source data 2 . Hypercapnia EP3R data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 06010 . 7554/eLife . 14170 . 061Figure 13—source data 3 . Optogenetics data . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 06110 . 7554/eLife . 14170 . 062Figure 13—figure supplement 1 . Optogenetic silencing of Ptger3-expressing cells decreases respiration-related activity . Transduced slices show the expression of EGFP in the pFRG/RTN , localized through Phox2b staining ( a , left ) . Phox2b-positive cells express EGFP after transduction ( a , right , arrowheads ) . Silencing Ptger3 cells with Halo57 stimulation decreased the frequency of the entire pFRG/RTN ( b ) . General depolarization caused by an increase in the potassium concentration to 9 mM increased the average network frequency in both the preBötC and pFRG/RTN during optogenetic inhibition ( N=6; c ) . Red line: Halo57 activation in response to 625 nm light . N: number of slices . DIV: days in vitro . Scale bars: 100 µm ( a , left ) , 10 µm ( a , right ) . Data are presented as means ± SD . *p<0 . 05 . Source data are available in a separate source data file . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 06210 . 7554/eLife . 14170 . 063Table 7 . Silencing of Ptger3-expressing cells inhibits the response to hypercapnia . Mean frequencies of the pFRG/RTN network during the control period and during exposure to hypercapnia with and without Halo57 stimulation are shown . N . S . : not significant . N: number of slices . Data are presented as mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 063N=41Mean frequency ( mHz ) ControlHypercapniaControl22 . 9 ± 9 . 0*34 . 0 ± 4 . 3*p<0 . 05Halo57 stimulation9 . 0 ± 1 . 710 . 3 ± 1 . 1N . S . Based on these findings , we suggest that the PGE2-EP3R pathway is an important mechanism in the hypercapnic response and a modulator of respiratory activity .
Here , we present two novel breathing brainstem organotypic cultures in which the respiration-related preBötC and pFRG/RTN regions maintain their functional organization , activity , and responsiveness to environmental cues . Using these cultures , we show that PGE2 is involved in the control of sigh activity and the response to hypercapnia via EP3R in the preBötC and the pFRG/RTN , respectively . These findings provide novel insights into central respiratory central pattern generation , its modulation , and the mechanisms underlying breathing disorders during the neonatal period . Due to the complexity of the respiratory mechanisms , it is difficult to create optimal in vitro model systems that represent in vivo conditions while allowing sufficient depth in detailed mechanisms and their manipulation . The majority of previous studies were performed on brainstem-spinal cord preparations ( en bloc ) ( Onimaru , 1995 ) or acute slices ( Ruangkittisakul et al . , 2006 ) . However , these preparations remain active only for hours , making it difficult to study development and long-term effects on respiratory rhythm . Organotypic slice cultures provide a bridge between cell cultures and animals in vivo ( Yamada and Cukierman , 2007 ) . Their preserved three dimensional structure allows functional circuits to be studied and manipulated over time under microenvironmental control ( Gähwiler et al . , 1997; Gogolla et al . , 2006; Yamada and Cukierman , 2007; Preynat-Seauve et al . , 2009 ) . First used with hippocampal tissue ( Gähwiler , 1988 ) , the organotypic culturing method has since expanded to research on the cerebellum ( Lu et al . , 2011 ) as well as on the brainstem auditory circuits ( Thonabulsombat et al . , 2007 ) . Recently , Phillips and colleagues ( Phillips et al . , 2016 ) presented an organotypic model system of the preBötzinger complex with respiration-related neuronal rhythm that persists for a month . Here , we characterize this new type of brainstem slice culture further , and also provide details on respiratory network structure and functional respiratory-related motor output . In addition we show that also the pFRG/RTN retains respiration-related rhythmic activity and chemosensitivity . As with all model systems , it has its limitations , e . g . , the slices lose several respiratory-related regions ( Smith et al . , 2009 ) . Nonetheless , in contrast to acute slices and the brainstem-spinal cord preparation , our new experimental model system allows long-term studies and manipulation of respiratory networks . This enables the use of different techniques and methods , and significantly reduces the number of procedures that otherwise need to be performed on live animals , as well as the total number of experimental animals . We have exploited this advantage by transfecting the brainstem slice cultures in vitro to be suitable for optogenetic techniques . Using a newly developed cross-correlation analysis algorithm ( Smedler et al . , 2014 ) , we revealed in the brainstem slice culture , a clustering of cells within the two central pattern generators , a small-world network . A small-world network is characterized by a mean clustering coefficient exceeding that in random networks , but has a mean shortest path-length as short as that in random networks ( Watts and Strogatz , 1998; Malmersjo et al . , 2013 ) . Furthermore , the presence of the connective nodes and hubs gives the network a scale-free organization . This finding is in line with a previous topological analysis based on neuronal staining in the preBötC ( Hartelt et al . , 2008 ) . The present insights into the network structure of the pFRG/RTN have not been achieved previously with other methods . Notably , scale-free and small-world networks have been suggested to have evolutionary advantages ( Barabasi and Oltvai , 2004; Malmersjo et al . , 2013 ) . Subsequently we examined how the networks and individual cells were connected . Early in development , gap junctions connect the respiration-related fetal neural networks ( Thoby-Brisson et al . , 2009 ) . During development , gap junction-mediated Ca2+-transients stimulate the proliferation of neural progenitor cells ( Malmersjo et al . , 2013 ) and form a template for chemical synapses to coordinate more mature neural networks ( Jaderstad et al . , 2010 ) . Using CBX and 18-α-GA , we demonstrated that intercellular connections still play a role in postnatal preBötC network activity . This is in line with previous findings ( Elsen et al . , 2008 ) . Notably , even though fewer cells remained active , respiratory neuron frequency and network structure were not affected . Although both CBX and 18-α-GA are commonly used as gap junction inhibitors ( Solomon et al . , 2003; Elsen et al . , 2008; Véliz et al . , 2008; Jaderstad et al . , 2010 ) , these drugs have side effects ( Rekling et al . , 2000; Schnell et al . , 2012 ) . We used GZA as a control substance because it is structurally similar to CBX but does not have any gap junction inhibiting properties ( Solomon et al . , 2003; Li and Duffin , 2004; Elsen et al . , 2008 ) . However , it mimics many of the side effects of CBX , e . g . the initial stimulatory effect seen in the present study . These limitations need to be kept in mind when interpreting our results on gap junction functions , and further studies are needed to confirm them , preferably using more specific methods of connexin blockage , such as RNAi . However , our findings do suggest the presence of a neuron-specific subnetwork , connected by chemical synapses , that is able to maintain the network structure . Furthermore , another subnetwork , likely a glial one ( Giaume et al . , 2010; Okada et al . , 2012 ) driven by the electrical connections that modulate network output also seems to be present . Thus , neonatal preBötC synchronization is both gap junction-and synaptic signal-dependent ( Feldman and Kam , 2015 ) , and it probably contains both neuronal and glial subnetworks . The pFRG/RTN , by contrast , requires gap junctions for its establishment in rodents but is not dependent on them postnatally for rhythmic , correlated network activity ( Fortin and Thoby-Brisson , 2009 ) . The main mechanism that drives activity in the pFRG/RTN is glutamatergic ( Guyenet et al . , 2013 ) . By contrast , pFRG/RTN gap junctions seem here to be involved in the hypercapnic response ( Figure 10 and 11 ) . It has been suggested that Cx26 is directly modulated by CO2 , independent of H+ , through the formation of carbamate bridges ( Meigh et al . , 2013 ) . Our data do not distinguish between intracellular pH-dependent and -independent mechanisms . However , since PGE2 can pass through connexins ( Reyes et al . , 2014 ) , the present data are in line with a CO2-induced , connexin-mediated , release of PGE2 ( Figure 14 ) . 10 . 7554/eLife . 14170 . 064Figure 14 . Model of how PGE2 modulates respiration and sighs in the preBötC and pFRG/RTN . Systemic inflammation , through the proinflammatory cytokine IL-1β and hypoxia , induces the production of PGE2 in blood brain barrier ( BBB ) endothelial cells ( Hofstetter et al . , 2007 ) . PGE2 subsequently induces respiratory depression and increases sigh activity via the inhibitory G-protein coupled receptor EP3Rα in the preBötC . In the pFRG/RTN , PGE2 plays a role in the response to elevated pCO2 . CO2 directly modulates connexin 26 ( Cx26 ) hemichannels , leading to ATP release . The results in this study suggest that Cx26 also releases PGE2 , possiblyfrom mPGEs-1+ astrocytes . PGE2 increases respiratory activity via the stimulatory G-protein coupled receptor EP3Rγ on pFRG/RTN neurons . Thus , inflammation , hypoxia , and hypercapnia alter respiratory neural network and motor output and breathing activity through distinct effects of PGE2 in the pFRG/RTN and the preBötC , respectively . Chronically elevated PGE2 levels , as observed during ongoing inflammation , may decrease the central pattern generators’ ability to respond to hypoxic and hypercapnic events . In extreme cases , this decrease may have fatal consequences . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 064 Prostaglandins are important regulators of autonomic functions in mammals . In many disease states , acute inflammatory responses are initially protective but become harmful under chronic conditions . In our previous reports , we demonstrated how the pro-inflammatory cytokine interleukin ( IL ) -1β impairs respiration during infection by inducing a PGE2 release in the vicinity of respiratory centers . We also showed that infection is the main cause of respiratory disorders in preterm infants ( Hofstetter et al . , 2007 , 2008 ) and , in the case of apneas , bradycardias and desaturations ( ABD ) events in neonates ( Siljehav et al . , 2015 ) . PGE2 is also a key component in the regulation of sigh frequency ( Ramirez , 2014; Koch et al . , 2015 ) . During and immediately after birth , PGE2 levels are increased ( Mitchell et al . , 1978 ) . Indeed , the first breaths of extrauterine life are deep and sigh-like , facilitating alveolar recruitment and CO2 removal ( Mian et al . , 2015 ) . In the brainstem slice cultures , PGE2 had a direct EP3R-dependent effect on both respiratory centers . Notably , PGE2 increased pFRG/RTN but inhibited preBötC frequency ( Video 3 ) . This finding might be explained by the different distributions of EP3R subtypes in the different regions ( Figure 12 ) . The coupling to inhibitory or stimulatory G proteins depends on the alternative post-transcriptional splicing of the C-terminal tail of the EP3R preprotein ( Namba et al . , 1993 ) . Furthermore , PGE2 caused a longer Ca2+ transient and a higher relative amplitude in an INaP-dependent manner , mimicking the PGE2-based induction of sighs that we observe in vivo and that were recently reported by Koch and colleagues in acute preBötC slices ( Koch et al . , 2015 ) . 10 . 7554/eLife . 14170 . 065Video 3 . Parallel display of Ca2+ oscillations visualized with Fluo-4 in the pFRG/RTN ( left ) and preBötC ( right ) . After 15 s , PGE2 ( 10 nM ) is added . This increases the activity of the pFRG/RTN network while the preBötC activity is inhibited . fps: frames per second . For high-resolution versions of the videos , please follow this link to the Karolinska Institutet Cloud Storage system ( Box ) : https://ki . box . com/s/abzuei0yzl4dzbn99995382va6btsq4l . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 065 Recent data reveal a role of neuromedin B ( NMB ) and gastrin-related peptide ( Grp ) and NMB-GPR-expressing preBötC neurons in sighing ( Li et al . , 2016 ) . In addition to these peptidergic pathways , the present and recent data from Koch and colleagues ( Koch et al . , 2015 ) suggest that low concentrations of the inflammation-associated PGE2 induce sighs , acting through modulation of the persistent sodium current in preBötC neurons . The preBötC results presented in this study provide evidence for how the general respiratory depression induced by inflammatory signaling , previously reported in vivo and in vitro ( Hofstetter et al . , 2007 ) and in human neonates ( Hofstetter et al . , 2007; Siljehav et al . , 2015 ) , is mediated by a direct effect of PGE2 on EP3R ( Siljehav et al . , 2012 ) in the preBötC . The present data may help to further explain the mechanism underlying apneas that occur during infectious periods in neonates ( Hofstetter et al . , 2007 , 2008; Di Fiore et al . , 2013; Siljehav et al . , 2015 ) . Another common respiratory problem in neonates , particularly premature infants , is an inability to respond adequately to hypoxia and hypercapnia . This may cause recurrent hypoxia , leading to cognitive disabilities later in life ( Greene et al . , 2014 ) . A disruption of central CO2 chemosensitivity is commonly seen in children with bronchopulmonary dysplasia ( Di Fiore et al . , 2013 ) , leading to chronic hypoventilation , which may explain why these infants have an increased risk of sudden infant death syndrome ( Martin et al . , 2011 ) . Therefore , we investigated the role of the pFRG/RTN in chemosensitivity ( Guyenet et al . , 2013 ) and found that the response to hypercapnia is dependent on functioning gap junctions . This is in line with previous findings showing that Cx26 is directly modified by CO2 ( Meigh et al . , 2013 ) . These CO2-sensitive connexin hemichannels can release ATP , and indeed the hypercapnic response is partly mediated by purinergic type 2 receptors ( Erlichman et al . , 2010; Gourine et al . , 2010; Guyenet et al . , 2013 ) . In addition to these purinergic pathways , we suggest that EP3R-dependent signaling is involved in the response to altered pCO2 . Genetic ablation of Ptger3 reduced the hypercapnic response both in vivo and in vitro , as did pharmacological blockage in vitro , in line with our previous experiments ( Siljehav et al . , 2014 ) . Moreover , the optogenetic inhibition of Ptger3-expressing cells in the pFRG/RTN revealed that these cells are essential for the CO2 response . We also demonstrated that PGE2 is released during hypercapnic exposure , likely through Cx26 or other CO2-sensitive connexins ( Huckstepp et al . , 2010b ) . Thus , part of the CO2 response seems to be mediated by a gap junction-dependent release of PGE2 . Generation of active expiration is another important function of the pFRG/RTN ( Feldman et al . , 2013 ) . It is possible that PGE2 stimulates both chemosensitive neurons and neurons important for active expiration . Such neuronal populations could overlap , but the ventral part pFRG/RTN seems to have a more chemosensitive character while the lateral part displays rhythmic activity and enforces active expiration when stimulated ( Pagliardini et al . , 2011; Feldman et al . , 2013; Huckstepp et al . , 2015 ) . The CO2-sensing of the pFRG/RTN slice remains functional . Whether the rhythmic activity we observe in the pFRG/RTN is generated by “active expiration-neurons” is outside the scope of the present study . Future studies should aim to investigate whether PGE2 also may affect active expiration . The pFRG/RTN is the best-recognized central chemosensitive region . However , in our pFRG/RTN brainstem slice culture , neurons of the raphe nucleus should be present ( Smith et al . , 2009 ) . Such neurons may also have chemosensing properties ( Richerson , 2004 ) , though this has not been shown conclusively ( Depuy et al . , 2011 ) . From the raphe nucleus there are evidence of projections to the pFRG/RTN ( Guyenet et al . , 2009 ) , and we cannot exclude the possibility that these are preserved in the brainstem slice culture . The effects of CO2 in the present study are based on a change in carbamylation of specific proteins , e . g . Cx26 ( Meigh et al . , 2013 ) , or intracellular pH , but testing these alternatives goes beyond the scope of the present work . In our experimental setup the extracellular pH remained stable while the dissolved CO2 increased . This specific approach was selected because CO2 has a direct modulating effect on connexins , allowing passage of small molecules ( Huckstepp et al . , 2010a; Huckstepp and Dale , 2011; Meigh et al . , 2013 ) , and our hypothesis was that PGE2 is released through such connexins . What still remains to be determined the exact source of the PGE2 released during hypercapnia . The indication of a gap junction-dependent release of PGE2 together with the presence of mPGEs-1 in pFRG/RTN astrocytes suggests that the PGE2 is of astrocytic origin . This would be in line with previous findings of astrocytic ATP release during hypercapnia ( Gourine et al . , 2010; Huckstepp et al . , 2010a ) . The astrocytic involvement in the CO2 response is also evident in a Rett syndrome model ( methyl-CpG-binding protein 2 ( MeCP2 ) knockout ) , in which conditional MeCP2 knockout in astroglia blunts the CO2 response ( Turovsky et al . , 2015 ) . We think that mPGEs-1-expressing astrocytes are the likely source , even though alternative sources of PGE2 , such as endothelial cells or microglia , remain to be investigated with regards to their possible involvement in the pFRG CO2 response . Nonetheless , CO2-mediated PGE2 release introduces a novel chemosensitive pathway ( Figure 14 ) . As PGE2 and the EP3R are directly involved in and modulate both the respiratory rhythm-generating preBötC and the Phox2b chemosensitive neurons , PGE2 from other sources , such as endothelial cells during hypoxia and inflammation ( Hofstetter et al . , 2007 ) , will alter the hypercapnic and the hypoxic responses . PGE2 has prominent respiratory depressant effects in humans , sheep , pigs , and rodents ( Guerra et al . , 1988; Long , 1988; Ballanyi et al . , 1997; Hofstetter et al . , 2007; Siljehav et al . , 2015 ) . The PGE2-induced attenuation of these vital brainstem neural networks , e . g . , during an infectious response , could result in gasping , autoresuscitation failure and ultimately death . However , how chronic PGE2 release associated with ongoing inflammation alters plasticity and the responsiveness to CO2 must be further investigated . To conclude , we identified a novel pathway in the hypercapnic response of brainstem neural networks that control breathing . This pathway depends on EP3R and gap junctions and is partly mediated by the release of PGE2 , linking chemosensitivity control to the inflammatory system . The present findings have important implications for understanding why and how ventilatory responses to hypoxia and hypercapnia are impaired and inhibitory reflexes exaggerated in neonates , particularly during infectious episodes .
C57 black ( C57BL/6J ) inbred mice ( Charles River , Wilmington , MA ) were utilized in the experiments . The eicosanoid prostanoid 3 receptor ( EP3R ) gene ( Ptger3 ) was selectively deleted in knockout mice ( Ptger3−/− ) with a C57BL/6J background , as described preciously ( Fabre et al . , 2001 ) . C57BL/6J mice were then used as experimental controls for Ptger3−/− mice . As results from Ptger3−/− mice were consistent with pharmacological and optogenetic inhibition of EP3Rs , we can confirm the lost EP3R function in the mice . To determine the location of mPGEs-1 , mice expressing green fluorescent protein ( GFP ) under the GFAP promoter were used . Frozen sperm from the GFAP-tTA ( Lin et al . , 2004; Pascual et al . , 2005 ) and tetO-Mrgpra1 ( Fiacco et al . , 2007 ) mouse strains were purchased from the Mutant Mouse Regional Resource Centers supported by NIH ( MMRRC ) . The strains were re-derived by Karolinska Center for Transgene Technologies ( KCTT ) , and the offspring was crossed as previously described ( Fiacco et al . , 2007 ) . Double transgenics were identified by PCR according to MMRRC's instructions . All mice were reared by their mothers under standardized conditions with a 12:12-hr light-dark cycle . Food and water was provided ad libitum . The studies were performed in accordance with European Community Guidelines and approved by the regional ethic committee . The animals were reared and kept at the Department of Comparative Medicine , Karolinska Institutet , Stockholm , Sweden . Ventilatory measurements were made using dual-chamber plethysmography in 9-day old ( P9 ) mice . Mice were cooled on ice for 2–3 min and then prostaglandin E2 ( PGE2 , 1 µM; Sigma-Aldrich , St . Louis , MO , USA , cat no . P5640 ) or vehicle ( artificial cerebrospinal fluid , aCSF , containing in mM: 150 . 1 Na+ , 3 K+ , 2 Ca2+ , 2 Mg2+ , 135 Cl− , 1 . 1 H2PO4− , 25 HCO3- and 10 glucose ) was slowly injected into the lateral ventricle by using a thin pulled glass pipette attached to polyethylene tubing ( Siljehav et al . , 2014 ) . The mouse was then immediately placed into the plethysmograph chamber . After a 10-min recovery period , confirming stable respiration and body temperature , respiratory parameters in normocapnia ( air ) was established followed by a hypercapnic challenge ( 5% CO2 and 20% O2 in N2 ) for 5 min . This was followed by 5 min of normocapnia . Skin temperature was measured throughout experimentation and remained stable . After experimentation , the mice were anesthetized with 100% CO2 and decapitated . The brain was dissected and examined at the injection site and for the presence of any intracranial hemorrhage . Three of 28 animals had visible intracranial bleeding and were excluded from analysis . P3 mice pups were used for the establishment of brainstem organotypic slice cultures . The pups were decapitated at the cervical C3–C4 level . The heads were washed with cold dissection medium consisting of 55% Dulbecco’s modified Eagle’s medium ( Invitrogen , Paisley , UK ) , 0 . 3% glucose ( Sigma-Aldrich , St . Louis , MO , USA ) , 1% HEPES buffer ( Invitrogen , UK ) and 1% Antibiotic-Antimycotic ( Invitrogen , UK ) . After washing , the heads were moved to fresh dissection medium on ice . The entire brain was dissected . During dissection , extra caution was taken around the cerebellopontine angle to ensure that the respiratory regions of the brainstem were not damaged . Nerves were cut with microscissors . The brain was sectioned into 300-µm-thick transverse slices by using a McIlwain Tissue Chopper ( Ted Pella , Inc . , Redding , CA , USA ) . Slices were selected by using anatomical landmarks , such as the shape and size of the entire slice and the fourth ventricle . For location of the preBötzinger complex ( preBötC ) , the presence of nucleus hypoglossus , nucleus spinalis nervi trigemini , pyramis medullae oblongatae and nucleus tractus solitarius ( not always clearly seen ) , together with the absence of the anterior horn for the nucleus cochlearis , according to online references ( Ruangkittisakul et al . , 2006 , 2011 , 2014 ) . For location of the parafacial respiratory group/retrotrapezoid nucleus ( pFRG/RTN ) , the presence of the nucleus facialis was used . On the slices , the preBötC is located within ventrolateral regions , and the pFRG/RTN is located at the ventrolateral edge . Selected slices were washed by moving them to brain slice medium ( 55% Dulbecco’s modified Eagle’s medium , 32 . 5% Hank’s balanced salt solution , 0 . 3% glucose , 10% fetal bovine serum , 1% HEPES buffer and 1% Antibiotic-Antimycotic [Invitrogen , UK] ) , after which they were carefully placed on insert membranes ( Millicell Culture Plate Inserts; Millipore , Billerica , MA , USA ) in six-well plates . The membranes were coated in advance with poly-L-lysine ( 0 . 3 ml; 0 . 1 mg/ml , Sigma-Aldrich , St . Louis , MO , USA ) . Brain slice medium ( 1 ml ) was placed underneath the membrane , and all fluid on top of the membrane was removed . It is important not to cover the slices with medium , because this may impair oxygenation ( Frantseva et al . , 1999 ) . The brainstem slice cultures were maintained in an incubator ( 37°C , 5% CO2 ) , and the brain slice medium was changed every second day . The brainstem slices were kept in culture for 7–21 days in vitro ( DIV ) before fixation or live imaging experiments . For a detailed protocol , see Herlenius and colleagues ( Herlenius et al . , 2012 ) . For immunohistochemistry , brainstem slice cultures were fixed with cold paraformaldehyde ( 4% ) in PBS for 1 hr at 4°C and 20% ice-cold methanol in PBS for 10 min . Permeabilization was conducted by using 0 . 2% Triton X-100 ( Roche Diagnostics , Hofgeismar , Germany ) and 0 . 1% Tween 20 ( Invitrogen , UK ) in PBS for 40 min at room temperature ( RT ) . Thereafter , slices were blocked in 5% bovine serum albumin ( BSA; Invitrogen , UK ) and 0 . 05% Tween 20 in PBS for 2 hr at RT . The Millicell insert membranes were carefully cut with a scalpel and placed back into the wells . The primary antibodies were diluted 1:200 in 0 . 05% Tween 20/PBS and incubated at 4°C for 48 hr . Next , the slices were washed 3 × 10 min with PBS and incubated for 1 . 5 hr at RT with Alexa Fluor-conjugated secondary antibodies ( Invitrogen , UK ) diluted 1:200 in 0 . 05% Tween 20/PBS . The slices were then washed 3 × 10 min with PBS and mounted in ProLong Gold Antifade Reagent with DAPI ( Invitrogen , UK , cat . no . P36931 ) . Primary antibodies used were mouse anti-microtubule associated protein 2 ( MAP2; Invitrogen , cat . no . P11137 ) , rabbit anti-neurokinin 1 receptor ( NK1R; Sigma-Aldrich , St . Louis , MO , USA , cat no . S8305 ) , mouse anti-GFAP ( Chemicon , Temecula , CA , USA , cat no . MAB360 ) , rabbit anti-S100β ( Millipore; cat . no . 04–1054 ) , mouse anti-neuron-specific class III β-tubulin ( Tuj1; Covance , Princeton , NJ , USA , cat no . MMS-435P ) , rabbit anti-K+/Cl− cotransporter 2 ( KCC2; Millipore , cat no . 07–432 ) , rabbit anti-vesicular glutamate transporter 2 ( VGLUT2; Synaptic Systems , Goettingen , Germany , cat no . 135–402 ) , mouse anti-connexin 26 ( Cx26; Invitrogen , Inc . , San Francisco , CA , cat no . 13–8100 ) , rabbit anti-connexin 32 ( Cx32; Invitrogen , cat . no . 71–0600 ) , mouse anti-connexin 43 ( Cx43; Zymed , cat no 13–8300 ) , goat anti-Phox2b ( Santa Cruz Biotechnology , Santa Cruz , CA , USA , cat no 13224 ) , goat Phox2b antibody ( R & D Systems , Minneapolis , MN , USA ) , and rabbit anti-caspase 3 ( Cell Signaling Technology , Beverly , MA , USA , cat no . 9661 ) . Negative controls with only secondary antibodies showed no staining . For EP3R staining , a different protocol was used . Initially , brains were fixed with 4% paraformaldehyde overnight followed by 10% sucrose overnight and then frozen to -80% . The frozen brainstems were cryosectioned and blocked in blocking buffer ( 1% BSA , 5% donkey serum , 5% dimethyl sulfoxide ( DMSO ) , 1% Triton X-100 in Tris-buffered saline ( TBS , consisting of 6 mM Tris-HCl , 1 mM Tris base and 9 mM NaCl in ddH2O ) for 1 hr at RT . After blocking , the slices were incubated with polyclonal rabbit anti-EP3R antibody ( Cayman Chemical Co . , Ann Arbor , MI , USA ) diluted 1:50 in 10% DMSO containing 0 . 2% Triton X-100 in TBS at RT overnight . Next , slices were washed 3 × 15 min with TBS with agitation , followed by incubation for 1 hr in the dark with Alexa Fluor 488-conjugated donkey anti-rabbit secondary antibody ( Life Technologies , Grand Island , NY , USA ) diluted 1:1000 in 1% BSA , 2% donkey serum , 2% DMSO and 5% Triton X-100 in TBS . The slices were then washed 3 × 15 min with TBS with agitation , and blocked again for 1 hr at RT in the same blocking buffer as used previously . After blocking , the slices were incubated with the second primary antibody , diluted 1:200 in 10% DMSO containing 0 . 2% Triton X-100 in TBS at 4°C overnight . Following overnight incubation , the slices were washed 3 × 15 min with TBS with agitation and incubated with Alexa Flour 647-conjugated donkey anti-goat secondary antibody ( Life Technologies , Grand Island , NY , USA ) diluted 1:1000 in 1% BSA , 2% donkey serum , 2% DMSO and 5% Triton X-100 in TBS . Finally , the slices were washed 3 × 15 min with TBS with agitation , and mounted in ProLong Gold Antifade Reagent with DAPI . Antibody binding was controlled by including an irrelevant rabbit polyclonal IgG isotype control ( Bioss , Woburn , MA , USA ) . EP3R staining was controlled by including an EP3R blocking peptide reconstituted in distilled water mixed with EP3R antibody at a 1:1 ( v/v ) ratio . A pre-incubation of EP3R antibody with the blocking peptide for 1 hr at RT was necessary before the antibody was added to the slice . The peptide was used in conjunction with the antibody to block protein-antibody complex formation during immunohistochemical analysis for the EP3Rs . These controls showed no staining . Double immunofluorescence staining was also performed according to Westman and colleagues ( Westman et al . , 2004 ) using polyclonal rabbit anti-human microsomal prostaglandin E synthase 1 antiserum ( mPGES-1; Cayman chemicals , cat . no . 160140 ) and monoclonal anti-mouse glial fibrillary acidic protein antibody ( GFAP; Chemicon , Temecula , CA , USA , cat no . MAB360 ) . PBS supplemented with 0 . 1% saponin ( PBS-saponin ) was used as a buffer through the experiment . Endogenous peroxidase activity was blocked using PBS containing 1% H2O2 and 0 . 1% saponin for 60 min in darkness . Endogenous biotin was blocked using an avidin-biotin blocking kit ( Vector Laboratories , Burlingame , CA ) supplemented with 0 . 1% saponin . The sections were incubated with primary antibodies overnight , in PBS-saponin containing 3% BSA antibody solution . Thereafter , they were blocked with 1% normal goat serum , or normal donkey serum ( depending on the host of secondary antibody ) in PBS-saponin for 15 min , followed by 1-hr incubation with secondary antibody , donkey anti-rabbit alexa fluorophore 488 or goat anti-mouse Alexa Fluor 546 . Propidium iodide ( 1 ml/L , Invitrogen , UK ) was added to brain slice medium ( dilution 1:1000 ) . Staining solution ( 1 ml ) was added on top of the membrane with the brainstem slice cultures and incubated at 37°C ( 5% CO2 ) for 3 hr . Immediately after incubation , the brainstem slice cultures were fixed in 4% paraformaldehyde for 1 hr . Positive controls were made by first treating the brainstem slice culture for oxygen glucose deprivation ( OGD ) for 1 h , as described by Montero Dominguez and colleagues ( Montero Domínguez et al . , 2009 ) . Whole-cell patch recordings were obtained from brainstem slice cultures at a temperature of 34°C . Cells were visualized by using IR-differential contrast microscopy ( Axioskop FS , Carl Zeiss , Jena , Germany ) . Recorded cells were selected visually , and paired recordings were obtained for neurons with lateral somatic distances of <100 µm . Recordings were amplified by using 700B amplifiers ( Molecular Devices , Sunnyvale , CA , USA ) , filtered at 2 kHz , digitized at 5–20 kHz by using ITC-18 ( Instrutech , Longmont , CO , USA ) , and acquired by using Igor Pro ( Wavemetrics , Lake Oswego , OR , USA ) . Patch pipettes were pulled with a P-97 Flamming/Brown micropipette puller ( Sutter Instruments , Novato , CA , USA ) and had an initial resistance of 5–10 MΩ in a solution containing in mM: 110 K-gluconate , 10 KCl , 10 HEPES , 4 Mg-ATP , 0 . 3 GTP and 10 phosphocreatine . Recordings were performed in current-clamp mode , with access resistance compensated throughout the experiments . Recordings were discarded when access resistance increased beyond 35 MΩ . To characterize the electrical properties of the recorded cells , depolarizing and hyperpolarizing current steps and ramps were injected , enabling the extraction of properties such as input resistance , membrane time constant and action potential threshold . Electrophysiological properties were presented as box plots , with maximum and minimum values . For recording of hypoglossal nerve activity and hypoglossal nucleus neuronal population discharge , an extracellular suction electrode was used together with a Model 1700 AC amplifier ( A-M systems , Carlsborg , WA , USA ) and AxoScope software , version 9 . 2 ( Axon Instruments , Union City , CA , USA ) . Recordings were made with a sampling interval of 0 . 3 ms . For Ca2+ imaging , Fluo-4 AM ( Invitrogen , UK ) dissolved in DMSO ( Invitrogen , UK ) was used at 10 µM in serum free brain slice medium or artificial cerebrospinal fluid ( aCSF , containing in mM: 150 . 1 Na+ , 3 K+ , 2 Ca2+ , 2 Mg2+ , 135 Cl− , 1 . 1 H2PO4− , 25 HCO3- and 10 glucose ) together with 0 . 02% pluronic acid ( Invitrogen , UK ) . We did not observe any differences in Ca2+activity between HEPES-free brain slice medium and aCSF during Ca2+imaging , despite slight differences in [K+] and [Ca2+] , which both affect the rhythm of the slice ( Ballanyi and Ruangkittisakul , 2009 ) . A higher [Ca2+] or an increase in [K+] from 3 mM to 4 . 8 mM did not affect the network properties in our system . To localize the preBötC or the pFRG/RTN , tetramethylrhodamine-conjugated Substance P ( TMR-SP; Biomol , Oakdale , NY , USA ) was used at a final concentration of 3 µM in brain slice medium or aCSF . The TMR-SP solution was placed on top on the brainstem slice and incubated for 10–12 min at 37°C in an atmosphere of 5% CO2 . The TMR-SP solution was then replaced with 1 ml of 10 µM Fluo-4 solution . The Fluo-4 solution was incubated for 30–40 min ( 37°C , 5% CO2 ) . Before imaging , the slice was washed with brain slice medium/aCSF for 10 min ( 37°C , 5% CO2 ) . During time-lapse imaging , slices were kept in an open chamber perfused with HEPES-free brain slice medium ( containing in mM: 132 Na+ , 4 . 8 K+ , 1 . 4 Ca2+ , 0 . 74 Mg2+ , 112 Cl- , 0 . 76 H2PO4- , 25 . 6 HCO3- and 16 . 8 glucose ) or aCSF ( 2 . 5 ml/min ) by using a peristaltic pump . A Chamlide Inline Heater ( Live Cell instruments , Seoul , Korea , cat no . IL-H-10 ) was used for temperature control , and a Chamlide AC-PU perfusion chamber for 25-mm coverslips ( Live Cell instruments , Seoul , Korea , cat no . ACPU25 ) was used for perfusion . HEPES-free medium was used to minimize the risk for hydrogen peroxide formation ( Lepe-Zuniga and Gery , 1987 ) . The medium or aCSF was constantly bubbled with 5% CO2 and 95% O2 . The temperature of the chamber was set to 32°C , which Hartelt and colleagues ( Hartelt et al . , 2008 ) showed to be well tolerated by neurons . Images were captured by using a Zeiss AxioExaminer D1 microscope equipped with 20× and 40× water immersion objectives ( N . A . 1 . 0 ) , a monochromatic Zeiss MrM CCD-camera , a Photometrics eVolve EMCCD-camera and filter sets 38HE ( Zeiss ) , 43 ( Zeiss ) , and et560/hq605 ( Chroma , Bellows Falls , VT , USA ) . For live imaging , a frame interval of 0 . 1–2 s was used . Exposure time was set to 100–300 ms . Substances added during imaging were [D-Ala2 , N-Me-Phe4 , Gly5-ol]-enkephalin ( DAMGO , 0 . 5 µM; Sigma-Aldrich , St . Louis , MO , USA , cat no . E7384 ) , carbenoxolone ( CBX 50 , µM; Sigma-Aldrich , St . Louis , MO , USA , cat no . C4790 ) , 18α-glycyrrhetinic acid ( 18-α-GA , 25 µM; Sigma-Aldrich , St . Louis , MO , USA , cat no . G10105 ) , glycyrrhizic acid ( GZA , 50 µM; Sigma-Aldrich , St . Louis , MO , USA , cat no . 50531 ) , tetrodotoxin ( TTX , 20 nM , Abcam , Cambridge , UK , cat . no . 120055 ) , riluzole ( 10 µM , Sigma-Aldrich , St . Louis , MO , USA , cat . no . R116 ) , flufenamic acid ( FFA , 50 µM , Sigma-Aldrich , St . Louis , MO , USA , cat . no . F9005 ) , Suramin ( 100 µM; Sigma-Aldrich , St . Louis , MO , USA , cat no . S2671 ) , TNP-ATP ( 20 nM; Sigma-Aldrich , St . Louis , MO , USA , cat . no . SML0740 ) , AH6809 ( Cayman Chemicals , cat . no . 33458-93-4 ) and prostaglandin E2 ( PGE2 , 10 nM; Sigma-Aldrich , St . Louis , MO , USA , cat no . P5640 ) . All substances were dissolved in brain slice medium/aCSF prior to experimentation and added to the chamber by using a continuous flow system . For each experiment , a control period with regular medium/aCSF was followed by drug application . GZA was used as a negative control for the gap junction inhibitors CBX and 18α-GA because it has non-gap junction-inhibiting properties , but similar side effects to those of CBX . Specificity was tested by using a second batch of medium or aCSF . During infections in neonatal children , PGE2 is present at a concentration of 15 pM in cerebrospinal fluid ( Hofstetter et al . , 2007 ) . A higher concentration ( 10 nM ) was used to compensate for the in vivo metabolism of the molecule . Exposure to isohydric hypercapnia was done by using aCSF adjusted with a high bicarbonate buffer concentration ( in mM: 150 . 1 Na+ , 3 K+ , 2 Ca2+ , 2 Mg2+ , 111 Cl− , 1 . 1 H2PO4− , 50 HCO3− and 10 glucose ) . This generated a hypercapnic carbon dioxide partial pressure ( pCO2 ) of 6 . 6 kPa at pH 7 . 5 when aCSF was saturated with 8% CO2 . A subgroup of 1-DIV-old brainstem slices were moved to a separate BSL-2 laboratory where they were transduced with a mouse prostaglandin E receptor 3 ( subtype EP3 ) lentivirus ( Ptger3 ) containing Halo57 , developed in collaboration with Dr Robert Finney ( Xactagen , Shoreline , WA , USA ) , by applying 0 . 2 µl of virus suspension on top of the slice . The brainstem slice cultures were then placed in an incubator for 5 days , and after washing with warm brain slice medium at time points 2 and 5 days , the brainstem slice cultures were moved back to the original laboratory and placed in an incubator overnight . Ca2+ time-lapse imaging was performed on the slices as described above . Halo57 was stimulated continuously during Ca2+ time-lapse imaging by using a 625-nm LED in a custom-built system ( Thorlabs , Newton , NJ , USA ) . The optogenetically inhibited network and NK1R positive neurons retained their response to general depolarization induced by elevated [K+] ( Supplementary Fig . S6 ) . The release of PGE2 in aCSF during control and hypercapnic conditions was assessed by ELISA . The aCSF samples were collected through perfusion system , during control and hypercapnic period and either analyzed immediately or stored at -80°C . For the validation of the experiments , two different ELISA kits have been used . Prostaglandin E2 EIA monoclonal kit by Cayman Chemical ( Ann Arbor , MI , US ) was performed according to standard procedure . Firstly , the PGE2 EIA Standard was prepared from #1 to #8 . The 96-well plate was ready to use and contained a minimum of two blanks ( Blk ) , two non-specific binding wells ( NSB ) , two maximum binding wells ( B0 ) and an eight point standard curve run in duplicate . Each sample was assayed in triplicate . The 96-well plate was coated for 18 hr at 4°C with 50 μl of Prostaglandin E2 AChE Tracer and 50 μl of Prostaglandin E2 Monoclonal Antibody per well . Plate was washed three times with specific Wash Buffer and in consequence , it was developed in the dark at room temperature on a plate shaker for 60–90 min by adding 200 μl of Ellman’s Reagent to each well . Finally , the plate was read at 405 nm . PGE2 ELISA kit by Enzo Life Sciences ( Farmdale , NY , US ) was also used for the confirmation of the results . A similar process was followed but a bit shorter . Samples were assayed in duplicate . The 96-well plate was incubated at room temperature on a plate shaker for 2 hr with 50 μl of PGE2 conjugate and 50 μl of antibody solution per well . Then , the plate was washed three times with washing solution . After the wash , 200 μl of the pNpp substrate solution were added to every well and the plate was incubated at room temperature for 45 min . Finally , 50 μl of Stop Solution were added to every well in order to stop the reaction and the plate was read immediately at 405 nm . The preBötC and pFRG/RTN regions were cut out from brainstem slices with micro scissors . The samples were pooled together litterwise to minimize the effect of different tissue piece sizes , and provide enough cells for accurate analysis . RNA was isolated from the tissue samples using the miRCURY RNA isolation Kit ( Exiqon ) according to manufacturer’s instructions . cDNA was synthesized from 20 ng RNA using SuperScript VILO cDNA Synthesis Kit ( Invitrogen ) . The reverse transcription was performed according to the manufacturer’s protocol . Real-time PCR was run with Power SYBR Green PCR Master Mix ( Applied Biosystems ) and amplified in a 7500 Real Time PCR system ( Applied Biosystems ) . Primers are listed in Table 8 . As endogenous control , glucose-3 phosphate dehydrogenase ( GAPDH; Applied Biosystems ) was used . Relative quantification ( RQ ) values were calculated using the CT ( ΔΔCT ) method ( Livak and Schmittgen , 2001 ) . 10 . 7554/eLife . 14170 . 066Table 8 . Primers used for qRT-PCR . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 066Oligo nameSequenceEP3alfa forward EP3alfa reverseGCTTCCAGCTCCACCTCCTT CATCATCTTTCCAGCTGGTCACTEP3 sense EP3beta anti-sense5′-TGACCTTTGCCTGCAACCTG-3′ 5′-GACCCAGGGAAACAGGTACT-3′EP3gamma forward EP3gamma reverseAGTTCTGCCAGGTAGCAAACG GCCTGCCCTTTCTGTCCAT From in vivo plethysmograph recording ( LabChart Pro , v 8 . 0 . 10 , AD Instruments , Dunedin , New Zealand ) , periods of calm respiration without movement artifacts were selected for analysis based upon visual observations during experimentation as in previous studies ( Hofstetter and Herlenius , 2005 ) . Mean respiratory frequency ( FR; breaths/min ) , tidal volume ( VT ) and minute ventilation ( VE ) during normocapnic and hypercapnic periods were calculated as described previously ( Hofstetter and Herlenius , 2005 ) . Sighs were excluded from the analysis . VT and VE were divided by body weight ( BW ) and expressed as milliliters per gram and milliliters per gram per min , respectively . The number of sighs , defined as breath with larger amplitude and a biphasic inspiratory phase , was calculated manually and expressed as sighs per min . Immunohistochemical staining was analyzed in a Zeiss AxioExaminer D1 microscope ( 10× , 20× and 40× water immersion objectives ) or a Zeiss LSM700 confocal ( 40× and 63× oil-immersion objectives ) , and captured images were processed by adjusting contrast in ImageJ ( 1 . 42q , National Institutes of Health , Bethesda , MD , USA ) to reduce background staining . Ca2+ imaging time traces were analyzed with a recently published method ( Malmersjo et al . , 2013; Smedler et al . , 2014 ) . Regions of interest were marked for all cells based on the standard deviation of fluorescence intensity over time , by using a semiautomatic-adapted ImageJ script kindly provided by Dr . John Hayes ( The College of William and Mary , Williamsburg , VA , USA , http://physimage . sourceforge . net/ ) . The mean intensity value and coordinates were measured using ImageJ . Average intensities of regions of interest were quantified for each frame , and dynamic fluorescence signals were normalized to baseline values . The linear similarity ( Pearson correlation ) was calculated ( Figure 4—figure supplement 1 ) between pairs of Ca2+ traces with a custom-made script in MATLAB ( version 7 . 9 . 0 . 529 R2009b; MathWorks , Natick , MA , USA ) and by using the mic2net toolbox ( Smedler et al . , 2014 ) ( version 6 . 12; MathWorks ) . Calculating the pairwise correlation coefficients resulted in a correlation matrix that was converted to an adjacency matrix by applying a cut-off level . The cut-off level was selected by calculating the mean of the 99th percentile of correlation coefficients for a set of experiments with scrambled signals . Scrambling was performed by randomly translating all traces in the time-domain . The network structure was visualized by plotting a line between pairs of cells , where the color of the lines was proportionate to the correlation coefficient . This was plotted on top of an image of the standard deviations of the fluorescence over time per pixel . Connectivity was defined as the number of cell pairs with a correlation coefficient larger than the cut-off value divided by the total number of the pairs of cells . This provided a measure of the degree of connections within a network . Small-world parameter , mean shortest path length ( λ ) and mean clustering coefficient ( σ ) were calculated by using the MATLAB BGL library ( http://www . mathworks . com/matlabcentral/fileexchange/10922 ) and compared to corresponding randomized networks . Many biological networks have a small-world structure , where the mean shortest path length is as short as in random networks and the mean clustering coefficient is higher . This signifies that the average number of nodes ( for example , neural cells ) that a signal has to pass is low , and that many of the nodes are connected in clusters ( Watts and Strogatz , 1998 ) . A small-world network structure creates the possibilities of regional specialization and efficient signal transfer , and is a common organization of networks within the brain ( Telesford et al . , 2011 ) . Data were further processed to produce graphs in OriginPro , version 9 . 1 ( OriginLab Corporation , Northamptom , MA , USA ) . Time-lapse Ca2+ imaging time traces were normalized individually through ΔF/F0 , where ΔF=F1−F0 . F1 is the specific fluorescence intensity at a specific time point , and F0 is the average intensity of 30 s before and after F1 . A previously published toolbox was used for the frequency analysis of time traces ( Uhlén , 2004 ) . Recordings of hypoglossal nerve activity were filtered ( 0 . 06-Hz low-pass ) , rectified and smoothed ( 1 s ) ( Talpalar et al . , 2011 ) by using OriginPro ( version 9 . 1 , OriginLab Corporation , USA ) . Statistical analysis of paired comparisons was performed by Student’s t-test . Full factorial two-way ANOVA was performed when there was more than one independent variable or multiple observations . Both tests were two-sided . The compared data was of equal variance and normally distributed . All calculations for the statistical tests were conducted with JMP ( v 11 . 1 . , SAS Institute Inc . , Cary , NC , US ) . In all cases , p<0 . 05 was considered statistically significant . Data are presented as means ± SD . All data sets were compared less than 20 times , which is why no statistical corrections were made . As these experiments were expended to provide new descriptive data , no explicit power analysis was performed . Instead sample sizes similar to previous publications with similar methods were used . Details on the statistics are presented in Tables 9 and 10 . 10 . 7554/eLife . 14170 . 067Table 9 . Successful experiments behind representative images . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 067FigurePanelNumber of experiments1c252b c d e f g31 23 9 19 14 1122 – S1112 , 23 , 12 33 , 23 , 10 27 , 12 , 11 22 , 15 , 8 15 , 8 , 82 – S252 – S3a b c12 5 203a b5 53 – S1a b5 54b c d g h12 13 8 9 54-S2840 , 621 , 4565a165 – S1a65 – S2a c11 56a b c d e f g h i j9 8 8 3 8 6 7 8 6 77a c d f5 , 4 9 12 5 , 47 – S14 , 48b c d e9 9 9 58 – S135 , 22 , 19 9 , 9 59b c d12 11 69 – S1315 , 429 , 1929 – S218 , 710a b7 7 , 7 and 511a12 , 411 – S1a b c11 11 612a c d e f5 , 4 16 44 6 1213a c f6 , 5 7 , 5 4113 – S1a b6 2710 . 7554/eLife . 14170 . 068Table 10 . Statistical details for presented figures . DOI: http://dx . doi . org/10 . 7554/eLife . 14170 . 068FigureTest usedExact p-valueDegrees of freedom&F/t/z/R value1dFull factorial two-way ANOVA0 . 0084F=0 . 612 , DFE=211eFull factorial two-way ANOVA0 . 0365F=0 . 284 , DFE=211fFull factorial two-way ANOVA0 . 0157F=0 . 329 , DFE=211gFull factorial two-way ANOVA0 . 0017 0 . 018 0 . 007F=0 . 547 F=0 . 332 F=1 . 618 DFE=232 – S1dStudent’s t-test0 . 35 0 . 45DF = 9 , 194eStudent’s t-test0 . 63 0 . 76 0 . 91DF=11 , 12 , 84fStudent’s t-test0 . 13 , 0 . 34 , 0 . 68 0 . 21 , 0 . 28 , 0 . 86 0 . 76 , 0 . 76 , 1 . 00DF=11 , 12 , 84 – S2Student’s t-test0 . 43 , 0 . 34 , 0 . 12 0 . 11 , 0 . 57 , 0 . 19DF=19164 – S2Paired t-test0 . 02 , 0 . 02 , 0 . 04 0 . 03 , 0 . 02 , 0 . 01DF=19165bStudent’s t-test0 . 42 , 0 . 51 , 0 . 80DF=155cStudent’s t-test0 . 50 , 0 . 62 , 0 . 98DF=155 – S2aPaired t-test0 . 0029DF=65 – S2bPaired t-test0 . 041DF=105 – S2dPaired t-test0 . 03DF=46kPaired t-test0 . 03125 0 . 00391 0 . 3125 0 . 625DF=236lPaired t-test0 . 01563 0 . 00781 0 . 28125 0 . 25DF=236iPaired t-test0 . 492 0 . 331 0 . 390 0 . 390DF=236mPaired t-test0 . 457 0 . 124 0 . 567 0 . 143DF=236 – S1aPaired t-test0 . 15625 0 . 46094 0 . 0625 0 . 625DF=236 – S1bPaired t-test0 . 15625 0 . 74219 0 . 15625 0 . 8125DF=236 – S1cPaired t-test0 . 1246 0 . 07813 0 . 3125 0 . 8125DF=237bPaired t-test0 . 018 0 . 034 0 . 047 0 . 079 0 . 132 0 . 084 0 . 028 0 . 063 0 . 067 0 . 012 0 . 077 0 . 90DF= 4 , 3 , 77gPaired t-test0 . 95 0 . 51DF=4 , 37-S1Paired t-test0 . 51861 0 . 15558 0 . 69733 0 . 51508 0 . 36415 0 . 2433DF=169eStudent’s t-test0 . 15 , 0 . 51 , 0 . 57 0 . 29 , 0 . 061 , 0 . 0081DF=11 , 10 , 79fStudent’s t-test0 . 38 , 0 . 51 , 0 . 75 0 . 59 , 0 . 66 , 0 . 99 0 . 10 , 0 . 10 , 0 . 95DF=11 , 10 , 79gPaired t-test0 . 43DF=39hPaired t-test0 . 0334DF=119iPaired t-test0 . 0141 0 . 0283 0 . 00475DF=6 , 5 , 39 – S1Student’s t-test0 . 77 , 0 . 51 , 0 . 92 0 . 28 , 0 . 07 , 0 . 60DF=9359 – S1Paired t-test0 . 04 , 0 . 04 , 0 . 01 0 . 03 , 0 . 02 , 0 . 01DF=9359 – S2Paired t-test0 . 00507 0 . 5745 0 . 22731 0 . 68788 0 . 81018 0 . 66252DF=1610cPaired t-test0 . 53 , 0 . 20 , 0 . 61 0 . 009 , 0 . 015 , 0 . 041 0 . 023 , 0 . 045 , 0 . 035 0 . 01 , 0 . 09 , 0 . 14DF=11 , 7 , 7 , 910 – S1aPaired t-test0 . 153 0 . 0848 0 . 388DF=610 – S1bPaired t-test0 . 59 0 . 43DF=611bFull factorial two-way ANOVA0 . 418 0 . 0161F=0 . 054 F=0 . 9712bPaired t-test0 . 00038 0 . 28DF= 6 , 413bPaired t-test0 . 13 , 0 . 56 0 . 16 0 . 24 , 0 . 12 , 0 . 012DF=713dPaired t-test0 . 00046 0 . 87DF=6 , 4 , 4 , 313gPaired t-test0 . 4112 0 . 0001DF=4013 – S1cPaired t-test0 . 0125 0 . 00098DF=5 | Humans and other mammals breathe air to absorb oxygen into the body and to remove carbon dioxide . We know that in a part of the brain called the brainstem , several regions work together to create breaths , but it is not clear precisely how this works . These regions adjust our breathing to the demands placed on the body by different activities , such as sleeping or exercising . Sometimes , especially in newborn babies , the brainstem’s monitoring of oxygen and carbon dioxide does not work properly , which can lead to abnormal breathing and possibly death . In the brain , cells called neurons form networks that can rapidly transfer information via electrical signals . Here , Forsberg et al . investigated the neural networks in the brainstem that generate and control breathing in mice . They used slices of mouse brainstem that had been kept alive in a dish in the laboratory . The slice contained an arrangement of neurons and supporting cells that allowed it to continue to produce patterns of electrical activity that are associated with breathing . Over a three-week period , Forsberg et al . monitored the activity of the cells and calculated how they were connected to each other . The experiments show that the neurons responsible for breathing were organized in a “small-world” network , in which the neurons are connected to each other directly or via small numbers of other neurons . Further experiments tested how various factors affect the behavior of the network . For example , carbon dioxide triggered the release of a small molecule called prostaglandin E2 from cells . This molecule is known to play a role in inflammation and fever . However , in the carbon dioxide sensing region of the brainstem it acted as a signaling molecule that increased activity . Therefore , inflammation could interfere with the body’s normal response to carbon dioxide and lead to potentially life-threatening breathing problems . Furthermore , prostaglandin E2 induced deeper breaths known as sighs , which may be vital for newborn babies to be able to take their first deep breaths of life . Future challenges include understanding how the brainstem neural networks generate breathing and translate this knowledge to improve the treatment of breathing difficulties in babies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"neuroscience"
] | 2016 | CO2-evoked release of PGE2 modulates sighs and inspiration as demonstrated in brainstem organotypic culture |
Assessing the danger of transition of HIV transmission from a concentrated to a generalized epidemic is of major importance for public health . In this study , we develop a phylogeny-based statistical approach to address this question . As a case study , we use this to investigate the trends and determinants of HIV transmission among Swiss heterosexuals . We extract the corresponding transmission clusters from a phylogenetic tree . To capture the incomplete sampling , the delayed introduction of imported infections to Switzerland , and potential factors associated with basic reproductive number R0 , we extend the branching process model to infer transmission parameters . Overall , the R0 is estimated to be 0 . 44 ( 95%-confidence interval 0 . 42—0 . 46 ) and it is decreasing by 11% per 10 years ( 4%—17% ) . Our findings indicate rather diminishing HIV transmission among Swiss heterosexuals far below the epidemic threshold . Generally , our approach allows to assess the danger of self-sustained epidemics from any viral sequence data .
Epidemics of HIV and other blood-borne and sexually transmitted diseases ( for instance syphilis , HBV and HCV ) can be subdivided into concentrated and generalized epidemics . While for the former , the rapid infectious agent transmission is restricted to core transmission groups involved in high-risk behaviors ( such as men who have sex with men and injecting drug users ) , the generalized epidemic refers to fast pathogen spreading in the heterosexual ( general ) population resulting in higher overall disease prevalence . Mechanistically , the key factor explaining whether the HIV transmission is concentrated or generalized , is the ability of HIV to spread among heterosexuals . If the epidemic in this population is not self-sustained , the HIV epidemic remains concentrated; otherwise the virus is spreading rapidly in the broad population leading to a generalized HIV epidemic . In most resource-rich settings HIV transmission is concentrated , that is , driven mostly by transmission among men who have sex with men ( MSM ) and injecting drug users ( IDU ) , whereas the limited transmission among heterosexuals is maintained by either imported infections or spillovers from other transmission groups ( Kouyos et al . , 2010; von Wyl et al . , 2011; Ragonnet-Cronin et al . , 2016; Xiridou et al . , 2010; Esbjörnsson et al . , 2016; Sallam et al . , 2017 ) . This suggests that in most Western European countries and similar epidemiological settings the basic reproductive number R0 among heterosexuals is below 1 . However , it is not clear how far away from self-sustained the epidemic is in heterosexuals . Moreover , the change in HIV transmission among heterosexuals over time is another important , yet unknown , factor , especially with evidenced increasing risky sexual behavior ( Kouyos et al . , 2015 ) . It is therefore crucial to assess both the transmission and its time trend in order to obtain meaningful insights into the epidemic . Assessing the subcritical transmission of HIV in the general population shares some methodological similarities with the analysis of stage III zoonoses , for instance , monkeypox ( Wolfe et al . , 2007 ) , which also exhibit stuttering transmission chains . Both cases follow a source-sink dynamics , i . e . , a flux of infections from a subpopulation in which the disease is self-sustained to a population where it is not . For the case of stage III zoonoses and tuberculosis , it has been shown that the distribution of outbreak sizes can be used to quantify the pathogen spread ( Blumberg and Lloyd-Smith , 2013b; Blumberg and Lloyd-Smith , 2013a; Borgdorff et al . , 1998 ) . The fundamental approach of our study is to apply this concept to transmission of HIV in the general population . However , there are two key differences between emerging zoonotic pathogens and human-to-human infectious agents . Firstly , while the contact tracing data are not available for many sexually transmitted infections ( STI ) , the viral sequences carry valuable information about the transmission chain size distribution . Thus , the approach of quantifying transmissibility from chain size distributions needs to be combined with a tool to derive clusters from viral sequences . Compared to the animal-human transmission the delayed introduction of the index case of an STI or blood-borne virus to the subpopulation of interest plays an important role , especially in viruses like HIV with long infectious periods in the absence of treatment and higher transmissibility during the acute phase ( Marzel et al . , 2016; Powers et al . , 2011; Rieder et al . , 2010; Rodger et al . , 2016; Hollingsworth et al . , 2008; Cohen et al . , 2011b; Cohen et al . , 2011a; Cohen et al . , 2016 ) . This is especially important because a considerable fraction of HIV cases in heterosexuals is found in migrants ( Del Amo et al . , 2004; von Wyl et al . , 2011; European Centre for Disease Prevention and Control/WHO Regional Office for Europe , 2016 ) . If , for example , a migrant infected with HIV abroad moves to Switzerland in the chronic stage of the infection , he/she has ( from the perspective of the Swiss population ) lost some transmission potential upon entering Swiss heterosexual transmission network . In order to quantify the subcritical transmission we combine phylogenetic cluster analysis with an adapted version of a branching process model based estimator that derives the basic reproductive number R0 from the size distribution of transmission chains . We further extend this approach to determine the impact of calendar time and other potential determinants on R0; especially in order to assess whether R0 exhibits an increasing time trend or is high in particular subgroups . Applying this method to the phylogenetic transmission clusters among heterosexuals in the Swiss HIV Cohort Study ( SHCS ) , we can assess transmission of HIV in this population and in particular the risk of a generalized HIV epidemic together with the main determinants of transmission .
To obtain an overall estimate for the R0 of HIV transmission in Swiss heterosexuals , the baseline model was fitted to all of the previously described transmission chain data . In this baseline model the R0 was estimated to be 0 . 44 ( 95%-confidence interval ( CI ) 0 . 42—0 . 46 ) . The fact that R0 was clearly below 1 ( p-value <0 . 001 from one-sided Wald hypothesis testing H0:R0=1 against the alternative HA:R0<1 ) indicated that HIV transmission is far away from a self-sustained epidemic . Although the overall R0 estimate was clearly below 1 , individual subtypes represent different epidemiological settings and hence individual subtypes may have R0 closer to the epidemic threshold . The subtype-stratified analyses indeed yielded lower R0 of 0 . 35 ( 95%-CI 0 . 33—0 . 39 ) for subtype B as compared to the non-B subtypes ( Figure 1 ) . The recombinant form CRF02_AG had the highest estimated R0 of 0 . 62 ( 95%-CI 0 . 56—0 . 69 ) . Despite these differences among the R0 estimates for different subtypes they were all significantly below 1 ( with all p-values from the one-sided test smaller than 0 . 001 ) . Therefore , we concluded that there is no danger of a self-sustained HIV epidemic in Swiss heterosexuals of any HIV subtype . Despite consistently low R0 estimates , an increasing time trend for R0 would impose a potential concern , especially if the time trend would predict a crossing of the epidemic threshold in the near future . To investigate this , we fitted a univariate model with log ( R0 ) as a linear function of the establishment date of the transmission chain . We found that overall the R0 is decreasing at a factor 0 . 89 per 10 years ( 95%-CI 0 . 83—0 . 96 ) . The per subtype-stratified analyses showed the consistently decreasing time trend among the subtypes ranging from factor 0 . 65 per 10 years for subtype A to 0 . 89 for B-subtype . To better capture the changes of R0 over time we included higher-order polynomials of the establishment date to our model ( Figure 2 ) . With the reference date on the 1st of January 1996 ( which corresponds to the median estimated date of infection - see Table 2 ) a cubic spline ( without the linear term ) was identified as the optimal model according to the Bayesian information criterion ( BIC ) . This model exhibits a mild increase of the R0 from the mid 1980’s to the mid 1990’s , with a peak-R0 of 0 . 49 ( 95%-CI 0 . 46—0 . 53 ) reached in 1996 and followed by a steep and monotonic decrease . It is noteworthy that the time of peak-R0 coincided with the introduction of highly active antiretroviral therapy . Shortly after the R0 started to rapidly decrease and has never rebounded . This extrapolation should be , however , taken with a grain of salt and seen more as a trend rather than a prognosis , since only a few transmission chains have been observed for the recent years ( which is reflected by wide confidence intervals ) . Finally , we identified the characteristics associated with higher R0 and therefore potential focal subpopulations , in which the basic reproductive number R0 could be above 1 . The simplest model containing only the linear terms of risk factors showed that the R0 is decreasing with the establishment date of the transmission chain and that all non-B subtypes have higher R0 compared to subtype B , which was consistent with the findings from the univariate model and per-subtype stratified analyses . Moreover , we found that reporting sex with occasional partners and longer time to HIV diagnosis of the index case are associated with higher R0 , whereas the earliest CD4 cell count and the age do not have significant effects ( Figure 3 ) . These trends remained robust ( Figure 4 ) when allowing the covariables to enter the model non-linearly ( for instance as polynomials like in the case of the time trend above ) . The final multivariate model identified subtype , establishment date of the transmission chain , frequency of reporting sex with occasional partner and time to diagnosis of the index case as the significant risk factors associated with R0 ( see Selection of the predictive models ) . Allowing nonlinear terms for the time to diagnosis provided better goodness-of-fit than the linear model . The steep increase of R0 in the early/acute phase ( see Figure 4 ) of the infection indicates the importance of early diagnosis ( which is nowadays closely related to early treatment initiation ) while the time becomes less relevant in the cases diagnosed late in the chronic phase .
Generally , our approach allows the assessment of the danger of a concentrated epidemic to become generalized based on the viral sequence data . We demonstrated this approach for the case of heterosexual HIV transmission in Switzerland . In particular , even though the study highlighted some heterogeneity between the HIV subtypes , our findings indicate that there is no imminent danger of a self-sustained epidemic among Swiss heterosexuals , but rather diminishing HIV transmission far below the epidemic threshold . Hence , the HIV epidemic in Switzerland is and most likely will remain restricted to high risk core groups , especially MSM . Moreover , the results suggest that integrated prevention measures in Switzerland taken over time were successful within the heterosexual population .
The SHCS is a multicenter , nationwide , prospective observational study of HIV infected individuals in Switzerland , established in 1988 ( Swiss HIV Cohort Study et al . , 2010 ) . The SHCS was approved by the ethics committees of the participating institutions ( Kantonale Ethikkommission Bern , Ethikkommission des Kantons St . Gallen , Comite Departemental d’Ethique des Specialites Medicales et de Medicine Communataire et de Premier Recours , Kantonale Ethikkommission Zürich , Repubblica e Cantone Ticino–Comitato Ethico Cantonale , Commission Cantonale d’Étique de la Recherche sur l’Être Humain , Ethikkommission beiderBasel; all approvals are available on http://www . shcs . ch/206-ethic-committee-approval-and-informed-consent ) , and written informed consent was obtained from all participants . Up to December 2016 over 19 , 500 patients have been enrolled . The SHCS is highly-representative as it covers more than 75% HIV-positive individuals on antiretroviral therapy ( ART ) in Switzerland ( Swiss HIV Cohort Study et al . , 2010 ) . In addition to the extensive demographic and clinical data collected at biannual/quarterly follow-up ( FUP ) visits , for approximately 60% of the patients at least one partial pol sequence from the genotypic resistance testing is available ( in total 22 , 036 sequences from the SHCS resistance database until August 2015 ) . The patients with heterosexual contact as the most likely transmission route comprise about one third of all SHCS participants . The phylogenetic tree was constructed from the Swiss HIV sequences of the SHCS patients and non-Swiss background sequences exported from the Los Alamos National Laboratory , 2016 database ( 241 , 783 HIV-1 viral sequences of any subtype and including the circulating recombinant forms 01–74 retrieved on February 23rd , 2016 spanning over the protease and RT regions with fragments of at least 250 nucleotides; the HXB2 sequence and sequences from Switzerland were removed afterwards ) . The sequences of 8 HIV-1 subtypes and circulating recombinant forms ( B , C , CRF01_AE , CRF02_AG , A ( 1-2 ) ) , G , D and F ( 1-2 ) ) were pairwise aligned to the reference genome HXB2 ( accession number K03455 ) using Muscle v3 . 8 . 31 ( Edgar , 2004 ) . Sequences with insufficient sequencing quality of the protease region ( coverage of less than 200 nucleotides between the positions 2253 and 2549 of HXB2 ) or reverse transcriptase region ( less than 500 nucleotides between positions 2550 and 3869 ) were excluded . Using the earliest available of the remaining sequences for each patient , the phylogenetic tree was built with the FastTree algorithm under the generalized time-reversible model of nucleotide evolution ( Price et al . , 2009 ) including 10 , 840 SHCS and 90 , 933 background sequences . The Swiss heterosexual transmission chains were defined as clusters in the phylogenetic tree containing exclusively Swiss HIV sequences from individuals with heterosexual contact as the most likely route of the transmission , regardless of the respective genetic distances and local support values ( see Sensitivity analyses and Appendix 1—figure 8 for alternative definition ) . The transmission chains and the patients enrolled in the SHCS forming them were identified with custom written functions in R ( version 3 . 3 . 2 ) . For each transmission chain we determined if it was introduced to the Swiss HIV heterosexuals either as an imported infection from abroad or from other HIV transmission groups within Switzerland . The geographic origin for a given chain was obtained as the country of the closest sequence , which did not belong to Swiss heterosexuals . Specifically , we considered the smallest clade that contained both the transmission chain and either a non-Swiss or non-heterosexual sequence , and chose the sequence with the smallest pairwise genetic distance to the transmission chain ( with respect to the Jukes and Cantor ( JC69 ) model ) . Additionally , in each extracted transmission chain the observed index case was identified as the patient with the earliest estimated date of infection in the chain . The date of HIV infection for each single individual was imputed with the model described by Taffé et al . ( 2008 ) if the patient had enough CD4 cell count measurements before the ART initiation and the estimated date of infection fell within the seroconversion window; otherwise the midpoint of the seroconversion window was used . The demographic characteristics ( Table 2 ) of the index case were extracted from the SHCS , including age at infection , time to diagnosis , first available CD4 cell count and sexual risk behavior . The latter was quantified as the fraction of semiannual follow-up visits at which the patient reported sex with occasional partners . The patients with no available questionnaire regarding the sexual risk behavior were assumed to have never reported on having sex with occasional partner ( see Sensitivity analyses and Appendix 1—figure 9 for the corresponding sensitivity analysis ) . The characteristics of the index case were then used to define the features of each corresponding transmission chain . Our model is based on the basic discrete-time branching process . The basic reproductive number R0 was inferred from the model as the expected number of offsprings , therefore the offspring distribution represents the crucial component of the chain size distribution model . In the following sections we describe the main extensions of the basic branching process theory , which were implemented in our model . The detailed derivations can be found in Appendix 3 . The maximum likelihood ( ML ) estimator for 𝜷 , the predictor for R0 and the corresponding statistics ( confidence intervals , p-values , etc . ) were implemented in the R package PoisTransCh ( Turk , 2017 , https://github . com/tejaturk/PoisTransCh; copy archived at https://github . com/elifesciences-publications/PoisTransCh ) . The provided confidence intervals are the Wald-type 95%-confidence intervals ( see Sensitivity analyses for the comparison against different types ) and the p-values are based on the Wald statistic . Initially , we assessed the impact of covariables potentially associated with HIV transmission . Specifically , we considered HIV-1 subtype , establishment date of the transmission chain ( i . e . , the earliest estimated date of infection in the transmission chain ) , reported sex with occasional partner , age at infection , first measured CD4 cell count and time to diagnosis of the index case . Final model selection was carried out by the forward selection and backward elimination algorithms based on the Akaike and Bayesian information criterion ( AIC and BIC , respectively ) . The detailed steps are provided in Selection of the predictive models . Previously published datasets from Kouyos et al . ( 2010 ) and von Wyl et al . ( 2011 ) were used in this study . As previously discussed in these publications , due to the large sampling density this data would , in principle , allow for the reconstruction of entire transmission networks and could thereby endanger the privacy of the patients . This is especially problematic because HIV-1 sequences frequently have been used in court cases . Therefore , a random subset of 10% of the sequences are accessible via GenBank . These accession numbers are as follows: GU344102-GU344671 , EF449787 , EF449788 , EF449796 , EF449798 , EF449828 , EF449829 , EF449838 , EF449844 , EF449852 , EF449853 , EF449854 , EF449860 , EF449880 , EF449883 , EF449889 , EF449895 , EF449901 , EF449904 , EF449905 , EF449917 , EF449921 , EF449928 , EF449930 , EF449943 , EF449950 , EF449960 , EF449971 , EF449980 , EF449987 , EF450004 , EF450005 , EF450011 , EF450024 , EF450026 , GQ848113 , GQ848120 , GQ848140 , GQ848145 , GQ848149 , JF769777-JF769851 | In epidemiology , the “basic reproductive number” describes how efficiently a disease is transmitted , and represents the average number of new infections that an infected individual causes . If this number is less than one , many people do not infect anybody and hence the transmission chains die out . On the other hand , if the basic reproductive number is larger than one , an infected person infects on average more than one new individual , which leads to the virus or bacteria spreading in a self-sustained way . Turk et al . have now developed a method to estimate the basic reproductive number using the genetic sequences of the virus or bacteria , and have used it to investigate how efficiently HIV spreads among Swiss heterosexuals . The results show that the basic reproductive number of HIV in this group is far below the critical value of one and that over the last years this number has been decreasing . Furthermore , the basic reproductive number differs for different subtypes of the HIV virus , indicating that the geographical region where the infection was acquired may play a role in transmission . Turk et al . also found that people who are diagnosed later or who often have sex with occasional partners spread the virus more efficiently . These findings might be helpful for policy makers as they indicate that the risk of self-sustained transmission in this group in Switzerland is small . Furthermore the method allows HIV epidemics to be monitored at high resolution using sequence data , assesses the success of currently implemented preventive measures , and helps to target subgroups who are at higher risk of an infection – for instance , by supporting frequent HIV testing of these people . The method developed by Turk et al . could also prove useful for assessing the danger of other epidemics . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"microbiology",
"and",
"infectious",
"disease"
] | 2017 | Assessing the danger of self-sustained HIV epidemics in heterosexuals by population based phylogenetic cluster analysis |
Biosensors for small molecules can be used in applications that range from metabolic engineering to orthogonal control of transcription . Here , we produce biosensors based on a ligand-binding domain ( LBD ) by using a method that , in principle , can be applied to any target molecule . The LBD is fused to either a fluorescent protein or a transcriptional activator and is destabilized by mutation such that the fusion accumulates only in cells containing the target ligand . We illustrate the power of this method by developing biosensors for digoxin and progesterone . Addition of ligand to yeast , mammalian , or plant cells expressing a biosensor activates transcription with a dynamic range of up to ~100-fold . We use the biosensors to improve the biotransformation of pregnenolone to progesterone in yeast and to regulate CRISPR activity in mammalian cells . This work provides a general methodology to develop biosensors for a broad range of molecules in eukaryotes .
Biosensors capable of sensing and responding to small molecules in vivo have wide-ranging applications in biological research and biotechnology , including metabolic pathway regulation ( Zhang et al . , 2012 ) , biosynthetic pathway optimization ( Raman et al . , 2014; Tang and Cirino , 2011 ) , metabolite concentration measurement and imaging ( Paige et al . , 2012 ) , environmental toxin detection ( Gil et al . , 2000 ) , and small molecule-triggered therapeutic response ( Ye et al . , 2013 ) . Despite such broad utility , no single strategy for the construction of biosensors has proven sufficiently generalizable to gain widespread use . Current methods typically couple binding of a small molecule to a single output signal , and use a limited repertoire of natural protein- ( Tang et al . , 2013 ) or nucleic acid aptamer-binding ( Yang et al . , 2013 ) domains , which narrows the scope of small molecules that can be detected . A general solution to small molecule biosensing should be adaptable to a range of small molecules and responses . A promising approach to biosensor design in eukaryotes uses conditionally stable ligand-binding domains ( LBDs ) ( Banaszynski et al . , 2006; Tucker and Fields , 2001 ) . In the absence of a cognate ligand , these proteins are degraded by the ubiquitin proteasome system ( Egeler et al . , 2011 ) . Binding of the ligand stabilizes the LBD and prevents its degradation . Fusing the destabilized LBD to a suitable reporter protein , such as an enzyme , fluorescent protein , or transcription factor , renders the fusion conditionally stable and generates sensor response ( Figure 1a ) . Naturally-occurring LBDs can be engineered to be conditionally stable ( Banaszynski et al . , 2006; Miyazaki et al . , 2012; Iwamoto et al . , 2010 ) , making it possible in principle to convert them into biosensors for target ligands . Designed LBDs can also be used , especially in cases for which natural binding proteins do not exist or lack sufficient specificity or bioorthogonality . 10 . 7554/eLife . 10606 . 003Figure 1 . A general method for construction of biosensors for small molecules . ( a ) Modular biosensor construction from a conditionally destabilized LBD and a genetically fused reporter . The reporter is degraded in the absence but not in the presence of the target small molecule . ( b ) yEGFP fluorescence of digoxin LBD-GFP biosensors upon addition of 250 μM digoxin or DMSO vehicle . ( c ) yEGFP fluorescence of progesterone LBD-GFP biosensors upon addition of 50 μM progesterone or DMSO vehicle . ( d ) Positions of conditionally destabilizing mutations of DIG0 mapped to the crystal structure of the digoxin LBD ( PDB ID 4J9A ) . Residues are shown as colored spheres and key interactions highlighted in insets . In b-c , fold activation is shown above brackets , ( - ) indicates cells lacking biosensor constructs , and error bars represent the standard error of the mean ( s . e . m . ) of three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 00310 . 7554/eLife . 10606 . 004Figure 1—figure supplement 1 . Population responses to cognate ligand for cells bearing LBD-biosensors . ( a ) yEGFP fluorescence of DIG0-GFP upon addition of 250 μM digoxin or DMSO vehicle . ( b ) yEGFP fluorescence of DIG1-GFP upon addition of 250 μM digoxin or DMSO vehicle . ( c ) yEGFP fluorescence of PRO0-GFP upon addition of 50 μM progesterone or DMSO vehicle . ( d ) yEGFP fluorescence of PRO1-GFP upon addition of 50 μM progesterone or DMSO vehicle . The data for each condition are presented for a single representative biological replicate with 20 , 000 events measured by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 00410 . 7554/eLife . 10606 . 005Figure 1—figure supplement 2 . Characterization of mutations conferring progesterone-dependent stability . ( a ) Single-mutant deconvolutions of mutations conferring progesterone sensitivity . The parental biosensor appears in the leftmost column of each panel . ( b–e ) Positions of mutations in PRO1 ( b ) , PRO2 ( c ) , and PRO3 ( d ) are mapped to the crystal structure of the digoxin LBD ( PDB ID 4J9A ) and are shown in colored spheres . ( e ) Fold activation of PRO0-GFP biosensors with digoxin biosensor mutations upon addition of 50 μM progesterone . In a and e , error bars represent s . e . m . of three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 005 Here , we convert a single designed LBD scaffold into multiple highly specific biosensors for the clinically relevant steroids digoxin and progesterone . We engineer LBDs fused with fluorescent reporters to be conditionally stable in the budding yeast Saccharomyces cerevisiae . Attaching these conditionally-stabilized LBDs to transcription factors ( TFs ) yields biosensors that respond to their target ligands with greater signal induction than observed with fusions to fluorescent proteins . We use TF-biosensors to improve the biosynthetic yield of progesterone in yeast . The biosensors retain function when ported directly into mammalian cells , with up to 100-fold activation over background , allowing us to develop a method for tight control of CRISPR/Cas9 genome editing . The biosensors also show up to 50-fold activation by ligand in Arabidopsis thaliana . The method presented here enables the rapid development of eukaryotic biosensors from natural and designed binding domains .
LBDs intended for biosensor development should recognize their targets with high affinity and specificity . We began with the computationally-designed binding domain DIG10 . 3 ( Tinberg et al . , 2013 ) , hereafter DIG0 , which binds the plant steroid glycoside digoxin and its aglycone digoxigenin with picomolar affinities . Introduction of three rationally-designed binding site mutations into DIG0 resulted in a progesterone binder ( PRO0 ) with nanomolar affinity ( Tinberg et al . , 2013 ) . We constructed genetic fusions of DIG0 and PRO0 to a yeast-enhanced GFP ( yEGFP , LBD-biosensors DIG0-GFP and PRO0-GFP ) and constitutively expressed them in S . cerevisiae ( Supplementary file 1 ) . The fusions showed little change in fluorescence in response to digoxin or progesterone , respectively ( Figure 1b , c and Figure 1—figure supplement 1 ) . Work by Wandless and co-workers has shown that mutagenesis of LBDs can be used to identify variants that are stable only in the presence of a target ligand ( Banaszynski et al . , 2006 ) . We randomly mutagenized the LBDs of DIG0-GFP and PRO0-GFP by error-prone PCR and subjected libraries of 105 ( Gil et al . , 2000 ) integrants to multiple rounds of FACS , sorting alternately for high fluorescence in the presence of the ligand and low fluorescence in its absence . We isolated LBD variants having greater than 5-fold activation by cognate ligand ( Figure 1b , c and Figure 1—figure supplement 1 ) . By making additional variants that contain only one of the up to four mutations found in the progesterone biosensors , we showed that some mutations are additive , while others predominately contribute to sensitivity ( Figure 1—figure supplement 2a ) . Many of the conditionally-destabilizing mutations identified in DIG0 involve residues that participate in key dimer interface interactions ( Figure 1d ) . The conditionally-destabilizing mutations of PRO0 are located throughout the protein ( Figure 1—figure supplement 2b–d ) ; the DIG0 interface mutations also rendered PRO0-GFP conditionally stable on binding progesterone ( Figure 1—figure supplement 2e ) . To improve the dynamic range and utility of the biosensors , we built conditionally-stable LBD-transcription factor fusions ( TF-biosensors ) by placing an LBD between an N-terminal DNA-binding domain ( DBD ) and a C-terminal transcriptional activation domain ( TAD , Figure 2a ) . The use of TFs serves to amplify biosensor response and allows for ligand-dependent control of gene expression ( Shoulders et al . , 2013; Beerli et al . , 2000; Louvion et al . , 1993 ) . Our initial constructs used the DBD of Gal4 , the destabilized LBD mutant DIG1 ( E83V ) , and either VP16 or VP64 as a TAD to drive the expression of yEGFP under the control of a GAL1 promoter . The dynamic range of TF-biosensor activity was maximal when the biosensor was expressed using a weak promoter and weak activation domain , because of lower yEGFP expression in the absence of ligand ( Figure 2—figure supplement 1a , b ) . 10 . 7554/eLife . 10606 . 006Figure 2 . Ligand-dependent transcriptional activation . ( a ) TF-biosensor construction from a conditionally destabilized LBD , a DNA-binding domain and a transcriptional activation ( TAD ) domain . ( b ) Positions of conditionally destabilizing mutations of Gal4 mapped to a computational model of Gal4-DIG0 homodimer . Residues are shown as colored spheres and key interactions are highlighted in insets . The TAD is not shown . ( c ) Concentration dependence of response to digoxin for digoxin TF-biosensors driving yEGFP expression . ( d ) Concentration dependence of response to progesterone for progesterone TF-biosensors driving yEGFP expression . ( e ) Time dependence of response to 250 μM digoxin for digoxin TF-biosensors . ( f ) Time dependence of response to 50 μM progesterone for progesterone TF-biosensors . ( g ) Time-dependent response to withdrawal of 250 μM digoxin for digoxin TF-biosensors . ( h ) Time-dependent response to withdrawal of 50 μM progesterone for progesterone TF-biosensors . In c-f , ( - ) indicates cells lacking biosensor plasmids and error bars represent s . e . m . of three biological replicates . Marker symbols in e and g are the same as in c . Marker symbols in f and h are the same as in d . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 00610 . 7554/eLife . 10606 . 007Figure 2—figure supplement 1 . Improvements to TF-biosensor response . Digoxin-dependent expression of yEGFP by G-DIG1-V TF-biosensors either ( a ) containing VP64 or VP16 as the TAD and expressed from a CYC1 promoter or ( b ) containing a VP16 TAD and expressed from a CYC1 , ADH1 , or TEF1 promoter . ( c ) Individual mutations identified in a FACS analysis of an error-prone PCR library of G-DIG-V biosensors were tested for their effect on biosensor function using digoxigenin . Transformants were analyzed in an yEGFP yeast reporter strain containing a deletion of PDR5 ( PyE14 ) . Improvements in fold activation relative to parental sequences were localized to mutations in Gal4 . ( d ) R60S and L77F mutations found in Gal4 were introduced into G-DIG1-V , G-DIG2-V , and G-PRO1-V . In each case , the Gal4 mutations had the effect of lowering the amount of luciferase activity in the absence of the relevant ligand . In a-d , error bars represent s . e . m . of three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 00710 . 7554/eLife . 10606 . 008Figure 2—figure supplement 2 . Population responses to cognate ligand for cells bearing TF-biosensors . ( a ) yEGFP fluorescence of DIG0-GFP upon addition of digoxin or DMSO vehicle ( left ) and yEGFP fluorescence of DIG1-GFP upon addition of digoxin or DMSO vehicle ( right ) . ( b ) yEGFP fluorescence of PRO0-GFP upon addition of progesterone or DMSO vehicle ( left ) and yEGFP fluorescence of PRO1-GFP upon addition of progesterone or DMSO vehicle ( right ) . Data for each condition are presented for a single representative biological replicate with 20 , 000 events measured by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 008 We chose Gal4-DIG1-VP16 ( hereafter G-DIG1-V ) for further TF-biosensor development because it has both a large dynamic range and maximal activation by ligand . A FACS-based screen of an error-prone PCR library of G-DIG0-V , G-DIG1-V , and G-DIG2-V variants identified mutations L77F and R60S in the Gal4 dimer interface ( hereafter GL77F , GR60S ) that further increased TF-biosensor response by lowering background activity in the absence of ligand ( Figure 2b and Figure 2—figure supplement 1c ) . Although these Gal4 mutations were identified by screening the libraries of digoxin-dependent TF-biosensors , they also increased progesterone-dependent activation of the G-PRO-V series of biosensors , indicating a shared mechanism of conditional stability in both systems ( Figure 2—figure supplement 1d ) . Combining mutations in Gal4 and DIG0 or PRO0 led to activations of up to 60-fold by cognate ligand , a ten-fold improvement over the most responsive LBD-biosensors ( Figure 2c , d and Figure 2—figure supplement 2a ) and a dynamic range that has been challenging to achieve with stability-based biosensors in yeast ( Rakhit et al . , 2011 ) . The TF-biosensors were also rapidly activated , showing a five-fold increase in signal after 1 hr of incubation with ligand and full activation after ~14 hr ( Figure 2e , f and Figure 2—figure supplement 2b ) . In contrast to the LBD-biosensors , the TF-biosensors exhibited a broad range of fluorescence levels across single cells , as well as a population of nonfluorescent cells in the presence of ligand ( Figure 2—figure supplement 2 ) . We used FACS to isolate cells from the nonfluorescent population and found those cells to be inviable , possibly indicating plasmid loss or toxicity from biosensor activation . Upon withdrawal of ligand , strains expressing TF-biosensors rapidly exhibited reduction in signal , reaching half of their maximum yEGFP fluorescence after approximately 5 hr and nearly undetectable fluorescence after 10–15 hr ( Figure 2g , h ) . The response of the TF-biosensors to the withdrawal of ligand is likely much faster than observed by fluorescence , as the reduction in fluorescence signal is dependent on both the degradation of the TF-biosensors as well as the degradation and dilution of previously expressed yEGFP . An attractive feature of the TF-biosensors is that the constituent parts – the DBD/promoter pair , the LBD , the TAD , the reporter , and the yeast strain – are modular , such that the system can be modified for additional applications . To demonstrate tunability , we replaced the DBD of G-DIG1-V with the bacterial repressor LexA and replaced the Gal4 DNA-binding sites in the GAL1 promoter with those for LexA . LexA-based TF-biosensors with DIG1 and a weak TAD ( B42 ) showed a strong response to digoxin ( nearly 40-fold ) only when the promoter-driving reporter expression contained LexA-binding sites ( Figure 3a ) . These results demonstrate that the biosensors can function with different combinations of DBDs and TADs , which could produce diverse behaviors and permit their use in eukaryotes requiring different promoters . Furthermore , the reporter gene can be swapped with an auxotrophic marker gene to enable growth selections . The TF-biosensors drove the expression of the HIS3 reporter most effectively when steroid was added to the growth media , as assessed by the growth of a histidine auxotrophic strain in media lacking histidine ( Figure 3b , d ) . Fusion of the Matα2 degron to the biosensor improved dynamic range by reducing the growth of yeast in the absence of ligand . Finally , the yeast strain could be modified to improve biosensor sensitivity toward target ligands by the deletion of the gene for a multidrug efflux pump ( Ernst et al . , 2005 ) , thereby increasing ligand retention ( Figure 3c–d ) . 10 . 7554/eLife . 10606 . 009Figure 3 . Tuning TF-biosensors for different contexts . ( a ) The TAD and DBD of the TF-biosensor and the corresponding binding site for the DBD in the reporter promoter can be swapped for a different application . Expression of a plasmid-borne luciferase reporter was driven by TF-biosensors containing either a LexA or Gal4 DBD and either a VP16 or B42 TAD . Promoters for the reporter contained DNA-binding sites for either Gal4 or LexA . ( b ) TF-biosensors were transformed into the yeast strain PJ69-4a and tested for growth on this minimal media containing 1 mM 3-aminotriazole ( 3-AT ) and the indicated steroid . To determine the effect of including an additional destabilization domain , the degron from Matα2 was cloned into one of four positions . ( c ) G-DIG1-V biosensor response to digoxigenin in yEGFP reporter strain PyE1 either with or without a deletion to the ORF of PDR5 . ( d ) Ligand and TF-biosensor-dependent growth on this media in yeast strains containing deleted ORFs for efflux-related transcription factors ( PDR1 and PDR3 ) or ABC transporter proteins ( YOR1 , PDR5 , SNQ2 ) . In a and c , error bars represent s . e . m . of three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 009 Improving bioproduction requires the ability to detect how modifications to the regulation and composition of production pathways affect product titers . Current product detection methods such as mass spectrometry or colorimetric assays are low-throughput and are not scalable or generalizable . LBD- and TF-biosensors could be coupled with fluorescent reporters to enable high throughput library screening or to selectable genes to permit rapid evolution of biosynthetic pathways ( Tang and Cirino , 2011; Dietrich et al . , 2010; Chou and Keasling , 2013 ) . Yeast-based platforms have been developed for the biosynthesis of pharmaceutically relevant steroids , such as progesterone and hydrocortisone ( Duport et al . , 1998; Szczebara et al . , 2003 ) . A key step in the production of both steroids is the conversion of pregnenolone to progesterone by the enzyme 3β-hydroxysteroid dehydrogenase ( 3β-HSD ) . We aimed to use a progesterone biosensor to detect and improve this transformation . An important feature of biosensors intended for pathway engineering is their ability to detect a product with minimal activation by substrate or other related chemicals . TF-biosensors built from PRO1 showed the greatest dynamic range and selectivity for progesterone over pregnenolone when driving yEGFP expression or when coupled with a HIS3 reporter assay ( Figure 4a , b and Figure 4—figure supplement 1a ) . We investigated whether this sensor could be used to detect the in vivo conversion of pregnenolone to progesterone by episomally-expressed 3β-HSD ( Figure 4c ) . Using GL77F-PRO1-V driving a yEGFP reporter , we could detect progesterone production , with biosensor response greatest when 3β-HSD was expressed from a high copy number plasmid and from a strong promoter ( Figure 4d ) . 10 . 7554/eLife . 10606 . 010Figure 4 . Application of biosensors to metabolic engineering in yeast . ( a ) Fold activation of GL77F-PRO1-V by a panel of steroids in yEGFP reporter strain PyE1 . Data are represented as mean ± SEM . ( b ) Growth of degron-G-PRO1-V in HIS3 reporter strain PJ69-4a is stimulated by progesterone but not pregnenolone . ( c ) Schematic for directed evolution of 3β-HSD using TF-biosensors for the conversion of pregnenolone to progesterone . ( d ) Fold activation of GL77F-PRO1-V by a panel of plasmids expressing wild-type 3β-HSD under varying promoter strengths in yEGFP reporter strain PyE1 when incubated in 50 μM pregnenolone . Data for plasmids containing CEN/ARS and 2 μ ( 2 micron ) origins are shown . ( e ) Fold activation of GL77F-PRO1-V by a panel of evolved 3β-HSD mutants expressed under the TDH3 promoter on a CEN/ARS plasmid and incubated in 50 μM pregnenolone . ( f ) Progesterone titer in 1 OD of cells produced by strains expressing 3β-HSD mutants . Progesterone became toxic at levels of 100 μM and above , leading to substantial cell death . β-estradiol and hydrocortisone were not soluble in yeast growth media at levels above 25 μM . In a and d-f , data are presented as mean ± s . e . m . of three biological replicates . In d and e , ( - ) indicates cells lacking 3β-HSD . *indicates significance with a threshold of p < 0 . 05 using 2-tailed Student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 01010 . 7554/eLife . 10606 . 011Figure 4—figure supplement 1 . Specificity of PRO biosensors enables selection for auxotrophy complementation . Specificity for progesterone ( PRO ) over digoxigenin ( DIG ) , digoxin ( DGX ) , digitoxigenin ( DTX ) , pregnenolone ( PRE ) , β-estradiol ( B-EST ) , and hydrocortisone ( HYD ) for ( a ) G-PRO0-V ( b ) G-PRO1-V ( c ) G-PRO2-V and ( d ) G-PRO3-V . ( e ) Growth response of yeast strain PyE1 transformed with 3β-HSD on CEN/ARS plasmids under various promoters and plated on SC –his ( and –ura –leu for plasmid maintenance ) containing titrations of 3-AT and either 0 . 5% DMSO ( upper panels ) or 50 μM pregnenolone ( lower panels ) . Progesterone became toxic at levels of 100 μM and above , leading to substantial cell death . Beta-estradiol and hydrocortisone were not soluble in yeast growth media at levels above 25 μM . In a-d , error bars represent s . e . m . of three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 011 We then sought to use the biosensor to improve this enzymatic transformation . To select for improved progesterone production , we required a growth assay in which wild-type 3β-HSD could no longer complement histidine auxotrophy when the yeast were grown on plates supplemented with pregnenolone . To this end , the selection stringency was tuned by adding the His3 inhibitor 3-aminotriazole ( Figure 4—figure supplement 1e ) . We mutagenized the 3β-HSD coding sequence using error-prone PCR and screened colonies that survived the HIS3 selection for their yEGFP activation by pregnenolone . By transforming evolved 3β-HSD mutations into a fresh host background , we showed that the mutations in the enzyme , and not off-target plasmid or host escape mutations , were responsible for increased biosensor response ( Figure 4e ) . Two of the mutants , 3β-HSD N139D and 3β-HSD F67Y , were assayed for progesterone production using gas chromatography and mass spectrometry and were found to produce two-fold more progesterone per OD than cells bearing the wild-type enzyme ( Figure 4f ) . Yeast is an attractive platform for engineering in vivo biosensors because of its rapid doubling time and tractable genetics . If yeast-derived biosensors function in more complex eukaryotes , the design-build-test cycle in those organisms could be rapidly accelerated . We first assessed the portability of yeast TF-biosensors to mammalian cells . Single constructs containing digoxin and progesterone TF-biosensors with the greatest dynamic ranges ( without codon optimization ) were stably integrated into human K562 cells using PiggyBac transposition . We characterized the dynamics of the TF-biosensors in human cells by dose response and time course assays similar to the yeast experiments ( Figure 5a–d ) . As with yeast , the human cells demonstrated greater sensitivity to digoxin , with fluorescence activation increasing up to 100 nM of cognate ligand for digoxin biosensors and 1 mM for progesterone biosensors . We observed >100-fold activation for the most sensitive progesterone biosensor GL77F-PRO1-V . The increase in mammalian dynamic range over yeast may arise from more aggressive degradation of destabilized biosensors or greater accumulation of target-stabilized biosensors or reporters resulting from larger cell sizes and slower doubling times . The time course data show that fluorescence increased four-fold within 4 hr of target introduction and rose logarithmically for 24–48 hr . We next assessed whether these biosensors could drive more complex mammalian phenotypes . The CRISPR/Cas9 system has proved to be an invaluable tool for genome editing ( Mali et al . , 2013; DiCarlo et al . , 2013; Gratz et al . , 2013; Hwang , 2013 ) . Despite the high programmability and specificity of Cas9-mediated gene editing achieved to date , unchecked Cas9 activity can lead to off-target mutations and cytotoxicity ( Fu et al . , 2013; Mali et al . , 2013; Pattanayak et al . , 2013 ) . Further , it may be desirable to tightly regulate Cas9 activity such that gene editing occurs only under defined conditions . To facilitate inducible gene editing , we fused human codon-optimized versions of the DIG3 and PRO1 LBDs to the N-terminus of Cas9 from S . pyogenes . We integrated this construct into a reporter cell line containing an EGFP variant with a premature stop codon that renders it non-functional . Upon separate stable integration of the DIG-Cas9 and PRO-Cas9 fusions , we transfected a guide RNA targeting the premature stop codon as well as a donor oligonucleotide containing the sequence to restore EGFP activity via homologous recombination . After a 48-hr incubation period , we observed an ~18-fold increase in GFP positive cells with digoxigenin relative to the mock control ( Figure 5e ) . 10 . 7554/eLife . 10606 . 012Figure 5 . Activation of biosensors in mammalian cells and regulation of CRISPR/Cas9 activity . ( a ) Concentration dependence of response to digoxin for constructs containing digoxin TF-biosensors and Gal4 UAS-E1b-EGFP reporter individually integrated into K562 cells . GR60S , L77F-PRO1-V serves as a digoxin insensitive control . ( b ) Concentration dependence of response to progesterone for constructs containing progesterone TF-biosensors and Gal4 UAS-E1b-EGFP reporter individually integrated into K562 cells . GR60S-DIG1-V serves as a progesterone insensitive control . ( c ) Time dependence of response to 100 nM digoxin for constructs containing digoxin TF-biosensors and Gal4 UAS-E1b-EGFP reporter individually integrated into K562 cells . GR60S , L77F-PRO1-V serves as a digoxin insensitive control . ( d ) Time dependence of response to 25 μM progesterone for constructs containing progesterone TF-biosensors and Gal4 UAS-E1b-EGFP reporter individually integrated into K562 cells . GR60S-DIG1-V serves as a progesterone insensitive control . ( e ) DIG3 and PRO1 fused to the N-terminus of S . pyogenes Cas9 were integrated into a K562 cell line containing a broken EGFP . EGFP function is restored upon transfection of a guide RNA and donor oligonucleotide with matching sequence in the presence of active Cas9 . The data are presented as mean fluorescence ± s . e . m . of three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 012 To assess generalizability of these sensors to multicellular organisms , we engineered G-DIG1-V to function as an environmental biosensor in plants . The DIG1 sequence was codon optimized for expression in Arabidopsis thaliana . We tested biosensor fusions to two different degrons , Matα2 from yeast and DREB2a from Arabidopsis ( Sakuma et al . , 2006 ) , and we used the VP16 and VP64 variants as the TAD . We initially tested the G-DIG1-TAD variants with a transient expression assay using Arabidopsis protoplasts and a reporter gene consisting of firefly luciferase under the control of a Gal4-activated plant promoter ( pUAS::Luc ) . The biosensor containing the Matα2 degron and VP16 TAD showed the highest fold activation of luciferase in the presence of digoxigenin ( Figure 6—figure supplement 1a ) . We next inserted the genes encoding G-DIG1-V-Matα2 and the Gal4-activated pUAS::Luc into a plant transformation vector and stably transformed them into Arabidopsis plants . Primary transgenic plants were screened in vivo for digoxigenin-dependent luciferase production , and responsive plants were allowed to set seed for further testing . Second generation transgenic plants ( T1 , heterozygous ) were tested for digoxin- or digoxigenin-dependent induction of luciferase expression . After 42 hr , we observed 30-50-fold induction of luciferase activity in digoxin-treated plants compared to the uninduced control ( Figure 6 ) . Both digoxin and digoxigenin were capable of inducing the biosensor . Digoxigenin-dependent luciferase induction was observed in multiple independent transgenic T1 lines ( Figure 6—figure supplement 1b ) , and a rising dose response to digoxigenin was observed in the transgenic plants ( Figure 6—figure supplement 1c ) . The specificity of the digoxigenin biosensor in plants parallels that in yeast cells ( Figure 6—figure supplement 1d ) . 10 . 7554/eLife . 10606 . 013Figure 6 . Application of biosensors in plants . ( a ) Activation of luciferase expression in transgenic Arabidopsis plants containing the G-DIG1-V biosensor in the absence ( left ) or presence ( right ) of 100 μM digoxin . Luciferase expression levels are false colored according to scale to the right . Relative luciferase units corresponding to 1 min of image pixel integration ( to avoid saturating pixels ) are shown above each individual plant . ( b ) Brightfield image of plants shown in a . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 01310 . 7554/eLife . 10606 . 014Figure 6—figure supplement 1 . Characterization of DIG biosensor in plants . ( a ) Test of DIG1 variants engineered for plant function in Arabidopsis protoplasts . Two transcriptional activation domains , VP16 ( V ) and VP64 ( VP64 ) , as well as two degrons , yeast MATα2 and Arabidopsis DREB2a , were added to DTF-1 ( G-DIG1 ) , and the proteins were constitutively expressed from the CaMV35S promoter . The Gal4-activated pUAS promoter controls expression of a luciferase reporter . Transformed protoplasts were incubated with digoxigenin at 0 , 100 , and 500 μM for 16 hr . ( b ) Digoxigenin-dependent activation of luciferase expression in three independent transgenic Arabidopsis lines . Plants were incubated in the absence ( Control ) or presence ( DIG ) of 100 μM digoxigenin for 42 hr and imaged . Quantification of luciferase expression is presented as mean relative luciferase units ± s . d . of ten plants . ( c ) Specificity of luciferase activation in transgenic Arabidopsis plants . All inducers were tested at 100 μM concentration . DIG , digoxigenin; DIGT , digitoxigenin; β-EST , β-Estradiol . Data are presented as mean fold activation relative to the control ± s . e . m . of ten technical replicates . ( d ) Digoxigenin dose response curve in transgenic Arabidopsis plants . Concentrations are expressed in micromolar . The data are presented as mean fold induction relative to the control ± s . e . m . of ten biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 014
In vivo biosensors for small molecules enable the regulation and detection of cellular responses to endogenous metabolites and exogenous chemicals . Here , we show that LBDs can be conditionally destabilized to create biosensors that function in yeast , mammalian cells , and plants , and we demonstrate the use of these biosensors in metabolic engineering and genome editing applications . While this method requires a high-affinity ligand-binding domain as a starting point , nearly all small molecules of interest have a natural protein interactor . Furthermore , the use of de novo-designed binders opens the possibility of generating biosensors for ligands with unsuitable or unknown binding proteins . By incorporating standard mutagenesis and screening , our method constitutes a simple platform for sensor development that can be applied to many areas of biotechnology . These sensors act either at the level of post-translational control over protein function or at the level of transcription ( Figure 7a ) , and they can be tuned by altering any of their components ( Figure 7b ) or by modifying efflux of the target ligand in the host organism . These tunable features should make the biosensors useful in many different cellular and environmental contexts . 10 . 7554/eLife . 10606 . 015Figure 7 . Schematic of biosensor platform . ( a ) Biosensors for small molecules are modularly constructed by replacing the LBD with proteins possessing altered substrate preferences . ( b ) Activity of the biosensor can be tuned by 1 ) introducing destabilizing mutations ( red Xs ) , 2 ) adding a degron , 3 ) altering the strength of the TAD or DNA binding affinity of the TF , 4 ) changes in the number of TF-binding sites or sequences , and 5 ) titrating 3-aminotriazole , an inhibitor of His3 . ( c ) Yeast provide a genetically tractable chassis for biosensor development before implementation in more complex eukaryotes , such as mammalian cells and plants . DOI: http://dx . doi . org/10 . 7554/eLife . 10606 . 015 Our results suggest a general mechanism of conditional stabilization for LBDs , allowing the rational development of biosensors for other targets . Furthermore , the portability of the mutations we identified suggests a structural basis for conferring conditional stability to the DIG0 LBD scaffold . Both the DIG0 LBD and Gal4 are homodimers , and the majority of the conditionally-stabilizing mutations are located at the dimer interfaces . A computational model of the Gal4-DIG0 complex indicates that the orientation of the two domains allows a homodimeric fusion to form ( Figure 2b ) . These results suggest an allosteric interplay between ligand binding and dimer formation: weakening of the dimer interface , in either the DIG0 or the Gal4 domain , is compensated by ligand binding . This LBD scaffold is derived from a member of the nuclear transport factor 2 family , a fold class that typically has a large dimer interface ( ~1200 Å2 ) that facilitates the large and open ligand-binding site ( ~600 Å2 ) . These protein folds are well suited for de novo design of other LBDs because of their large binding pocket and natural substrate diversity ( Todd et al . , 2002 ) . Exploiting dimer interfaces to modulate stability without impairing ligand binding may be a general mechanism to confer conditional stability on LBDs . This possibility is supported by the observation that interface mutations in DIG0 and Gal4 conferring digoxin-dependent stability lead to progesterone-dependent stability in a progesterone biosensor ( Figure 1—figure supplement 2a ) . A long-standing challenge in metabolic engineering is to rapidly detect and control how changes to the regulation and composition of biosynthetic pathways affect product titers . Transcriptional control by a product or intermediate ( Zhang et al . , 2012; Raman et al . , 2014; Tang and Cirino , 2011 ) and directed evolution of constituent pathway elements ( Agresti et al . , 2010; Alper et al . , 2005; Dietrich et al . , 2013 ) have emerged as promising strategies towards this goal . These approaches require high selectivity against intermediates ( Zhang and Keasling , 2011 ) , a feature demonstrated here that can be explicitly considered during the computational design and screening process . Our method allows biosensors to be generated that are highly selective for a small molecule , facilitating a simple directed evolution strategy without requiring prior structural or bioinformatic knowledge about the targeted enzyme ( s ) or pathway ( s ) . Because the biosensors are TF-based , sophisticated systems of optimizing metabolic output , such as dynamic control of gene expression ( Zhang et al . , 2012 ) and feedback-regulated genome evolution ( Chou and Keasling , 2013 ) , are possible . A reliance on the general principles of protein stability and ligand binding allows the development of biosensors that function in any organism with similar protein quality control machinery . Here , we engineered biosensors based on a designed scaffold derived from a bacterial protein . These biosensors required only minimal modifications to retain high levels of sensitivity when developed in yeast and deployed across mammalian and plant species , demonstrating unprecedented portability for biosensors . In some cases , these biosensors showed a greater dynamic range in mammalian cells relative to yeast , possibly due to larger cell volume and variations in protein degradation machinery . Further work using biosensors based on multiple different LBD scaffolds and introduced into diverse organisms will allow us to better understand the principles by which biosensor variability across hosts arises . Small molecule biosensors with the modularity incorporated here enable diverse cellular responses to a variety of exogenous and endogenous signals ( Banaszynski et al . , 2008 ) . Gene editing is an area that requires particularly tight coupling of cell response to activation signals . The CRISPR/Cas9 system provides a facile and robust genome-editing platform , but it can result in off-target genetic changes ( Fu et al . , 2013; Mali et al . , 2013; Pattanayak et al . , 2013 ) . Proposed solutions include optimizing guide RNA sequences ( Fu et al . , 2014; Cho et al . , 2014 ) , building chimeric Cas9 fusions requiring the presence of two Cas9 molecules in close proximity ( Mali et al . , 2013; Tsai et al . , 2014; Guilinger et al . , 2014; Ran et al . , 2013 ) , and regulating Cas9 activity by chemical or light-based inducers ( Dow et al . , 2015; Zetsche et al . , 2015; Polstein and Gersbach , 2015 ) . While small molecule inducers including doxycycline and rapamycin have been used , these molecules may confer leaky expression and cytotoxicity ( Xie et al . , 2008 ) . Thus , an expanded chemical repertoire is needed for tightly regulated gene editing and gene therapy applications . By exploiting the low background of the LBD-biosensors , we produced biosensor-Cas9 fusions with tightly controlled activation ( Figure 5e ) . This switch-like control over CRISPR/Cas9 activity could reduce background activity and off-target editing , a critical feature for safer gene therapies ( Mandal , et al . , 2014; Wu et al . , 2013; Schwank et al . , 2013 ) . Our biosensor design approach should have numerous applications in agriculture . For example , biosensors could be developed to enable plants to monitor the environment for pollutants , toxins or dangerous compounds . Coupling biosensors with a phytoremediation trait could enable plants to both sense a contaminant and activate a bioremediation gene circuit . When paired with an agronomic or biofuel trait , such biosensors could serve as triggers for bioproduction . In the transgenic Arabidopsis plants , we observed ligand-dependent activation in all cells , tissues and organs examined ( Figure 6 ) . The technology introduced here operates at either the transcriptional or post-translational level . These biosensors can be developed in yeast and readily transferred with minimal modification to other eukaryotic cell types , where they retain a high level of sensitivity ( Figure 7c ) . The generality of our approach arises from the universality of the transcriptional activation and protein degradation machinery across eukaryotes , together with the modularity and tunability of the constituent parts . These biosensors should find broad application , including improving metabolically engineered pathway flux and product titers , exerting ligand-dependent control over genome editing , and detecting exogenous small molecules or endogenous metabolites .
Biological replicates are defined as samples inoculated from distinct colonies . Growth media consisted of YPAD ( 10 g/L yeast extract , 20 g/L peptone , 40 mg/L adenine sulfate , 20 g/L glucose ) and SD media ( 1 . 7 g/L yeast nitrogen base without amino acids , 5 g/L ammonium sulfate , 20 g/L glucose and the appropriate amount of dropout base with amino acids [Clontech , Mountain View , CA] ) . The following selective agents were used when indicated: G418 ( 285 mg/L ) , pen/strep ( 100 U/mL penicillin and 100 μg/mL streptomycin ) . The DIG10 . 3 sequence ( Tinberg et al . , 2013 ) was cloned by Gibson assembly ( Gibson et al . , 2009 ) into a pUC19 plasmid containing yeast enhanced GFP ( yEGFP , UniProt ID B6UPG7 ) and a KanMX6 cassette flanked by 1000 and 500 bp upstream and downstream homology to the HO locus . The DIG10 . 3 sequence was randomized by error-prone PCR using a Genemorph II kit from Agilent Technologies . An aliquot containing 100 ng of target DNA ( 423 bp out of a 7 . 4 kb plasmid ) was mixed with 5 μL of 10X Mutazyme buffer , 1 μL of 40 mM dNTPS , 1 . 5 μL of 20 μM forward and reverse primer containing 90 bp overlap with the pUC19 plasmid ( oJF70 and oJF71 ) , and 1 μL of Mutazyme polymerase in 50 μL . The reaction mixture was subject to 30 cycles with Tm of 60°C and extension time of 1 min . Vector backbone was amplified using Q5 polymerase ( NEB , Ipswich , MA ) with oJF76 and oJF77 primers with Tm of 65°C and extension time of 350 s . Both PCR products were isolated by 1 . 5% agarose gel electrophoresis and the randomized target was inserted as a genetic fusion to yEGFP by Gibson assembly ( Gibson et al . , 2009 ) . Assemblies were pooled , washed by ethanol precipitation , and resuspended in 50 μL of dH2O , which was drop dialyzed ( EMD Millipore , Billerica , MA ) and electroporated into E . cloni supreme cells ( Lucigen , Middleton , WI ) . Sanger sequencing of 16 colonies showed a mutation rate of 0–7 mutations/kb . The library was expanded in culture and maxiprepped ( Qiagen , Valencia , CA ) to 500 μg/μl aliquots . 16 μg of library was drop dialyzed and electrotransformed into yeast strain Y7092 for homologous recombination into the HO locus . Integrants were selected by growth on YPAD solid media containing G418 followed by outgrowth in YPAD liquid media containing G418 . Libraries of DIG0-yEGFP and PRO0-yEGFP integrated into yeast strain Y7092 were subject to three rounds of fluorescence activated sorting in a BD FACSAria IIu . For the first round , cells were grown overnight to an OD600 of ~1 . 0 in YPAD containing steroid ( 500 μM digoxigenin or 50 μM progesterone ) , and cells showing the top 5% of fluorescence activation were collected and expanded overnight to an OD600 of ~1 . 0 in YPAD lacking steroid . In the second sort , cells displaying the lowest ~3% fluorescence activation were collected . Cells passing the second round were passaged overnight in YPAD containing steroid to an OD600 of ~1 . 0 and sorted once more for the upper 5% of fluorescence activation . The sorted libraries were expanded in YPAD liquid culture and plated on solid YPAD media . Ninety-six colonies from each library were clonally isolated and grown overnight in deep well plates containing 500 μL of YPAD . Candidates were diluted 1:50 into two deep well plates with SD-complete media: one plate supplemented with steroid and the other with DMSO vehicle . Cells were grown for another 4 h , and then diluted 1:3 into microtitre plates of 250 μL of the same media . Candidates were screened by analytical flow cytometry on a BD LSRFortessa cell analyzer . The forward scatter , side scatter , and yEGFP fluorescence ( 530 nm band pass filter ) were recorded for a minimum of 20 , 000 events . FlowJo X software was used to analyze the flow cytometry data . The fold activation was calculated by normalizing mean yEGFP fluorescence activation for each steroid to the mean yEGFP fluorescence in the DMSO only control . Highest induction candidates were subject to Sanger sequencing with primers flanking the LBD sequence . Reporter genes were cloned into the integrative plasmid pUG6 or the CEN plasmid pRS414 using the Gibson method ( Gibson et al . , 2009 ) . Each reporter ( either yEGFP or firefly luciferase ) was cloned to include a 5’ GAL1 promoter ( S . cerevisiae GAL1 ORF bases ( -455 ) - ( -5 ) ) and a 3’ CYC1 terminator . For integration , linearized PCR cassettes containing both the reporter and an adjacent KanMX antibiotic resistance cassette were generated using primers containing 50 bp flanking sequences of homology to the URA3 locus . Integrative PCR product was transformed into the yeast strain PJ69-4a using the Gietz method ( Gietz and Schiestl , 2007 ) to generate integrated reporter strains . G-DIG/PRO-V fusion constructs were prepared using the Gibson method ( Gibson et al . , 2009 ) . Constructs were cloned into the plasmid p416CYC ( p16C ) . Gal4 ( residues 1–93 , UniProt ID P04386 ) , DIG10 . 3 ( Tinberg et al . , 2013 ) , and VP16 ( residues 363-490 , UniProt ID P06492 ) PCR products for were amplified from their respective templates using Phusion high-fidelity polymerase ( NEB , Waltham , MA ) and standard PCR conditions ( 98°C 10 s , 60°C 20 s , 72°C 30 s; 30 cycles ) . The 8-residue linker sequence GGSGGSGG was used between Gal4 and DIG10 . 3 . PCR primers were purchased from Integrated DNA technologies and contained 24–30 5’ bases of homology to either neighboring fragments or plasmid . Clones containing an N-terminal degron were similarly cloned fusing residues 1–67 of Matα2 ( UniProt ID P0CY08 ) to the 5’- end of G-DIG-V . Plasmids were transformed into yeast using the Gietz method ( Gietz and Schiestl , 2007 ) , with transformants being plated on synthetic complete media lacking uracil ( SD -ura ) . Mutations were introduced into DIG10 . 3/pETCON ( Tinberg et al . , 2013 ) or the appropriate G-DIG/PRO-V construct using Kunkel mutagenesis ( Kunkel , 1985 ) . Oligos were ordered from Integrated DNA Technologies , Inc . For mutants constructed in pETCON/DIG10 . 3 , the mutagenized DIG10 . 3 gene was amplified by 30 cycles of PCR ( 98°C 10 s , 61°C 30 s , 72°C 15 s ) using Phusion high-fidelity polymerase ( NEB ) and 5’- and 3’- primers having homologous overlap with the DIG10 . 3-flanking regions in p16C-G-DIG-VP64 ( Gal4_DIG10 . 3_VP64_hr_fwd and Gal4_DIG10 . 3_VP64_hr_rev_rc ) . Genes were inserted into p16C-Gal4- ( HE ) -VP16 by Gibson assembly ( Gibson et al . , 2009 ) using vector digested with HindIII and EcoRI-HF . The gene for DIG10 . 3 Y34F/Y99F/Y101F was amplified from the appropriate DIG10 . 3/pETCON ( Tinberg et al . , 2013 ) construct by 30 cycles of PCR ( 98°C 10 s , 59°C 30 s , 72°C 15 s ) using Phusion high-fidelity polymerase ( NEB ) and 5’- and 3’- primers having homologous overlap with the DIG10 . 3-flanking regions in p16C-G-DIG-VP64 ( DIG_fwd and DIG_rev ) . Genes were inserted into p16C-GDVP16 by Gibson assembly ( Gibson et al . , 2009 ) using p16C-Gal4- ( HE ) -VP16 vector digested with HindIII and EcoRI-HF . A randomized G-DIG-V library was constructed by error-prone PCR using a Genemorph II kit from Agilent Technologies ( Santa Clara , CA ) . An aliquot containing 20 ng p16C GDVP16 , 20 ng p16C GDVP16 E83V , and 20 ng p16C Y36H was mixed with 5 μL of 10X Mutazyme buffer , 1 μL of 40 mM dNTPS , 1 . 5 μL of 20 μM forward and reverse primer containing 37- and 42-bp overlap with the p16C vector for homologous recombination , respectively ( GDV_ePCR_fwd and GDV_ePRC_rev ) , and 1 μL of Mutazyme polymerase in 50 μL . The reaction mixture was subjected to 30 cycles of PCR ( 95°C 30 s , 61°C 30 s , 72°C 80 s ) . Template plasmid was digested by adding 1 μL of DpnI to the reaction mixture and incubating for 3 hr at 37°C . Resulting PCR product was purified using a Quiagen PCR cleanup kit , and a second round of PCR was used to amplify enough DNA for transformation . Gene product was amplified by combining 100 ng of mutated template DNA with 2 . 5 μL of 10 μM primers ( GDV_ePCR_fwd and GDV_ePRC_rev ) , 10 μL of 5X Phusion buffer HF , 1 . 5 μL of DMSO , and 1 μL of Phusion high-fidelity polymerase ( NEB , Waltham , MA ) in 50 μL . Product was assembled by 30 cycles of PCR ( 98°C 10 s , 65°C 30 s , 72°C 35 s ) . Following confirmation of a single band at the correct molecular weight by 1% agarose gel electrophoresis , the PCR product was purified using a Quaigen PCR cleanup kit and eluted in ddH2O . Yeast strain PyE1 ΔPDR5 was transformed with 9 μg of amplified PCR library and 3 μg of p16C Gal4- ( HE ) -VP16 triply digested with SalI-HF , BamHI-HR , and EcoRI-HF using the method of Benatuil ( Benatuil et al . , 2010 ) , yielding ~106 transformants . Following transformation , cells were grown in 150 mL of SD -ura media . Sanger sequencing of 12 individual colonies revealed an error rate of ~1–6 mutations per gene . An error-prone library of G-DIG0/DIG1/DIG2/-V transformed into yeast strain PyE1 ΔPDR5 was subjected to three rounds of cell sorting using a Cytopeia ( BD Influx ) fluorescence activated cell sorter . For the first round , cells displaying high fluorescence in the presence of digoxin ( on-state ) were collected . Transformed cells were pelleted by centrifugation ( 4 min , 4000 rpm ) and resuspended to a final OD600 of 0 . 1 in 50 mL of SD -ura media , pen/step antibiotics , and 5 μM digoxin prepared as a 100 mM solution in DMSO . The library was incubated at 30°C for 9 hr and then sorted . Cells displaying the highest fluorescent values in the GFP channel were collected ( 1 , 747 , 058 cells collected of 32 , 067 , 013 analyzed; 5 . 5% ) , grown up at 30°C in SD -ura , and passaged twice before the next sort . For the second round of sorting , cells displaying low fluorescence in the absence of digoxin ( off-state ) were collected . Cells were pelleted by centrifugation ( 4 min , 4000 rpm ) and resuspended to a final OD600 of 0 . 1 in 50 mL of SD -ura media supplemented with pen/strep antibiotics . The library was incubated at 30°C for 8 hr and then sorted . Cells displaying low fluorescent values in the GFP channel were collected ( 1 , 849 , 137 cells collected of 22 , 290 , 327 analyzed; 11 . 1% ) , grown up at 30°C in SD -ura , and passaged twice before the next sort . For the last sorting round , cells displaying high fluorescence in the presence of digoxin ( on-state ) were collected . Cells were prepared as for the first sort . Cells displaying the highest fluorescent values in the GFP channel were collected ( 359 , 485 cells collected of 31 , 615 , 121 analyzed; 1 . 1% ) . After the third sort , a portion of cells were plated and grown at 30°C . Plasmids from 12 individual colonies were harvested using a Zymoprep Yeast miniprep II kit ( Zymo Research Corporation , Irvine , CA ) and the gene was amplified by 30 cycles of PCR ( 98°C 10 s , 52°C 30 s , 72°C 40 s ) using Phusion high-fidelity polymerase ( NEB ) with the T3 and T7 primers . Sanger sequencing ( Genewiz , Inc . , South Plainfield , NJ ) was used to sequence each clone in the forward ( T3 ) and reverse ( T7 ) directions . Of 12 sequenced clones from the library sorts , two showed significantly improved ( >2-fold ) response to DIG over the input clones ( clone 3 and clone 6 ) . Clone 3 contains the following mutations: Gal4_T44T ( silent ) , Gal4_L77F , DIG10 . 3_E5D , DIG10 . 3_E83V , DIG10 . 3_R108R ( silent ) , DIG10 . 3_L128P , DIG10 . 3_I137N , DIG10 . 3_S143G , and VP16_A44T . Clone 6 contains the following mutations: Gal4_R60S , Gal4_L84L ( silent ) , VP16_G17G ( silent ) , VP16_L48V , and VP16_H98H ( silent ) . To identify which mutations led to the observed changes in DIG response , variants of these clones with no silent mutations and each individual point mutant were constructed using Kunkel mutagenesis ( Kunkel , 1985 ) . Oligos were ordered from Integrated DNA Technologies , Inc . Sequence-confirmed plasmids were transformed into PyE1 ΔPDR5f and plated onto selective SD -ura media . Individual colonies were inoculated into liquid media , grown at 30°C , and passaged once . Cells were pelleted by centrifugation ( 4 min , 1700 × g ) and resuspended to a final OD660 of 0 . 1 in 1 mL of SD -ura media supplemented 50 μM DIG prepared as a 100 mM solution in DMSO . Following a 6 hr incubation at 30°C , cells were pelleted , resuspended in 200 μL of PBS , and cellular fluorescence was measured on an Accuri C6 flow cytometer using a 488 nm laser for excitation and a 575 nm band pass filter for emission . FlowJo software version 7 . 6 was used to analyze the flow cytometry data . The data are given as the mean yEGFP fluorescence of the single yeast population in the absence of DIG ( off-state ) and the mean yEGFP fluorescence of the higher fluorescing yeast population in the presence of DIG ( on-state ) . A model of the Gal4-DIG10 . 3 fusion was built using Rosetta Remodel ( Huang et al . , 2011 ) to assess whether the linker between Gal4 and the DIG LBD , which are both dimers , would allow for the formation of a dimer in the fusion construct . In the simulation , the Gal4 dimer was held fixed while the relative orientation of the DIG LBD monomers were sampled symmetrically using fragment insertion in the linker region . Constraints were added across the DIG LBD dimer interface to facilitate sampling . The lowest energy model satisfied the dimer constraints , indicating that a homodimer configuration of the fusion is possible . Yeast strain PyE1 transformed with p16C plasmids containing G-LBD-V variants were inoculated from colonies into SD –ura media supplemented and grown at 30°C overnight ( 16 h ) . 10 μL of the culture was resuspended into 490 μL of separately prepared media each containing a steroid of interest ( SD –ura media supplemented the steroid of interest and DMSO to a final concentration of 1% DMSO ) . Resuspended cultures were then incubated at 30°C for 8 hr . 125 μL of incubated culture was resuspended into 150 μL of fresh SD –ura media supplemented with the steroid of interest and DMSO to a final concentration of 1% . These cultures were then assayed by analytical flow cytometry on a BD LSRFortessa using a 488 nm laser for excitation . The forward scatter , side scatter , and yEGFP fluorescence ( 530 nm band pass filter ) were recorded for a minimum of 20 , 000 events . FlowJo X software was used to analyze the flow cytometry data . The fold activation was calculated by normalizing mean yEGFP fluorescence activation for each steroid to the mean yEGFP fluorescence in the DMSO only control . G-PRO0-V was assayed on a separate day from the other TF biosensors under identical conditions . Yeast strain PyE1 was transformed with p16C plasmids containing G-LBD-V variants were inoculated from colonies into SD –ura media and grown at 30°C overnight ( 16 hr ) . 5 μL of each strain was diluted into 490 μL of SD –ura media in 2 . 2 mL plates . Cells were incubated at 30°C for 8 hr . 5 μL of steroid was then added for a final concentration of 250 μM digoxin or 50 μM progesterone . For each time point , strains were diluted 1:3 into microtitre plates of 250 μL of the same media . Strains were screened by analytical flow cytometry on a BD LSRFortessa cell analyzer . The forward scatter , side scatter , and yEGFP fluorescence ( 530 nm band pass filter ) were recorded for a minimum of 20 , 000 events . FlowJo X software was used to analyze the flow cytometry data . The fold activation was calculated by normalizing mean yEGFP fluorescence activation for each time point to the mean yEGFP fluorescence at T = 0 hr . Yeast strains containing either a plasmid-borne or integrated luciferase reporter were transformed with p16C plasmids encoding TF-biosensors . Transformants were grown in triplicate overnight at 30°C in SD –ura media containing 2% glucose in sterile glass test tubes on a roller drum . After ~16 hr of growth , OD600 of each sample was measured and cultures were back diluted to OD600 = 0 . 2 in fresh SD –ura media containing steroid dissolved in DMSO or a DMSO control ( 1% DMSO final ) . Cultures were grown at 30°C on roller drum for 8 hr prior to taking readings . Measurement of luciferase activity was adapted from a previously reported protocol ( Leskinen et al . , 2003 ) . 100 μL of each culture was transferred to a 96-well white NUNC plate . 100 μL of 2 mM D-luciferin in 0 . 1 M sodium citrate ( pH 4 . 5 ) was added to each well of the plate and luminescence was measured on a Victor 3V after 5 min . Genomic deletions were introduced into the yeast strains PJ69-4a and PyE1 using the 50:50 method ( Horecka and Davis , 2014 ) . Briefly , forward and reverse primers were used to amplify an URA3 cassette by PCR . These primers generated a product containing two 50 bp sequences homologous to the 5’ and 3’ ends of the ORF at one end and a single 50 bp sequence homologous to the middle of the ORF at the other end . PCR products were transformed into yeast using the Gietz method ( Gietz and Schiestl , 2007 ) and integrants were selected on SD –ura plates . After integration at the correct locus was confirmed by a PCR screen , single integrants were grown for 2 days in YEP containing 2 . 5% ethanol and 2% glycerol . Each culture was plated on synthetic complete plates containing 5-fluoroorotic acid . Colonies were screened for deletion of the ORF and elimination of the Ura3 cassette by PCR and confirmed by Sanger sequencing . Yeast strains expressing the TF-biosensors and yEGFP reporter ( either genetically fused or able to be transcriptionally activated by the TAD ) were grown overnight at 30°C in SD –ura media for 12 hr . Following overnight growth , cells were pelleted by centrifugation ( 5 min , 5250 rpm ) and resuspended into 500 μL of SD –ura . 10 μL of the washed culture was resuspended into 490 μL of separately prepared media each containing a steroid of interest ( SD –ura media supplemented with the steroid of interest and DMSO to a final concentration of 1% DMSO ) . Steroids were tested at a concentration of 100 μM digoxin , 50 μM progesterone , 250 μM pregnenolone , 100 μM digitoxigenin , 100 μM beta-estradiol , and 100 μM hydrocortisone . Stock solutions of steroids were prepared as a 50 mM solution in DMSO . Resuspended cultures were then incubated at 30°C for 8 hr . 125 μL of incubated culture was resuspended into 150 μL of fresh SD –ura media supplemented the steroid of interest , and DMSO to a final concentration of 1% . These cultures were then assayed by analytical flow cytometry on a BD LSRFortessa using a 488 nm laser for excitation . The forward scatter , side scatter , and yEGFP fluorescence ( 530 nm band pass filter ) were recorded for a minimum of 20 , 000 events . FlowJo X software was used to analyze the flow cytometry data . The fold induction was calculated by normalizing mean yEGFP fluorescence activation for each steroid to the mean yEGFP fluorescence in the DMSO only control . Yeast strain PJ69-4a transformed with p16C plasmids containing degron-G-DIG-V variants were first inoculated from colonies into SD -ura media and grown at 30°C overnight ( 16 hr ) . 1 mL of each culture was pelleted by centrifugation ( 3000 rcf , 2 min ) , resuspended in 1 mL of fresh SD -ura and the OD660 was measured . Each culture was then diluted in SD -ura media to an OD660 = 0 . 2 and incubated at 30°C for 4–6 hr . 1 mL of each culture was pelleted and resuspended in sterile , distilled water and the OD660 measured again . Each transformant was then diluted to an OD660 = 0 . 1 . Four 1/10 serial dilutions of each culture were prepared in sterile water ( for a total of 5 solutions ) . 10 μL of each dilution was spotted in series onto several SD –ura –his agar plates containing 1 mM 3-aminotriazole and the indicated steroid . Steroid solutions were added to agar from 200x steroid solutions in DMSO ( 0 . 5% DMSO final in plates ) . The 3β-HSD ORF was synthesized as double-stranded DNA ( Integrated DNA Technologies , Inc . , Coralville , IA ) and amplified using primers oJF325 and oJF326 using KAPA HiFi under standard PCR conditions and digested with BsmBI to create plasmid pJF57 . 3β-HSD expression plasmids ( pJF76 through pJF87 ) were generated by digesting plasmid pJF57 along with corresponding plasmids from the Yeast Cloning Toolkit ( Lee et al . , 2015 ) with BsaI and assembled using the Golden Gate Assembly method ( Engler et al . , 2008 ) . The 3β-HSD sequence was randomized by error-prone PCR using a Genemorph II kit from Agilent Technologies . An aliquot containing 100 ng of target DNA was mixed with 5 μL of 10X Mutazyme buffer , 1 μL of 40 mM dNTPS , 1 . 5 μL of 20 μM forward and reverse primer containing 90-bp overlap with the 3β-HSD expression plasmids and 1 μL of Mutazyme polymerase in 50 μL . The reaction mixture was subject to 30 cycles with Tm of 60°C and extension time of 1 min . Vector backbone was amplified using KAPA HiFi polymerase with oJF387 and oJF389 ( pPAB1 ) or oJF387 and oJF389 ( pPOP6 ) with Tm of 65°C and extension time of 350 s . PCR products were isolated by 1 . 5% agarose gel electrophoresis and assembled using the Gibson method ( Gibson et al . , 2009 ) . Assemblies were pooled , washed by ethanol precipitation , and resuspended in 50 μL of dH2O , which was drop dialyzed ( Millipore ) and electroporated into E . cloni supreme cells ( Lucigen ) . Sanger sequencing of 16 colonies showed a mutation rate of 0–4 mutations/kb . The library was expanded in culture and maxiprepped ( Qiagen ) to 500 μg/μL aliquots . 16 μg of library was drop dialyzed and electrotransformed into yeast strain PyE1 . PyE1 transformed with libraries of 3β-HSD were seeded into 5 mL of SD –ura –leu media supplemented and grown at 30°C overnight ( 24 hr ) . Cultures were measured for OD600 , diluted to an OD600 of 0 . 0032 , and 100 μL was plated onto SD –ura –leu –his plates supplemented 35 mM 3-AT and either 50 μM pregnenolone or 0 . 5% DMSO . Production strains were inoculated from colonies into 5 mL SD –ura media and grown at 30°C overnight ( 16 hr ) . 1 mL of each culture was washed and resuspended into 50 mL of SD –ura with 250 μM of pregnenolone and grown at 30°C for 76 hr . OD600 measurements were recorded for each culture before pelleting by centrifugation . Cells were lysed by glass bead disruption , and lysates and growth media were extracted separately with heptane . Extractions were analyzed by GC/MS . The K562 cell line was obtained from the ATCC . The cell line was not authenticated and not tested for mycoplasma contamination . For each TF-biosensor , 1 μg of the PiggyBac construct along with 400 ng of transposase were nucleofected into K562 cells using the Lonza Nucleofection system as per manufacturer settings . Two days post-transfection , cells underwent puromycin selection ( 2 μg/mL ) for at least eight additional days to allow for unintegrated plasmid to dilute out and ensure that all cells contained the integrated construct . An aliquot of 100 , 000 cells of each integrated population were then cultured with 25 μM of progesterone , 1 μM of digoxigenin , or no small molecule . Forty-eight hours after small molecule addition , cells were analyzed by flow cytometry using a BD Biosciences Fortessa system . Mean EGFP fluorescence of the populations was compared . The PiggyBac transposase system was employed to integrate biosensor constructs into K562 cells . Vector PB713B-1 ( Systems Biosciences , Mountain View , CA ) was used a backbone . Briefly , this backbone was digested with NotI and HpaI and G-LBD-V , Gal4BS-E1b-EGFP ( EGFP; enhanced GFP ref or UniProt ID A0A076FL24 ) , and sEF1-Puromycin were cloned in . Gal4BS represents four copies of the binding sequence . For hCas9 , the PiggyBac system was also employed , but the biosensors were directly fused on the N-terminus of Cas9 and were under the control of the CAGGS promoter . Cas9 from S . pyogenes was used . Construct integration was carried out as for the Cas9 experiments for EGFP assays , except that the constructs were integrated into K562 containing a broken EGFP reporter construct . Introduction of an engineered nuclease along with a donor oligonucleotide can correct the EGFP and produce fluorescent cells . Upon successful integration ( ~10 days after initial transfection ) , 500 , 000 cells were nucleofected with 500 ng of guide RNA ( sgRNA ) and 2 μg of donor oligonucleotide . Nucleofected cells were then collected with 200 μL of media and 50 μL aliquots were added to wells containing 950 μL of media . Each nucleofection was split into four separate wells containing 1 μM of digoxigenin , 25 μM of progesterone , or no small molecule . Forty-eight hours later , cells were analyzed using flow cytometry and the percentage of EGFP positive cells was determined . Digoxin transcriptional activators were initially tested in a transient expression assay using Arabidopsis protoplasts according previously described methods ( Yoo et al . , 2007 ) , with some modifications . Briefly , protoplasts were prepared from 6-week old Arabidopsis leaves excised from plants grown in short days . Cellulase Onozuka R-10 and Macerozyme R-10 ( Yakult Honsha , Inc . , Japan ) in buffered solution were used to remove the cell wall . After two washes in W5 solution , protoplasts were re-suspended in MMg solution at 2 × 105 cells/mL for transformation . Approximately 104 protoplasts were mixed with 5 μg of plasmid DNA and PEG4000 at a final concentration of 20% , and allowed to incubate at room temperature for 30 min . The transformation reaction was stopped by the addition of two volumes of W5 solution , and after centrifugation , protoplasts were re-suspended in 200 μL of WI solution ( at 5 × 105/mL ) and plated in a 96-well plate . Digoxigenin ( Sigma-Aldrich , St . Louis , MO ) was added to the wells , and protoplasts were incubated overnight at room temperature in the dark , with slight shaking ( 40 rpm ) . For luciferase imaging , protoplasts were lysed using Passive Lysis Buffer ( Promega , Madison , WI ) and mixed with LARII substrate ( Dual-Luciferase Reporter Assay System , Promega , Madison , WI ) . Luciferase luminescence was collected by a Stanford Photonics XR/MEGA-10 ICCD Camera and quantified using Piper Control ( v . 2 . 6 . 17 ) software . G-DIG1-V was recoded to function as a ligand-dependent transcriptional activator in plants . Specifically , an Arabidopsis thaliana codon optimized protein degradation sequence from the yeast MATα2 gene was fused in frame in between the Gal4 DBD and the DIG1 LBD . The resulting gene sequence was codon-optimized for optimal expression in Arabidopsis thaliana plants and cloned downstream of a plant-functional CaMV35S promoter to drive constitutive expression in plants , and upstream of the octopine synthase ( ocs ) transcriptional terminator sequence . To quantify the transcriptional activation function of DIG10 . 3 , the luciferase gene from Photinus pyralis ( firefly ) was placed downstream of a synthetic plant promoter consisting of five tandem copies of a Gal4 Upstream Activating Sequence ( UAS ) fused to the minimal ( -46 ) CaMV35S promoter sequence . Transcription of luciferase is terminated by the E9 terminator sequence . These sequences were cloned into a pJ204 plasmid and used for transient expression assays in Arabidopsis protoplasts . After confirmation of function in transient tests , the digoxin biosensor genetic circuit was transferred to pCAMBIA 2300 and was stably transformed into Arabidopsis thaliana ecotype Columbia plants using a standard Agrobacterium floral dip method ( Clough and Bent , 1998 ) . Transgenic plants were selected in MS media ( Murashige and Skoog , 1962 ) containing 100 mg/L kanamycin . Transgenic plants expressing the digoxin biosensor genetic circuit were tested for digoxigenin-induced luciferase expression by placing 14–16 day-old plants in liquid MS ( - sucrose ) media supplemented with 0 . 1 mM digoxigenin in 24-well plates , and incubated in a growth chamber at 24°C , 100 μE . m2 . s-1 light . Luciferase expression was measured by imaging plants with a Stanford Photonics XR/MEGA-10 ICCD Camera , after spraying luciferin and dark adapting plants for 30 min . Luciferase expression was quantified using Piper Control ( v . 2 . 6 . 17 ) software . Plants from line KJM58-10 were used to test for the specificity of induction by incubating plants , as described above , in 0 . 1 mM digoxigenin , 0 . 1 mM digitoxigenin , and 0 . 02 mM β-estradiol . All chemicals were obtained from Sigma-Aldrich ( St . Louis , MO ) . | Small molecules play essential roles in organisms , and so methods to sense these molecules within living cells could have wide-ranging uses in both biology and biotechnology . However , current methods for making new “biosensors” are limited and only a narrow range of small molecules can be detected . One approach to biosensor design in yeast and other eukaryotic organisms uses proteins called ligand-binding domains , which bind to small molecules . Here , Feng , Jester , Tinberg , Mandell et al . have developed a new method to make biosensors from ligand-binding domains that could , in principle , be applied to any target small molecule . The new method involves taking a ligand-binding domain that is either engineered or occurs in nature and linking it to something that can be readily detected , such as a protein that fluoresces or that controls gene expression . This combined biosensor protein is then engineered , via mutations , such that it is unstable unless it binds to the small molecule . This means that , in the absence of the small molecule , these proteins are destroyed inside living cells . However , the binding of a target molecule to one of these proteins protects it from degradation , which allows the signal to be detected . Feng , Jester , Tinberg , Mandell et al . use this method to create biosensors for a human hormone called progesterone and a drug called digoxin , which is used to treat heart disease . Further experiments used the biosensors to optimize the production of progesterone in yeast and to regulate the activity of a gene editing protein called Cas9 in human cells . The biosensors can be also used to produce long-term environmental sensors in plant cells . This approach makes it possible to produce a wide variety of biosensors for different organisms . The next step is to continue to explore the ability of various proteins to be converted into biosensors , and to find out how easy it is to transfer a biosensor produced in one species to another . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | A general strategy to construct small molecule biosensors in eukaryotes |
Ecdysis ( moulting ) is the defining character of Ecdysoza ( arthropods , nematodes and related phyla ) . Despite superficial similarities , the signalling cascade underlying moulting differs between Panarthropoda and the remaining ecdysozoans . Here , we reconstruct the evolution of major components of the ecdysis pathway . Its key elements evolved much earlier than previously thought and are present in non-moulting lophotrochozoans and deuterostomes . Eclosion hormone ( EH ) and bursicon originated prior to the cnidarian-bilaterian split , whereas ecdysis-triggering hormone ( ETH ) and crustacean cardioactive peptide ( CCAP ) evolved in the bilaterian last common ancestor ( LCA ) . Identification of EH , CCAP and bursicon in Onychophora and EH , ETH and CCAP in Tardigrada suggests that the pathway was present in the panarthropod LCA . Trunk , an ancient extracellular signalling molecule and a well-established paralog of the insect peptide prothoracicotropic hormone ( PTTH ) , is present in the non-bilaterian ctenophore Mnemiopsis leidyi . This constitutes the first case of a ctenophore signalling peptide with homology to a neuropeptide .
Ecdysis or moulting , which describes the process of shedding the outer integument , the cuticle , is a defining feature of Ecdysozoa ( arthropods , tardigrades , onychophorans , nematodes and related phyla ) ( Aguinaldo et al . , 1997; Schmidt-Rhaesa et al . , 1998; de Rosa et al . , 1999; Dunn et al . , 2008; Telford et al . , 2008 ) . Despite superficial similarities of the ‘moulting behaviour’ within Ecdysozoa , the neuroendocrine components underlying this process remain elusive for the majority of the ecdysozoans outside of Arthropoda . This includes well-established model organisms such as the nematode Caenorhabditis elegans , for which the gene regulatory network responsible for ecdysis remains to be fully resolved ( Frand et al . , 2005; reviewed by Page et al . , 2014 and Lažetić and Fay , 2017 ) . In arthropods , ecdysis can be divided into three distinct stages , pre-ecdysis , ecdysis and post-ecdysis . Each of these stages correlates with major behavioural , molecular and cellular changes and encompasses a series of specific muscular contractions controlled by a cascade of hormones and neuropeptides ( Truman , 2005 ) . Studies in insects have revealed that the major components of this peptidergic signalling pathway are ecdysis-triggering hormone ( ETH ) , eclosion hormone ( EH ) , crustacean cardioactive peptide ( CCAP ) and bursicon ( Gammie and Truman , 1997a; Gammie and Truman , 1997b; Zitnan et al . , 1999; Clark et al . , 2004; Kim et al . , 2006a; Kim et al . , 2006b; Arakane et al . , 2008; Lee et al . , 2013 ) . The process begins with the release of prothoracicotropic hormone ( PTTH ) from neurohemal organs . PTTH initiates a signalling cascade that results in the biosynthesis of ecdysteroids ( i . e . , steroid hormones synthesised from ingested cholesterol ) , including ecdysone ( E ) and 20-hydroxyecdysone ( 20E ) ( Figure 1 ) . The decline of the ecdysone titre due to the ecdysone-inactivating enzyme cytochrome P450 protein Cyp18a1 ( Guittard et al . , 2011; reviewed by Rewitz et al . , 2013 ) triggers the release of ETH that , in turn , causes the release of EH . These two hormones mutually enhance one another in a positive feedback loop to control and regulate pre-ecdysis behaviour ( Figure 1 ) . With the ensuing release of CCAP , caused by EH , pre-ecdysis ceases and the ecdysis motor program is initiated . Finally , bursicon responds to the increasing levels of CCAP and initiates post-ecdysis behaviour and cuticle tanning ( Figure 1 ) . Comparative biochemical , genomic and transcriptomic analyses revealed that ecdysteroids and the required genes responsible for their biosynthesis are present outside of Ecdysozoa , showing that some key molecular players of moulting predate the origin of Ecdysozoa ( Mendis et al . , 1984; Nolte et al . , 1986; Garcia et al . , 1989; Barker et al . , 1990; Schumann et al . , 2018 ) . Such integrative and comparative analyses have so far not been conducted on the major components of the peptidergic signalling system underlying moulting . To fill this gap in knowledge , we explored the distribution of PTTH , ETH , EH , CCAP and bursicon ligand-receptor pairs across Metazoa .
We show that key peptidergic components of the arthropod ecdysis pathway emerged prior to the protostome-deuterostome split , and thus considerably earlier than commonly assumed . EH , CCAP and the bursicon signalling systems are more widespread among non-moulting animals than previously appreciated . The presence of the eth-receptor ortholog in ecdysozoans , lophotrochozoans and deuterostomes , in combination with the restriction of its known ligand to insects , arachnids and tardigrades , suggests a scenario in which promiscuous ligand/receptor relationships can lead to novel signalling interactions that provide new opportunities for natural selection to generate novel functions ( Figure 2B ) . The identification of the near complete suite of the peptidergic arthropod ecdysis pathway components in Onychophora and Tardigrada strongly suggests that the entire pathway was at least functional in the last common ancestor of Panarthropoda and maybe as early as in the ur-ecdysozoan ( Figures 2B and 4B ) . However , considering the crucial role of the ETH and bursicon signalling systems in insect moulting , together with the apparent secondary loss of ETH in Onychophora and bursicon in Tardigrada ( Figure 4B ) , the consequences of harbouring only the partial set of the ecdysis signalling genes should be the focus of future assessments . Independent recruitment of novel peptidergic components into insect ecdysis has been shown ( cf . Kim et al . , 2004; Kim et al . , 2006a; Kim et al . , 2006b; extensively reviewed by Zitnan et al . , 1996 ) , illustrating the evolutionary plasticity of this signalling pathway and calling for more detailed functional investigations into the role of individual components during moulting of the various ecdysozoan lineages .
To obtain a comprehensive sampling across Metazoa , ecdysozoan , deuterostome and non-bilaterian protein-coding sequence ( CDS ) databases were downloaded from publicly available sites and combined with previous lophotrochozoan transcriptomes ( see De Oliveira et al . , 2019 ) . The acoel transcriptomic data were pre-processed and assembled as described in De Oliveira et al . ( 2019 ) . The databases include representatives from the following phyla: Porifera , Ctenophora , Cnidaria , Placozoa , Xenacoelomorpha , Echinodermata , Hemichordata , Chordata , Annelida , Brachiopoda , Ectoprocta , Entoprocta , Gastrotricha , Mollusca , Nemertea , Phoronida , Platyhelminthes , Rotifera , Arthropoda , Tardigrada , Onychophora and Nematoda . The choanoflagellate Monosiga brevicollis was used as outgroup . Supplementary file 1 summarises the databases and the publicly available repositories from which they were obtained . Sequence read archive ( SRA ) accession numbers for xenacoelomorph databases are also shown . Sensitive probabilistic iterative similarity searches based on profile hidden Markov models ( HMMs ) were performed with jackhmmer ( Johnson et al . , 2010 ) against the respective metazoan and choanoflagellate databases . Insect eh , eth ccap , ptth and bursicon orthologs were retrieved from NCBI ( National Center for Biotechnology Information ) and their respective receptors from Vogel et al . ( 2013 ) . These sequences were used as queries in the similarity searches . The searches were performed under the default parameters using varying e-value thresholds ( 1 to 1e-06 ) controlled by the options –E and –domE , as defined in jackhmmer . The best hits found in the metazoan and choanoflagellate databases were stored in fasta format and used in the subsequent analyses . EH , ETH , CCAP , PTTH and bursicon ligand candidates retrieved from the metazoan and choanoflagellate databases were used as input , together with their respective insect orthologs , in the program clans ( Frickey and Lupas , 2004 ) under different e-value thresholds ( 0 . 1 to 1e-06 ) and blast programs , that is blastp or psiblast ( Camacho et al . , 2009 ) . Singleton sequences ( isolated unconnected sequences ) were excluded from the map . To further improve the orthology assessment , multiple sequence alignments were performed with mafft ( Katoh and Standley , 2013 ) and the presence of shared conserved amino acid regions and residues were investigated with aliview ( Larsson , 2014 ) . The final 3D maps were collapsed into 2D after the clustering for easier visualisation . Putative EH , ETH , CCAP , PTTH and bursicon receptor candidates retrieved from the metazoan and choanoflagellate databases were aligned with mafft together with their respective orthologs , when found , and subsequently trimmed with BMGE software under the following parameters: –h 1 –b 1 –m BLOSUM30 –t AA ( Criscuolo and Gribaldo , 2010 ) . Outgroups for the phylogenetic analyses were defined according to Vogel et al . ( 2013 ) . Phylogenetic analyses were performed using RAxML ( Stamatakis , 2014 ) , PhyML ( Guindon et al . , 2010 ) and mrbayes ( Ronquist et al . , 2012 ) softwares using the appropriate best-fit model of amino acid substitution . RaxML was executed under default parameters and rapid bootstrap . PhyML was executed under the default parameters and an optimised starting tree ( -o tlr option ) . The number of bootstrap values was set to 1 . 000 in RaxML and PhyML and the number of generations used in mrbayes was determined using a convergence diagnostic . All runs in mrbayes were performed with the samplefreq and relative burn-in defined as 1000 and 25% , respectively . The three final phylogenetic trees obtained for each of the four different receptors were visualised and combined with TreeGraph2 ( Stöver and Müller , 2010 ) . All data generated in the course of this study are included in this article ( Figure 2—source datas 1–5 and Figure 3—source data 1 ) . The accession numbers for the publicly available datasets used in this work are available in Supplementary file 1 . The 3D cluster peptide maps can be visualised and manipulated using the program clans ( Frickey and Lupas , 2004 ) ; see ftp://ftp . tuebingen . mpg . de/pub/protevo/CLANS/ ) . The multiple sequence alignment files can be viewed with aliview ( Larsson , 2014 ) . The phylogenetic tree files can be viewed using Figtree ( http://tree . bio . ed . ac . uk/software/figtree/ ) or TreeGraph2 ( Stöver and Müller , 2010 ) . | Animals such as insects , crabs and spiders belong to one of the most species-rich animal groups , called the arthropods . These animals have exoskeletons , which are hard , external coverings that support their bodies . Arthropods shed their exoskeletons as they grow , a process called ecdysis or moulting , and this behaviour is controlled by a set of hormones and small protein-like molecules called neuropeptides that allow communication between neurons . Other animals , such as roundworms , also moult; and together with arthropods they are classified into a group called the Ecdysozoa . Since moulting is a common behaviour in ecdysozoans , it was previously assumed that its signalling components had evolved in the common ancestor of roundworms and arthropods , although differences in the moulting machinery between both groups exist . Here , De Oliveira et al . investigate the evolutionary origins of the arthropod moulting machinery and find that some of the hormones and neuropeptides involved appeared long before the arthropods themselves . Database searches showed that important hormones and neuropeptides involved in arthropod moulting can be found in diverse animal groups , such as jellyfish , molluscs and starfish , confirming that these molecules evolved before the last common ancestor of roundworms and arthropods . These animals must therefore use the hormones and neuropeptides in many processes unrelated to moulting . De Oliveira et al . also found that roundworms have lost most of these molecules , and that moulting in these animals must be driven by a different complement of hormones and neuropeptides . These results invite research into the role of moulting hormones and neuropeptides in animals outside the Ecdysozoa . They also show that signalling pathways and the processes they regulate are highly adaptable: two animals can use the same hormone in entirely different processes , but conversely , the same behaviour may be regulated by different molecules depending on the animal . This means that the evolution of a process and the evolution of its regulation can be decoupled , a finding that has important implications for the study of signalling pathways and their evolution . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"short",
"report",
"genetics",
"and",
"genomics"
] | 2019 | Ancient origins of arthropod moulting pathway components |
Neuropeptides are essential for the regulation of appetite . Here we show that neuropeptides could regulate feeding in mutants that lack neurotransmission from the motor neurons that stimulate feeding muscles . We identified nlp-24 by an RNAi screen of 115 neuropeptide genes , testing whether they affected growth . NLP-24 peptides have a conserved YGGXX sequence , similar to mammalian opioid neuropeptides . In addition , morphine and naloxone respectively stimulated and inhibited feeding in starved worms , but not in worms lacking NPR-17 , which encodes a protein with sequence similarity to opioid receptors . Opioid agonists activated heterologously expressed NPR-17 , as did at least one NLP-24 peptide . Worms lacking the ASI neurons , which express npr-17 , did not response to naloxone . Thus , we suggest that Caenorhabditis elegans has an endogenous opioid system that acts through NPR-17 , and that opioids regulate feeding via ASI neurons . Together , these results suggest C . elegans may be the first genetically tractable invertebrate opioid model .
Animal feeding behavior has been well studied because of its health relevance as well as its fundamental physiological and ecological importance . Caenorhabditis elegans is a useful model system for studying the genetic control of feeding ( Avery , 2012 ) . In C . elegans , feeding is dependent on the pumping motion of the pharynx , which results in the ingestion of food , and whose rate is an easily quantified measure of feeding . The pathways that control feeding are at least partly conserved from worms to mammals . Control of animal feeding is not simple . Multiple peripheral and central signals , neuropeptides among them , modulate food intake . For example , α-melanocyte-stimulating hormone , a peptide which activates MC3 and MC4 melanocortin receptors and inhibits food intake ( Meister , 2007 ) , putative satiety signal peptide YY3–36 ( PYY3–36 ) ( Batterham et al . , 2002 ) , and cholecystokinin ( Gibbs et al . , 1973 ) participate in the termination of meals in mammals . Galanin is a peptide that stimulates feeding in satiated rats after intraventricular or intrahypothalamic injection . Opioids are also involved in controlling food intake , and they have many other biological effects ( e . g . , analgesia , drug addiction , and immune response ) . For example , the opioid agonist morphine increases food intake in mammals while the opioid blocker , naloxone , significantly decreases food intake ( Martin et al . , 1963 ) . The opioid system is composed of μ-opioid receptors ( MORs ) , δ-opioid receptors ( DORs ) , and κ-opioid receptors ( KORs ) , and endogenous ligands for these receptors . Enkephalins , dynorphins , and β-endorphin peptides are produced by proteolytic cleavage of large protein precursors known as preproenkephalin , preprodynorphin , and proopiomelanocortin ( POMC ) , respectively . These peptides form the opioid family . All opioid peptides share a common N-terminal YGGF signature sequence , which interacts with opioid receptors ( Holtzman , 1974; Akil et al . , 1998 ) . Many researchers have shown that the opilioid system modulates food intake . β-endorphin stimulates food intake when administrated directly into the VMH ( ventromedial hypothalamus ) ( Grandison and Guidotti , 1977 ) . Selective agonists for the μ receptor ( DAMGO ) , the δ receptor ( DADLE ) , and the κ receptor ( U50448 ) also increase food intake ( Tepperman and Hirst , 1983; Gosnell et al . , 1986; Jackson and Cooper , 1986 ) . Furthermore , β-endorphin levels are associated with overeating in genetically obese mice and fa/fa rats ( Margules et al . , 1978 ) . There is abundant pharmacological evidence that the opioid system is present and controls feeding in both invertebrates and vertebrates ( Harrison et al . , 1994 ) . A simple invertebrate model system would be a useful starting point for understanding more complex ones . Furthermore , it might shed light on the evolutionary origin of the opioid system . However , no opioid system has yet been molecularly identified in invertebrates . Peptide hormones and peptide neurotransmitters are well conserved from humans to invertebrates ( Kimura et al . , 1997; Janssen et al . , 2008b ) . Among the most conserved are those that signal nutritional state and control feeding and metabolism—insulin is the best-known example . C . elegans has 115 neuropeptide genes encoding over 250 distinct neuropeptides ( Li and Kim , 2008 ) . However , the functions of most neuropeptides remain unknown . Here we show that , like humans , C . elegans has opioid peptide signals that regulate feeding . We found that morphine and naloxone , respectively , stimulate and inhibit feeding in starved wild-type worms , but not in worms lacking the G-protein coupled receptor NPR-17 . Our genetic screens identified NLP-24-derived peptides as possible endogenous opioid ligands on which feeding motions that occur in the absence of food may depend and NPR-17 as the opioid receptor on which they act . To our knowledge this is the first genetically tractable invertebrate opioid model .
In well-fed C . elegans , pharyngeal pumping , the motion of the feeding muscles , is mostly determined by MC motor neuron activity in the pharynx ( Raizen et al . , 1995 ) . The food stimulates the MC neuron , which fires at the beginning of each pump , triggering a muscle action potential that makes the muscle contract . But a starved worm continually pumps at a low rate even in the absence of food . This pumping does not require MC neurons . This MC-independent pumping makes worms viable without MC , although they pump slowly , grow slowly , and are small and pale . We were curious how worms continue to pump without MC . In mammals , neuropeptides modulate food intake . Over 250 neuropeptides encoded by 115 genes have been identified in C . elegans ( Li and Kim , 2008 ) . Most of their functions are still unknown . We hypothesized that MC-independent pumping might depend on neuropeptide signaling . To test our hypothesis , we made eat-2; egl-3 mutants , in which signals from both MC and neuropeptides are disrupted . eat-2 encodes an acetylcholine receptor subunit specific to the MC → muscle synapse; thus the mutants are functionally MC-minus ( McKay et al . , 2004 ) . egl-3 encodes a protease necessary for neuropeptide processing , and thus the mutant lacks most neuropeptides ( Husson et al . , 2006 ) . egl-3 mutants pumped at the wild-type rate . However , eat-2; egl-3 mutants pumped only 1/3 as fast as eat-2 mutants ( Figure 1A ) . This result indicated that neuropeptides regulate pumping in the absence of MC neuron activity . 10 . 7554/eLife . 06683 . 003Figure 1 . MC neurons and neuropeptides redundantly control pumping rate . ( A ) Pumping rate in eat-2 , egl-3 and eat-2; egl-3 mutants . egl-3 reduced pumping in eat-2 mutants . ***p < 0 . 001 . ( ANOVA + Tukey tests ) . ( B ) nlp-3 suppressed pumping and nlp-24 stimulated pumping in eat-2 mutants . ***Different from eat-2 , p < 0 . 001 ( ANOVA + Dunnett tests ) . In this and all figures , errors bars represent standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 003 To identify neuropeptides that regulate pumping in the absence of MC function , we performed an RNAi ( RNA interference ) screen of neuropeptide genes . Because RNAi works less efficiently in the nervous system than in other tissues , we first tested egl-3 RNAi efficiency in two genetic backgrounds that enhance RNAi in neurons , rrf-3 and eri-1; lin-15b ( Simmer et al . , 2002; Wang et al . , 2005 ) . eat-2; egl-3 mutants grow slowly compared to eat-2 due to their slow pumping rate , showing that we could use growth rate as an indicator for the pumping defect . eat-2; eri-1; lin-15b mutants grew slowly after egl-3 RNAi , whereas rrf-3; eat-2 did not . This suggested that eat-2; eri-1; lin-15b mutants are more sensitive to RNAi in the cells that produce neuropeptides relevant to MC-independent pumping . Thus we used eat-2; eri-1; lin-15b for the RNAi screen . From the RNAi screen of 115 neuropeptide genes we found several that changed the growth rate of eat-2; eri-1; lin-15b mutants . RNAi of five NLP ( Neuropeptide-Like Proteins ) family genes ( nlp-2 , nlp-12 , nlp-14 , nlp-24 , and nlp-34 ) and 2 INS ( INSulin related ) family genes ( ins-34 and ins-39 ) decreased growth rates compared to control ( Supplementary file 1 ) . Among these seven genes RNAi of nlp-24 delayed growth most . To validate the RNAi results , we constructed eat-2; nlp-24 double mutants and measured their pumping rate . As expected from RNAi results , the eat-2; nlp-24 mutants had decreased pumping rates compared to eat-2 ( Figure 1B ) . Interestingly , RNAi of nlp-3 and nlp-13 improved the growth rate of eat-2 mutants , suggesting that peptides encoded by these genes might antagonize MC-independent pumping . Indeed , the eat-2; nlp-3 mutants have increased pumping rates compared to eat-2 . These results demonstrate that nlp-24 increases pumping and nlp-3 suppresses pumping in the absence of MC neuron activity . Because eat-2; nlp-24 mutants do not pump as slowly as eat-2; egl-3 mutants ( Figure 1A , B ) , we suggest that there may be other neuropeptides that act together with nlp-24 to stimulate pumping in the absence of MC neuron activity . Based on these results we hypothesize that nlp-24 and nlp-3 regulate pumping rate in the absence of food when MC is inactive . In mammals , one major mode of neuropeptide regulation is control of mRNA levels ( Brady et al . , 1990; Challis et al . , 2003 ) . Thus we measured the mRNA levels of nlp-3 and nlp-24 in well-fed and starved worms . nlp-3 mRNA levels were unaffected , but the nlp-24 increased after a 3 hr starvation , suggesting it plays a role in regulation of pumping during starvation ( Figure 2A ) . Wild type worms initially pump slowly when taken off food , but increase pumping rate gradually during several hours of starvation ( Avery and Horvitz , 1990 ) . In confirmation , nlp-24 mutants showed a reduced pumping rate compared to wild type after 1 hr of starvation , and this defect was rescued by a transgene that encoded the wild-type copy of nlp-24 gene under control of its own promoter . Moreover , nlp-24 overexpression further increased the pumping rate in starved worms compared to non-transgenic worms ( Figure 2B ) . These results indicate that nlp-24 stimulates pumping during starvation . egl-3 mutants , which lack most active neuropeptides , also had decreased pumping during starvation ( Figure 2—figure supplement 1 ) , showing that neuropeptides contribute to pumping during starvation . 10 . 7554/eLife . 06683 . 004Figure 2 . nlp-24 stimulates pumping in starved worms . ( A ) nlp-3 and nlp-24 mRNA levels in well-fed and starved worms . qRT-PCR was performed to assess mRNA expression . Normalized with ama-1 . ***Different from well-fed , p < 0 . 001 ( two-way ANOVA on Ct with Bonferroni correction ) . ( B ) Pumping rate after 1 hr starvation . nlp-24 stimulated pumping during starvation . ***p < 0 . 001 , N . S . not significant ( ANOVA + Tukey tests ) . ( C ) NLP-24 and human endogenous opioid peptides . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 00410 . 7554/eLife . 06683 . 005Figure 2—figure supplement 1 . Neuropeptides stimulate pumping in starvation . egl-3 mutants , lacking most active neuropeptides , pumped less than wild type worms during starvation . ( ***p < 0 . 001 , Student's t-test . ) DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 00510 . 7554/eLife . 06683 . 006Figure 2—figure supplement 2 . Phylogenetic analysis of opioid receptors . Branch length is proportional to evolutionary distance . The diagram was constructed using Phylogeny . fr ( phylogeny . lirmm . fr/ ) ( Dereeper et al . , 2008 ) . Accession numbers of sequences used in the alignment: NP_001186948 . 1 ( nociceptin receptor Homo sapiens ) , NP_000902 . 3 ( delta-type opioid receptor ( DOR-1 ) Homo sapiens ) , NP_000903 . 2 ( kappa-type opioid receptor ( KOR-1 ) Homo sapiens ) , NP_000905 . 3 ( mu-type opioid receptor ( MOR-1 ) Homo sapiens ) , NP_498743 . 2 ( NPR-17 Caenorhabditis elegans ) , NP_490815 . 2 ( NPR-23 Caenorhabditis elegans ) , NP_497125 . 2 ( NPR-30 Caenorhabditis elegans ) , NP_499038 . 1 ( NPR-29 Caenorhabditis elegans ) , XP_001900553 . 1 ( ORL1-like opioid receptor Brugia malayi ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 00610 . 7554/eLife . 06683 . 007Figure 2—figure supplement 3 . nlp-3 pumping rate after 1 hr starvation . nlp-3 did not change pumping rate in starved worms ( N . S . not significant , Student's t-test . ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 00710 . 7554/eLife . 06683 . 008Figure 2—figure supplement 4 . Well-fed nlp-24 mutant pumping rate . nlp-24 mutants pumped at a rate similar to wild type on food . ( N . S . not significant , Student's t-test . ) DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 008 NLP-24 is predicted to be processed into peptides contain YGGY sequences either internally or at the N-terminus ( Nathoo et al . , 2001; Janssen et al . , 2008a ) ( Figure 2C ) . Comparing the sequence of NLP-24 peptides with that of human neuropeptides , we noted that the YGGY sequences are similar to the conserved YGGF sequence motif of endogenous opioids in humans ( Snyder and Pasternak , 2003 ) . Thus we hypothesize that nlp-24 potentially encodes endogenous opioids in C . elegans . Although no C . elegans opioid signaling system has been reported , a study identified NPR-17 , a G protein coupled receptor , as most closely related to a predicted Brugia malayi ORL1 opioid receptor-like protein with 68% identity ( Harris et al . , 2010 ) . They found that nlp-3 , a neuropeptide gene , and npr-17 regulated avoidance behavior , suggesting NLP-3 and NPR-17 together could mediate a signal to control pain in C . elegans ( Harris et al . , 2010 ) . In BLAST searches , we verified their finding further; NPR-17 is similar to human nociceptin receptor ( 28% identity ) , MOR ( 23% identity ) , DOR ( 23% identity ) and to the KOR ( 24% identity ) ( Figure 2—figure supplement 2 ) . Together these results suggested that NLP-24 peptides and NPR-17 receptor constitute an endogenous opioid system in C . elegans . Because NLP-24 stimulated C . elegans pumping under starvation conditions and because previous work and sequence homology suggested NPR-17 could be an opioid receptor , we tested whether opiates affect pumping , and whether NPR-17 mediates these effects . First , we treated starved wild type and npr-17 mutants with opioid agonist morphine and blocker naloxone and measured the pumping rate . As predicted , morphine stimulated pumping whereas naloxone suppressed pumping in wild type . Drugs get into intact worms inefficiently , both because the wild-type cuticle is poorly permeable ( Partridge et al . , 2008 ) and because foreign chemicals are actively pumped out ( Broeks et al . , 1996; Ardelli and Prichard , 2013 ) . For instance , although 1 μM serotonin maximally stimulates pumping in dissected pharynxes ( Niacaris and Avery , 2003 ) , roughly 10 mM is necessary to produce similar effects in intact animals ( Horvitz et al . , 1982; Hobson et al . , 2006; Raizen et al . , 2012 ) . Thus we were not surprised that high concentrations of morphine ( 0 . 5 mM ) and naloxone ( 10 mM ) were necessary to produce these effects . Specificity was demonstrated by the observation that these effects were completely abolished by npr-17 mutations , showing that NPR-17 mediated the effects , suggesting it could be an opioid receptor in C . elegans ( Figure 3A , B ) . We tested two different alleles ( tm3210 and tm3225 ) of npr-17 and both alleles gave the same results ( Figure 3B and Figure 3—figure supplement 1 ) . Taken together these results indicate that opioids stimulate C . elegans pumping and that NPR-17 might be a worm opioid receptor . 10 . 7554/eLife . 06683 . 009Figure 3 . Opioids control pumping through npr-17 . ( A ) Effect of morphine on pumping . ( B ) Effect of naloxone on pumping rate . Adult hermaphrodites were starved for 1 hr and tested . Morphine stimulates pumping and naloxone inhibits it in wild-type worms; neither drug affects npr-17 ( tm3210 ) mutants . ***Different from 0 mM , p < 0 . 001; N . S . not significantly different from 0 mM ( two-way ANOVA , concentration effect ) . The effects of morphine and naloxone on wild-type are significantly different from their effects on npr-17 , p < 0 . 001 ( two-way ANOVA , genotype × concentration interaction ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 00910 . 7554/eLife . 06683 . 010Figure 3—figure supplement 1 . Effect of naloxone on pumping rate in npr-17 ( tm3225 ) mutant worms . Adult hermaphrodites were starved for 1 hr and tested . Naloxone did not affect npr-17 ( tm3210 ) mutants . ***Different from 0 mM , p < 0 . 001; N . S . not significantly different from 0 mM ( two-way ANOVA , concentration effect ) . The effect of naloxone on wild-type is significantly different from its effect on npr-17 ( tm3225 ) , p < 0 . 001 ( two-way ANOVA , genotype × concentration interaction ) . Wild-type data are the same as in Figure 3B . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 01010 . 7554/eLife . 06683 . 011Figure 3—figure supplement 2 . Naloxone reduces the pumping rate of eat-2 mutants . Naloxone reduced the pumping rate in eat-2 mutants , which lack MC → muscle neurotransmission . ( ***p < 0 . 001 , Student's t-test . ) DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 011 To further test whether npr-17 mediates nlp-24 action , we overexpressed nlp-24 in npr-17 mutants . nlp-24 overexpression in the wild type background caused faster pumping under starvation conditions as shown above . However , nlp-24 overexpressed in the npr-17 background did not increase the pumping rate ( Figure 4A ) . nlp-24; npr-17 double mutants also pumped at the same rate as npr-17 mutants ( Figure 4B ) . These results indicate that NPR-17 is downstream of NLP-24 and are consistent with the hypothesis that NPR-17 is a receptor for NLP-24-derived peptides to regulate pumping during starvation . 10 . 7554/eLife . 06683 . 012Figure 4 . nlp-24 regulates pumping through npr-17 . ( A ) Overexpressed nlp-24 in npr-17 mutants did not stimulate pumping . ( B ) In the npr-17 background , nlp-24 mutations have no effect on pumping rate in starved worms . ( N . S . not significant , Student's t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 012 Next we examined whether NLP-24 peptides indeed function as opioids and NPR-17 as an opioid receptor using heterologous expression in HEK293 cells . We coexpressed NPR-17 in HEK293 cells with the promiscuous G protein Gα15 and administered synthetic NLP-24 peptides . In this heterologous system , only one of these peptides , GPYGYGamide , activated NPR-17 ( Figure 5A ) , suggested that NLP-24 encodes at least one NPR-17 ligand . We also tested agonists specific for mammalian opioid receptor subtypes: loperamide for the MORs ( DeHaven-Hudkins et al . , 1999 ) , SB205607 for the DORs ( Fujii et al . , 2004 ) , and U69593 for the KORs ( Towett et al . , 2006 ) . NPR-17 was also activated by loperamide and U69593 ( Figure 5A ) . GPYGYGamide , loperamide , and U69593 activated NPR-17 in a dose-dependent manner , and these responses were abrogated by naloxone ( Figure 5B ) . These results indicate that NPR-17 is an opioid receptor and NLP-24 peptide GPYGYGamide is a C . elegans opioid . 10 . 7554/eLife . 06683 . 013Figure 5 . Activities of NLP-24 peptides and opioid agonists on NPR-17 . Opioid agonists and NLP-24 peptide-mediated changes in intracellular calcium were measured with the calcium detector Fluo-4 . ( A ) Calcium responses in human embryonic kidney 293 ( HEK-293 ) cells transfected with NPR-17 and Gα15 . All compounds were tested at 1 μM . Loperamide ( μ agonist ) , SB205607 ( δ agonist ) , U69593 ( κ agonist ) and GPYGYGamide activated NPR-17 . Different from 0 , *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 , two-way ANOVA + sequential Bonferroni correction . ( B ) Dose-response curves of HEK-293 cells transfected with NPR-17 and Gα15 . Loperamide , U69593 and NLP-24 peptide GPYGYGamide induced NPR-17 activation in a dose dependent manner , and these responses were suppressed by opioid blocker naloxone ( 10 μM ) . Responses were normalized to 1 μM loperamide . The loperamide effect is significant at p < 0 . 001 , and the effects of U69593 and GPYGYGamide at p < 0 . 01 , two-way ANOVA + sequential Bonferroni correction . Pooling the results from A and B , the effect of GPYGYGamide is significant at p < 0 . 001 ( Stouffer's weighted Z-score test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 013 We then tested whether NLP-24 peptides can activate human opioid receptors . We individually coexpressed each human opioid receptor in HEK293 cell with Gα15 . We tested an agonist specific for each receptor subtype in addition to NLP-24 peptides . Mu-type opioid receptor ( MOR-1 ) and kappa-type opioid receptor ( KOR-1 ) responded only to YGGYGamide peptide , while delta-type opioid receptor ( DOR-1 ) did not respond to any NLP-24 peptide ( Figure 6 ) . Thus C . elegans has a peptide that can activate human opioid receptors , predicted to be produced from the same precursor that produces at least one ligand for the C . elegans opioid receptor NPR-17 , which can also be activated by vertebrate receptor-specific opioid agonists . These results suggested that C . elegans and vertebrates share a conserved opioid system . 10 . 7554/eLife . 06683 . 014Figure 6 . Activities of NLP-24 peptides and agonists on MOR-1 , DOR-1 and KOR-1 . Opioid agonist and NLP-24 peptide-mediated changes in intracellular calcium were measured by the calcium detector Fluo-4 . Agonists and peptides were tested at 1 μM . ( A ) Calcium responses in HEK-293 cell transfected with MOR-1 and Gα15 . ( B ) Calcium responses in HEK-293 cell transfected with DOR-1 and Gα15 . ( C ) Calcium responses in HEK-293 cell transfected with KOR-1 and Gα15 . ***p < 0 . 001 , **p < 0 . 01 , *p < 0 . 05 . ( Two-way ANOVA with Bonferroni correction . ) ( D ) Dose-response curves of HEK-293 cells transfected with MOR-1 or KOR-1 and Gα15 . NLP-24 peptide YGGYGamide activated MOR-1and KOR-1 , and these responses were suppressed by the opioid blocker naloxone ( 10 μM ) . MOR-1 responses were normalized to 1 μM loperamide and KOR-1 to 1 μM U69593 . The effects of YGGYGamide on MOR-1 and KOR-1 are significant at p < 0 . 001 ( two-way ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 014 npr-17 was shown to be expressed in ASI neurons ( Harris et al . , 2010 ) , a pair of head sensory neurons that regulate multiple food-related effects such as quiescence , dauer formation , and caloric-restriction-induced life span increase ( Inoue and Thomas , 2000; Bishop and Guarente , 2007; Gallagher et al . , 2013b ) . Because the pumping rate regulated by nlp-24 is also food-related and the receptor NPR-17 is known to be expressed in ASI , we tested whether ASI neurons are important for opioid signaling . When we treated worms whose ASI neurons were genetically ablated with naloxone , there was no naloxone-induced reduction in pumping ( Figure 7A ) . Thus the opioid blocker controlled pumping through ASI neurons . Next , to examine whether ASI neurons are the site of action of NPR-17 , we expressed it in ASI using the ASI specific promoter , gpa-4 . Expression of NPR-17 only in the ASI neurons completely restored the pumping reduction induced by naloxone treatment ( Figure 7B and Figure 7—figure supplement 1 ) . npr-17 is also expressed in intestine . We therefore also tested the effect of naloxone on worms engineered to express NPR-17 only in the intestine ( using the ges-1 promoter ) . These worms were not rescued ( Figure 7B and Figure 7—figure supplement 1 ) . Thus NPR-17 in ASI mediated opioid signaling to regulate pumping during starvation . 10 . 7554/eLife . 06683 . 015Figure 7 . ASI neurons are required for opioid-mediated feeding control . ( A ) Effect of naloxone on pumping . Worms genetically engineered to lack ASI neurons were not affected . ASI neurons were genetically ablated by the recCaspase method ( Chelur and Chalfie , 2007; Beverly et al . , 2011 ) ( ASI− ) ***Different from 0 mM , p < 0 . 001; N . S . not significantly different from 0 mM ( two-way ANOVA , concentration effect ) . The effect of naloxone on wild-type is significantly different from its effect on npr-17 , p < 0 . 001 ( two-way ANOVA , genotype × concentration interaction ) . ( B ) npr-17 expression in ASI neurons is required for opioid-mediated feeding control . npr-17 mutants were not affected by naloxone , but this phenotype was rescued with by expression of npr-17 under control of the ASI-specific promoter gpa-4 . ***Different from 0 mM , p < 0 . 001 , N . S . , not significantly different from 0 mM ( two-way ANOVA , concentration effect ) . The interaction between genotype and concentration is significant at p < 0 . 001 for Pges-1::npr-17 rescued vs WT , but not Pnpr-17::npr-17 or Pgpa-4::npr-17 rescued . The interaction between genotype and concentration is significant at p < 0 . 001 for Pnpr-17::npr-17 and Pgpa-4::npr-17 rescued vs npr-17 , but not Pges-1::npr-17 rescued . ( C ) Chronic effect of naloxone on pumping . Worms grown to adulthood with or without 10 mM naloxone plate were then tested with or without 10 mM naloxone . ***p < 0 . 001; N . S . not significantly different . The stimulatory effect of taking naloxone away from worms grown in the presence of naloxone is significant at p < 0 . 05 . ( ANOVA + Tukey tests ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 01510 . 7554/eLife . 06683 . 016Figure 7—figure supplement 1 . npr-17 in ASI neurons are required for opioid-mediated feeding control . npr-17 ( tm3225 ) mutants were not affected by naloxone . This phenotype was rescued by npr-17 expression under the ASI specific promoter , gpa-4 . ***Different from 0 mM , p < 0 . 001 , N . S . , not significantly different from 0 mM ( two-way ANOVA , concentration effect ) . The interaction between genotype and concentration is significant at p < 0 . 001 for Pges-1::npr-17 rescued vs WT , but not Pnpr-17::npr-17 or Pgpa-4::npr-17 rescued . The interaction between genotype and concentration is significant at p < 0 . 001 for Pnpr-17::npr-17 and Pgpa-4::npr-17 rescued vs npr-17 , but not Pges-1::npr-17 rescued . Wild-type and npr-17 ( tm3225 ) data reproduced from Figure 3B and Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 01610 . 7554/eLife . 06683 . 017Figure 7—figure supplement 2 . nlp-24 expression . ( A ) nlp-24::GFP transcriptional reporter expression . nlp-24 was expressed in hypodermis . ( B ) nlp-24::mCherry translational reporter expression . NLP-24::GFP was found in coelomocytes ( white arrows ) . ( C ) NLP-24::SL2::GFP translational reporter expression . NLP-24 was expressed in ASI neurons and intestine . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 01710 . 7554/eLife . 06683 . 018Figure 7—figure supplement 3 . Starvation induces nlp-24 expression in intestine . NLP-24::SL2::GFP expression is increased by starvation . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 018 While the insensitivity of npr-17 mutants to naloxone is consistent with NPR-17 being the naloxone receptor , the sign of the effect was initially surprising . In the absence of drug , npr-17 mutants pumped at the same rate as wild type , while they pumped faster than wild-type worms in the presence of naloxone . This would suggest that naloxone is an agonist rather than a blocker , but this is inconsistent with the known effect of naloxone of opioid receptors , as well as the measured effect on heterologously expressed NPR-17 ( Figure 5B ) . An alternate hypothesis is that npr-17 worms , having lacked the receptor all their lives , had adapted to its absence . One of these adaptations was an increase in the basal pumping rate . In wild-type worms , in contrast , NPR-17 had always been around supporting basal pumping . When we suddenly removed this support by adding naloxone , the pumping rate went down . This hypothesis predicts that the chronic effects of naloxone on wild type should resemble those of the mutation . We thus grew wild-type worms to adulthood on naloxone , then measured the pumping rate after starvation either with or without naloxone . As predicted , wild-type worms grown and tested in the presence of naloxone pumped at the same rate as worms grown and tested without ( Figure 7C ) . Furthermore , after chronic naloxone treatment worms tested in its absence showed a withdrawal effect: they pumped slightly faster . These results are consistent with the hypothesis that , in vivo as in vitro , naloxone acts as an NPR-17 blocker . npr-17 is expressed in ASI neurons and nlp-24 is also expressed in ASI neurons ( Figure 7—figure supplement 2C ) ( Nathoo et al . , 2001 ) . This suggests that NLP-24 peptides may be secreted by and act on ASI neurons in an autocrine manner . We also suspect that NLP-24 is secreted by other cells and acts hormonally on ASI neurons , because , NLP-24::GFP expression is increased in intestine after starvation ( Figure 7—figure supplement 3 ) . An NLP-24::mCherry fusion protein accumulates in coelomocytes ( Figure 7—figure supplement 2B ) , specialized cells for endocytosis and degradation of secreted proteins ( Fares and Grant , 2002 ) , which have been found to accumulate other secreted peptide signals ( Sieburth et al . , 2007; Lee and Ashrafi , 2008 ) . Our result is thus consistent with the idea that NLP-24 is a precursor for secreted peptides , supporting the idea that NLP-24 is secreted and sensed by NPR-17 in ASI neurons . Together our results demonstrate that C . elegans has an endogenous opioid system , which modulates pumping during starvation . But what is the purpose of pumping in the absence of food ? Following a suggestion by David Raizen ( personal communication ) , we suggest that this pumping helps worms discover food in the environment , even when their external senses are unable to reliably detect it . According to this hypothesis , opioids help worms survive starvation by increasing the probability of catching food . In essence , food-independent pumping is ( or facilitates ) exploration . This is consistent with the finding that in many animals , including C . elegans , starvation increases exploration ( Dallman et al . , 1999; Torres et al . , 2002 ) . This hypothesis suggested that , in addition to increasing pumping to enhance the chance of catching food , opioids might increase locomotion to enhance the chance of finding food . C . elegans locomotion is tightly related to nutritional status ( Shtonda and Avery , 2006; Ben Arous et al . , 2009 ) . There are three potentially distinct locomotive states: roaming , dwelling , and quiescence ( Fujiwara et al . , 2002; You et al . , 2008 ) . Roaming is the food-seeking behavior; worms roam when the food is scarce . We therefore tested the effect of nlp-24 mutation on locomotion in starved worms . Wild type spent 93% of the time roaming , but in nlp-24 mutants this was reduced to 69% ( Figure 8 ) . An nlp-24 transgene rescued the defect . Based on these results , we suggest that by increasing pumping and increasing exploration opioids help worms survive starvation . 10 . 7554/eLife . 06683 . 019Figure 8 . nlp-24 stimulates roaming behavior in starved worms . Behavioral states of adult worms on plates without food . A single worm was starved for 30 min , then recorded for 1 hr . Locomotion was analyzed as described previously ( Gallagher et al . , 2013 ) . Starvation increased roaming behavior in wild type . nlp-24 mutants had reduced roaming behavior compared to wild type , and an nlp-24 transgene rescued this phenotype . N = 22 worms for wild type , 27 for nlp-24 mutants , and 19 for nlp-24 rescued worms . All nlp-24 state probabilities are significantly different from the corresponding probabilities for wild type and the rescued strain , p < 0 . 001 , with the exception of the Q state for nlp-24 vs wild type , for which p < 0 . 01 . There are no significant differences between wild type and nlp-24 rescued . ( Mann–Whitney U-test with sequential Bonferroni correction ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 019
We hope as a result of these studies to establish C . elegans as an invertebrate genetic model for the study of opioid signaling . Our results confirm previous indications that opioids have a long evolutionary history . Several researchers have found evidence that the opioid system might exist in invertebrates: peptides encoding opioids or opioid receptors were detected , and drugs such as naloxone that act on opioid receptors induced biological responses ( Harrison et al . , 1994 ) . We confirm these results in C . elegans and add a crucial control: a mutant lacking the receptor fails to respond . Furthermore , like vertebrate opioids , worm opioids have a role in feeding behavior . nlp-24 encodes endogenous opioids and npr-17 an opioid receptor , and together they act to regulate feeding behavior . Opioids also participate in the immune system and pain mechanisms in vertebrates ( McCarthy et al . , 2001 ) . nlp-24 and npr-17 may have related functions in C . elegans . nlp-24 expression is down regulated by a pathogenic bacterium , Pseudomonas aeruginosa , and a fungus , Drechmeria coniospora ( Troemel et al . , 2006; Pujol et al . , 2008 ) . npr-17 mediates ASH-mediated aversive behavior ( Harris et al . , 2010 ) , a C . elegans model for pain . Harris et al . ( 2010 ) found that nlp-3 affects ASH-mediated aversion upstream of NPR-17 , and suggested therefore that NPR-17 might be an NLP-3 peptide receptor . We too found that nlp-3 mutation affects MC-independent pumping , although it had no effect on starved wild-type worms ( Figure 2—figure supplement 3 ) . While we cannot exclude the possibility that NLP-3 peptides are NPR-17 ligands , Harris et al . 's results are equally well explained by a model in which NLP-3 peptides affect NPR-17 indirectly by affecting the release of NLP-24 peptides . We showed that opioids stimulate pumping during starvation . We started with the hypothesis that neuropeptides regulate pumping in the absence of MC activity . Thus , we performed experiments in starved worms . Opioid agonists and antagonists modulated pumping in starved worms , and the opioid mutant nlp-24 had decreased pumping in starvation . In the presence of food , in contrast , both nlp-24 mutants ( Figure 2—figure supplement 4 ) and egl-3 mutants ( Figure 1A ) showed the same pumping rate as wild type . Nutritional state affected opioid activity: nlp-24 expression was increased by starvation ( Figure 2A ) . Likewise , food deprivation and feeding changes endogenous opioid activity in mammals . In rats , food deprivation induces complex changes in the level of brain endogenous opioids , especially in the hypothalamus ( Vaswani and Tejwani , 1986 ) . In addition , opioid signaling may be involved in altering the hedonic taste/palatability of food in mammals ( Bodnar , 2004 ) . Why does C . elegans need opioid signaling to control feeding during starvation ? In a well-fed worm , pumping depends on MC motor neurons , which are active only in the presence of food . But opioids modulated pumping independent of MC neurons: naloxone decreased pumping in eat-2 mutants , which lack MC neuromuscular transmission ( McKay et al . , 2004 ) ( Figure 3—figure supplement 2 ) . We speculate that , like a baby whose first response to any new object is to put it in its mouth , the worm cannot be sure if there is food around without taking some in . A well-fed worm avoids this dangerous test , but a starving worm continually pumps at a low rate even in the absence of food in order to detect any that might appear . Consistent with this , we found that opioids motivate worms to seek food by stimulating locomotion as well as pumping . Thus opioids make worms willing to search for food despite the high risk . Animals live in dynamic environments where food availability is variable . For survival in this environment , food-seeking behavior is important . Opioids help worms find food through risky pumping and exploration behaviors . Thus endogenous opioids may help worms survive in dynamic environments . Opioids stimulated pumping through ASI neurons ( Figure 7 ) . ASI neurons are activated by nutrition and release DAF-7 TGF-β when the environment is favorable . daf-7 expression in ASI neurons regulates fat accumulation and food intake ( Greer et al . , 2008 ) . ASI neurons also stimulate roaming behavior ( Flavell et al . , 2013 ) . These results suggest that ASI neurons are important for modulating energy balance and feeding behavior . In mammals opioid receptors are expressed in the hypothalamus , which plays a central role in the control of food intake and the regulation of energy . Injection of opioid agonists or antagonists into the hypothalamus modulates food intake ( Jenck et al . , 1987; Li et al . , 2006; Naleid et al . , 2007 ) . We suggest that opioids modulate C . elegans feeding through ASI neurons , similar to the way vertebrate opioids modulate feeding through the hypothalamus . Which pharyngeal neurons are involved in opioid signaling ? Our lab did several suppressor screens with slow pumping . From these screens , the most commonly mutated gene was slo-1 . slo-1 mutations enhance neurotransmitter release ( Wang et al . , 2001 ) . This enhancement is particularly striking for M4 and M5 , the motor neurons that synapse on TB ( terminal bulb ) muscle ( Albertson and Thomson , 1976; Chiang et al . , 2006 ) . We also screened for suppressors of the slow pumping and growth phenotypes of eat-2; egl-3 mutants . 7/15 suppressors were new slo-1 mutations . These results are consistent with the possibility that M4 might be involved in the opioid signaling response . We do not know yet how ASI neurons communicate with pharyngeal neurons . The pharyngeal and extrapharyngeal nervous systems are connected only by a bilateral pair of gap junctions between the extrapharyngeal RIP neurons and the pharyngeal I1 neurons ( Albertson and Thomson , 1976 ) , and I1s are not necessary for MC-independent pumping ( Avery and Horvitz , 1989 ) . This , together with the fact that ASI secretes several different peptides , suggests that communication is likely to be humoral . Heterologously expressed NPR-17 was activated by specific MOR-1 and KOR-1 agonists . These results suggested that NPR-17 might be more similar to μ and κ type receptors than δ . We also tested pumping with these agonists . Among the three , loperamide induced nose muscle contraction similar to that caused by fluoxetine ( Choy and Thomas , 1999 ) , so that we could not measure the pumping rate . U69593 increased pumping in wild-type worms but not npr-17 mutants ( Figure 9 ) . These results suggest that NPR-17 might be similar to the KOR but do not exclude similarity to μ as well . Vertebrate opioid receptors are about 60% identical to each other ( Chen et al . , 1993 ) . In BLAST results , NPR-17 showed 23% identity with MOR-1 and DOR-1 and 24% identity with KOR-1 . The three vertebrate receptors almost certainly diverged from each other well after any common ancestor of nematodes and vertebrates , so NPR-17 is probably an ortholog of all three . Clearly , NPR-17 is not a close homolog of any particular vertebrate opioid receptor . Rather , we suggest that NPR-17 and the vertebrate receptors collectively derive from a common ancestor in which the outlines of opioid system function had already appeared . Further , NPR-17 appears to be functionally closer to vertebrate opioid receptors than any other molecularly characterized invertebrate receptor is . 10 . 7554/eLife . 06683 . 020Figure 9 . Opioid agonist effects in C . elegans . The δ agonist SB205607 did not affect pumping . κ agonist U69594 stimulated pumping in starved wild-type worms but not npr-17 mutants . **p < 0 . 01 . N . S . not significant . ( Two-way ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06683 . 020 We originally thought that NLP-24 might be related to opioids because it encodes peptides containing YGGY , similar to the YGGF sequence common to vertebrate opioids . In fact , of the six predicted NLP-24 peptides , YGGYGamide alone activated MOR-1 and KOR-1 expressed in human embryonic kidney 293 ( HEK-293 ) cells . But , to our surprise , YGGYGamide did not detectably activate NPR-17 in the same system . In fact , the only peptide that worked was GPYGYGamide . By itself , this result might cast doubt on the idea that NLP-24 and NPR-17 are components of the C . elegans opioid system . The extreme version of this doubt would argue that NLP-24 , like POMC , is a precursor for multiple peptides that act on distinct unrelated receptors , and that it is essentially coincidence that the NPR-17 ligand GPYGYGamide shares a precursor with the opioid YGGYGamide . But this idea fails to explain multiple lines of data suggesting NPR-17 is a bona fide opioid receptor , namely: ( 1 ) Heterologously expressed NPR-17 responds to opioid receptor agonists and the blocker naloxone . ( 2 ) Naloxone blocks the response of NPR-17 to GPYGYGamide . ( 3 ) npr-17 is necessary for the behavioral response of worms to opioid agonists and naloxone . ( 4 ) NPR-17 has sequence similarity to opioid receptors . A more parsimonious explanation of the discrepancy is that in the HEK-293 expression system , where we asked a single npr-17 product to work entirely through vertebrate proteins , we failed to completely reconstitute its in vivo activity . Of course , these ideas are not mutually exclusive . It may be that in the worm some form of NPR-17 does respond to YGGY-containing NLP-24 products , and that NLP-24 produces products that act on other receptors as well . Our study confirms with molecular genetic data the previous pharmacological evidence that an invertebrate homolog of the vertebrate opioid system exists and has related functions . In the future , we hope this will enable the discovery of new molecular pathways through which opioids act and perhaps help to find ways to mitigate the side-effects that limit the usefulness of opiate drugs .
Worms were cultured routinely on NGMSR plates ( Avery , 1993 ) . All worms were maintained at 20°C on Escherichia coli strain HB101 unless indicated otherwise . The wild-type strain was C . elegans variant Bristol , strain N2 . The following mutant strains were used in this study: DA465 eat-2 ( ad465 ) II , GR1328 egl-3 ( nr2090 ) V , KP1873 egl-3 ( nu349 ) V , DA2434 eat-2 ( ad465 ) II; egl-3 ( nu349 ) V , DA2435 eat-2 ( ad465 ) II; egl-3 ( nr2090 ) V , DA2467 eat-2 ( ad465 ) II; nlp-3 ( tm3023 ) X , DA2456 nlp-24 ( tm2105 ) V , DA2466 eat-2 ( ad465 ) II; nlp-24 ( tm2105 ) V , DA2457 npr-17 ( tm3210 ) III , DA2458 npr-17 ( tm3225 ) III , DA2561 adEx2561[npr-17::GFP Punc-122::RFP] , DA2582 npr-17 ( tm3210 ) III; adEx2561[Pnpr-17::npr-17::GFP Punc-122::RFP] , DA2583 npr-17 ( tm3225 ) III; adEx2561[Pnpr-17::npr-17::GFP Punc-122::RFP] , DA2586 npr-17 ( tm3210 ) III; adEx2586[Pgpa-4::npr-17::SL2::RFP Punc-122::RFP] , DA2587 npr-17 ( tm3210 ) III; adEx2587[Pges-1::npr-17::SL2::RFP Punc-122::RFP] , DA2588 npr-17 ( tm3225 ) III; adEx2588[Pgpa-4::npr-17::SL2::RFP Punc-122::RFP] , DA2589 npr-17 ( tm3225 ) III; adEx2589[Pges-1::npr-17::SL2::RFP Punc-122::RFP] , DA2596 npr-17 ( tm3210 ) III; adEx2596[Pnlp-24::nlp-24::GFP Punc-122::RFP] , DA2597 npr-17 ( tm3225 ) III; adEx2597[Pnlp-24::nlp-24::GFP Punc-122::RFP] , DA2557 adEx2557[Pnlp-24::nlp-24::mCherry Punc-122::GFP] , DA2590 adEx2590[Pnlp-24::nlp-24::GFP Punc-122::RFP] , DA2591 adEx2591[Pnlp-24::nlp-24::GFP Punc-122::RFP] , DA2592 adEx2592[Pnlp-24::GFP Punc-122::RFP] , DA2593 nlp-24 ( tm2105 ) V; adEx2593[Pnlp-24::nlp-24::GFP Punc-122::RFP] , PY7505 oyIs84[Pgpa-4::TU813 Pgcy-27::TU814 Pgcy-27::eGFP Punc-122::DsRed] ( Beverly et al . , 2011 ) PY7005 was gift from Piali Sengupta . Most DNA constructs were made using overlap-extension PCR ( Hobert , 2002 ) . Briefly , to fuse two or more individual PCR products ( PfuUltra High-Fidelity DNA Polymerase , stratagene , La Jolla , CA ) , we amplified with oligonucleotides that included 5′ extensions complementary to the fusion target . Primers for making these constructs are listed in Supplementary file 2 . A transcriptional reporter for nlp-24 was made by fusing the promoter of nlp-24 ( −2262 to +1 ) to the coding sequence of GFP amplified from pPD95 . 77 ( Fire et al . , 1990 ) . For an nlp-24 translational reporter ( Pnlp-24::nlp-24::GFP ) , the genomic sequence of nlp-24 ( −2760 to +316 ) was fused in frame to GFP followed by the unc-54 3′ UTR from pPD95 . 77 . For nlp-24 rescue and overexpression , we used an nlp-24 translational reporter . To make NLP-24 fused with mCherry , the genomic sequence of nlp-24 ( −2760 to +316 ) was fused in frame to mCherry followed by the unc-54 3′ UTR from pAV1997 ( Miedel et al . , 2012 ) . To make NLP-24::SL2::GFP , we used a pJG7-psm-SL2-GFP vector ( a gift from Dr Cori Bargmann ) . The genomic sequence of nlp-24 ( −3151 to +460 ) was ligated into the pJG7-psm-SL2-GFP vector . To make an npr-17 translational reporter ( Pnpr-17::npr-17::GFP ) , we amplified GFP from pPD95 . 77 and inserted this PCR product in the genomic sequence of npr-17 ( −5110 to +5545 ) before the stop codon , thus fusing GFP with the npr-17 3′ UTR ( +5548 to +6524 ) . For site specific expression of npr-17 , we used a vector from Douglas S Kim based on pSM into which was inserted SL2::RFP and the unc--54 3′ UTR . First , npr-17 cDNA was amplified and ligated into the pSM-SL2::RFP vector as an XmaI/NheI fragment . Then , to express npr-17 in the ASI neurons , 2 . 5 kb of gpa-4 promoter amplified from N2 genomic DNA was cloned into this pSM npr-17 SL2::RFP construct using BamHI and NotI . To express npr-17 in the intestine , 3 kb of ges-1 promoter was amplified from N2 genomic DNA and inserted into the pSM npr-17 SL2::RFP construct using BamHI and NotI sites . To express in mammalian cell culture , npr-17 cDNA was amplified from N2 cDNA using MyTaq DNA polymerase ( Bioline , Taunton , MA ) to generate PCR products with a 5′ Kozak consensus sequence and 3′-A overhang . This product was inserted into the PCDNA3 . 3 TOPO TA cloning vector ( Invitrogen , Carlsbad , CA ) according to the manufacturer's instructions . Transgenic animals were generated by microinjection using a Zeiss Axio Observer A1 DIC microscope equipped with an Eppendorf Femtojet microinjection system . For the injection marker , we used Punc-122::GFP or Punc-122::RFP . 50 ng/μl of the fused construct or the plasmid was injected together with 50 ng/μl of the injection marker . Worms were grown on NGMSR plates seeded with E . coli HB101 until the first day of adulthood , then transferred to NGMSR plates without food , After 1 hr at 20°C , we measured the pumping rate by counting pharyngeal grinder movements for 1 min using a Wild M410 dissecting microscope at 64× magnification . For drug experiments , we prepared drug-containing plates by spreading 300 μl of a 10× aqueous drug solution on a 35 mm diameter plate , then waiting at least 3 hr for the drug to diffuse into the agar . First-day adults were then transferred to these plates and incubated for 1 hr at 20°C before counting . RNAi was induced by feeding as described ( Timmons et al . , 2001 ) , with the following modifications . Standard NGMSR agar was supplemented with 25 µg ml−1 carbenicillin and 1 mM isopropyl β-D-1-thiogalactopyranoside and poured into 12-well plates . E . coli HT115 carrying the appropriate RNAi clones was grown in LB containing 100 µg ml−1 carbenicillin at 37°C overnight , then 50 μl drops were seeded onto plates , making sure the culture dried within 1–2 hr , and induced at 37°C for 12 hr . Fourth stage larval ( L4 ) worms were then transferred to the plates ( 3 worms per well ) and allowed to grow to young adulthood at 20°C for approximately 4 days . We tested a total of 115 different neuropeptide genes . We used the Ahringer RNAi library for 91 neuropeptide clones ( Kamath and Ahringer , 2003 ) . For 24 neuropeptides , we cloned PCR-amplified exons into the L4440 vector . Primers for making these constructs are listed in Supplementary file 3 . Total RNA was isolated using TRIzol ( Life Technologies , Carlsbad , CA ) , and reverse transcribed using the Tetro cDNA Synthesis Kit ( Bioline ) . The cDNA was quantified using Nanodrop ( ND-1000 ) and used in qRT-PCR reactions . These reactions were performed on a CFX96 Real-Time PCR Detection System using the SensiMixPlus SYBR & Fluorescein Kit ( Bioline ) . We used 500 ng of cDNA per sample in a total volume of 25 µl . Target genes were amplified using specific primers . Amplification and expression analysis were performed in triplicate . mRNA levels in the tested strains were normalized with ama-1 , a housekeeping gene . For quantification , we used the Delta–delta Ct method . Primers were nlp-3 Forward TGTGTCTACTCTGCTCCCTATG , nlp-3 Backward TGATCATGTCTGGACGGAAAG for nlp-3 and nlp-24 Forward ACGGAGGTGGACGTTATGGA , nlp-24 Backward GAGACCGCCTCCTCCGTAT for nlp-24 . HEK-293 cells ( ATCC , Manassas , VA ) were cultured in DMEM supplemented with 10% FBS at 37°C in a humidified atmosphere containing 5% CO2 . For stable cell line generation , we transfected with a 1:1 ratio of pcDNA3 . 3 containing 1 μg npr-17 and Gα15 cDNA ( promiscuous G protein 15 , Missouri S&T cDNA Resource Center ) using X-tremeGENE HP DNA transfection reagent ( Roche , Switzerland ) . Cells were seeded into 10 cm dishes and antibiotic ( 500 μg/ml G418 ) was added to the culture medium the next day . The selection medium was changed every 3 days until colonies formed . A single colony was picked , expanded , and tested by calcium imaging . Stable cell lines expressing human opioid receptor MOR-1 , DOR-1 and KOR-1 were constructed similarly . MOR-1 , DOR-1 and KOR-1 plasmids were provided by the Duke University GPCR Assay Bank . HEK-293 cells cotransfected with NPR-17 , MOR-1 , DOR-1 or KOR-1 receptors and Gα15 were plated onto 35 mm MatTek glass bottom dishes for calcium imaging . When they reached 70–80% confluency calcium imaging was conducted using the Fluo-4 Direct Calcium assay kit ( Molecular Probes , Carlsbad , CA ) . The cells were loaded with Fluo-4 Direct Calcium reagent and incubated at 37°C for 50 min . For naloxone treatment , 10 μM naloxone was added in the Fluo-4 Direct Calcium reagent before incubation . Synthetic NLP-24 peptides ( all synthesized with amidated C-termini ) , opioid agonists , and blockers diluted in HBSS ( Hank's Balanced Salt Solution ) buffer were applied to the cell for 10 s , and fluorescence at 530 nm was monitored ( excitation wavelength 470 nm ) with a Zeiss Axio Observer A1 microscope and ZEN software . Experiments were conducted on three plates for each condition on three different days . Fluorescence was quantified using ImageJ . We imported images using the ImageJ Bio-Formats plugin and subtracted background , then measured the mean Integrated Density . L4 worms were picked to an HB101-seeded NGMSR plate and given 12 hr to develop to adulthood . Adult worms were transferred to 6 cm NGMSR plates without food and incubated for 30 min . After 30 min , a single starved worm was then transferred to an individual 3 . 5 cm plate and its behavior recorded as described previously ( Gallagher et al . , 2013b ) . We used a 10 mm copper ring to constrain the worm . The light was then turned on and video capture proceeded at 1 frame/s for 1 hr . Recordings were made on a modified version of the nine-worm recording station described by Shtonda and Avery ( 2006 ) in which the worms were imaged through Computer MLM3X-MP macro zoom lenses onto Pointgrey GRAS-14S5M-C digital cameras . We used ImageJ to get x , y coordinates . We inverted each image and subtracted background . The light/dark threshold was adjusted to find the outline of the worm then we determined the coordinates of the worm centroid using the ‘Analyze Particles’ command in ImageJ . Finally , we used HMM analysis to define locomotive behavioral states ( Gallagher et al . , 2013a ) . Pumping rates were tested for statistically significant effects by ANOVA and corrected for multiple testing by the Dunnett , Tukey , or sequential Bonferroni methods , as appropriate . Interactions ( for instance , the difference between the effects of morphine and naloxone on wild-type and npr-17 in Figure 2 ) , were tested by two-way ANOVA . Gene expression ( Figure 2A ) was tested by the interaction between gene ( ama-1 , nlp-3 , or nlp-24 ) and nutritional state ( well-fed or starved ) , with well-fed ama-1 as the base case . These tests were carried out on Ct measurements , not gene expression levels , since gene expression measurements are typically heteroscedastic . Calcium imaging results were more complicated . In each experiment , a series of measurements was made of the effects of different drugs or drug concentrations . All measurements in a series were related by the use of cells sampled from the same population . Each series also included a blank in which buffer was added . For each condition , we first measured base fluorescence ( f0 ) , then added drug and measured fluorescence after 1 min ( f1 ) . We then pooled all the series measuring a given effect and tested log ( f1/f0 ) in a two-way ANOVA with series and drug as the factors , no interaction , and the blank as the base case . This effectively normalizes each measurement to the blank in its own series and tests the residues for drug effects . Particularly for Figure 5B and Figure 6D this test is conservative , since it makes no use of the normalization to the maximal effect measured with loperamide or U69593 . | When and how much an animal eats is controlled by a complex web of signals that are produced by the animal's body and brain . Molecules called opioid neuropeptides are among these signals , and act to control eating in mammals by binding to receptors in the brain and body . These receptors can also bind to similar molecules called opiates ( such as morphine ) ; opiates are amongst the oldest drugs used by humans and have diverse effects ranging from pain relief to addiction . While the activities of opiates and opioid neuropeptides have been studied in mammals , relatively little is known about opioid signaling in simpler animals . The mechanisms behind many biological processes have been investigated using a worm called C . elegans as a model system because it has a simple body plan and its genes can be altered easily . The feeding behavior of C . elegans is no exception . This worm feeds by contracting and relaxing its pharyngeal muscle to move food into its gut . When the worms sense that food is available , this ‘pharyngeal pumping’ is regulated by one type of nerve cell . Slow pharyngeal pumping also continues in starved worms when food is not available , possibly to encourage them to eat new potential sources of food . However , this slow pumping does not require the same type of nerve cell . Cheong et al . hypothesized that the slow pumping in starved worms might depend on neuropeptide signaling instead , and have now tested this idea using engineered worms that made lower levels of a number of these molecules . The experiments uncovered a molecule called NLP-24 that promotes the slow pharyngeal pumping . This molecule is similar to opioid neuropeptides found in mammals . Worms that made less NLP-24 than normal grew more slowly; this suggests that they had problems feeding . Moreover , the levels of NLP-24 were found to increase in normal worms soon after they were deprived of food . Further experiments revealed the identity of the receptor for this molecule , which is also similar to mammalian opioid receptors . The discovery that opioid signaling is involved in C . elegans' feeding behavior may well , in future , also help to identify new molecular players involved in opioid signaling . Further studies might also help the search for ways to reduce the problematic side-effects that limit the usefulness of opiate drugs as medicines . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | An opioid-like system regulating feeding behavior in C. elegans |
The circuit structure and function underlying post-coital male behaviors remain poorly understood . Using mutant analysis , laser ablation , optogenetics , and Ca2+ imaging , we observed that following C . elegans male copulation , the duration of post-coital lethargy is coupled to cellular events involved in ejaculation . We show that the SPV and SPD spicule-associated sensory neurons and the spicule socket neuronal support cells function with intromission circuit components , including the cholinergic SPC and PCB and the glutamatergic PCA sensory-motor neurons , to coordinate sex muscle contractions with initiation and continuation of sperm movement . Our observations suggest that the SPV and SPD and their associated dopamine-containing socket cells sense the intrauterine environment through cellular endings exposed at the spicule tips and regulate both sperm release into the hermaphrodite and the recovery from post-coital lethargy .
Persistence in performing a goal-orientated behavior must be balanced by behavioral termination cues once the task is completed . One such behavior , mating , is important for species propagation and can improve an individual's ability to cope with stress ( Neumann , 2009 ) . In humans , rats , and other animals , a period of disinterest and mating inability follows ejaculation in males ( Beach and Holz-Tucker , 1949; Masters and Johnson , 1966; Barfield and Geyer , 1972; Oomura et al . , 1983; Ureshi et al . , 2002 ) . Primarily studied in rodents , sexual disinterest and inability following mating are described in two ways: the refractory period , defined by the short term duration between consecutive ejaculations ( Levin , 2009 ) , and sexual satiation or exhaustion , a period of time following repeated copulations where the male rats require 6–14 days to regain sexual potency ( Beach and Jordan , 1956 ) . If a male rat is considered to be sexually satiated , he cannot sire progeny even if he engages in copulatory activity ( Tlachi-Lopez et al . , 2012; Lucio et al . , 2014 ) . While the behavioral phenomenon has been described , little is understood about the molecular and cellular mechanisms controlling both satiation and the refractory period . Neurotransmitters and hormones such as serotonin and prolactin may extend the period of inactivity , while others such as dopamine and norepinephrine may shorten it ( McIntosh and Barfield , 1984a , 1984b , 1984c; Buvat et al . , 1985; Marson and McKenna , 1992 ) . However , the basic structure and function of mating circuits that exhibit a period of inactivity are still being elucidated ( Levin , 2009; Turley and Rowland , 2013 ) . The well-defined structural components of the nervous system in the Caenorhabditis elegans hermaphrodite have facilitated a detailed understanding of how circuits function to produce behaviors ( White et al . , 1986; Varshney et al . , 2011; Cohen and Sanders , 2014 ) . Combining the anatomical information with optogenetics , cell ablations and calcium imaging have uncovered information on how C . elegans responds to both attractive and repulsive stimuli ( Cohen and Sanders , 2014 ) . For example , Li et al . ( 2011 ) identified the sensory neurons and their direct downstream targets that regulate response to the noxious stimuli of a harsh touch ( Li et al . , 2011 ) . Hendricks et al . ( 2012 ) determined which neurons controlled head movement in response to the chemo attractant isoamyl alcohol ( Hendricks et al . , 2012 ) . Additionally , several studies highlight the role that extrasynaptic neuromodulation plays in regulating behavioral responses , adding another layer to neuromuscular circuit control of behavior ( Flavell et al . , 2013; Leinwand and Chalasani , 2013 ) . The tool set used to deconstruct the circuits in hermaphrodites can be applied to study the most complex behavior exhibited by the nematode , male mating . Previous work on the mating steps that precede ejaculation provides a foundation for understanding the circuit structure and function that produces copulation-induced inactivity . Reconstruction of serial electron microscopy images provides detailed information of the structure and connectivity of the male tail that is not available in other species ( Sulston et al . , 1980; Jarrell et al . , 2012 ) . The connectome has then been utilized as a tool to determine how circuits allow the flexibility necessary for executing a multi-step goal-oriented behavior . We and others have undertaken multiple studies to elucidate how the connectome functions to produce male mating ( Liu and Sternberg , 1995; Barr and Sternberg , 1999; Hurd et al . , 2010; Wang et al . , 2010; Koo et al . , 2011; Miller and Portman , 2011; Siehr et al . , 2011; Barrios et al . , 2012; Garrison et al . , 2012; Sherlekar et al . , 2013 ) . C . elegans males intromit by initially prodding the tightly closed hermaphrodite vulva slit with their two copulatory spicules ( Figure 1A , B ) . After the spicules breach the vulval slit and fully penetrate , the males transfer sperm ( Figure 1B; Liu and Sternberg , 1995 ) . Coupling proper spicule position with prodding is coordinated via cholinergic and glutamatergic signaling from the left-right bilateral post cloacal sensilla ( p . c . s . ) ( Figure 1C ) . These neurons sense the vulva using sensory processes that project posteriorly from the cloacal opening . They stimulate the sex-specific oblique and gubernaculum muscles that , via gap junctions , induce twitch contractions in the spicule protractor muscles ( Figure 1C ) . The protractor contractions are transduced into spicule movements through their hemidesmosome attachments to the spicule cuticle ( Figure 1C; Sulston et al . , 1980; Liu et al . , 2011 ) . Appropriate prodding is limited to the vulva slit via dopamine ( DA ) signaling through the sensory ray neurons ( Correa et al . , 2012 ) . The left-right bilateral cholinergic SPC proprioceptive neurons innervate the protractor muscles and have a sensory projection that is attached to the base of the spicules ( Figure 1C ) . When the spicules partially penetrate the vulval slit , the SPC neurons induce the protractor muscles to contract tonically , likely through sensing the change in spicule position ( Garcia et al . , 2001; Garcia and Sternberg , 2003; Liu et al . , 2011 ) . ∼14 s following spicule insertion , sperm moves from the seminal vesicle to the vas deferens ( a process termed initiation ) ( Figure 2A , B ) , followed by drainage into the hermaphrodite uterus ∼3 s later ( a process termed release ) ( Figure 1B ) ( Schindelman et al . , 2006 ) . 10 . 7554/eLife . 02938 . 003Figure 1 . Conceptual diagram of structures and connectivity in the C . elegans male tail . ger = gubernaculum erector muscle , grt = gubernaculum retractor muscle , adp = anal depressor muscle , ob = oblique muscle . ( A ) Diagram of the male tail positioned at the hermaphrodite vulva . The positions of the copulatory spicules and associated muscles are indicated . The oval structures in the hermaphrodite depict eggs . ( B ) Diagram of the male during spicule insertion and sperm release . ( C ) Abridged connectivity in the male tail , adapted from Jarrell et al . ( 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 00310 . 7554/eLife . 02938 . 004Figure 2 . Conceptual diagrams of the structure and connectivity involved in ejaculation . ( A ) Diagram of the males’ reproductive tract . dpm = dorsal protractor muscle , vpm = ventral protractor muscle . ( B ) Diagram of the initiation step of ejaculation . When the valve region separating the seminal vesicle from the vas deferens opens , sperm cells move toward the cloaca . ( C ) Connectivity of the spicule associated cells . Adapted from Jarrell et al . ( 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 004 Since mating termination must occur during or following the ejaculation step , we hypothesize that cells which regulate ejaculation might contribute to how the males behave after coitus . Little is known of which cells initiate and control ejaculation . The SPC and p . c . s . neurons innervate the gonad , suggesting that they might regulate sperm movement ( Figure 2C ) . Additionally , the SPV and SPD sensory neurons are proposed to inhibit premature ejaculation , but their modes of action remain unclear ( Liu and Sternberg , 1995; Schindelman et al . , 2006 ) . These neurons send a sensory process down the two spicules' shafts and have sensory endings that are exposed to the environment at the spicule tips ( Figure 2C; Sulston et al . , 1980 ) . They do not directly innervate the gonad , but they do have connections to the SPC and p . c . s . , suggesting that they might indirectly regulate ejaculation through these neurons ( Figure 2C; Jarrell et al . , 2012 ) . In this study , we report that acetylcholine , dopamine , and glutamate secreting cells are stimulated upon successful intromission and promote sperm movement and release . Ejaculation , induced from the coordinated activities of the p . c . s . , SPC , SPD , SPV , spicule socket cells , sex-muscles , and the gonad , leads to a refractory period consisting of reduced mating drive and ability . The duration of the refractory period allows the male time to recover , prevents him from mating multiple times with the same mate , and resets the complex neuromuscular network required for mating . Disruption of these cues leads to a shortened refractory period .
Similar to other species , C . elegans males exhibit a period of reduced activity following ejaculation ( Barfield and Geyer , 1972; Oomura et al . , 1983; Ureshi et al . , 2002 ) . To develop metrics for measuring how the male's behavior changes immediately after mating , we first determined how fast C . elegans males re-copulate following ejaculation . Within 2 min after being placed on a 5 mm diameter bacterial lawn containing 15 immobile hermaphrodites , 1-day-old virgin males commenced mating and inserted their spicules into the hermaphrodites ( Figure 1B , Figure 3A ) . However following ejaculation , ∼12 min ( SD = ±6 min 30 s ) passed before they intromit their spicules again ( Figure 3A ) . The males obviously regained the ability to mate prior to the 2nd intromission , but their individual mating behaviors were variable , preventing us from establishing a metric to determine effectively when their full ability to mate returned . Thus , we conservatively refer to the interval from the 1st spicule insertion to the 2nd spicule insertion as the ‘refractory period’ . We then asked if the refractory period duration was due to mating disinterest , reduced motor ability , or a combination of both . 10 . 7554/eLife . 02938 . 005Figure 3 . The refractory period is regulated by sperm release . Line represents median . ( A ) 1st insert: the time required for a male to insert his copulatory spicules into the hermaphrodite vulva from the time he was placed with the hermaphrodites . 2nd insert: the time from 1st insert to finding a second hermaphrodite and repeating the mating process ( refractory period ) . y-axis , the time it takes a male to insert his spicules , x-axis , insertion number . ( B ) Mating drive . 1st commencement: the time it takes the male to begin mating with a hermaphrodite after being placed on a mating lawn . 2nd commencement: the time from retraction to the next beginning of mating . ( C ) The total time the male spent at the vulva prior to insert . ( D ) The number of vulva passes prior to insertion . ( A–D ) n = 17 . *p<0 . 05 , **p<0 . 005 , ***p<0 . 0001 , paired t test . ( E ) The number of progeny sired following successive couplings . Males are grouped into two categories: males that re-copulated between 3 and 8:59 min following the 1st insertion and males that re-copulated between 9 and 30 min following the first insertion . The letters identify the same male for each insertion . x-axis indicates the insert as well as what refractory period group each male was placed in . y-axis is the number of progeny each male sired for the indicated insert . **p<0 . 05 , Mann–Whitney test . ( F and G ) The first insert ( F ) and refractory period ( G ) for males with the indicated cell ( s ) removed . *p<0 . 05 , Mann–Whitney test . n is indicated below the x-axis . The x-axis indicates the cells removed during the operation , and the y-axis indicates the time it took for males to insert their spicules into the hermaphrodite . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 005 To determine what behaviors were affected during the refractory period , we digitally recorded and measured the time males required to execute the different mating steps . Within 34 s after introduction to the mating lawn , virgin males placed their tail on the hermaphrodite cuticle and began searching for the vulva ( Figure 3B ) ; following intromission , they kept their spicules inserted for an average of 94 s . After spicule retraction , ∼5 min passed before the males re-executed copulation ( Figure 3B ) , indicating that male sexual drive was temporarily suppressed . However , even after males recommenced mating , the executions of subsequent motor steps were also slowed . The duration that the males spent positioned over the vulva prior to intromission was slightly longer , but not statistically different between the 1st and 2nd mating bouts ( Figure 3C ) ; however , the number of times they passed over the vulva was significantly increased ( Figure 3D ) . These observations indicate that various male sensory-motor circuits recover differentially to the source ( s ) that attenuate mating behavior after ejaculation . This raised the question of why the male would need an extended time period between copulations . In mammals , since the amount of sperm decreases in the male reproductive tract after ejaculation , the refractory period may provide an opportunity to re-establish sperm count ( Judd et al . , 1997; Levin , 2009; Tlachi-Lopez et al . , 2012 ) . To determine if , in C . elegans , a longer refractory period might increase the amount of sperm transferred , we counted the number of sired progeny as a proxy for successful ejaculation . Each male was allowed to mate two consecutive times , and the refractory period was recorded . Additionally , the cross-progeny from each mated hermaphrodite was counted to determine if the male had successfully ejaculated . We found that we could bin the males into two groups , the first of which took 3 to 8:59 min to copulate twice , and the second of which took 9 to 30 min to copulate twice . We found that males with a longer refractory period were more successful in siring progeny than their counterparts , whose refractory periods were shorter than 9 min ( Figure 3E ) . Importantly , males from both groups did not display a difference in the amount of progeny they sired during the 1st mating ( Figure 3E ) . This indicates that the reduced number of progeny sired during the 2nd mating , by the group of males with a shorter refractory period , is not due to a lack of ability . Thus , a longer refractory period allows males to recover their ability to sire progeny . We next asked how the various cellular components involved in ejaculation influence the refractory period . We used the refractory period ( the time between the 1st and 2nd spicule insertion ) as a metric to identify the cellular source ( s ) that promotes post-coital mating attenuation . The male gonad was one of our candidates for regulating male sexual drive , since it is implicated in modifying whether males fed or searched for a mate ( Lipton et al . , 2004 ) . The mature gonad consists of the germ line , seminal vesicle , and vas deferens ( Figure 2A ) . We measured the initial insertion time ( Figure 3F ) and refractory period ( Figure 3G ) of males that had different cellular constituents of their gonad laser-ablated . We found that males lacking all of these structures required longer time to commence mating ( Figure 3F ) , implying that the gonad sends a signal to promote this behavior . After the operated males inserted and retracted their spicules , their refractory period was similar to the mock-ablated control and coincidentally similar to the time required for their 1st spicule insertion ( Figure 3G ) . The behavioral similarity between gonad-ablated males and post-coital intact males led us to speculate that a pro-mating gonadal signal might be repressed or depleted by an aspect of ejaculation; re-establishing the pro-mating signal might occur during the refractory period . We next asked if the somatic gonad , consisting of the seminal vesicle ( the storage area of mature sperm ) and the vas deferens ( a tubular seminal fluid-producing conduit for sperm movement ) , was sufficient to promote the initial mating drive , or if the germ line was also required . We found that germ line-ablated males behaviorally resembled the gonad-ablated males ( Figure 3F , G ) , indicating that the germ line is required for the gonad to promote mating . Since the germ line produces sperm , we next asked if sperm production or sperm movement during ejaculation affects the initial mating drive and/or the refractory period duration . To uncouple sperm production from sperm release , we laser-ablated the linker cell in early larval stage 4 ( L4 , the final stage before adulthood ) males . The linker cell guides the developing gonad to the posterior region of the male . Removal of the linker cell late in development prevents the mature vas deferens from connecting to the cloacal opening ( Sulston et al . , 1980 ) . The somatic gonad and germ line are still functional and sperm moves from the seminal vesicle , but are retained in the vas deferens upon spicule insertion . Interestingly , the operation did not affect the initial mating drive ( Figure 3F ) . However , as a population of operated males , preventing sperm release increased the duration of the refractory period , relative to the mock-ablated control ( Figure 3G ) . Under closer inspection , the laser operation gave rise to two populations: males that displayed a normal refractory period and males that took much longer to re-mate ( 5 min 40 s ( n = 6 ) vs 24 min 40 s ( n = 5 ) , p value=0 . 0043 , Mann–Whitney test ) ( Figure 3G ) . In equal proportions , some males were able to compensate for the lack of sperm movement , while others failed to do so . This suggests that the duration of the refractory period is plastic and could be modified by cellular components that directly or indirectly sense sperm release . This raised the question of what senses the transfer of sperm from the male into the hermaphrodite . The male contains bilateral schlerotic copulatory spicules that he inserts into the hermaphrodite . The spicules serve to anchor the male to his mate and to widen the vulval channel to facilitate the flow of sperm . In addition , each spicule contains the sensory dendrites of two neurons , the bilateral left/right SPV and SPD ( Figure 2C ) . These neurons send a process down the shaft of the spicule and have ciliated sensory endings that are exposed to the environment at the spicule tip . The processes of the neurons are surrounded by the structural sheath and socket cells ( Sulston et al . , 1980 ) . When we laser-ablated these neurons during L4 , the resulting males had variable mating problems . These defects include failure to follow through the steps of mating , failure to insert their spicules , and failure to ejaculate or prematurely ejaculating sperm prior to intromission ( Figure 4A ) ; some of these defects have been previously reported ( Liu and Sternberg , 1995; Schindelman et al . , 2006 ) . These pleiotropic defects made measuring the refractory period difficult . We speculated that removing these neurons too early in larval development resulted in the variable re-wiring of the remaining circuits . To examine SPV and SPD's role during the refractory period , we used the laser to cut off the spicule tips in virgin adult males ( Figure 4—figure supplement 1 ) . This operation causes the cytoplasm of both SPV and SPD to leak out , killing the cells . Unlike the larval SPD and SPV-ablated males , spicule tip cut adult males did not display variable mating defects ( Figure 4A ) . Both the mock-cut control and operated males displayed normal initial response times to hermaphrodites ( Figure 3F ) , but interestingly , the spicule tips-cut males displayed a significantly shorter refractory period than control males ( 296 ± 217 s vs 499 ± 224 s , p value=0 . 0121 , Mann–Whitney test ) ( Figure 3G ) . This suggests that in wild-type males , the spicule sensilla modify the duration of the refractory period . 10 . 7554/eLife . 02938 . 006Figure 4 . The spicule sensilla promote insertion behaviors . ( A ) Sperm transfer ranking for males lacking the SPV and SPD sensory neurons and their age-matched controls . A male received a 0 if he ejaculated into the uterus , a 1 if he inserted but didn't ejaculate , a 2 if he ejaculated without inserting into the vulva , and a 3 if he was unable to ejaculate or insert . Additionally , each male was given up to 5 min to insert and transfer sperm . That time was divided by 300 s and added to the number he received for ejaculating , giving him a final ranking . Filled circles represent males that inserted and transferred sperm into the hermaphrodite . Open circles represent males that inserted but did not transfer sperm . Stars represent males that ectopically ejaculated . Xs represent males that neither inserted nor ejaculated within 5 min . Line indicates mean . **p<0 . 005 , Mann–Whitney test . ( B ) Mating potency . The number on each column is the % of males that sired progeny . **p<0 . 005 , Fisher's exact test . n = 12 . ( C ) Time the males remained inserted the first time they mated . ( D ) Following the second intromission , spicule tips cut males often did not leave their spicules inserted very long . Reported here is the total amount of time the males’ spicules were inserted into the hermaphrodite for the first 30 s following the 2nd insert . *p<0 . 05 , Mann–Whitney test . ( C and D ) n = 14 . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 00610 . 7554/eLife . 02938 . 007Figure 4—figure supplement 1 . Cells affected by cutting off the spicule tips with a laser . ( A ) Simplified diagram of the neuronal connectivity present in the spicule circuit in the male tail . ( B ) The same diagram as ( A ) representing what cells ( with an ‘X’ over them ) and connections are lost when the spicule tip is cut using a laser . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 00710 . 7554/eLife . 02938 . 008Figure 4—figure supplement 2 . Spicule tips cut males display defects in mating . ( A ) Mating drive in spicule tip cut males . The y-axis is the time it takes a virgin male to commence backing along a hermaphrodite cuticle . ( B ) Total time at vulva prior to insert in operated spicule tips cut males and intact males . ( C ) Effect of surgical environment on refractory period . The stage males are exposed to azide and/or an agar pad is given on the x-axis . The y-axis is the refractory period . *p<0 . 05 , Mann–Whitney . For ( A–C ) , line is the median . ( D ) The number of times a male successfully inserts his spicules in the 60 s following the 2nd insert . The median for both mock ablated and Spicule tips is 0 . ( E ) The number of times a male retracts his spicules in the 60 s following the 2nd insert . *p<0 . 05 , Mann–Whitney . For ( A–C and E ) , line is the median . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 008 Although spicule tips-cut males lacking functional SPV and SPD could insert their spicules , many of the males were defective in ejaculation ( Figure 4A ) . Analysis of sperm movement revealed defects in the initiation and release steps of ejaculation . Under standard conditions , sperm moves from the seminal vesicle through the valve to the vas deferens ( initiation ) ( Figure 2B ) , and then drains out the cloaca into the hermaphrodite uterus ( release ) ( Figure 1B; Schindelman et al . , 2006 ) . In approximately half of the spicule tips-cut males , the valve region opened ( 12/13 control vs 8/15 ablated , p=0 . 0377 , Fisher's exact test ) , but sperm rarely passed out of the cloaca into the hermaphrodite ( 10/14 control vs 1/15 ablated , p=0 . 0005 , Fisher's exact test ) . To verify if sperm transferred into the hermaphrodites , we moved the mated partner to individual plates after each successful insertion . The mated hermaphrodites contained a mutation that disrupts locomotion . We counted the number of plates that had moving cross-progeny , as a proxy for successful sperm release . 67% of control males sired progeny at the 1st mating and 33% of males sired progeny at the second mating ( n = 12 ) ( Figure 4B ) . No operated males were successful at the 1st insert , and only two were successful the 2nd time ( n = 12 ) ( Figure 4B; Liu and Sternberg , 1995; Schindelman et al . , 2006 ) . Thus , the putative sensory function of the spicules promotes both sperm movement and affects the refractory period . This raised the possibility that these two phenomena are coupled . The refractory state follows the ejaculatory behavioral state . Normally , the males should maintain the ejaculatory mating state , keeping their spicules inserted until all sperm are transferred ( Schindelman et al . , 2006 ) . Therefore , we used spicule insertion time as a proxy metric to ask if the spicule sensilla regulate the duration of the ejaculatory mating state . To determine if the spicule sensilla influence insertion duration , we recorded the 1st and 2nd copulations of virgin mock-cut and spicule tips-cut males . We did not measure any differences in mating drive or ability between the two populations ( Figure 4—figure supplement 2 ) . The average spicule insertion duration during the 1st mating was also not significantly different between the two treatments , although we observed a larger variability for spicule tips-cut males ( F value = 0 . 0009 ) ( Figure 4C ) . However , we noticed this difference became exaggerated during the 2nd mating . Generally , a male maintains spicule insertion throughout the sperm transfer process , which normally takes longer than 30 s . In contrast , many of the mating-experienced cut males inserted and retracted their spicules multiple times , with each penetration lasting only a few seconds ( Figure 4D , Figure 4—figure supplement 2 ) . This observation indicated that the spicule tips-cut males were defective in either entering or maintaining the ejaculatory state . Therefore , we were interested in identifying what additional cells or circuits function temporally with the spicule sensilla to transition the male between intromission , ejaculation , and refractory behaviors . We utilized males that co-expressed the Ca2+ sensor G-CaMP and the mDsRed fluorescent proteins to determine if cells , previously implicated through connectome wiring and laser-ablation/behavioral analyses , function during spicule penetration , sperm transfer , and post ejaculation behaviors ( Nakai et al . , 2001; Tian et al . , 2009; Correa et al . , 2012 ) . As the males were mating , we digitally recorded the G-CaMP fluorescence changes and later measured and corrected them using the mDsRed fluorescence as a non-fluctuating reference . The data were plotted as a %ΔF/F0 over time and correlated with the male’s behavior . A representative male is shown for each cell ( s ) of interest . Figure supplements contain four additional males that inserted their spicules and one male that prod at the vulva but did not insert . To measure Ca2+ transients in the SPV and SPD neurons during mating , we expressed G-CaMP and mDsRed in these cells using the gpa-1 promoter ( Pgpa-1 ) ( Jiang and Sternberg , 1999a ) . In many males , individual SPV and SPD fluorescence could not be isolated , since their cell bodies overlapped . Thus we measured their combined fluorescence during data analysis . After spicule insertion , Ca2+ transients in these cells gradually increased; the average peak activity occurred ∼4 s post intromission . The Ca2+ transients began to decline during the sperm's movement through the vas deferens and continued to decrease after sperm release ( n = 5 , Figure 5A; Table 1; Figure 5—figure supplement 1 ) . Although the data indicate that the SPV and SPD are active post intromission and during the initial stages of ejaculation , we surveyed additional cells to see if their activities are also correlated with the timing of spicule insertion and ejaculation . 10 . 7554/eLife . 02938 . 009Figure 5 . Ca2+ transients in SPV , SPD , SPC neurons , and gonad muscles during intromission and ejaculation . ( A , B , D ) . % ΔF/F0 trace including insertion and ejaculation . One representative recording is reported for each cell type . x-axis is time ( seconds ) , y-axis is % ΔF/F0 . The time of mating behaviors is indicated in color . Additional traces given in Figure 4—figure supplement 1 and Figure 4—figure supplement 3 . ( A ) Trace of the SPV/SPD neurons , G-CaMP expressed from Pgpa-1 . Scale bar = 20 µM . ( B ) Trace of the SPC neuron , G-CaMP expressed from Pgar-3B . Scale bar = 10 µM . ( C ) Graph of the largest calcium change ( ΔF/F0 ) from time point 0 . The pictures display fluorescence in the valve region at time 0 and 1 min later for males exposed and not exposed to OxoS . Males express G-CaMP in the gonadal valve from Ptry-5 . *p value<0 . 05 , Mann–Whitney test . Scale bar = 10 µM . ( D and E ) Traces of the valve region of the gonad , G-CaMP expressed from Ptry-5 . Scale bar = 100 µM . SpTipsCut = spicule tips cut . ( F ) Slope of the line extrapolated from the highest Ca2+ transient plus 15 s . This time was chosen to include sperm release . *p value<0 . 05 , Mann–Whitney test . Line represents median . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 00910 . 7554/eLife . 02938 . 010Figure 5—figure supplement 1 . Ca2+ transient changes in cells that regulate ejaculation . % ΔF/F0 trace for mating from insertion through sperm release and retraction , except for the first trace in each column which represents vulva prodding behavior . Arrows indicate when the behavior occurred . ( A ) Trace of SPC , G-CaMP expressed from Pgar-3B . ( B ) Trace of SPV/SPD , G-CaMP expressed from Pgpa-1 . ( C ) Traces of the valve region of the gonad , G-CaMP expressed from Ptry-5 . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 01010 . 7554/eLife . 02938 . 011Figure 5—figure supplement 2 . Removing the spicule sensilla impacts the Ca2+ changes in the gonadal valve . % ΔF/F0 trace for mating from insertion ( 1 ) through valve opening ( 2 ) and sperm release ( 3 ) . G-CaMP is expressed in the valve via Ptry-5 . ( A ) Intact males and ( B ) spicule tip ablated males . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 01110 . 7554/eLife . 02938 . 012Table 1 . Ca2+ transients following spicule insertionDOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 012Cell↑Ca2+ insertion→peak ( sec ) *SignificanceSlope of initial Ca2+ increase ( ΔF/F0%/sec ) SignificanceSPC1 . 3 ± 0 . 52a45 ± 13Valve1 . 8 ± 0 . 79a60 ± 14SPV/SPD4 . 0 ± 1 . 0p<0 . 05 to ‘a’21 ± 14p<0 . 05 to SPC and valvePCA6 . 3 ± 0 . 56p<0 . 05 to ‘a’15 ± 2 . 9p<0 . 05 to SPC and valveSocket cells1 . 7 ± 0 . 36a94 ± 18p<0 . 005 to PCASex muscles 1st peak1 . 3 ± 1 . 7a155 ± 82p<0 . 005 to socket cellsSex muscles 2nd peak11 ± 3 . 9p<0 . 05 to sex muscles 1st peak46 ± 19†p<0 . 0001 to sex muscles 1st peakSPC control1 . 2 ± 0 . 3060 ± 26SPC ablated2 . 2 ± 0 . 74p=0 . 022 to SPC control29 ± 12p=0 . 035 to SPC controlMean and standard deviation reported . For non-operated cell types , n = 5 . Results of ANOVA: Newman–Keuls Multiple Comparison Test are shown . SPC–control and ablated , Mann–Whitney t test . SPC–control n = 6 , SPC–ablated n = 7 . *Time ( sec ) required for Ca2+ to increase from spicule insertion to Ca2+ peak . †Slope determined from where the Ca2+ begins to increase for a second time to Ca2+ peak following this second increase . aPeak times that are significantly different from the SPV/SPD and PCA cells . Previous work found that sustained spicule insertion requires the bilateral cholinergic SPC sensorimotor neurons . From their structural morphology , the neurons are speculated to sense spicule movements through their proprioceptive attachments to the base of the spicules . Via acetylcholine ( ACh ) -mediated synaptic transmission , the SPC neurons can then induce tonic spicule protractor muscle contraction ( Figure 1C ) . Laser-ablation of these cells impairs sustained spicule intromission and consequently , the ability to ejaculate . The SPC makes chemical synapses to the gonad , raising the possibility that these neurons might also contribute to sperm movement ( Figure 2C; Sulston et al . , 1980; Liu and Sternberg , 1995; Garcia et al . , 2001; Jarrell et al . , 2012 ) . We expressed G-CaMP in the SPC using the Pgar-3B promoter and recorded fluorescent changes during mating ( Liu et al . , 2007 ) . Coincident with full spicule penetration , Ca2+ transients rapidly increased in the SPC . The peak G-CaMP fluorescence intensity occurred faster than in the SPV and SPD neurons ( 1 . 3 s vs 4 . 0 s , p<0 . 05 , ANOVA , Figure 4B; Table 1; Figure 4—figure supplement 1 ) . However , similar to the SPV and SPD , the Ca2+ transients in the SPC began to decrease during sperm movement into the gonadal vas deferens ( Figure 5B ) . The somatic gonad stores sperm until receiving appropriate cues ( Figure 2A , B ) . Upon spicule insertion , this organ facilitates sperm transfer to the hermaphrodite ( Figure 1B ) . The gonad's active role in ejaculation is not well understood , although it requires the cellular secretion machinery to promote sperm release ( Schindelman et al . , 2006 ) . Video recordings of the sperm-activating protease TRY-5 highlight the fluid movement in the gonad . First , fluid containing TRY-5 within the vas deferens transfers to the hermaphrodite , and then additional TRY-5 release coincides with sperm movement from the valve ( Smith and Stanfield , 2011 ) . Since the cholinergic SPC makes chemical synapses to the somatic gonad , we asked if ACh signaling can stimulate gonadal Ca2+ transients . The potentially non-specific ACh agonist oxotremorine S ( OxoS ) induces males to protrude their spicules and some males to ejaculate . We expressed G-CaMP in the somatic gonadal valve using the try-5 promoter ( Smith and Stanfield , 2011 ) and asked if exogenously applied OxoS can induce valve Ca2+ transients . We saw increases in gonadal Ca2+ transients in males ( n = 5 ) that were exposed to the agonist ( Figure 5C ) , indicating that the somatic gonad activity can be stimulated by ACh . We then asked when during mating can we detect somatic gonad activity . Prior to spicule insertion , the gonadal valve keeps the sperm in the seminal vesicle . Immediately upon spicule insertion , we measured that the valve’s Ca2+ transients reached their peak ( ∼1 . 8 s ) with kinetics similar to the SPC neurons ( Figure 5D; Table 1; Figure 5—figure supplement 1 ) . This coordination in cellular activities is consistent with the idea that signaling from the cholinergic SPC partially contributes to the initiation of ejaculation ( Garcia et al . , 2001; Emmons , 2005 ) . However , although Ca2+ transients in the valve increased upon intromission , the contractions that open the valve did not immediately occur . Ca2+ transients remained high in the valve until it opened , on average 6 . 9 ± 1 . 5 s after insertion ( n = 7 ) . When sperm moved through the vas deferens , the level of Ca2+ transients in the valve began to slowly decrease ( Figure 5D ) . The delay between spicule insertion and the movement of sperm out of the seminal vesicle suggests that additional cells , which are active during this interval , might contribute to the opening of the valve . Since the Ca2+ transient levels in the SPV and SPD neurons peaked after the SPC neurons , we asked if damaging these sensory cells , by using a laser to cut off the spicule tips , would interfere with the opening of the valve or perturb the profile of the valve's Ca2+ transients . Similar to the mock-cut males , the valve in spicule-tips cut males displayed a rapid increase in Ca2+ transients ( Figure 5E , Figure 5—figure supplement 2 ) . But instead of a distinct decrease in Ca2+ transients that coincide with the opening of the valve and sperm movement , the decrease in Ca2+ transients was more gradual and protracted ( Figure 5E , F ) . Coincident with the altered activity profiles , the valve did not open in 10/19 males ( compared to 18/19 controls , p value=0 . 0078 , Fisher's exact test ) , and in the remaining males , sperm was not transferred to the hermaphrodite . In contrast , nearly all mock-cut controls released sperm into their mates ( 1/19 ablated vs 18/19 controls , p value<0 . 0001 , Fisher's exact test ) . Taken together , the Ca2+ imaging suggests that immediately upon intromission , the SPC primes the gonad for the initiation of ejaculation , and the spicule sensory neurons provide further stimulation necessary for sperm movement and release . The calcium imaging and laser-ablation experiments suggested that during mating , the SPV , SPD , and SPC neurons facilitated ejaculation . Therefore , we wanted to confirm if stimulating a limited number of cells that included these neurons is sufficient to induce ejaculation in non-mating males . To address this question , we utilized the short wavelength light ( blue , 475 nm ) -activated ion channel channelrhodopsin2 ( ChR2 ) ( Nagel et al . , 2003 , 2005 ) . We expressed ChR2 from the promoters: Pgar-3b ( expresses in multiple cells including the SPC ) ( Liu et al . , 2007 ) , Punc-103B ( many cells including the post cloacal sensilla and spicule sensilla and sex muscles ) ( Reiner et al . , 2006 ) , Punc-17 ( all cholinergic neurons ) ( Alfonso et al . , 1993; Garcia et al . , 2001 ) , and Punc-17S ( head cholinergic neurons only ) . We found that none of the transgenic males expressing ChR2 initiated ejaculation or released sperm upon stimulation ( Figure 6A ) . This result indicated that stimulation of specific groups of cells is not sufficient to induce the ejaculatory state . However , transgenic males simultaneously expressing Punc-103B:ChR2 , Punc-17S:ChR2 , and Pgar-3B:ChR2 did ejaculate upon light stimulation ( Figure 6B ) . This indicated that a larger number of cells than we expected , which includes the SPD , SPC , PCA , and PCB neurons , as well as sex-shared head and ventral cord neurons and sex muscles , must be stimulated to initiate ejaculation and release sperm ( Figure 6C ) . 10 . 7554/eLife . 02938 . 013Figure 6 . Cholinergic and glutamatergic neurons promote ejaculation . ( A ) Chart of promoter , expression pattern , and ability of the promoter driving ChR2 to induce ectopic ejaculation ( Ejac ) . ( B ) Graph of ectopic ejaculation in response to 475 nm wavelength light in rgIs6[Punc-17 small:ChR2 , Punc-103B:ChR2 , Pgar-3B:ChR2] males . x-axis indicates the cells ablated , y-axis is the percent of males that ectopically ejaculated in response to 475 nm wavelength light stimulation . n numbers are indicated at the bottom of each bar . *p<0 . 05 , **p<0 . 005 , Fisher's exact test . ( C ) Fluorescent , confocal image of a rgIs6[Punc-17 small:ChR2 , Punc-103B:ChR2 , Pgar-3B:ChR2] L4 male tail . Dorsal is up , posterior to the right . Cells important in mating behavior are indicated by the arrows . Scale bar = 10 µM . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 01310 . 7554/eLife . 02938 . 014Figure 6—figure supplement 1 . % of ChR2-expressing males that ejaculate in response to 475 wavelength light stimulation . ( A ) Rate of ectopic ejaculation in response to 475 nm wavelength light exposure in males expressing ChR2 in the spicule and post cloacal sensilla . x-axis lists the cells ablated , y-axis is the % of males that ejaculated . *<0 . 05 , **<0 . 005 , Fisher's exact test . These numbers are normalized in Figure 4B . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 014 Evoking ejaculation in non-mating males via intense 475 nm wavelength light and ChR2 is artificial , but we used the assay to help us identify additional cells that might be involved in ejaculation behavior . We laser-ablated various ChR2-expressing cells to determine if artificial ejaculation could be perturbed . We killed the SPV/SPD and SPC neurons singly and found that these treatments slightly reduced the ejaculation efficiency , but not enough to be statistically significant ( p value=0 . 78 and 0 . 42 , respectively , Fisher's exact test ) ( Figure 6B , Figure 6—figure supplement 1 ) . Since removing the spicule sensilla and SPC did not significantly reduce artificial ejaculation , we next asked if other mating-associated cells in the male tail , specifically the post cloacal sensilla , can contribute to the behavior . The bilateral PCA , PCB , and PCC neurons constitute the post cloacal sensilla . They redundantly function to sense the vulva and execute repetitive intromission attempts ( Liu and Sternberg , 1995; Liu et al . , 2011 ) . The cholinergic PCB and PCC make chemical synapses to the gonad ( Figure 2C; Garcia et al . , 2001; Jarrell et al . , 2012 ) . We initially focused on the PCB , since it expresses ChR2 and PCC does not . We found that ablating the PCB singly reduced the ejaculation efficiency , but like the SPC neurons , not enough to be statistically significant ( p value=0 . 2 , Fisher's exact test ) ( Figure 6B ) . However , compared to the mock-ablated control , ablating both SPC and PCB significantly reduced the number of males that released sperm ( 42% , p=0 . 013 , Fisher's exact test ) ( Figure 6B ) . This suggested an additive role for these cells in ejaculation behavior . However , some males still ejaculated in response to 475 nm light stimulation . This indicates that ejaculation-promoting neurons remain; this could include the post cloacal sensilla PCC and PCA . In the transgenic males , the PCC neurons did not express ChR2 , but nonetheless , we laser-ablated the PCC with SPC and PCB . We reasoned that PCC could be indirectly activated by light-induced ChR2 stimulation through their connectivity with other cells in the circuit ( Figure 2C; Jarrell et al . , 2012 ) . However , we did not see any obvious decrease in ejaculation behavior when comparing SPC/PCB/PCC to SPC/PCB ablated males ( p value=0 . 37 , Fisher's exact test ) ( Figure 6B ) . This raised the possibility that the remaining post cloacal sensilla , the PCA neurons , could be involved in promoting ejaculation . Unlike the SPC , PCB , and PCC , the PCA neurons do not make chemical synapses with the gonad and are not cholinergic . However , similar to the PCB and PCC , PCA does make motor synapses to the posterior sex muscles ( Figure 1C; Jarrell et al . , 2012 ) . To our surprise , we observed that few PCA-ablated males were able to ejaculate after stimulation ( 16% , p value=0 . 0003 , Fisher's exact test ) ( Figure 6B ) . This uncovered a role for the PCA in promoting ejaculation . The PCA , PCB , and PCC neurons make chemical synapses to the male specific oblique and gubernaculum muscles ( Figure 1C ) . The oblique muscles , when contracted , help the male press the posterior region of his tail over the vulva during intromission attempts . During ejaculation , the contracted gubernaculum muscles pull the male proctodeum/protracted spicules posteriorly to allow the vas deferens' opening access to the cloacal opening ( Figure 7A ) . Laser-ablation of the gubernaculum muscles causes the protracted spicules to block the vas deferens , thus inhibiting sperm release ( Liu et al . , 2011 ) . We speculated that if the PCA neurons facilitate sperm release , then the gubernaculum muscles and the PCA might display correlated activity . We used the Punc-103E promoter to express G-CaMP in the male sex muscles , and focused our region-of-interest on the anal depressor , gubernaculum erector , and gubernaculum retractor ( Reiner et al . , 2006 ) . As expected , the sex muscles displayed a peak in Ca2+ transients immediately upon spicule intromission ( Table 1; Figure 7A , B , Figure 7—figure supplement 1 ) . However , the Ca2+ transients slightly decreased after the initial spicule penetration . Coincident with the initiation of ejaculation , ∼11 s later , the gubernaculum and anal depressor muscles displayed an increase in Ca2+ transients and became hyper-contracted ( Table 1 ) . After ∼15 to 20 s , we detected sperm in the hermaphrodite uterus . When sperm transfer was completed , the Ca2+ transients in the muscles began to oscillate as they dampened ( Figure 7B ) . 10 . 7554/eLife . 02938 . 015Figure 7 . PCA contributes to the sex muscle-controlled spicule movement required for sperm to drain from the cloaca into the uterus . ( A ) Images of sex muscle activity during intromission and ejaculation . Sc = spicule , ger = gubernaculum erector , adp = anal depressor , grt = gubernaculum retractor . The gubernaculum is located posteriorly to the spicule and assists in spicule movement . After insertion and prior to sperm release , the gubernaculum erector is required to adjust spicule position to allow sperm to drain into the hermaphrodite . The image on the left indicates G-CaMP fluorescence , the central image of diagram of male tail position at the hermaphrodite vulva , and the right image is mDsRed fluorescence . Time is indicated on the left image and spicule location on the right . ( B ) % ΔF/F0 in the anal depressor and the gubernaculum erector and retractor . G-CaMP expressed from Punc-103E . ( C ) % ΔF/F0 in the post cloacal sensilla PCA . G-CaMP expressed from Peat-4 . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 01510 . 7554/eLife . 02938 . 016Figure 7—figure supplement 1 . Ca2+ transient changes in sex muscles and the PCA neurons during ejaculation . ( A ) Ca2+ transient changes in the muscles during mating . % ΔF/F0 trace for mating from insertion ( 1 ) through sperm release ( 3 ) , and retraction ( 4 ) . ( A ) G-CaMP is expressed in the sex muscles via Punc-103E . ( B ) Nomarski image overlaid with fluorescent image of Peat-4:YFP expression in the PCA post cloacal sensilla in the adult male tail . Scale bar = 10 µM . ( C ) %ΔF/F0 in PCA using Peat-4 to drive G-CaMP . ( A and C ) The first graph is extended prodding at the vulva , the other four include insertion and ejaculation . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 016 We used the Peat-4 promoter to express G-CaMP in the PCA neurons to determine if the activity of these cells coincided with ejaculatory gubernaculum muscle contractions ( Figure 7—figure supplement 1 ) . The PCA is likely glutamatergic , since it expresses the eat-4 encoded glutamate vesicular transporter ( Lee et al . , 1999 , 2008; Serrano-Saiz et al . , 2013 ) . We found that the PCA Ca2+ transients gradually increased after spicule insertion with their first peak at 6 . 3 ± 0 . 56 s post intromission ( n = 5 , Table 1; Figure 7C , Figure 7—figure supplement 1 ) ; but unlike the SPC and the SPV/SPD , elevated Ca2+ transients lasted after sperm release ( Figure 7C ) . The timing of the peak PCA Ca2+ transients ( 9 . 2 ± 2 . 9 s ) proceeded and overlapped with the second Ca2+ increase in the gubernaculum erector and retractor and anal depressor muscles ( 11 ± 3 . 9 s , p value=0 . 8 , Mann–Whitney t test compared to PCA peak ) , suggesting PCA is promoting these changes . However , laser-ablation of the PCA did not have gross effects on vulva location behavior , spicule intromission , or ejaculation . Likely , the PCB and PCC neurons can compensate for the damaged cells in operated males ( Liu and Sternberg , 1995; Liu et al . , 2011 ) . The ChR2 experiments revealed a role for the PCA , SPC , and PCB neurons in regulating ejaculation . Surprisingly , ablating the SPV/SPD neurons in this experiment did not have an effect , suggesting ChR2 is expressed in downstream targets of SPV/SPD . These neurons share chemical and electrical synapses with SPC and electrical synapses with PCB , and could indirectly regulate PCA ( Figure 2C; Jarrell et al . , 2012 ) . To test this hypothesis , we cut off the spicule tips in males expressing G-CaMP in the PCA , sex muscles , SPC , and PCB and asked if removing the SPV/SPD neurons impacted Ca2+ transients during mating . We utilized Peat-4 and Punc-103E to drive G-CaMP expression in the PCA neurons and the dorsal sex muscles , respectively . Removing the SPV/SPD neurons did not affect PCA or dorsal sex muscle activity ( Figure 8A , B , Figure 8—figure supplements 1 , 2 ) . Thus , PCA and its targets are not downstream of SPV/SPD . To determine what could be a downstream target , we analyzed the SPC motor neuron , which shares chemical and electrical synapses with SPV and electrical synapses with SPD ( Figure 2C; Jarrell et al . , 2012 ) . 10 . 7554/eLife . 02938 . 017Figure 8 . Calcium imaging in spicule tip cut males . ( A ) % ΔF/F0 in the PCA . ( B ) % ΔF/F0 in the dorsal protractor , anal depressor , and gubernaculum erector and retractor . ( C and D ) % ΔF/F0 in the SPC . ( E ) % ΔF/F0 in the PCB . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 01710 . 7554/eLife . 02938 . 018Figure 8—figure supplement 1 . Removing the spicule sensilla has no effect on the PCA neuron Ca2+ transients . % ΔF/F0 for 30 s following insertion in PCA neurons . G-CaMP expressed in PCA using Peat-4 . Males were either operated on by having their spicule tips removed ( Spicule tips cut column ) or exposed to the operating conditions ( Spicule tips cut control column ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 01810 . 7554/eLife . 02938 . 019Figure 8—figure supplement 2 . Removing the spicule sensilla has no effect on the sex muscle Ca2+ transients . Calcium imaging in the dorsal sex muscles ( dorsal protractor , gubernaculum erector and retractor , anal depressor ) in spicule tip cut ( right column ) and non-cut controls ( left column ) . G-CaMP expressed using Punc-103E . % ΔF/F0 from insertion through 30 s of intromission . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 01910 . 7554/eLife . 02938 . 020Figure 8—figure supplement 3 . Ca2+ transients increase at a slower rate in the SPC neurons when the spicule sensilla are removed . Calcium imaging in the SPC in spicule tips cut males ( right column ) and non-operated control males ( left column ) . G-CaMP expressed from Pgar-3B . % ΔF/F0 reported from spicule insertion into hermaphrodite vulva through 30 s . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 02010 . 7554/eLife . 02938 . 021Figure 8—figure supplement 4 . Removing the spicule sensilla has no effect on the PCB neuron Ca2+ transients . Calcium imaging from Pdop-2:G-CaMP expressed in PCB . The first column shows the % ΔF/F0 for insertion and sperm release for five males . The final graph shows calcium imaging in the PCB while the male spicules are rhythmically prodding at the hermaphrodite vulva slit . The next two columns represent males that have been either surgically altered through removal of their spicule tips ( Spicule tips cut ) or males that have been placed under the operating conditions ( Mock cut ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 021 We utilized Pgar-3B:G-CaMP-expressing males to examine the effect of removing SPV/SPD on Ca2+ transients in the SPC neurons . Upon insertion , the Ca2+ transients in the SPC increased rapidly , followed by a decline to near-baseline levels around the time of sperm release ( Table 1; Figure 8C , D , Figure 8—figure supplement 3 ) . In contrast , when SPV/SPD were removed , the increases occurred at a slower rate ( Table 1; Figure 8C , D , Figure 8—figure supplement 3 ) . This suggests that the SPV/SPD are promoting SPC activity upon insertion . However , the ChR2 experiments revealed that removing the SPC alone is insufficient to reduce 475 nm wavelength light-induced ejaculation , suggesting that SPV/SPD may have more than one target . Removing the PCB neurons in conjunction with the SPC neurons was able to reduce 475 nm wavelength light-induced ejaculation in ChR2-expressing males . Thus , we analyzed how removing the SPV/SPD impacted Ca2+ transients in the PCB . We utilized Pdop-2 to drive G-CaMP expression in the PCB ( Correa et al . , 2012 ) . Ca2+ transients in the PCB displayed minor increases upon insertion , and their activity quickly dissipated ( Figure 8—figure supplement 4 ) . This activity was unaffected by SPV/SPD removal ( Figure 8E , Figure 8—figure supplement 4 ) . This result suggests that PCB does not play a significant role after insertion . However , it does synapse to the gonad and can respond to SPD activity through its electrical coupling . We propose that in SPC-ablated males , the PCB can compensate for the loss of the SPC and promote 475 nm wavelength light-induced ejaculation in ChR2-expressing males . The severing of the spicule tips , which removes the SPV and SPD neurons in adult C . elegans males , also affects the neuronal support sheath and socket cells ( Figure 4—figure supplement 1 , Figure 9—figure supplement 1 ) . In adult males , the socket cells encase the neuronal processes and sheath cells inside the spicule ( Sulston et al . , 1980 ) . Four socket cells are located on each side of the male tail , and during L4 larval development they secrete the sclerotic material that hardens to make the spicules ( Jiang and Sternberg , 1999b ) . We asked if removing these socket cells negatively impacted the male's ability to mate . We laser-ablated all eight spicule socket cell nuclei in adult males to determine how this operation impacted mating behavior . We found that insertion was unaffected ( 11 out of 27 control mock-ablated males could insert , compared to 7/22 ablated males , p value=0 . 6 , Fisher's exact test ) . However , only 1/7 ablated males released sperm into the hermaphrodite for ≥10 s; the other ablated males showed variable dysfunctions in executing or sustaining sperm release . In contrast , 73% ( n = 8/11 ) of mock-ablated males released sperm into the hermaphrodite for ≥10 s ( p value=0 . 0498 , Fisher's exact test ) . Thus , similar to cutting off the spicule tips , ejaculation was significantly reduced , indicating that the socket cells promote sperm release . However , ablating the socket cells was not equivalent to cutting off the spicule tips . In socket cell ablations , 6 out of 7 ( 86% ) males released sperm from the valve ( vs 12/12 , 100% for mock-ablated ) , compare to only 8 out of 15 ( 53% ) spicule tips-cut males ( vs 12/13 , 92% mock cut control , p value=0 . 038 , Fisher's exact test ) . Thus , removing the SPV and SPD in conjunction with the socket cells had a more significant impact on mating behavior . We favored the idea that the spicule socket cells directly function in ejaculation behaviors , but we also considered the hypothesis that perturbing these structural cells might indirectly affect the function of neurons or other cells that regulate sperm release . A prediction from this hypothesis is that during intromission and ejaculation behavior , these cells are inert and will display no obvious changes in their cellular activities . To explore these ideas , we expressed G-CaMP in the socket cells using the Pbas-1 promoter and observed the socket cells during mating behavior . bas-1 encodes the aromatic amino acid decarboxylase that converts L-DOPA to the neurotransmitter dopamine ( DA ) in neurons ( Hare and Loer , 2004 ) . We found that socket cell Ca2+ transients increased rapidly upon insertion , with a profile similar to the SPC and the gonadal valve ( Table 1; Figure 9A , Figure 9—figure supplement 2 ) . Ca2+ transients then declined to about half their peak intensity at the time of sperm release ( Figure 9A ) . This supports the hypothesis that the socket cells are directly involved in ejaculatory behaviors . We next asked through what mechanism might the neuronal support cells be promoting sperm release . 10 . 7554/eLife . 02938 . 022Figure 9 . Socket cell dopamine ( DA ) promotes sperm release . ( A ) % ΔF/F0 in the socket cells . G-CaMP is expressed from Pcat-2 . ( B ) % ΔF/F0 in the dopaminergic rays 5 , 7 , 9A . G-CaMP is expressed from Pdat-1 . ( C ) % of males that transferred sperm for ≥10 s into the uterus . ( D ) Ectopic protraction exhibited when males are placed with hermaphrodites . ( C and D ) cat-2 ( + ) indicates in what cells cat-2 was rescued . *p value<0 . 05 , **p value<0 . 005 , ***p value<0 . 0001 , Fisher's exact test . n values below the x-axis , percentages above the bars . ( E ) Time to 1st insert ( 1st ) and refractory period ( 2nd ) for cat-2 ( lf ) rescued males . cat-2 ( + ) is in all DA ray neurons . *p value<0 . 05 , Mann–Whitney test . ( F ) The time required for virgin males to insert their spicules into a hermaphrodite . x-axis is the concentration of DA on which the males mated . y-axis is the amount of time ( sec ) the males took from being placed with hermaphrodites until they inserted their spicules into the uterus . Closed symbols indicate time from placement with hermaphrodites to insertion . Open symbols indicate the male did not insert within 5 min of being placed with hermaphrodites . Bars represent the median . *p value<0 . 05 , **p value<0 . 005 , ***p value<0 . 0001 , one-way ANOVA with Bonferroni's correction . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 02210 . 7554/eLife . 02938 . 023Figure 9—figure supplement 1 . Socket cell DA regulates male mating . ( A ) Dopamine synthesis pathway . Two DA synthesis genes , bas-1 and cat-2 , are expressed in the male socket cells . Left images are DIC , right fluorescence . Sp = spicule , So = socket cells . Dorsal is up , anterior is to the left . Scale bar = 10 µM . ( B ) Images of a male expressing Pcat-2:YFP . Dorsal is up , anterior is to the right . Right , DIC images of the male tail . Left , fluorescent images of the male tail . Images of the spicule tips cut off are the same male 1 day later . Scale bar = 10 µM . ( C ) cat-2 ( lf ) males insert at a low efficiency that is rescued by cat-2 ( + ) in the ray neurons . Numbers above the bars are percentages , numbers at the x-axis indicate n . **p<0 . 005 , ***p<0 . 0001 , Fisher’s exact test . ( D ) Commencement times for males placed on the indicated concentrations of DA . Each dot represents one male and is a measurement of the time it took him , from the moment he was placed next to a 10 µM pad + hermaphrodites , to start backing along the hermaphrodite cuticle . n = 20 for each concentration . *p<0 . 05 , one-way ANOVA with Bonferroni's correction . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 02310 . 7554/eLife . 02938 . 024Figure 9—figure supplement 2 . Ca2+ transients in the DA-expressing cells in the male tail . ( A and B ) The first panel is the % ΔF/F0 in a male continuously prodding at the hermaphrodite vulva . All subsequent panels are % ΔF/F0 in males during insertion and sperm transfer into the uterus . ( A ) Ca2+ transient changes in the socket cells . G-CaMP was expressing using Pbas-1 . ( B ) Ca2+ transient changes in the DA ray neurons 5 , 7 , 9A . G-CaMP was expressing using Pdat-1 . The lines under the x-axis indicate where the male moved off the vulva . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 024 Intriguingly , the enzyme necessary to make DA , tyrosine hydroxylase ( cat-2 ) , in addition to aromatic amino acid decarboxylase ( bas-1 ) , is also expressed in these neuronal support cells ( Figure 9—figure supplement 1; Lints and Emmons , 1999; Hare and Loer , 2004 ) . Thus , we analyzed the contribution of the neurotransmitter DA in socket cell function during mating behavior . Previous work identified the importance of DA in the sex-specific bilateral sensory ray neurons R5A , R7A , and R9A in modulating tail posture , backward locomotion , and intromission attempts ( Sulston et al . , 1975; Koo et al . , 2011; Correa et al . , 2012 ) . Additionally , DA deficient mutants exhibit difficulty transferring sperm , but the DA source was not identified ( Correa et al . , 2012 ) . To determine if the DA-expressing rays might potentially promote sperm movement in addition to the socket cells , we expressed G-CaMP using the Pdat-1 promoter . We found that in contrast to the socket cells , these neurons exhibited no gross changes in Ca2+ transients after spicule insertion ( Figure 9B , Figure 9—figure supplement 2 ) . This suggests that if DA promotes ejaculation , its source would be the socket cells , not the ray neurons . To determine if DA regulates ejaculation , we looked at the behavior of tyrosine hydroxylase deficient cat-2 ( lf ) mutant males . We previously reported that these mutants execute ectopic intromission attempts on non-vulval areas of the hermaphrodite and have difficulty inserting their spicules into the vulva ( Correa et al . , 2012 ) . Additional work reported a role for DA in switching between behavioral states ( Sawin et al . , 2000; Vidal-Gadea et al . , 2011 ) . In this study , we observed that the few cat-2 ( lf ) males capable of spicule insertion displayed additional defects . Many mutant males had difficulty releasing sperm or sustaining sperm release into the hermaphrodite; some males even displayed ‘coitus interruptus’ , by pulling out their spicules from the hermaphrodite and then erratically releasing sperm onto the media . When we measured if the males can sustain sperm release into the hermaphrodite for at least 10 s , 79% of wild-type males successfully transferred sperm , compared to only 31% of cat-2 ( lf ) males ( p=0 . 02 , Fisher's exact test , Figure 9C ) . To verify that the mating defects were due to the lack of tyrosine hydroxylase and not because of some undescribed developmental defect , we asked if treating mutant adults with exogenous DA can ameliorate the cat-2 ( lf ) mating defects . We exposed cat-2 ( lf ) adult males to exogenous DA for at least 3 hr and asked if mating ability was restored . The dopaminergic rays express the DA reuptake transporter ( Jayanthi et al . , 1998; Koo et al . , 2011 ) , but the socket cells do not , leaving open the possibility that exogenous DA might preferentially restore different aspects of mating performance . As expected , adult cat-2 ( lf ) males exhibit reduced ectopic intromission attempts in response to exogenous DA ( Figure 9D ) . This result is consistent with our previous study suggesting that the dopaminergic ray neurons regulate the spicule protraction circuit . However interestingly , exogenously applied DA did not restore the ability to transfer sperm ( Figure 9C ) . To determine which dopaminergic cells promoted sperm release , we expressed cat-2 ( + ) in the dopaminergic rays and socket cells using the Pcat-2 promoter; this construct rescued both ectopic intromission attempts and ejaculation defects ( Figure 9C , D ) . We next used the Pdat-1 promoter to express cat-2 in ray neurons , but not in the spicule socket cells . Similar to exogenous DA exposure , ectopic intromission attempts were suppressed to wild-type levels ( Figure 9D ) , but sperm transfer remained abnormal ( Figure 9C ) . Thus , DA synthesis in the socket cells is necessary for sperm release in C . elegans males . Since our previous experiments established a link between sperm release and the refractory period , we then asked if the socket cell DA could also regulate the post-ejaculatory refractory period . The cat-2 ( lf ) male's intromission efficiency is severely compromised by his erratic ectopic intromission attempts . This made it difficult for us to obtain enough males to compare their refractory periods ( the interval between the 1st spicule insertion and the 2nd spicule insertion ) to wild-type males with statistical confidence . However , Pdat-1 expression of cat-2 ( + ) in the ray neurons restored the male’s intromission ability , and allowed us to score the spicule socket cell's contribution to the refractory period ( Figure 9—figure supplement 1 ) . We found that during the first mating , transgenic cat-2 ( lf ) males expressing cat-2 ( + ) in the ray neurons and spicule socket cells , or just in the ray neurons commenced mating in a similar time frame ( 51 s vs 121 s , p value=0 . 08 , Mann–Whitney test ) ( Figure 9E ) . However , transgenic males that lacked cat-2 ( + ) in the socket cells had shorter refractory periods , relative to males with cat-2 ( + ) -containing socket cells ( 518 s vs 1083 s , p value=0 . 0127 , Mann–Whitney test ) ( Figure 9E ) . Thus , DA contributes to the refractory period length . This result is consistent with our previous observations that the refractory period and the efficiency of sperm release correlate with the functions of the spicule-associated cells . Since endogenous DA extends the period of time between copulations , we asked if exogenous DA is able to extend the time required for a virgin male to mate successfully the 1st time . Exogenous DA is capable of paralyzing hermaphrodites after 20 min of exposure on an NGM plate supplemented with 25–30 mM DA ( Chase et al . , 2004 ) . To avoid paralyzing the males , we gave them 5 min to mate on plates supplemented with various concentrations of DA . We did not know which concentration would extend a virgin male’s first mating experience and therefore tested a variety of concentrations . We found that while concentrations of 15 and 20 mM DA slowed down the time it takes a male to insert his spicules , the difference was not statistically significant ( Figure 9F ) . However , 25 mM DA induces a significant increase in the time required for males to insert their spicules ( Figure 9F ) . Indeed , few males were successful in inserting their spicules within this time frame . This decreased mating ability was not due to sluggish movement or paralysis , as they still commenced mating in a similar , but slightly longer timeframe ( Figure 9—figure supplement 1 ) and their backing velocity along the hermaphrodite remained the same ( 171 ± 43 µm/s n = 6 control , 178 ± 35 µm/s n = 5 25 mM DA ) . Thus , exogenous DA is capable of interfering with males' mating ability .
Much remains to be discovered concerning the molecular and structural pathways involved in post ejaculatory behavioral activity ( Levin , 2009; Turley and Rowland , 2013 ) . Mating is energy-costly and precludes participation in other behaviors such as feeding ( Schneider et al . , 2013 ) . Thus , a period of reduced activity following successful copulation is advantageous to organisms . In rats , the post ejaculatory interval represents the period ( ∼6–10 min ) between ejaculations . After at least five successive ejaculations , males achieve a state of sexual satiation . During satiation , male rats will not mate again for 6–14 days ( Beach and Jordan , 1956; Phillips-Farfan and Fernandez-Guasti , 2009 ) . In this work , we also measured the post ejaculatory period in C . elegans males , referred to as the refractory period . However , we do not know if C . elegans males achieve a state of sexual satiation as exhibited by rodents . C . elegans males display an average lifespan of 12 days , show significant decline in sexual function at day 3 , and are unable to sire progeny by day 5 ( Gems and Riddle , 1996; Guo et al . , 2012 ) . Therefore , the advantages C . elegans males would have by exhibiting an extended satiation time after multiple intromissions are uncertain , since their functional reproductive span is very short . We discovered that a longer refractory period allows males to recover their ability to transfer sperm . When males attempt to mate again too quickly , their ability to sire progeny is reduced . Although males recover from the refractory period whether they can sire progeny or not , we favor the hypothesis that the refractory period ensures that a male does not continuously copulate with the same mate . Following ejaculation , the males' mechanosensory and chemosensory neurons are still in immediate proximity to the hermaphrodite cues . A male that reinitiates the mating program , only to intromit his spicules into the same partner , is at a competitive disadvantage to disseminate his genetic material . Thus , the period of reduced activity and ability we observed following ejaculation may help insure that the male copulates with different partners ( Figure 10A ) . 10 . 7554/eLife . 02938 . 025Figure 10 . Socket cell DA and neuronal ACh regulate ejaculation and the refractory period . ( A ) The steps of C . elegans male mating behavior . The order of the individual steps is given by the numbers . The refractory period is a period of reduced activity and mating ability following ejaculation . ( B ) Neurons , muscles , and support cells that are activated by spicule insertion . Mating behavior steps are represented by blue boxes . GER = gubernaculum erector muscle , GRT = gubernaculum retractor muscle , ADP = anal depressor muscle , DA = dopamine , ACh = acetylcholine . Glutamate is a possible PCA neurotransmitter . DOI: http://dx . doi . org/10 . 7554/eLife . 02938 . 025 We determined that the refractory period is dependent upon the males' ability to produce and sense successful ejaculation . These interconnected activities partly depend on the SPV and SPD neurons . These neurons' sensory endings are exposed to the intrauterine environment through the spicules tips ( Figure 2C; Sulston et al . , 1980 ) , allowing them to respond after insertion and during sperm release . When we cut off the spicule tips ( Figure 4—figure supplement 1 ) , removing the neuronal and support cell function , males displayed a reduced ability to transfer sperm and recovered from mating quicker . We propose that SPV and SPD sense the intrauterine environment , promote sperm release , and regulate the refractory period . The tight coupling of ejaculation and the refractory period led us to examine how sperm initiation and release are controlled in the male ( Figure 10A ) . Calcium imaging data support a role for the SPV and SPD in regulating sperm release . However , these cells do not synapse the gonad ( Figure 2C ) . Rare males lacking these cells are able to initiate sperm movement from the valve , but they poorly control the release of sperm into the hermaphrodite , and generally spill their ejaculate onto the mating lawn . Therefore , the SPV and SPD contribute to both sperm initiation and release ( Figure 10B ) , providing control of the timing of sperm movement ( Liu and Sternberg , 1995 ) . Our data indicate that the SPC neurons , with support from the SPV/SPD neurons and the post cloacal sensilla , trigger the initiation of ejaculation ( Figure 10B ) . The initial ejaculation step occurs when the valve opens , permitting sperm to move from the seminal vesicle to the vas deferens ( Figure 2B; Schindelman et al . , 2006 ) . The cholinergic SPC was previously identified as the trigger for full spicule intromission , once the hermaphrodite vulva is breached . However , since ablating the SPC results in males that cannot insert their spicules , evidence for their role in ejaculation was lacking ( Liu and Sternberg , 1995; Garcia et al . , 2001 ) . Our calcium imaging and optogenetic data reveal SPC’s role as a trigger of sperm initiation . The SPC displays a rapid increase in Ca2+ transients immediately upon insertion . Additionally , ablating SPC in conjunction with the post cloacal sensilla neuron PCB results in ChR2-expressing males that do not ejaculate in response to light activation . We propose a model where the SPC , via direct cholinergic signaling to the gonad ( Jarrell et al . , 2012 ) , triggers the initiation of ejaculation ( Figure 10B ) . This interaction allows coupling of full spicule insertion into the uterus with the initiation of sperm movement . Interestingly , we found that the rapid increase in SPC Ca2+ transients upon insertion is dependent on the SPV and SPD neurons , despite the fact that the SPV and SPD do not reach their peak activity as fast as the SPC . The SPC shares chemical synapses with the SPV and electrical synapses with the SPV and SPD ( Figure 2C; Jarrell et al . , 2012 ) . We propose that the SPV and SPD help maintain a lower threshold in the SPC , allowing the SPC to trigger spicule intromission without inducing premature ejaculation . When the spicules tips enter the uterine environment , the SPV and SPD neurons further facilitate the SPC activity to begin sperm initiation . Thus , without the SPV and SPD present , the SPC cannot reach its peak activity as quickly as normal . This model accounts for the role of SPV and SPD in SPC activity and initial sperm movement . Once sperm initiation has occurred , and sperm has traveled the length of the vas deferens to the cloaca , it must be released into the hermaphrodite uterus ( Figure 1B ) . Calcium imaging data presented in this work and ablation data in previous work suggest that the gubernaculum erector and retractor muscles , and possibly the anal depressor , are involved in the penultimate step of male mating ( Figure 10B ) . The gubernaculum erector and retractor are located dorsally to the spicules and attach the gubernaculum to the body wall ( Figure 1A ) . The gubernaculum is a thin strip of cuticle material located dorsally from the spicules and is thought to assist in positioning the copulatory structures ( Figure 1A , Figure 9—figure supplement 1; Sulston et al . , 1980 ) . Our calcium imaging of the gubernacular and anal depressor muscles reveal that they undergo contraction and display increased Ca2+ transients immediately prior to sperm transfer , suggesting this is a mechanism by which sperm is released . Previous work showed that males lacking the gubernaculum erector and retractor had difficulty transferring sperm , with sperm being stuck at the distal end of the vas deferens , unable to drain into the hermaphrodite ( Liu et al . , 2011 ) . Together , these data indicate that the gubernaculum muscles contract to facilitate sperm release , likely by re-adjusting the spicules' position . What promotes gubernacular muscle contraction and , ultimately , sperm release ? Our data suggest the post cloacal sensilla PCA is partially responsible ( Figure 10B ) . This neuron sends a process to the posterior of the cloacal opening , allowing it to sense and maintain vulva position , and also makes synapses with the gubernaculum erector and retractor ( Figure 1C; Liu and Sternberg , 1995; Liu et al . , 2011; Jarrell et al . , 2012 ) . We identified glutamate as a potential PCA neurotransmitter and showed that PCA's peak activity following intromission corresponds with gubernacular muscle contractions . Additionally , we showed that ChR2-expressing males were unable to ectopically ejaculate when PCA was removed . While ablating the PCA in a wild type mating background does not affect sperm transfer ( Liu and Sternberg , 1995; Liu et al . , 2011 ) , we propose this is due to the ability of the other p . c . s . neurons , the PCB and PCC , which make similar muscular connections ( Jarrell et al . , 2012 ) , to compensate for the loss of the PCA . Both redundancy and the ability to compensate when one function has been impaired is reported in the male mating circuit in general and the p . c . s . neurons specifically ( Liu and Sternberg , 1995; Koo et al . , 2011; Liu et al . , 2011; LeBoeuf and Garcia , 2012 ) . Additionally , p . c . s . function likely contributes to the ability of exceptional males lacking the SPV/SPD/socket cells to initiate sperm release . The synaptic connections they make to the gonad could function to prime this organ when the p . c . s . are active during prodding behavior , increasing the likelihood that impaired SPC function could nevertheless stimulate sperm initiation . Thus , the sex muscle contractions facilitated by the PCA and other neurons are a necessary component of sperm release . However , PCA activity and gubernaculum sex muscle contractions are insufficient for sperm release into the uterus . Males that display normal activity in these cells still fail to transfer sperm in the absence of the SPV/SPD and socket cells . Our data indicate that the SPV/SPD neurons are involved in sperm initiation and possibly release , while the DA-expressing socket cells play no role in initiation but are required for sperm release ( Figure 10B ) . Removal of the socket cells themselves or their ability to produce DA results in impaired sperm release . These data support an evolutionarily conserved role for DA in promoting ejaculation ( Peeters and Giuliano , 2008 ) . This role of DA could represent a general theme of monoaminergic modulation of sexual motor acts , as there is also peripheral modulation of the corpus cavernosum in mammals by monoamines , including DA and serotonin ( Angulo et al . , 2001; Hayes and Adaikan , 2002 , d'Emmanuele di Villa Bianca et al . , 2005; Senbel , 2011 ) . Interestingly , the socket cells represent a non-neuronal source of DA . The socket cells in C . elegans are proposed to be invertebrate glia ( Ward et al . , 1975; Oikonomou and Shaham , 2011 ) . Since impairing glia function is known to negatively impact C . elegans neuronal function ( Bacaj et al . , 2008; Felton and Johnson , 2011 ) , we cannot rule out that disrupting socket cell function impairs the SPV/SPD . To our knowledge , a role for neuronal support cells in either DA production or ejaculatory behavior has not been previously identified . Additionally , socket cell DA may function hormonally , as these cells do not express the DA re-uptake transporter . Vesicles have been identified in socket cells in the hermaphrodite head , suggesting this cell type is secretory ( Doroquez et al . , 2014 ) . Extrasynaptic roles for DA , as important modulatory components of neurotransmission , have been identified in both C . elegans and other organisms , and DA receptors appear on tissues throughout animals ( Sargent et al . , 1977; Chase et al . , 2004; Asano et al . , 2012; Fuxe et al . , 2012 ) . Our results suggest a functional role for extraneuronal and extrasynaptic DA in coordinating ejaculation . Additionally , our experiments show that DA promotes the refractory period , the time between successful mating attempts , coupling sperm release with mating-related behavioral inhibition . We propose a model where the DA that promotes ejaculation , either humorally or synaptically , acts to reduce the activity of the post-copulatory male . Once the effect of DA is slowly extinguished , the male is then able to mate again . The refractory period could indirectly cause males to mate with different hermaphrodites . However , the refractory period could also provide the gonad and male mating circuitry time to recover , as we observed that mating drive returns before mating potency ( Figure 10A ) . This raises the question of what does the male mating circuitry need to recover from ? One possibility is that different circuits involved in mating need varying times to re-set following the activities necessary for successful ejaculation . Strikingly , we noticed that a large number of neurons and other cells show a dramatic increase in calcium transients upon spicule insertion . This suggests that a large amount of neurotransmitters , neuropeptides , and hormones are released at intromission . While the large amount of information contained within these signals is required for sperm release , over a period of time they may have a dampening effect on the circuitry . Many receptors undergo periods of desensitization , and the large amounts of signals released near-simultaneously might compound this effect . The re-setting of circuits could require the removal of signaling molecules that inundate the male's nervous system during copulation . Thus , the receptors could return to functional levels required for subsequent mating . Additionally , calcium transients oscillate in the sex muscles following the initiation of sperm movement . While we do not know the purpose for these fluctuations , they could be in response to the signals released upon intromission , and their dampening is likely required for re-insertion . Thus , the high amount of circuit activity required for ejaculation might necessitate a similar period of reduced activity to restore proper circuit function . We propose that the SPV and SPD neurons not only promote sperm initiation but regulate the length of the refractory period . When these neurons' sensory endings enter the uterus , their activity increases and consequently sperm initiation occurs . However , their activity decreases when they sense sperm release ( Figure 10B ) . Consequently , the activity of the ejaculation circuit , which includes the SPC and socket cells , is also reduced . If the SPV and SPD do not sense intrauterine sperm , then their extended activity subsequently promotes the continued release of DA and other neurotransmitters . This could result in the lengthening of the refractory period . This hypothesis is based on the interesting observation we obtained when we prevented sperm from draining into the hermaphrodite but did not interfere with SPV and SPD function . This experiment resulted in a subpopulation of males that displayed an extended refractory period . Thus in intact wild-type males , if the SPV and SPD do not efficiently sense sperm in the uterine environment , a possible cause could be due to a low ejaculate sperm count . Consequently , lengthening the refractory period would allow time for the males to produce sperm . In mammals , post-ejaculatory regulation of the refractory period occurs in networks in the hypothalamus , amygdala , and septal nuclei ( Gogate et al . , 1995; Parfitt and Newman , 1998; Dominguez and Hull , 2010 ) . In contrast , we identified that the refractory period in C . elegans is modulated by the circuitry located in the male tail . However , we cannot rule out a role for the head ganglia in regulating both ejaculation and the refractory period . We were unable to achieve optogenetically controlled ejaculation unless ChR2 was expressed in cholinergic head neurons . This suggests that , similar to other species , males' sexual activities are modulated by neuronal networks that are not directly associated with the sex organs . Additional parallels can be drawn between the circuits regulating ejaculation in mammals and the circuits we have discovered in C . elegans . The dorsal nerve in rats and humans that is part of the copulatory organ elicits ejaculatory responses ( Pescatori et al . , 1993; Wieder et al . , 2000 ) . This signal is sent to lumbar spinothalamic ( LSt ) neurons in the spinal cord ( Truitt and Coolen , 2002 ) . LSt cells then signal the neurons that control the emission and expulsion phases of ejaculation . The emission phase involves the secretion and movement of seminal fluids to the proximal urethra via seminal vesicle and vas deferens contraction . Once the ejaculate is in position , it is ejected from the penis in the expulsion phase that includes the rhythmic contractions of smooth muscle ( Coolen , 2005 ) . C . elegans ejaculatory circuitry is similarly set up ( Figure 10 ) and allows us to expand the general understanding of circuit control of ejaculatory and post-ejaculatory behaviors .
Animals were maintained on NGM agar plates with E . coli strain OP50 and contain him-5 ( e1490 ) for their high instance of males ( Brenner , 1974; Hodgkin et al . , 1979 ) . Strains used in this study were: let-23 ( sy1 ) ( Aroian and Sternberg , 1991 ) on LGII , pha-1 ( e2123 ) ( Schnabel and Schnabel , 1990 ) and unc-64 ( e426 ) ( Brenner , 1974 ) on LGIII , rgIs6[Punc-17 small:ChR2::YFP , Punc-103B:ChR2::YFP , Pgar-3B:ChR2::YFP] on LGIV , and lite-1 ( ce314 ) on LGX ( Edwards et al . , 2008 ) . Transgenic strains include: rgEx494[Punc-17 small:ChR2::YFP , Punc-103B:ChR2::YFP] , rgEx501[Punc-17 small:ChR2::YFP , Pgar-3B:ChR2::YFP] , rgEx485[Punc-17 small:ChR2::YFP] , rgEx496[Punc-17 small:ChR2::YFP , Punc-103B:ChR2::YFP , Pgar-3B:ChR2::YFP TL1] , rgEx506[Punc-17 small:ChR2::YFP , Punc-103B:ChR2::YFP , Pgar-3B:ChR2::YFP TL2] , rgEx551[Punc-17 small:ChR2::YFP] , rgEx480[Punc-103B:ChR2::YFP TL5] , rgEx481[Punc-103B:ChR2::YFP TL2] , rgEx502[Punc-103B:ChR2::YFP , Pgar-3B:ChR2::YFP TL2] , rgEx498[Punc-103B:ChR2::YFP , Pgar-3B:ChR2::YFP TL1] , rgEx354[Pgar-3B:ChR2::YFP TL3] , rgEx509[Pgar-3B:G-CaMP3::SL2:::mDsRed] , rgEx576[Pgpa-1:G-CaMP3::SL2:::mDsRed] , rgEx546[Ptry-5:G-CaMP3::SL2:::mDsRed] , rgEx575[Peat-4:G-CaMP3::SL2:::mDsRed] , rgEx578[Pocr-2:G-CaMP3::SL2:::mDsRed] , rgEx560[Ppkd-2:G-CaMP3::SL2:::mDsRed] , rgEx513[Pdat-1:G-CaMP3::SL2:::mDsRed] , rgEx567[Punc-103E:G-CaMP3::SL2:::mDsRed] , rgEx430[Plev-11:G-CaMP1 , Plev-11:mDsRed] , rgEx623[Pbas-1:G-CaMP6M::SL2:::mDsRed] , rgEx590[Pbas-1:CFP] , rgEx624[Pcat-2:YFP] , rgEx658[Peat-4:G-CaMP6M::SL2:::mDsRed] , rgEx642[Pdop-2:G-CaMP6M::SL2:::mDsRed] . All transgenic strains include pha-1 ( lf ) ; lite-1 ( lf ) and pha-1 ( + ) rescue . The strains rgEx629 , rgEx630 , rgEx628[Pcat-2:cat-2::SL2:::GFP] , rgEx641[Pcat-2:cat-2::SL2:::GFP] , rgEx637[Pdat-1:cat-2::SL2:::GFP] , and rgEx654[Pdat-1:cat-2::SL2:::GFP] include cat-2 ( e1112 ) ;pha-1 ( lf ) and pha-1 ( + ) rescue . All males contain a mutation in lite-1 , which encodes a short wavelength ( 475 nm , blue ) light receptor ( Edwards et al . , 2008 ) . Without this mutation , 475 nm wavelength light activates an avoidance pathway that reduces the efficiency of ChR2 stimulation . In non-integrated lines , brightly expressing L4 males were selected and incubated overnight on plates containing all-trans retinol ( A . G . Scientific , San Diego , CA ) . Plates were spread with freshly prepared 50 µM all-trans retinol in OP50 . For integrated lines , no pre-selection for the brightest males was performed; L4 males were selected and placed on plates containing all-trans retinol . The next day , one male at a time was placed on a fresh plate containing retinol . The 475 nm wavelength light was turned on , and the amount of time the males protract their spicules over 60 s was recorded . In males that fully protracted their spicules for 60 s , the 475 nm wavelength light was left on until they ejaculated , retracted their spicules , or 5 min had passed , whichever came first . At least two independently obtained transgenic lines were analyzed for each promoter:ChR2 construct ( s ) . 475 nm wavelength stimulated ejaculation is a rare event . To properly analyze this response , we utilized binary categorical data that allowed us to determine how the number of ‘yes’ ( males did ejaculate ) and ‘no’ ( males did not ejaculate ) responses varied between genotypes . We employed Fisher's exact test to analyze our contingency tables . Fisher's exact test is used to determine the p value for small sample sizes . This test is employed on data throughout the manuscript where we wanted to determine if males of a given population could execute a specific behavior that depended only on the male being successful ( a yes/no question ) and not on the amount of time the male took . Plasmids were co-injected with pBX1 , which contains wild-type pha-1 ( Schnabel and Schnabel , 1990 ) , into pha-1 ( e2123ts ) ;lite-1 ( ce314 ) hermaphrodites unless otherwise indicated . Injected animals were incubated at 20°C; only animals containing a transgene expressing pha-1 were able to survive to adulthood . pBL240 , pLR289 , pLR283 , pBL255 , pLR160 , pJM1 , pBL285 , pBL267 , pPC88 , and pLR183 were injected at 50 ng/µl . pJM3 was injected with no other ChR2 expressing plasmid at 150 ng/µl . pJM3 was injected with pJM1 and pLR183 at 50 ng/µl . pBL247 , pBL256 , and pBL257 were injected at 100 ng/µl . pBL270 was co-injected with pBL279 at concentrations of 5 ng/µl and 0 . 5 ng/µl , respectively . All G-CaMP lines were screened for their ability to mate . pBL290 and pBL291 were injected at 50 ng/µl into cat-2 ( e1112 ) ;pha-1 ( e2123ts ) . At least two transmitting lines were scored for behavior . Transgenic lines containing pBL290 and pBL291 express GFP in the same cells as cat-2 ( + ) , allowing us to screen for and remove mosaic animals . rgEx496[Punc-17 small:ChR2::YFP , Punc-103B:ChR2::YFP , Pgar-3B:ChR2::YFP] was integrated using the procedure described in Mello and Fire ( 1995 ) . Briefly , late L4 to early adult worms were exposed to UV at 340 µW/cm2 after being incubated with 4 , 5 , 8-trimethylpsoralen for 15 min . Irradiated worms were then transferred to NGM agar plates plus Escherichia coli OP50 and allowed to lay progeny . ∼2000 individual F2 hermaphrodites were picked and progeny were screened for complete penetrance of fluorescence . Seven integrated lines were obtained , of which five ectopically ejaculated in response to 475 nm wavelength light . The line with the highest level of ejaculation in response to 475 nm wavelength light was rgIs6 , used for further studies . rgIs6 was mapped to LGIV using the procedure described in Wicks et al . ( 2001 ) . Cell ablations were performed as described in Bargmann and Avery ( 1995 ) . We used a Spectra-Physics VSL-337ND-S Nitrogen Laser ( Mountain View , CA ) attached to an Olympus BX51 microscope ( Olympus , PA ) . Ablations were done on 2 . 5% noble agar pads in 50% water and 50% M9 . Ablations at L1 were done using 4 mM NaN3 , at L4 using 12 mM NaN3 , and in adults using 15 mM NaN3 . Non-operated control males were placed on pads+NaN3 for the same amount of time . To remove the somatic gonad plus germ line , we ablated the gonad precursor cells Z1-4 in L1 worms , and to remove only the germ line we ablated the precursor cells Z2-3 ( Kimble and Hirsh , 1979 ) . The phenotypes we saw in ablated males are not the result of collateral laser damage to the genitalia structures that occurred during the operation . The operation was conducted at the L1 larval stage , and genital structures developed two to three days later during the L4 stage . Any animal that displayed development defects due to anesthetic toxicity or collateral laser damage was not used in the assays . In adult males that had their spicule tips cut , only the very tip of the spicules was removed with the laser . We expressed CFP in the SPV , SPD , and socket cells to determine that using the laser to remove the spicule tips of adult males resulted in fluorescent cytoplasm leaking out the surgically enlarged spicule opening . Fluorescence in the cell body does not recover ( Figure 9—figure supplement 1 ) . Additionally , we discovered that the operation procedure has a significant impact on the refractory period in adult males ( Figure 4—figure supplement 1 ) . Placing adult males on 2 . 5% noble agar pads without azide to paralyze the worms is sufficient to significantly reduce the refractory period ( Figure 4—figure supplement 1 ) . To account for this issue , we picked L4 males in the morning and ablated them when they became adults , ∼6 hr later , and allowed the males to recover from the operation overnight . This increased the refractory period length in mock ablated males . The morning of the experiment , 10 µl of a saturated E . coli OP50 culture was spotted on an NGM agar plate and allowed to dry . 15 two-days-old unc-64 ( lf ) hermaphrodites were transferred to this plate and allowed to incubate for at least 1 hr , to ensure the males would respond to the hermaphrodites . The E . coli + hermaphrodites section of the plate was cut out and placed on a microscope slide and transferred to an Olympus BX51 microscope . Recordings were then made using a Hamamatsu ImagEM Electron multiplier ( EM ) CCD camera ( Japan ) of the male mating . Each recording was then analyzed for the mating behavior of interest . Images were taken with either an Olympus BX51 microscope and Hamamatsu ImagEM Electron multiplier ( EM ) CCD camera or with a Olympus IX81 microscope , csu-xi Yokogawa spinning disk , and Andor iXon EM CCD camera . Recordings that managed to capture at least a few frames prior to insertion through intromission and sperm transfer were used for analysis . Using the Hamamatsu SimplePCI ( version 6 . 6 . 0 . 0 ) software , a region of interest ( ROI ) was centered on the cell ( s ) or tissues of interest in both the G-CaMP and mDsRed channels . Additional ROIs of the same size were placed on the E . coli lawn to record the background fluorescence levels for each image . The mean gray level of each ROI was calculated for each frame , with the ROIs being moved frame-by-frame as necessary to keep the cells or tissues of interest inside the ROI as the male moved . Even though the male was situated at the vulva of an unmoving hermaphrodite throughout the recording , we found that even slight changes of position in the male tail necessitated moving the ROI . We were able to correlate fluorescent changes with behavior since both the spicules and sperm are auto fluorescent , allowing us to determine spicule insertion and sperm entry into the vulva . The data were then transferred to Microsoft Excel to calculate the percent change in fluorescence ( ΔF ) over initial fluorescence ( F0 ) ( Correa et al . , 2012 ) . Photobleaching and other experimental artifacts were accounted for as in Correa et al . ( 2012 ) . While photobleaching is not significant in G-CaMP over the period of time utilized for the experiment , it is significant for mDsRed . mDsRed fluorescent decay was plotted with respect to time . A decay curve was fit to this line and used to adjust the mDsRed fluorescent levels to remove artifacts caused by bleaching . These adjusted red values were used to determine a green:red fluorescent ratio and the %ΔF/F0 for each cell or tissue of interest was then plotted with respect to time . attb1unc17pfullfor: GGGGACAAGTTTGTACAAAAAAGCAGGCTGTGTATGGTGGTGGAGCATTCGACATattb1unc17pfor: GGGGACAAGTTTGTACAAAAAAGCAGGCTTGCAGACTTTTCCCCAAACTAGCattb2unc17pbac: GGGGACCACTTTGTACAAGAAAGCTGGGTGACGGGCACGTTGAAGCCCAACTATTB1gpa1pro: gggg aca agt ttg tac aaa aaa gca ggc tcatgacttttggtatttctttgcagaaactcgcgATTB2gpa1prodwn: Ggg gac cac ttt gta caa gaa agc tgg gtctgaagtcttcgaataaatgacattgaataatattgFpcat-21 . 4: GGGGACAAGTTTGTACAAAAAAGCAGGCTGCTCAAAAAGAAAATCCGATTTAAATGTCTCPcat-2r: GGGGACCACTTTGTACAAGAAAGCTGGGTCTGATCGGTGAGCTGTTTTCGGTGTTGPeat-4 ( 5 kb ) : GGGGACAAGTTTGTACAAAAAAGCAGGCTCCCCTCAGGCAAGCACAAAGAAGAAGAATattb1Peat-4R: GGGGACCACTTTGTACAAGAAAGCTGGGTAGGTTTCTGAAAATGATGATGATGATGATGGFpbas-1: GGGGACAAGTTTGTACAAAAAAGCAGGCTGACTTCCGCGAATCCCCATCCPbas-1r: GGGGACCACTTTGTACAAGAAAGCTGGGTTATACCGAACTACTACTGAAAGTTCGACfcat2pyl30: TACAAAGTGGTGATCATGAGGTGTCAGAAGGTACAGTAATCCcat2pyl30r: TTGGGGGATCCTCATTCACATTGTAATCGATATTTTCATCCGATCfpyl30: ATGAGGATCCCCCAACAGAGTTGTTGATCpyl30r: GATCACCACTTTGTACAAGAAAGCTGAACGForUp GCAMP: ATGGTCGACTCATCACGTCGTAAGTGGAATAAGACAGGTCRevUp GCAMP: GACCTGTCTTATTCCACTTACGACGTGATGAGTCGACCATFordownGCAMP: AAACTACGAAGAGTTTGTACAAATGATGACAGCGAAGTGARevdownGCAMP: TCACTTCGCTGTCATCATTTGTACAAACTCTTCGTAGTTT | The nematode worm , C . elegans , is roughly 1 mm long , made up of around 1000 cells and has two sexes: male and hermaphrodite . Hermaphrodite worms produce both eggs and sperm and can self-fertilize to generate around 300 offspring each time . Fertilization by a male , on the other hand , results in three times as many progeny and introduces genetic diversity into the population . However , it also reduces the lifespan of the hermaphrodite . Mating also incurs a cost for males: it requires a lot of energy , which prevents male works from engaging in other activities , such as feeding , and it also increases their risk of predation . In many species , including C . elegans , the frequency with which a male can mate is limited by a period of reduced mating drive and ability that follows each instance of successful mating . However , the molecular and cellular basis of this ‘refractory period’ remains largely unclear . Using a range of techniques , LeBoeuf et al . have now identified the circuits that regulate male mating behavior in C . elegans . When male worms were introduced into a Petri dish containing 15 hermaphrodites , most males initiated mating within about 2 min . The length of the refractory period varied between worms , but averaged roughly 12 min . This consisted of a period of disinterest , in which males did not approach hermaphrodites , followed by a period in which males attempted mating but were slower and less efficient , suggesting that the neural circuits controlling mating behaviors had yet to recover completely . Males with longer refractory periods produced more progeny in their second mating than those with shorter refractory periods , suggesting that the interval also enables males to replenish their sperm levels . Further experiments revealed that a chemical transmitter called dopamine promotes ejaculation and then immediately reduces the worm's activity levels , giving rise to the refractory period . By enforcing a delay between matings , the refractory period may also increase the likelihood that successive matings will be with different hermaphrodites , helping to maximize the number and diversity of offspring . Some aspects of the neural circuitry that controls the refractory period in C . elegans resemble those seen in mammals , suggesting that insights gained from an animal with 1000 cells could also be relevant to more complex species . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | Caenorhabditis elegans male sensory-motor neurons and dopaminergic support cells couple ejaculation and post-ejaculatory behaviors |
Evolution has often copied and repurposed the mitogen-activated protein kinase ( MAPK ) signaling module . Understanding how connections form during evolution , in disease and across individuals requires knowledge of the basic tenets that govern kinase-substrate interactions . We identify criteria sufficient for establishing regulatory links between a MAPK and a non-native substrate . The yeast MAPK Fus3 and human MAPK ERK2 can be functionally redirected if only two conditions are met: the kinase and substrate contain matching interaction domains and the substrate includes a phospho-motif that can be phosphorylated by the kinase and recruit a downstream effector . We used a panel of interaction domains and phosphorylation-activated degradation motifs to demonstrate modular and scalable retargeting . We applied our approach to reshape the signaling behavior of an existing kinase pathway . Together , our results demonstrate that a MAPK can be largely defined by its interaction domains and compatible phospho-motifs and provide insight into how MAPK-substrate connections form .
The MAPK family of proteins is a ubiquitous signaling element in eukaryotes , and is essential to the function of a wide variety of cellular behaviors , from the regulation of differentiation and proliferation to stress responses and more ( Cargnello and Roux , 2011 ) ; this diversity of functions has been made possible by the evolutionary expansion of the MAPK repertoire ( Caffrey et al . , 1999; Manning et al . , 2002 ) . For the expansion of the MAPK signaling module to have been feasible , it needed to be amenable to forming new kinase-substrate regulatory links , while at the same time having the capacity to avoid unwanted crosstalk . However , it still remains unclear what information is sufficient to create an entirely new set of regulatory interactions . One way to understand how potentially large numbers of novel regulatory links can be established is by developing a scalable method to create such links ourselves ( Elowitz and Lim , 2010 ) . What are the core components necessary for the formation of a new – functional – kinase-substrate interaction ? Following the association of the kinase and substrate , the amino acids in the immediate vicinity of the phosphorylated residue – together making up the ‘phospho-motif’ – help to dictate whether the substrate is phosphorylated by the kinase ( Mok et al . , 2010; Howard et al . , 2014 ) . However , it is the site that is phosphorylated – rather than the kinase itself – that mediates the functional outcome of kinase regulation . In particular , the phosphorylated phospho-motif can be recognized by a regulatory protein bearing a phospho-motif binding domain and control protein localization or degradation among many other effects ( Seet et al . , 2006; Bhattacharyya et al . , 2006 ) . Even before the kinase has a chance to interact with the phospho-motif , the two proteins must be colocalized ( Ubersax and Ferrell , 2007 ) . Residues apart from the kinase active site are frequently responsible for recognizing a substrate; indeed , several studies have sought to modify or replace these residues in a variety of kinases to redirect them to new – but still related – targets ( Skerker et al . , 2008; Won et al . , 2011; Grewal et al . , 2006 ) . Adaptor proteins , such as synthetic scaffolds , have also been used to steer a kinase towards a particular native substrate ( Park et al . , 2003; Whitaker et al . , 2012; Hobert and Schepartz , 2012; Harris et al . , 2001 ) . Regulation of a modified native substrate by a kinase can also be rescued using a pair of completely heterologous interaction domains ( Yadav et al . , 2009 ) . These studies show that by controlling the colocalization of a kinase with a native – or closely related – substrate allows the functional regulation of that target . Taking it a step further , two groups have recently used native MAPK-interacting motifs – ‘docking domains’ – to allow several types of MAPKs in mammalian cells and yeast to regulate the nuclear localization of fluorescent reporters ( Regot et al . , 2014; Durandau et al . , 2015 ) . Although it is generally accepted that docking domains primarily control colocalization ( Sharrocks et al . , 2000 ) , several studies suggest that binding may also serve to allosterically regulate the MAPK ( Chang et al . , 2002; Heo et al . , 2004; Zhou et al . , 2006; Tokunaga et al . , 2014; Bhattacharyya et al . , 2006 ) . As such , the precise role of these interactions remains unclear . Regardless , the question of how completely new and orthogonal regulatory relationships are created remains . Like the signaling modules that have been expanded in natural systems , engineered genetic circuits also rely on components that are amenable to rewiring . The creation of novel transcription factors has been successful in a large part because the necessary functional characteristics have been identified . Importantly , these characteristics can be embodied in distinct modular DNA and protein domains , such as promoters , transcriptional-regulation domains , and DNA-binding domains – these domains can then be mixed and matched to yield the desired connectivity and regulation ( Khalil et al . , 2012; Stanton et al . , 2014; Kiani et al . , 2014; Zalatan et al . , 2015; Khakhar et al . , 2015 ) . Although hurdles to creating large genetic circuits remain ( Brophy and Voigt , 2014; Cardinale and Arkin , 2012 ) , these parts have allowed scientists to construct and interrogate more complex engineered and naturally occurring genetic systems ( Prindle et al . , 2012; Gilbert et al . , 2014 ) . Unfortunately , our understanding of how to assemble modular post-translational signaling proteins lags behind . At the same time , recent work with engineered modular receptors expressed on T-cells has shown the considerable power of the ability to rationally design even relatively simple post-translational signaling systems ( Wu et al . , 2015; Roybal et al . , 2016; Morsut et al . , 2016 ) . Targeting a kinase to a new substrate is an essential step towards creating modular kinase signaling systems . As discussed above , Regot et al . and Durandau et al . have described an approach wherein a kinase-specific docking domain can be used to direct a particular kinase to a new substrate—a powerful tool for interrogating natural kinase signaling systems ( Regot et al . , 2014; Durandau et al . , 2015 ) . However , the number of naturally occurring kinase-substrate docking interactions inherently limits the scalability of the approach . For example , a given kinase ‘module’ cannot be reused in parallel signaling pathways , because it would not be able to distinguish between downstream targets in one pathway versus another . To overcome this limitation , it would be useful to be able to tease apart the ‘targeting’ module of the kinase from the ‘enzymatic’ module—and likewise , the ‘targeting’ and ‘effector’ modules of the substrate . If these functions can be defined as separable parts , the enzymatic module of a kinase would be available for reuse in orthogonal pathways , just by pairing it with unique targeting domains . We have used simple , single-function modular protein domains to explicitly test the requirements for allowing a MAPK to regulate an arbitrary substrate protein . We utilized modular interaction domains to co-localize Fus3 – the terminal MAPK of the mating pathway of the yeast Saccharomyces cerevisiae – with a substrate of interest . To link phosphorylation of the substrate to a meaningful regulation event we utilized phosphorylation-activated ubiquitination-based signaling motifs—phosphodegrons . We re-targeted Fus3 to regulate several disparate proteins to determine the flexibility of the substrate design rules . Likewise , to determine whether this approach generalizes to other MAPKs , we retargeted a constitutively active version of the mammalian MAPK , ERK2 , to functionally regulate a fluorescent reporter in yeast . We explored the effect that synthetically implemented post-translational regulatory connections could have on the signaling of an endogenous kinase cascade in yeast . Our results demonstrate that these new connections can be used to alter the natural signaling behaviors , damping signal amplification and even yielding concentration-based band-pass filtering . Taken together , in this paper , we define a modular set of scalable components that can be utilized to rewire MAPKs to regulate proteins through ubiquitination . Attempting to rationally design new kinase-substrate regulatory links not only sheds light on the natural processes , but also serves as the foundation for the construction of synthetic kinase signaling pathways , and with them the control of cell behaviors in biomedical or biotechnological applications .
To test whether a direct interaction – along with a functional phospho-motif – can render an arbitrary protein a substrate for a MAPK , we used the yeast MAPK Fus3 to target and regulate a fluorescent reporter protein . Fus3 is easily triggered using the yeast mating pheromone , α-factor . α-factor signals to the central MAPK kinase cascade via a surface-associated receptor; signaling through the pathway activates Fus3 , which in turn mediates signaling to a myriad of downstream effectors , directly regulating protein function and gene expression ( Figure 1A ) ( Bardwell , 2004 ) . 10 . 7554/eLife . 15200 . 003Figure 1 . Rewiring the mating cascade MAPK , Fus3 , to regulate the degradation of YFP . ( A ) The core components of the yeast mating cascade . The yeast mating factor – α-factor – triggers the sequential activation of the kinases Ste11 and Ste7 ( rounded gray rectangles ) followed by the MAPK , Fus3 ( yellow ) . Arrows with red circles denote phosphorylation-mediated regulation . All three kinases are organized on the scaffold Ste5 ( also gray ) . Among other effectors , Fus3 activates the transcription factor Ste12 ( rounded gray box ) . ( B ) Fus3 targeted regulation of YFP ( green ) . The colocalization was controlled by the addition of the mPDZ domain to YFP and a PDZ ligand to Fus3 ( light blue ) . Degradation was mediated by the addition of a phosphodegron derived from the transcription factor Tec1 ( purple ) . Upon activation of the mating pathway , Fus3 phosphorylates the phosphodegron fused to YFP , resulting in the recruitment of an E3 ubiquitin ligase and the ubiquitination and subsequent degradation of YFP . ( C ) Cells bearing the modified Fus3 and either the fully functional system , a reporter construct with an inactivated phosphodegron , a Fus3 with its kinase activity knocked out or an unmatched interaction domain ( an SH3 domain instead of mPDZ ) were grown to log phase and induced with 10 μM α-factor ( blue histograms ) or un-induced ( gray histograms ) . Data shown are from 3 hrs post-induction . The vertical dashed black lines on the histograms represent medians of treated populations and solid black lines represent medians of untreated populations . In all figures , the fluorescence has been normalized to the cell size ( see Figure 1—figure supplement 1 ) . Full time-course experiments appear in the supplement to Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 00310 . 7554/eLife . 15200 . 004Figure 1—figure supplement 1 . Reducing the variability of single-cell fluorescence by accounting for cell-to-cell variation in cell size . In yeast constitutively expressing a single-copy fluorescent protein inserted in the genome , fluorescence ( FL1-A ) is strongly correlated with cell size ( approximated by FSC-A ) —shown by R2 values . This is true in both cells that are untreated ( top left ) or treated with 10 µM α-factor ( bottom left ) . The effect is likely due to the way that flow cytometers measure fluorescence , where cells with the same concentration of fluorescent protein – but with different volumes – will have different levels of fluorescence . For example , smaller cells that have just divided will have a lower fluorescence value than larger cells that are just about to divide . Normalizing by cells size – dividing FL1-A by FSC-A – reduces the coefficient of variation ( CV ) of the fluorescent signal by ~67% or ~41% in untreated or treated cells , respectively ( graphs on right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 00410 . 7554/eLife . 15200 . 005Figure 1—figure supplement 2 . Western Analyses of degradation assays . The representative western blots above show results from degradation assays on a test strain with the entirely functional system described in Figure 1 and a control strain that has a non-functional phosphodegron . We observe that upon treatment with α-factor , the test strain has a significantly fainter band as compared to the untreated lane , whereas the control strain does not . This is consistent with our flow cytometry observations . We also observe that when cells are treated with MG132 , a proteasome inhibitor , the alpha-factor triggered degradation in the test strain is prevented , implying that the degradation we observe is indeed due to the proteasome as hypothesized . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 00510 . 7554/eLife . 15200 . 006Figure 1—figure supplement 3 . Swapping interaction domains between kinase and substrate . Population histograms and medians of YFP fluorescence signal normalized by cell size for yeast strains that have either a Fus3 kinase fused to a PDZ ligand and YFP fused to a PDZ domain ( right ) or Fus3 kinase fused to a PDZ domain and YFP fused to a PDZ ligand . These strains were diluted down from saturated overnights and were grown for 5 hrs to log phase and then cytometry reads were performed after 2 , 3 and 4 hrs post induction with α-factor . The dashed black lines represent the median fluorescence at 4 hrs and the solid black lines are the media fluorescence at 0 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 00610 . 7554/eLife . 15200 . 007Figure 1—figure supplement 4 . Fusing interaction domain to the native copy of the kinase . Population histograms ( bottom ) and corresponding medians ( top ) of YFP fluorescence signal normalized by cell size for yeast strains where the native copy of the kinase is fused to a PDZ ligand . The left most panel describes a strain with just the YFP substrate , the middle one represents a strain where the endogenous copy of Fus3 had an interaction domain fused to it but the YFP substrate has a non-functional degron and the rightmost panel describes a strain that has both an interaction domain on the Fus3 and a functional degron on the YFP substrate . These strains were diluted down from saturated overnights and were grown for 5 hrs to log phase and then cytometry reads were performed at 15 min intervals post induction with α-factor . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 007 Given the important role ubiquitin-based degradation plays in signaling ( Hunter , 2007; Swaney et al . , 2013 ) , we decided to use a phosphodegron as the regulated phospho-motif . Upon phosphorylation , the phosphodegron interacts with a specialized F-box protein – Cdc4 – to recruit the E3 ubiquitin ligase machinery ( the SCF complex ) , which then marks the substrate for degradation by covalently attaching a poly-ubiquitin chain ( Figure 1A ) ( Skaar et al . , 2013 ) . A phosphodegron has the added benefit of making a functional phosphorylation event easy to observe: if the substrate protein is a fluorescent reporter , such as YFP , phosphorylation and subsequent ubiquitination is followed by a decrease in YFP fluorescence . Thus , this approach is amenable to high-throughput measurements in a way that changes in localization may not be . To start , we wanted to use a phosphodegron that was proven to be both functional and compatible with Fus3 . The transcription factor Tec1 fulfills these criteria , as it has been shown to be both a substrate for Fus3 and Cdc4 ( Chou et al . , 2004; Bao et al . , 2010 ) —thus , we chose a region of Tec1 that encompassed several residues up and downstream of the Cdc4 consensus sequence ( 37 residues , total ) ( Nash et al . , 2001; Orlicky et al . , 2003 ) . Also , since Cdc4 primarily acts in the nucleus , we added a nuclear localization signal derived from SV40 large T-antigen to the N-terminus of the protein ( Blondel et al . , 2000; Kalderon et al . , 1984 ) . To complete our synthetic substrate , we needed to control its interaction with an engineered kinase . To this end , we added the mPDZ domain to the YFP-degron fusion , a modular protein interaction domain that has been used in a variety of different contexts ( Dueber et al . , 2009; Moon et al . , 2010; Ryu and Park , 2015 ) . To target Fus3 to the new substrate , we fused the complementary interaction domain , the PDZ ligand , to its C-terminus ( Figure 1B ) . As in all the following experiments , these constructs were integrated as a single copy into the haploid yeast genome . Moreover , since we were only concerned with whether our modified Fus3 construct was able to functionally target our new YFP substrate – and not the behavior of other effectors downstream of the mating pathway – we did not remove the native FUS3 gene—thus , our modified Fus3 construct operated in parallel with the native Fus3 . Following the induction of the mating pathway with 10 µM α-factor , we measured the YFP fluorescence of the cells using flow cytometry—to account for variation caused by cell-to-cell differences in cell size , we normalized the fluorescent signal by cell size ( Figure 1—figure supplement 1 ) . We observed a ~3 . 7-fold drop in the yeast strain containing both the Fus3-mPDZ ligand fusion and our new YFP-degron-mPDZ construct ( Figure 1C ) . On the other hand , the drop in fluorescence was not observed when the phospho-acceptor residues in the degron ( two threonine residues ) were changed to methionine and alanine ( Bao et al . , 2010 ) , when the catalytic site of the targeted kinase was inactivated with a K42R mutation ( Gartner et al . , 1992 ) , or when the interaction domain fused to YFP was changed to an SH3 domain . The latter suggests that the Tec1 degron is not able to directly recruit Fus3 to the YFP construct on its own . Finally , we also found that the drop in YFP fusion protein level was sensitive to the presence of the proteasomal inhibitor MG132 , strongly suggesting that the construct was indeed being tagged and actively degraded ( Figure 1—figure supplement 2 ) . We also explored whether our rewiring approach was sensitive to which protein — the substrate or the kinase — the respective interaction domains were fused to . We built yeast strains in which the interaction domains were flipped—with the Fus3 kinase fused to the mPDZ domain and the YFP-degron fusion linked to a PDZ ligand . Following induction of the mating pathway with α-factor , we measured the YFP fluorescence of the cells using flow cytometry and observed qualitatively similar substrate degradation . However , the fold change observed three hours post induction for the swapped domains was approximately half that of the original orientation ( Figure 1—figure supplement 3 ) . This is likely due to the fusions affecting either protein expression or sterically interfering with the function of one of the involved enzymes . While these results demonstrate that this retargeting approach is largely modular , they also suggest that other characteristics of the fusions – such as how they affect translation or protein folding – may not be . We also asked whether endogenous Fus3 could be re-targeted in the fashion described above . We found that by inserting the sequence encoding the PDZ ligand downstream of the native copy of the FUS3 gene in the yeast genome , the native kinase could just as efficiently cause the degradation of the YFP substrate ( Figure 1—figure supplement 4 ) . These results – along with those discussed above – imply that an interaction domain and a phospho-motif are necessary and sufficient to target the regulation of a native signal transduction cascade to a substrate of choice . To determine how general our targeting approach is , we exchanged the mPDZ/PDZ ligand pair for unrelated pairs of modular protein interaction domains , both naturally derived and synthetic . We built variants of our Fus3-substrate pair with the naturally occurring SH3 domain or the synthetic SYNZIP domain ( Figure 2A ) ( Dueber et al . , 2009; Thompson et al . , 2012 ) . In both cases we observed significant reporter degradation , ~10 . 1-fold in the case of SH3 and ~4 . 7-fold for SYNZIP domains versus a control with a degron in which the two threonine residues in the Cdc4 binding site had been switched to a methionine or alanine ( Figures 2B and Figure 2—figure supplement 1 ) . These results confirm the flexible nature of the interactions that enable a productive kinase-substrate interaction . 10 . 7554/eLife . 15200 . 008Figure 2 . Demonstrating the flexibility and scalability of the system by varying interaction domains . ( A ) Variants of the different complementary interaction domains used . The constitutive interaction domains mPDZ , SH3 and SYNZIP are shown on the left; the ABA inducible ABI-PYL interaction domains appear on the right . ( B ) Comparison of YFP signal normalized by cell size from constructs bearing the indicated interaction domains along with either a functional ( blue histograms ) or non-functional ( gray histograms ) phosphodegron in yeast treated with 10 μM α-factor as in Figure 1C . The vertical dashed black lines on the histograms represent the medians of the populations with functional degrons whereas the solid black lines represent the median of the populations with non-functional degrons . ( C ) Median fluorescence – shaded regions cover the interquartile range – and population histograms of the YFP signal normalized to cell size from cells expressing the ABA inducible ABI-PYL interaction domains fused to Fus3 and YFP , respectively for a range of ABA concentrations . The raw time-course data corresponding to these endpoint observations can be found in Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 00810 . 7554/eLife . 15200 . 009Figure 2—figure supplement 1 . Time course characterization of different interaction domain variants post induction with α-factor . In this plot each subplot is labeled with the name of the interaction domain pair that was used to target the MAPK Fus3 to degrade a YFP reporter in the yeast strain . These yeast strains were grown up to log phase from saturated cultures for 5 hrs and then induced with 10 μM α-factor at time 0 to activate the kinase . The dashed lines are uninduced controls . In the case of the ABI-PYL strain two additional induction conditions were assayed , namely with ABA ( 100 μM ) and with ABA and α-factor . For these cultures the medium used to grow the yeast up to log phase had ABA in it . The fluorescence of these cultures was then assayed at regular intervals using flow cytometry . Raw data for two replicates performed on different days under identical conditions is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 009 We further tested our approach using a pair of inducible interaction domains derived from a plant hormone-sensing pathway . The association of the protein domains PYL and ABI can be controlled using the small-molecule plant hormone abscisic acid ( ABA ) ( Figure 2A , right side ) ( Liang et al . , 2011 ) . When we fused these domains to Fus3 and our YFP-phosphodegron reporter , we observed a change in the fluorescent signal only when the concentration of ABA was 1 μM or higher ( Figure 2C ) . These results provide additional evidence both that the kinase and substrate are indifferent as to the nature of their interaction , and that the targeting of the kinase to the substrate directly triggers the observed degradation , as the decrease in the YFP signal is correlated with ABA dose . However , it is important to note that the identity of the interaction domain fused to the YFP-phosphodegron target influenced the steady-state fluorescence of the reporter ( Figure 2—figure supplement 1 ) . Thus , even interaction domains with similar affinities may not have equivalent behaviors when used inside of cells . We next investigated whether synthetic interaction domains enable multiple MAPKs to target independent substrates in parallel and in an orthogonal manner . We targeted one copy of Fus3 to an mCherry-phosphodegron reporter using a constitutive mPDZ-PDZ ligand interaction and a second copy of Fus3 to a YFP-phosphodegron reporter via the ABA inducible ABI-PYL interaction ( Figure 3A ) . In the presence of α-factor alone only the mCherry signal was reduced , while the YFP value remained unchanged . Only when both α-factor and ABA were added , did we see a drop in the YFP signal ( Figure 3B and C ) . From this perspective , the two Fus3 variants are analogous to orthologous MAPKs , with each targeting its own substrate . 10 . 7554/eLife . 15200 . 010Figure 3 . Targeting of orthogonal substrates . ( A ) Cells expressed two distinct forms of modified Fus3 and used either a constitutive interaction domain ( left ) or the ABA inducible domains ( right ) to target mCherry or YFP , respectively . ( B ) Population histograms of mCherry ( left ) and YFP ( right ) fluorescence normalized by cell size for cells under the indicated conditions—i . e . untreated , treated with 10 μM α-factor , treated with 100 μM ABA or both . The solid vertical black lines on the histograms represent the medians of the untreated populations and the dashed black lines represent the medians of the treated populations . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 01010 . 7554/eLife . 15200 . 011Figure 3—figure supplement 1 . Competition between two Fus3 MAPKS with different interaction domains for MAPKK Ste7 . Population histograms of YFP fluorescence signal normalized by cell size for yeast strains that have either a single copy of Fus3 targeted to a YFP substrate via an ABA dependent PYL interaction domain ( top ) or two copies of Fus3 one with a PDZ ligand and another with a PYL domain ( bottom ) . These strains were diluted down from saturated overnights and were grown for 5 hrs to log phase with and without ABA and then cytometry reads were performed after 2 hrs post-induction with α-factor . Each plot corresponds to a specific induction condition , with the solid black lines indicating the median of the untreated controls and the dashed black lines indicating the median of that particular treatment . The magnitude of log differences between treated and untreated medians for each treatment is displayed the treatment label . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 01110 . 7554/eLife . 15200 . 012Figure 3—figure supplement 2 . Competition between mCherry and GFP when targeted by the same Fus3 . Population histograms of mCherry ( left ) and YFP ( right ) fluorescence signal normalized by cell size for yeast strains that have either a single YFP substrate targeted by Fus3 via a PDZ interaction ( top ) or two substrates , YFP and mCherry , both of which are targeted by Fus3 via a PDZ interaction ( bottom ) . These strains were diluted down from saturated overnights and were grown for 5 hrs to log phase . Measurement of fluorescence was performed with flow cytometry 2 hrs post-induction with α-factor . The treated populations are depicted as blue histograms and the untreated as gray histograms . The solid black lines indicating the median of the untreated controls and the dashed black lines indicating the median of α-factor treated cells . The magnitude of log differences between the median values of the treated and untreated samples are indicated on each plot . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 012 However , we noticed that when there were two parallel MAPK-substrate systems in the same cell the net fold change of the ABA-sensitive YFP-phosphodegron reporter was moderately reduced compared to when it was present on its own—from ~2 . 15 to ~1 . 85 fold . We tested whether this decrease in efficiency was due to competition—either between the two copies of Fus3 for the pool of the activated upstream MAPK kinase , Ste7 , or between the substrates for the ubiquitination/degradation machinery . To examine this question we constructed strains that expressed our standard system – one kinase targeting one substrate – and added either a competing copy of Fus3 or a competing substrate . In both experiments we observed a diminished response in YFP degradation in the presence of the competitor ( Figure 3—figure supplements 1 and 2 ) . Thus , it is likely that a confluence of factors – both saturation points as well as the less efficient ABA-induced interaction – contribute to the different levels of degradation observed for the mCherry and YFP reporters in this dual-targeting system . The parallel synthetic kinases mimic the behaviors of a natural pair of yeast MAPKs , Fus3 and Kss1 . Fus3 and Kss1 share many of the same targets , but also have distinct substrates , presumably as a result of the specialization of their preferences for related docking domains ( Reményi et al . , 2005 ) . Likewise , the engineered system described above also retains the native targeting of Fus3 , but uses distinct heterologous protein interaction domains to recognize unique targets . The ability to modulate the dynamics of MAPK-dependent degradation would be useful for reprogramming cell behaviors . We explored two strategies to modulate the degradation dynamics . First , we varied the number of phosphodegrons fused to the protein ( Figure 4A ) . As we increased the number of phosphodegrons from one to three , we observed a concurrent increase in the rate of degradation of the reporter; adding more than three phosphodegrons to the reporter did not seem to affect the rate of degradation further ( Figure 4B ) . In addition to changing the degradation dynamics , increasing the number of phosphodegrons also decreased the steady state expression of the reporter , possibly by multiplying the weaker interactions of the un-phosphorylated degron ( s ) with the degradation machinery ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 15200 . 013Figure 4 . Modulating regulation by altering the number and sequence of phosphodegrons . ( A ) We varied either the phosphodegron number ( left ) or the sequence ( right ) —differing residues are red , the phosphorylated residue is highlighted in blue . ( B ) Time-course data of strains induced with 10 μM α-factor and expressing Fus3 targeting YFP reporters with one to five phosphodegrons . The fluorescence of each strain was normalized to cell size and then to its initial fluorescence . Data normalized only to cell size can be found in Figure 3—figure supplement 1 . ( C ) Fus3 targeting of YFP substrates with the indicated phosphodegron sequence variants . As in B ) , the fluorescence of each strain is normalized to cell size and then against its initial fluorescence . Data normalized only to cell size can be found in Figure 3—figure supplement 2 . The curves indicate the median values , while the shaded regions cover the interquartile range . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 01310 . 7554/eLife . 15200 . 014Figure 4—figure supplement 1 . Time course data of reporter variants with different numbers of phosphodegrons normalized by cell size . Each subplot in this plot is a replicate of the experiment performed on different days under identical conditions . Each yeast strain assayed was diluted from a saturated culture , grown for 5 hours to reach log phase growth and then induced with 10 μM α-factor at time 0 to activate the kinase . The fluorescence of these cultures was then assayed at regular intervals using flow cytometry . Data for each variant is depicted in a different color with dark blue being one degron and light blue being five adjacent degrons . Solid lines indicate the median value , while shaded regions indicate the interquartile range . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 01410 . 7554/eLife . 15200 . 015Figure 4—figure supplement 2 . Time course data of reporter variants with different degron sequences normalized by cell size . Each yeast strain assayed was grown up to log phase from saturated cultures for 5 hrs and then induced with 10 μM α-factor at time 0 to activate the kinase . The fluorescence of these cultures was then assayed at regular intervals using flow cytometry . Solid lines indicate the median value , while shaded regions indicate the interquartile range . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 015 We found that altering the amino acid sequence of the phosphodegron itself also changed the dynamics of degradation . We constructed two additional variants of the phosphodegron motif that more closely mimicked the amino acid sequence of the published ‘consensus motif’ for the WD40 domain of Cdc4 ( Figure 3A ) ( Nash et al . , 2001; Orlicky et al . , 2003 ) . The sequences of the two variants only differ at one site — two residues N-terminal of the phosphorylated threonine — where the Cdc4 consensus leucine was changed to a proline , an amino acid that is supposed to be preferred by Fus3 ( Mok et al . , 2010 ) . Both variants had similar behaviors , with a similarly decreased rate of degradation relative to the phosphodegron derived from Tec1 . ( Figure 4C ) . These results suggest that phosphodegron design is flexible , and with more study it may become feasible to rationally tune their degradation dynamics . Moreover , the number of phosphodegrons is not limited to those found in nature . Taken together , these results demonstrate that our approach is applicable to several different phosphodegrons , and lays out a framework for how phosphodegrons may be used to alter degradation dynamics of a protein of interest . We next swapped out the kinase module to test whether other MAPKs are also amenable to rewiring in the same manner . We focused on the human MAPK , ERK2—a widely studied kinase implicated in several pathologies , which has also been previously studied in the context of yeast ( O'Shaughnessy et al . , 2011 ) . Native ERK2 has been shown to regulate protein stability via phosphodegrons; for example , a phosphodegron found in the protein MKP1 is targeted by ERK2 and subsequently tagged by the ubiquitin machinery and degraded ( Lin and Yang , 2006 ) . Our engineered substrate consisted of a 64 residue region surrounding the phosphodegron of MKP-1 fused to a YFP reporter . Rather than port the entire ERK2 signaling cascade into yeast , we used a constitutively active version of the MAPK created by fusing the upstream MAPK kinase – MEK1 – to ERK2 ( Robinson et al . , 1998 ) . To enable the kinase-substrate interaction we fused the mPDZ domain and PDZ ligand to the substrate and MAPK , respectively ( Figure 5A ) . We also included a construct missing the heterologous targeting domains to make sure that targeting was not simply due to direct interactions mediated by sequence elements surrounding the phosphodegron . Since the strains constitutively expressed both the engineered kinase and target , we measured the steady-state YFP fluorescence via flow cytometry . In strains with the active kinase targeting the functional YFP reporter , fluorescence did not rise above background levels ( Figure 5B ) —suggesting that the substrate is actively targeted , phosphorylated and then degraded . Fluorescence was significantly higher in control strains where the interaction domain , the phosphodegron or both were missing or inactivated ( Figure 5B ) . These results indicate that the interaction domains and the phosphodegron are necessary and sufficient for retargeting the regulation of ERK2 . Importantly , these results also strongly suggest that this rewiring approach is potentially applicable to a wide range of MAPKs . 10 . 7554/eLife . 15200 . 016Figure 5 . Retargeting the mammalian MAPK , ERK2 . ( A ) As with Fus3 , the human MAPK , ERK2 , was targeted to a YFP reporter ( green ) via an interaction between the mPDZ domain and the PDZ ligand . A phosphodegron ( yellow ) fused to the YFP reporter was derived from the mammalian MKP-1 . ERK2 was rendered constitutively active by fusing it to a constitutively active form of MEK1 ( purple ) . ( B ) Population histograms of YFP fluorescence normalized by cell size of yeast strains in log phase growth with active ERK2 targeted to YFP with a functional phosphodegron ( blue histogram ) . Controls strains with an inactive phosphodegron fused to YFP and/or an untargeted version of the kinase were also tested ( gray histograms ) . The solid vertical black lines on the histograms represent the medians of the first histogram – the untargeted kinase paired with the non-functional degron – and the dashed black lines represent the medians of each subsequent population . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 016 Thus far , we have described the retargeting of MAPKs to synthetic targets such as fluorescent proteins , which double as the readout for kinase activity . Next , we asked whether MAPKs could be targeted to arbitrary endogenous substrates and – more specifically – whether this approach can be used to modify the response of an existing signaling pathway . To answer these questions , we targeted Fus3 to up- and downstream elements in the yeast mating cascade , including the kinase Ste7 , the scaffold protein Ste5 , and the transcription factor Ste12 . We built a total of six yeast strains containing the synthetic kinase-substrate pairs . Three of these strains constitutively expressed Fus3 with a PDZ interaction domain , while the other three expressed a version of Fus3 with a non-matching interaction domain . All of the strains included one of the mating cascade proteins – Ste5 , Ste7 or Ste12 – fused to a complementary interaction domain and the Tec1 phosphodegron . The interaction domain and phosphodegron were inserted into the native genomic locus of the protein of interest . The Fus1 gene , whose expression is activated by the mating pathway upon induction with α-factor , was fused to YFP to provide an independent readout for pathway activation . We chose these specific target proteins because their regulation by Fus3 results in interesting regulatory topologies . Specifically , Fus3-mediated degradation of Ste7 and Ste5 are examples of negative feedback loops , while the degradation of Ste12 results in an incoherent feed-forward loop ( Figure 6A–C ) . Such regulatory links can be used to fundamentally alter the signal processing properties of native pathways ( Alon , 2007; Bashor et al . , 2008 ) . Of note , this is the first time that purely post-translational feedback loops have been used to re-engineer signaling . 10 . 7554/eLife . 15200 . 017Figure 6 . Implementation of negative feedback and feed-forward signaling topologies using a rewired MAPK . ( A–C ) Plots and schematics that depict the relationship between the α-factor input and the YFP reporter for yeast strains with synthetic post-translational negative feedback or feed-forward loops . Fus3 ( yellow ) was rewired to target ( A ) the scaffold Ste5 , ( B ) the kinase Ste7 or ( C ) the transcription factor Ste12 ( all depicted in light blue ) –in each case , the endogenous copies of these proteins were modified by inserting a phosphodegron and a complementary interaction domain at their C-terminus . Plots of the median fluorescence of the YFP reporter – under the control of the mating-specific pFUS1 promoter – normalized to cell size for increasing concentrations of α-factor . Data from control strains with an untargeted kinase – and thus no feedback/feed-forward control – are shown in dark blue . Points indicate the median values at each α-factor concentration , while the vertical bars cover the interquartile range of the data . The data from both the no feedback and feedback conditions were used to determine the parameter values used with the formula: A+ B[α]n1+ C[α]n – where C was fixed between the two data sets . n and [α] are the hill coefficient and the α-factor concentration , respectively . Fits are plotted as dashed lines . Time courses of the same strains treated with 10 µM α-factor are shown in Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 01710 . 7554/eLife . 15200 . 018Figure 6—figure supplement 1 . Time course charecterization of Negative feedback topologies . Timecourse data of YFP fluorescence for populations of yeast cells where the MAPK Fus3 is targeted to phosphorylate and cause the degradation of Ste12 , Ste5 or Ste7 ( from left to right ) after they were induced with 10 μM α-factor at time 0 . The blue solid lines depict medians of strains that had a negative feedback topology described in the corresponding cartoon and the gray solid lines are medians for control strains that had a non-frunctional phosphodegrons . The blue and gray ribbons describe the interquartile range . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 018 To determine the impact of negative feedback , we measured the fluorescence output of the pathway following induction with varying levels of α-factor . Relative to the untargeted controls , the negative feedback through Ste5 or Ste7 reduced the maximal pathway activation in those backgrounds by ~60% and ~45% , respectively . The apparent Hill coefficients ( nH ) were also moderately changed with negative feedback compared to the untargeted kinase controls—when Ste5 was the target , nH increased from 1 . 5 to 1 . 9 , while when Ste7 was negatively regulated nH remained 1 . 6 ( Figure 6A and B ) . These values are qualitatively consistent with but slightly higher than the sensitivities reported previously for a system with negative feedback realized through transcription and recruitment of a phosphatase in the yeast mating cascade ( Bashor et al . , 2008 ) . The increase in pathway sensitivity observed for negative feedback applied to Ste5 is surprising ( Kholodenko , 2000 ) . However , the response of a scaffolded signaling cascade is highly sensitive to the concentration of the scaffold protein—with a reduction of the scaffold concentration resulting in an increase in the sensitivity of the cascade ( Levchenko et al . , 2000 ) . Although a more detailed analysis is required , this observation suggests a potential explanation for the observed increase in the apparent Hill coefficient . However , we also note different fusion proteins are required for each experiment and that these protein modifications alone can result in changes of the pathway sensitivity—e . g . by changing the concentrations of pathway components ( O'Shaughnessy et al . , 2011 ) . For example , the non-feedback controls in the three experiments shown in Figure 6 have apparent Hill coefficients of nH =1 . 5 , 1 . 6 and 2 . 1 . In the incoherent feed-forward loop – created by having Fus3 both activate and inhibit the transcription factor Ste12 – we find that the inhibitory connection dominates at all levels of induction resulting in a complete elimination of the downstream response ( Figure 6C and Figure 6—figure supplement 1 ) . However , as we will show next , more interesting behaviors are possible in a slightly more complex incoherent feed-forward loop . Hybrid regulatory schemes that occur at the level of both transcription and translation are often observed in nature and further enrich the available behaviors in the design of engineered biological circuits ( Yeger-Lotem et al . , 2004; Mishra et al . , 2014 ) . By putting the YFP-phosphodegron-mPDZ domain fusion protein under the control of the mating pathway-controllable promoter – pFUS1 – we created a simple incoherent feed-forward circuit regulated at the level of both transcription and translation . Such a ‘type 3’ incoherent feed-forward loop design can produce pulses and other behaviors , depending on the design parameters ( Mangan and Alon , 2003 ) . A phenomenological model of a hybrid incoherent feed-forward loop is included in Appendix 1 . We performed time-course experiments over a range of α-factor concentrations ( Figure 7B and Figure 7—figure supplement 1 ) . In cells containing the feed-forward loop the fluorescent signal initially increased sharply as the α-factor concentration was increased from 0 . 1 μM to ~1 μM—however , induction with concentrations of α-factor higher than 1 μM resulted in decreasing levels of YFP fluorescence . The incoherent feed-forward loop thus created a concentration-based band-pass filter for the α-factor input . In a control where the phosphodegron fused to YFP was broken we observe the normal signal amplification behavior of the mating cascade ( Figure 7A ) —thus , it is the targeted regulation of YFP by Fus3 that caused the band-pass-like behavior . 10 . 7554/eLife . 15200 . 019Figure 7 . Conversion of a native amplifier to a band-pass filter . ( A , B ) The relationship between the α-factor input and YFP expression – driven by the mating pathway-specific promoter pFUS1 – for strains without and with a synthetic post-translational incoherent feed-forward loop . Induction of the mating pathway activated a modified Fus3 ( yellow ) that indirectly up-regulates the expression of a YFP reporter ( green ) fused to a phosphodegron . An interaction between the Fus3 and the YFP-degron reporter was enabled via complementary interaction domains . In one case ( A ) the phosphodegron was mutated and inactive , while in the other ( B ) it was fully functional . The points indicate the median YFP signal – normalized by cell size and then to the untreated condition – in yeast strains in log phase growth treated with the indicated concentration of α-factor . The error bars depict the interquartile range of the population data . Dashed lines are fits to the equation A+B[α]n1+C[α]n1+E[α]1+D[α] – model derivation and fitting are described in more detail in Appendix 1 . Time-course data is shown in Figure 7—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 01910 . 7554/eLife . 15200 . 020Figure 7—figure supplement 1 . Time course of dose response behavior to α-factor induction of yeast strains with ( top row ) and without ( bottom row ) the mating cascade modified with an incoherent feed forward loop . Each yeast strain assayed was grown up to log phase from saturated cultures for 5 hrs and then induced with a range of different α-factor concentrations . The fluorescence of these cultures was then assayed at regular intervals — indicated above each graph — using flow cytometry . Median data ( solid blue line ) with the interquartile range ( blue ribbon ) of the population for two replicates performed on different days under identical conditions is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 15200 . 020 Taken together , these results demonstrate that re-targeted kinases can be used to modulate the behavior of signaling cascades through variety of circuit designs , including negative feedback and incoherent feed-forward loops . The data also highlight the utility of using this rewiring approach to study the effects of kinase-directed ubiquitination-based regulation , which occur extensively in nature ( Swaney et al . , 2013; Beltrao et al . , 2012 ) and adds to the available tools for the study of this pervasive mode of signaling ( Schneekloth et al . , 2004; Melchionna and Cattaneo , 2007; Bonger et al . , 2011; Neklesa et al . , 2011 ) .
Here we have demonstrated that MAPK-directed ubiquitin-based signaling can be rewired to regulate a protein of choice . The addition of sets of two modular components is sufficient to rewire a MAPK to regulate any protein of interest—a complementary set of protein interaction domains and a phosphodegron . Natively , MAPKs are co-localized with their substrates via an interaction between the docking peptide of a substrate and a set of residues on the surface of the MAPK; it has been hypothesized that this interaction may be necessary to catalytically unlock the kinase ( Chang et al . , 2002; Heo et al . , 2004; Zhou et al . , 2006; Tokunaga et al . , 2014 ) . Our results suggest that while these domains may have some allosteric properties , simply co-localizing an active MAPK with a protein bearing a compatible amino acid motif that can be phosphorylated is sufficient for the functional regulation of the protein . One implication of our results is that the evolution of new connections within MAPK regulation networks is only constrained by the two criteria discussed above—namely the appearance of 1 ) an accessible phospho-motif; and , 2 ) a protein-protein interaction strong enough to co-localize the new kinase-substrate pair . The proteomes of Saccharomyces cerevisiae and humans have ~3500 and >50 , 000 phosphorylation sites , respectively ( Hornbeck et al . , 2015 , 2012; Beltrao et al . , 2009 ) . The amino acid composition of the surrounding phospho-motifs is constrained by the residues in and around the kinase active site; as such their length is generally fairly short—on the order of four amino acids on either side of the phosphorylated residue ( Ubersax and Ferrell , 2007 ) . With such a short length , and given the degeneracy of the recognition requirements ( Mok et al . , 2010 ) , the probability that new phospho-motifs will appear by chance is high . Indeed , many human SNPs – both those associated with disease as well as apparently healthy individual variation – have been observed to create and destroy verified phospho-motifs ( Hornbeck et al . , 2015; Reimand and Bader , 2013; Ryu et al . , 2009 ) . Many protein interactions occur between short , linear stretches of amino acids and protein domains , the classic examples being the PDZ and SH3 domains , but the binding of docking domains to the surfaces of MAPKs also belongs to this class ( Harris and Lim , 2001; Mayer , 2001; Reményi et al . , 2006; Van Roey et al . , 2014 ) . Like phospho-motifs , these short motifs can appear spontaneously during evolution ( Neduva and Linear motifs , 2005; Davey et al . , 2015; Beltrao and Serrano , 2007 ) . Given that both phospho-motifs and short , linear interaction peptides are degenerate , common and short it is interesting to consider what constrains the formation of a new , functional connection between a kinase and a substrate—i . e . whether it is the formation of phospho-motifs or of protein-protein interactions that is rate limiting . This may be addressed by future studies . Our creation of modular components for kinase signaling may help recapitulate the success modular transcriptional circuits have enjoyed ( Stanton et al . , 2014; Kiani et al . , 2014; Zalatan et al . , 2015; Prindle et al . , 2012 ) . However , while our approach is a powerful tool it does have certain limitations . For instance , our system requires that a phosphodegron be known , and its cognate F-box be expressed for ubiquitination to occur . We demonstrate one way in which this problem may be addressed , i . e . by the design of new phosphodegrons based on the consensus sequences of the MAPK and F-box . Another consideration in any protein-engineering endeavor is the effect that various protein fusions have on expression—indeed we noted in our experiments that fusion of additional domains to MAPKs or their substrates altered the expression levels . These altered expression levels affect the behavior of kinase cascades , and so a sufficiently diverse set of modules need to be defined and characterized to make the desired behavior ( s ) achievable . Thus , the scalability afforded by the use of modular interaction domains comes at the potential price of altered protein expression . In contrast , using docking domains for co-localization obviates engineering the kinase , but is not a scalable rewiring approach . Finally , more work is necessary to render other kinase families ‘engineerable’ . Still the flexibility and scalability of kinase-substrate interactions demonstrated through our work lays a comprehensive foundation for future attempts to understand and re-engineer the signaling behavior of cells .
All strains were built using a W303a background into which each synthetic construct was integrated at either the URA , HIS , TRP or LEU genomic loci . The plasmids used to generate the strains are listed in Supplementary file 1 . The YFP reporter constructs were built by fusing an SV40 nuclear localization tag , an interaction domain and a phosphodegron in tandem to the YFP protein separated by 12 amino acid long glycine-serine linkers . The retargeted kinase constructs were built by fusing a complementary interaction domain to the kinase , also separated by a 12 amino acid glycine-serine linker . The strong constitutive promoter derived from the native TDH3 gene was used to drive expression of the constructs . For all examples of the system that involved the yeast mating cascade , the kinase used was the MAPK FUS3 . For the system that demonstrated mammalian MAPK retargeting , the kinase utilized was a constitutively active version of MEK1 fused to ERK2 and an interaction domain . For the feedback and feed-forward strains YFP was fused in tandem with the FUS1 gene , whose expression was activated by the mating pathway , to act as a reporter . These strains also had a copy of FUS3 fused to an interaction domain integrated into the genome . In the case of the negative feedback and the feed-forward strains the genomic copies of Ste5 , Ste7 and Ste12 were fused to an interaction domain , a phosphodegron and an mCherry reporter . The incoherent feed-forward strains were identical except that the expression of the YFP-nuclear localization tag-phosphodegron-interaction domain fusion was driven from a FUS1 promoter . All cytometry measurements in experiments just measuring YFP expression were acquired with an Accuri C6 cytometer with attached CSampler apparatus using 488 nm and 640 nm excitation lasers and a 533 nm ( FL-1: YFP/GFP ) emission filter ( BD Biosciences ) . In those experiments that included mCherry , we used a MACSQuant VYB ( Miltenyi Biotec ) , with 405 , 488 and 561 nm excitation lasers and 561 nm ( FSC ) , 525 nm ( YFP ) and 615 nm ( mCherry ) emission filters . Synthetic complete growth medium was used in all experiments . Experiments involving time course data were taken during log phase via the following preparation: 16 hrs of overnight growth in the synthetic complete medium in a 30°C shaker incubator followed by 1:100 dilution into fresh , room-temperature medium . After 5 hrs of growth at 30°C , 100 µL aliquots were read periodically – with 10 thousand events collected for every read – until the completion of the experiment . In all cases where Fus3 was being retargeted , the yeast cultures were induced with α-factor 5 hrs post-dilution . For experiments involving dose response behavior , cultures were grown overnight , then diluted down in the morning 1:100 in fresh media and grown for 5 hrs to log phase . They were then induced with α-factor , as well as other inducers like ABA in some cases , and allowed to grow for between two to six hours depending on the experiment and then read on the cytometer . As the MEK-ERK2 fusion is constitutively active no inducer was necessary ( Robinson et al . , 1998 ) . Data were analyzed using custom python scripts and FCSParser and Seaborn libraries ( DOI: 10 . 5281/zenodo . 45133 ) using the following steps: ( 1 ) Anomalies – such as bubbles – were identified by plotting and visually inspecting the FSC-A value versus the time each cell was collected for each well . ( 2 ) To prevent the creation of NA values when the data was log transformed any 0 values in the data collected from the Accuri C6 cytometer were converted to 1 . Since data collected on the MACSQuant VYB can fall below 0 , all the data was normalized by adding the absolute value of the lowest value collected that day to the raw values and then adding 1 . ( 3 ) To control for the effects of cells size , the fluorescence values for each event were then divided by the FSC-A value for that event . All reported data is the aggregate of at least two technical replicates performed on consecutive days . The fits presented in Figures 6 and 7 were performed using custom python scripts . 10 mL cultures of yeast strains expressing untargeted control substrates or targeted test substrates were grown at 30°C in YEPD medium to approximately 1*107 cells/mL . Cells were incubated with DMSO or the proteasome inhibitor MG132 ( 25 µg/mL ) for 30 min prior to addition of α-factor or vehicle control for an additional 10 min . Cycloheximide was then added to a concentration of 50 µg/mL and cells were harvested by centrifugation at the denoted time points . Pellets were lysed in 200 µL SUMEB buffer ( 8 M urea , 10 mM MOPS , 10 mM EDTA , 1% SDS , 0 . 01% bromo- phenol blue , pH 6 . 8 ) by vortexing with acid washed beads for 5 min at 25°C . Lysate was clarified by centrifugation at 13000 rpm for 5 min and subjected to western analysis . Protein lysates were resolved by SDS-PAGE using 4–20% gradient gels ( Lonza ) . Western analyses were performed with rabbit anti-GFP ( 1:2500 ) or mouse anti-ubiquitin antiserum ( 1:10 ) . | Nature has evolved a number of ways to link signals from a cell’s environment , like the concentration of a hormone , to the behavior of that cell . These new connections often form by reusing certain common signaling components , such as mitogen-activated protein kinases . These enzymes – referred to as MAPKs for short – are activated by specific signals and alter the activity of target proteins in the cell by adding a phosphate group to them: a process called phosphorylation . These connections thus dictate how cells respond to their environments – and consequently , disruptions to the connections are a common source of disease . Groves , Khakhar et al . set out to understand how connections can be made between a MAPK and a new target protein to gain insights into how these links emerge through evolution and how they might break in disease . Their approach focused on one of the ways that phosphorylation can alter the activity of a target protein: marking it for degradation . Experiments with budding yeast showed that a MAPK could only achieve this if two conditions are met . First , the target protein and kinase need to bind to each other . Second , the target needs to contain a site that when phosphorylated is subsequently recognized by the cell’s protein degradation machinery . By engineering proteins so that they fulfilled these two criteria , Groves , Khakhar et al . created new connections between a yeast MAPK called Fus3 or a human MAPK called ERK2 and a variety of targets . The results showed that the parts of the proteins involved in the interaction step could be completely separate from the parts that are involved in the phosphorylation step . This suggests that connections between kinases and their targets can be rewired simple by mixing together parts of other existing proteins . Finally , Groves , Khakhar et al . confirmed that engineered connections between kinases and targets could predictably change how yeast cells responded to a hormone that normally controls the yeast’s reproductive cycle . Together these results bring us one step closer to understanding how cells assemble the signaling pathways that they use to process information . However further work is needed to see if these findings can be generalized to other signaling components , and if so , to explore if new connections can be built to yield more complicated cellular behaviors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"computational",
"and",
"systems",
"biology"
] | 2016 | Rewiring MAP kinases in Saccharomyces cerevisiae to regulate novel targets through ubiquitination |
Most amino acids can be encoded by several synonymous codons , which are used at unequal frequencies . The significance of unequal codon usage remains unclear . One hypothesis is that frequent codons are translated relatively rapidly . However , there is little direct , in vivo , evidence regarding codon-specific translation rates . In this study , we generate high-coverage data using ribosome profiling in yeast , analyze using a novel algorithm , and deduce events at the A- and P-sites of the ribosome . Different codons are decoded at different rates in the A-site . In general , frequent codons are decoded more quickly than rare codons , and AT-rich codons are decoded more quickly than GC-rich codons . At the P-site , proline is slow in forming peptide bonds . We also apply our algorithm to short footprints from a different conformation of the ribosome and find strong amino acid-specific ( not codon-specific ) effects that may reflect interactions with the exit tunnel of the ribosome .
Different synonymous codons are used in genes at very different frequencies , and the reasons for this biased codon usage have been debated for three decades ( Fitch , 1976; Hasegawa et al . , 1979; Miyata et al . , 1979; Bennetzen and Hall , 1982; Lipman and Wilbur , 1983; Sharp and Li , 1986; Bulmer , 1987; Drummond and Wilke , 2008 ) ( reviewed by Plotkin and Kudla ( 2011 ) ; Forster ( 2012 ) ; Novoa and Ribas de Pouplana ( 2012 ) ) . In particular , it has been suggested that the frequently-used codons are translated more rapidly than rarely-used codons , perhaps because tRNAs for the frequent codons are relatively highly expressed ( Plotkin and Kudla , 2011 ) . However , there have also been competing hypotheses , including the idea that frequently-used codons are translated more accurately ( Plotkin and Kudla , 2011 ) . Genes are often recoded to use frequent codons to increase protein expression ( Burgess-Brown et al . , 2008; Maertens et al . , 2010 ) , but without any solid understanding of why this manipulation is effective . There is little or no direct in vivo evidence as to whether the more common codons are indeed translated more rapidly than the rarer codons . Even if they are , the fact that translation is typically limited by initiation , not elongation , leaves the effectiveness of codon optimization a puzzle ( Plotkin and Kudla , 2011 ) . Ribosome profiling ( Ingolia et al . , 2009 ) allows the observation of positions of ribosomes on translating cellular mRNAs . The basis of the method is that a translating ribosome protects a region of mRNA from nuclease digestion , generating a 30 base ‘footprint’ . The footprint is roughly centered on the A-site of the ribosome . If some particular codon in the A-site were translated slowly , then the ribosome would dwell at this position , and so footprints generated from ribosomes at this position would be relatively common . Thus , if one looked at the number of ribosome footprints generated along an mRNA , there should be more footprints centered at every codon that is translated slowly and fewer centered at every codon translated rapidly; in principle , this is a method for measuring rates of translation of individual codons . Experimentally , there is dramatic variation in the number of footprints generated at different positions along any particular mRNA ( Ingolia et al . , 2011 ) ( Figure 1 ) . However , these large peaks and valleys do not correlate with particular codons ( Ingolia et al . , 2011; Charneski and Hurst , 2013 ) . It is still unclear what features of the mRNA cause the peaks and valleys , though there is evidence that prolines , or a poly-basic amino acid stretch , contribute to a slowing of the ribosome and a peak of ribosome footprints ( Ingolia et al . , 2011; Brandman et al . , 2012; Charneski and Hurst , 2013 ) . 10 . 7554/eLife . 03735 . 003Figure 1 . Two ribosome profiles of the TDH1 gene . Top profile is from the data of Ingolia et al . , 2009; bottom profile is from the SC-lys dataset ( ‘Materials and methods’ ) . The first ( leftmost ) peak in the profiles is at the ATG start codon; it may differ in relative height because the SC-lys dataset was generated using flash-freezing . DOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 003 Still , the fact that prolines and poly-basic amino acid stretches affect translation speed does not tell us whether different synonymous codons may also cause smaller effects . This question was investigated by Qian et al . ( 2012 ) and Charneski and Hurst ( 2013 ) using the yeast ribosome profiling data of Ingolia et al . ( 2009 ) . Neither group found any effect of different synonymous codons on translation rate—that is , perhaps surprisingly , each codon , rare or common , appeared to be translated at the same rate ( Qian et al . , 2012; Charneski and Hurst , 2013 ) . We have re-investigated this issue with two differences from these previous investigations . First , we have generated four yeast ribosome profiling datasets by optimized methods , including the flash-freezing of growing cells before the addition of cycloheximide ( ‘Materials and methods’ ) ; Ingolia et al . added cycloheximide before harvesting cells . Second , we have developed a novel method of analysis , designed with the knowledge that , at best , codon decoding rates could account for only a small portion of the variation in ribosome footprints across an mRNA ( ‘Materials and methods’ ) . The combination of optimized data and novel analysis reveals that different codons are decoded at different rates .
We tested this method of analysis using simulated and real positive and negative control data . For a simulated negative control , we assigned real footprint data from our SC-lys dataset to random codons and did RRT analysis . As expected , all codons at all positions show an RRT of about 1 , that is , no signal ( Figure 2A ) . For a simulated positive control , we generated a simulated data set of 2 million 10-codon reads over coding genes , but we biased these simulated reads to give more reads for the codon AAA at position 6 of the footprint . As expected , RRT analysis shows a peak for AAA at position 6 ( Figure 2B ) . 10 . 7554/eLife . 03735 . 004Figure 2 . Validation for ribosome residence time analysis . ( A ) Simulated data , negative control . Real footprint data from the SC-lys dataset were randomly assigned to codons , and RRT analysis was carried out . A flat line with an RRT value of 1 indicates no signal . ( B ) Simulated data , positive control . A dataset of 2 million simulated reads was generated but biased to give more reads over the codon AAA at position 6 . ( C ) Real data , negative control . RNA-seq data from naked fragments of RNA 30 nucleotides long , processed as if for ribosome profiling , were analyzed . ( D ) Real data , positive control . Real ribosome footprinting data from Li et al . were analyzed ( Li et al . , 2012 ) . In this experiment , E . coli were starved for serine . Note that the highest Ser peak is for TCA , which is the rarest Ser codon in E . coli , and the lowest Ser peak is for AGC , which is the most common Ser codon in E . coli . High values at position 9 as well as 8 may indicate that the A-site may be at position 8 in some fragments and position 9 in others . DOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 004 For a real-data negative control , we pooled the control mRNA-seq data for 30 bp fragments from our four experiments ( ‘Materials and methods’ ) and analyzed these mRNA fragments . Since this RNA came from a total naked RNA preparation , there were no ribosomes and no ribosome footprints , so there should not be any signal from translation , even though we are analyzing real 30 bp RNA fragments . Indeed , RRT analysis shows no peaks in positions 2 through 9 of these fragments ( Figure 2C ) . However , there are modest deviations from 1 at the termini , positions 1 and 10 . We attribute these to some base-specificity for the enzymatic reactions used to generate the fragment library ( Lamm et al . , 2011; Jackson et al . , 2014; Raabe et al . , 2014 ) . Supporting this interpretation , the same peaks and valleys at positions 1 and 10 ( i . e . , the same base-specificity ) were seen in real ribosome-footprint data ( see below ) . For a real data positive control experiment , we used the Escherichia coli data generated by Li et al . , who starved E . coli for serine , and did ribosome profiling ( Li et al . , 2012 ) . Because of the starvation for serine , there is an expectation that all six serine codons should be decoded slowly and so should have high RRT values . This proved to be the case ( Figure 2D ) . The six serine codons had 6 of the 7 highest RRT values at position 8 ( Figure 2D , Table 1 ) , which presumably represents the A-site in this experiment . Note that because these are E . coli ribosomes , the phase of the footprint ( i . e . , the position of the A-site in the footprint ) is different from its phase with regard to yeast ribosomes ( see below ) . The RRT analysis of E . coli footprints also showed interesting variation at positions 2 , 3 , and 4 ( Figure 2D ) , which we will consider elsewhere . 10 . 7554/eLife . 03735 . 005Table 1 . Top ten RRTs at position 8 in E . coli starved for serineDOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 005CodonAAUsageRRTTCASer8 . 11 . 98TCCSer9 . 01 . 90TCGSer8 . 81 . 73TCTSer8 . 71 . 71AGTSer9 . 41 . 57ATAIle5 . 51 . 42AGCSer16 . 01 . 25ATTIle29 . 71 . 18CCTPro7 . 21 . 15CCAPro8 . 41 . 13 Lareau et al . ( 2014 ) starved Saccharomyces cerevisiae for histidine using the His3 inhibitor 3-aminotriazole . This was another potential positive control , where the two His codons should be decoded slowly . We analyzed these ribosome profiling data . However , of the 11 million reads obtained in that experiment , about 10 . 6 million mapped to ribosomal RNA . The remaining ∼0 . 4 million reads mapped to mRNA , but gave only 10 ( ten ) total windows passing our quality filters for RRT analysis , and this is too few . However , when we relaxed the filters to obtain more ( albeit lower quality ) windows , we observed obvious peaks ( high RRT values ) for both histidine codons at position 6 specifically in the 3-aminotriazole experiment ( data not shown ) . Having found that RRT analysis gives the expected results in control experiments , we applied it to the analysis of four of our ribosome profiling experiments . Our experiments differ from those of Ingolia et al . and Lareau et al . , in that in those studies , cycloheximide was added to the growing yeast culture before harvesting ( Ingolia et al . , 2009; Lareau et al . , 2014 ) , whereas we harvest by flash-freezing and later add cycloheximide to the frozen cells ( ‘Materials and methods’ ) . The nature of our results is shown in Figure 3 using the rare Leu codon CTC as an example . In this example , 10 codon ( 30 nucleotide ) footprints that have CTC as the first codon have about the average relative frequency—that is , they have about the same relative frequency as footprints with any other codon at the first position . Similarly when CTC is in the 2nd , 3rd , 4th , 7th , 8th , 9th , and 10th positions . However , there is a relative over abundance of footprints that have CTC at the 6th position . In fact , for CTC at the 6th position , averaged over 451 windows ( in the case of this rare codon ) , there are 1 . 89-fold more footprints than at the baseline . This suggests that ribosomes move relatively slowly when CTC is at the 6th position , and , therefore , these ribosomes are more frequently captured as footprints . We say that CTC has a Ribosome Residence Time ( RRT ) of 1 . 89 at position 6 . 10 . 7554/eLife . 03735 . 006Figure 3 . Principle of ribosome residence time analysis . The ribosome protects a 30 nt ‘footprint’ of RNA centered around the A , P , and E sites ( positions 6 , 5 , and 4 ) . The rare Leu codon CTC has a high RRT at position 6 , which is likely the A-site . DOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 006 Figure 4 shows data for all 61 sense codons from one of four experiments , the ‘SC-lys’ experiment . In a large majority of cases , a codon has its highest or lowest footprint abundance when the codon is in position 6 . We interpret this to mean that the codon affects the rate of ribosome movement when the codon is in position 6 , which we believe to be the A-site of the ribosome ( see below for further support for this assignment ) . The behavior of the six Leu codons and the four Thr codons is highlighted in Figure 4B , C . Footprint frequencies also differ from the average in a specific way at positions 5 ( Figure 4D ) ( see below ) and 1 and 10 , the two ends of the footprint . We attribute variation at positions 1 and 10 to some base-specificity for the enzymatic reactions involved in generating and analyzing ribosome footprints ( Lamm et al . , 2011; Jackson et al . , 2014; Raabe et al . , 2014 ) ; the same variations are seen in reactions with naked RNA fragments . 10 . 7554/eLife . 03735 . 007Figure 4 . Results of Ribosome Residence Time analysis . ( A ) The pattern of RRTs for all codons at all positions . Most peaks are at position 6 , with some at position 5 . ( B ) The RRTs for the six leucine codons . CTC has the highest RRT of any codon at position 6 . ( C ) The RRTs for the four threonine codons . ACC has the lowest RRT of any codon at position 6 . ( D ) The RRTs for the four proline codons . Proline has peaks at position 5 , the P-site , as well as at position 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 007 Figure 5A shows the deduced rate of ribosome movement for each codon , plotted against the frequency of codon usage . There is a good correlation ( r = –0 . 52 ) ; that is , the ribosome moves faster over the more common codons . 10 . 7554/eLife . 03735 . 008Figure 5 . Correlation of ribosome residence times with codon properties . ( A ) Correlation of RRT with codon usage . RRT is plotted against the frequency of each codon per 1000 codons . ( B ) Correlation of RRT with the GC content of each codon . The codons were divided into quartiles by RRT ( Fastest–Slowest ) , and the GC content of those ∼15 codons is shown in a violin plot . DOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 008 There is also a correlation , albeit weaker , with the AT-richness of the codon . AT-rich codons are decoded somewhat faster than average , while GC-rich codons are decoded more slowly ( Figure 5B ) . The mean RRT of codons with 3 or 2 GC residues was 1 . 23 , while the mean RRT of codons with 1 or 0 GC residues was 1 . 01 , a statistically significant difference ( p < 0 . 003 by a two-tailed t test ) . Table 2 shows the Ribosome Residence Time at position 6 for each of the 61 sense codons . The slowest codon is the rare Leu codon CTC . Relatively , the ribosome spends about 1 . 9 times as long with a CTC codon in the A site as it does at the average codon . If the yeast ribosome spends 50 milliseconds ( Futcher et al . , 1999 ) on an average codon in the A-site , then the RRT suggests it spends about 95 milliseconds on CTC codons . The fastest codon is the relatively abundant Thr codon ACC ( Figure 4C , Table 2 ) , where it spends 0 . 70 times as long as average ( i . e . , about 35 milliseconds ) . 10 . 7554/eLife . 03735 . 009Table 2 . Ribosome residence time at position 6 ( A ) and 5 ( B ) DOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 009ACodonAAUsageRRTp valueCTCLeu5 . 41 . 89*0 . 0001CCCPro6 . 81 . 71*0 . 0001GGGGly61 . 61*0 . 0001AGGArg9 . 21 . 59*0 . 0001ATAIle17 . 81 . 57*0 . 0001GGAGly10 . 91 . 56*0 . 0001TGGTrp10 . 41 . 53*0 . 0001GTGVal10 . 81 . 52*0 . 0001CGCArg2 . 61 . 45*0 . 0001CGAArg31 . 45*0 . 0008CGGArg1 . 71 . 44*0 . 0010TCGSer8 . 61 . 43*0 . 0001CCAPro18 . 31 . 38*0 . 0001ACAThr17 . 81 . 35*0 . 0001CCGPro5 . 31 . 31*0 . 0001GTAVal11 . 81 . 31*0 . 0001GCAAla16 . 21 . 28*0 . 0001CCTPro13 . 51 . 27*0 . 0001TCASer18 . 71 . 26*0 . 0001TACTyr14 . 81 . 25*0 . 0001TATTyr18 . 81 . 25*0 . 0001GAGGlu19 . 21 . 25*0 . 0001CTALeu13 . 41 . 25*0 . 0001CTTLeu12 . 31 . 24*0 . 0001TGCCys4 . 81 . 23*0 . 0001GGCGly9 . 81 . 22*0 . 0001CAGGln12 . 11 . 15*0 . 0002ACGThr81 . 120 . 0069AGTSer14 . 21 . 100 . 0060AGCSer9 . 81 . 090 . 0213CACHis7 . 81 . 080 . 0098TTTPhe26 . 11 . 050 . 0529GAAGlu45 . 61 . 040 . 0538AGAArg21 . 31 . 010 . 3014TTCPhe18 . 41 . 000 . 4955GCGAla6 . 20 . 990 . 4650TCCSer14 . 20 . 990 . 3341TTALeu26 . 20 . 990 . 3166TCCSer23 . 50 . 980 . 2249CATHis13 . 60 . 930 . 0188GGTGly23 . 90 . 93*0 . 0003ATGMet20 . 90 . 920 . 0027ATTIle30 . 10 . 92*0 . 0005TTGLeu27 . 20 . 92*0 . 0001CTGLeu10 . 50 . 920 . 0139AATAsn35 . 70 . 88*0 . 0001AAALys41 . 90 . 88*0 . 0003CGTArg6 . 40 . 87*0 . 0002CAAGln27 . 30 . 87*0 . 0001GCCAla12 . 60 . 86*0 . 0001GACAsp20 . 20 . 85*0 . 0001TGTCys8 . 10 . 81*0 . 0001GCTAla21 . 20 . 81*0 . 0001ATCIle17 . 20 . 80*0 . 0001ACTThr20 . 30 . 78*0 . 0001GATAsp37 . 60 . 76*0 . 0001AACAsn24 . 80 . 76*0 . 0001GTTVal22 . 10 . 75*0 . 0001GTCVal11 . 80 . 75*0 . 0001AAGLys30 . 80 . 74*0 . 0001ACCThr12 . 70 . 70*0 . 0001BCodonAAUsageRRTp valueCCTPro13 . 51 . 80*0 . 0001CCCPro6 . 81 . 48*0 . 0001CCAPro18 . 31 . 48*0 . 0001AATAsn35 . 71 . 39*0 . 0001CGCArg1 . 71 . 340 . 0070CCGPro5 . 31 . 30*0 . 0001A . Usage of each codon per 1000 codons and the Ribosome Residence Time ( RRT ) at position 6 ( the A-site of the ribosome ) . The p-value for a difference between the calculated RRT value and an RRT value of 1 is shown . p-values less than or equal to 0 . 001 are marked with an asterisk . B . As for A , but for the six highest values at position 5 ( the P-site ) . There are also peaks at position 5 ( Figure 4A , D ) , which we interpret as the ribosome's P-site , where the peptide bond is formed . All four Pro codons are high at position 5: CCT , CCA , and CCC are the three slowest codons at position 5 , while CCG is 6th ( Figure 4D , Table 2 ) . Proline is a unique amino acid in having a secondary rather than a primary amino group , and so it is less reactive in peptide bond formation . Proline forms peptide bonds slowly ( Muto and Ito , 2008; Wohlgemuth et al . , 2008; Pavlov et al . , 2009; Johansson et al . , 2011 ) , and proline has been associated with slow translation in footprinting experiments ( Ingolia et al . , 2011 ) . Our result that the ribosome slows with proline at position 5 is consistent with this and tends to confirm our assignment of position 5 to the P-site and , therefore , position 6 to the A-site . A few other residues also seem slightly slow at position 5 ( e . g . , Asn , Gly , see Table 2 and Supplementary file 1 ) , possibly due to low reactivity in peptide bond formation ( Johansson et al . , 2011 ) . All four proline codons also have high RRTs at position 6 , the A-site ( Figure 4D , Table 2 ) . The dipeptide ProPro is translated very slowly ( Doerfel et al . , 2013; Gutierrez et al . , 2013; Peil et al . , 2013; Ude et al . , 2013 ) . We wondered whether the apparent slowness of proline at both positions 5 and 6 was an informatic artefact due to extreme slowness for ProPro dipeptides . We redid the original analysis after excluding all footprints encoding ProPro dipeptides . Results did not change significantly; Pro still appeared to be slow at both positions 5 and 6 ( Figure 6A ) . On the other hand , when we looked specifically at footprints containing a ProPro dipeptide , there was a very large peak at position 5 ( Figure 6B ) , consistent with the very slow peptide bond formation seen in studies cited above . 10 . 7554/eLife . 03735 . 010Figure 6 . Analysis of ProPro dipeptides . ( A ) RRT analysis of windows containing no ProPro dipeptides . ( B ) RRT analysis of windows containing ProPro dipeptides . DOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 010 To establish repeatability , we generated and analyzed three other ribosome profiling datasets and also re-analyzed previously published data ( Ingolia et al . , 2009 ) . All five data sets gave qualitatively similar results; pairwise correlations for RRTs at position 6 ranged from 0 . 22 to 0 . 96 between the datasets ( Table 3 ) . The poorest correlation ( 0 . 22 ) was a correlation with the previously published dataset , which was generated using significantly different methods than our datasets . In particular , that dataset was generated by adding cycloheximide to the growing culture , then harvesting ( Ingolia et al . , 2009 ) , whereas our data were generated by flash-freezing first , then adding cycloheximide to the frozen cells . Complete results for all five experiments are given in Supplementary file 1 . More recently , we also subjected the long footprint data of Lareau et al . ( 2014 ) to RRT analysis and obtained correlations at position 6 of 0 . 21 , 0 . 47 , 0 . 23 , and 0 . 27 , respectively , for their ‘untreated 1’ , ‘untreated 2’ , ‘untreated merge’ , and ‘cycloheximide 1’ experiments to our SC-lys experiment . Again , these experiments were carried out in a significantly different way from ours and it is not surprising that the correlations are modest . It is reassuring that a positive correlation can be seen even for experiments where no cycloheximide was used . 10 . 7554/eLife . 03735 . 011Table 3 . Correlations between experimentsDOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 011YPD1-HisYPD2Ingo . -Lys0 . 800 . 350 . 760 . 22YPD10 . 530 . 960 . 55-His0 . 580 . 37YPD20 . 53The pairwise Spearman correlations between the RRT values at position 6 are shown for five independent experiments , where the experiments are named YPD1 , YPD2 , SC-Lys , SC-His , and Ingolia . The SC-Lys and SC-His experiments were carried out by JG , and used flash-freezing as the initial method for stopping ribosome movement . The YPD1 and YPD2 experiments were carried out by YC ( Cai and Futcher , 2013 ) , and used addition of ice and cycloheximide to the culture as the initial method for stopping ribosome movement . The ‘Ingo’ experiment was that carried out by Ingolia et al . ( 2009 ) . Further details are given in ‘Materials and methods’ . Complete RRT values for each position in each experiment are provided in Supplementary file 1 . There are strong correlations between codon usage , the number of tRNA genes for the relevant tRNA , and tRNA abundance ( Ikemura , 1981 , 1982; Dong et al . , 1996; Tuller et al . , 2010; Novoa and Ribas de Pouplana , 2012 ) . Although one cannot determine causation from this correlation ( Plotkin and Kudla , 2011 ) , nevertheless it is consistent with the idea that the rate of decoding in translation is at least partly limited by tRNA concentration . Most of our results are consistent with this . However , there are some interesting exceptions . In yeast , the 61 sense codons are decoded by only 42 tRNAs . There are 12 pairs of codons that share a single tRNA ( e . g . , Phe TTC and TTT; Tyr TAT and TAC; etc ) ( Roth , 2012 ) . In many but not all cases , the RRT of the two codons is similar ( Table 2 ) , consistent with the ‘concentration’ hypothesis . However , there are also cases where the RRT appears to be significantly different for two codons sharing the same tRNA . For instance , the Cys codon TGC has an RRT of 1 . 23 , while TGT has an RRT of 0 . 81 ( Table 2 ) . Both codons are recognized by the same tRNA , which in this case is complementary for TGC , and wobble for TGT . Similarly , the Gly codon GGC has an RRT of 1 . 22 ( tRNA is complementary ) , while GGT has an RRT of 0 . 93 ( tRNA is wobble ) . Both these relationships ( RRT for TGC > TGT , and RRT for GGC > GGT ) were true in all five datasets ( Supplementary file 1 ) . In both the cases , the perfect match is decoded more slowly than the wobble match and in both cases , the slower , complementary pairing has a G:C match at the third ( i . e . , wobble ) position . These and other similar examples ( not shown ) suggest that the RRT depends on more than just the concentration of the relevant tRNA . Perhaps the long RRTs for these GC-rich codons are related to the time needed to eject incorrectly paired anti-codons of incorrect tRNAs , although this explanation is somewhat at odds with the literature ( Daviter et al . , 2006; Gromadski et al . , 2006 ) . Alternatively , it has been suggested that translocation can occur more quickly when the codon:anticodon interaction is weaker ( Semenkov et al . , 2000; Khade and Joseph , 2011 ) . Recently , Lareau et al . made the exciting discovery that ribosome profiling on cells that have not been treated with any drug yields two classes of footprints , long ( 28–30 nucleotides ) and short ( 20–22 nucleotides ) ( Lareau et al . , 2014 ) . It is the long class that is seen in cycloheximide experiments , and which we have characterized above . The short ( 20–22 nuc . ) footprints seem to represent a different conformation of the ribosome , perhaps one that occurs when the ribosome translocates along the mRNA . Furthermore , Lareau et al . found that treatment of cells with the elongation inhibitor anisomycin efficiently generates short footprints . Lareau et al . suggest that the long and short footprints are reporting on two different states of translation ( Lareau et al . , 2014 ) . We applied RRT analysis to the short footprints generated by Lareau et al . , with special focus on the footprints after anisomycin treatment . All three of their anisomycin datasets were studied , and the pairwise correlations between the RRT results for these three datasets were very high , ranging from 0 . 89 to 0 . 998 . Partial results are shown in Figure 7 and Table 4 , and complete results are shown in Supplementary file 2 . RRT analysis showed a series of peaks at different positions along the 7-codon footprint . The RRT values for the short footprints did not significantly correlate with RRT values for the long footprints , even when the phases of the footprints were shifted . This suggests , in agreement with Lareau et al . , that the short and long footprints are indeed reporting on different translational processes . Furthermore , for the short footprints the RRT values are amino acid-specific , while for the long footprints at position 6 , the RRT values are codon-specific ( Table 2; Table 4; Figure 4 , Figure 7 , Figure 8 ) . This again indicates that the two kinds of footprints are reporting on different translational processes . The amino acids in the peaks at positions 3 , 5 , and 6 are shown in Table 4: the peak at position 3 contains glycine; the peak at position 5 contains smallish hydrophobic amino acids ( Leu , Val , Ile , and to some extent Phe ) , and the peak at position 6 is dominated by the two basic amino acids , Arg and Lys . It has previously been shown that basic amino acids can cause a pause in elongation by interacting with the ribosome exit tunnel ( Lu et al . , 2007; Lu and Deutsch , 2008; Brandman et al . , 2012; Wu et al . , 2012; Charneski and Hurst , 2013 ) . The basis of the anisomycin arrest is partly but not fully understood ( Hansen et al . , 2003; Blaha et al . , 2008 ) , and so it is difficult to clearly interpret these results ( but see ‘Discussion’ ) . Nevertheless , the application of RRT analysis to the anisomycin-generated footprints gives strong specific signals that are unlikely to be explained by a random process . We note , however , that results from the short footprints from untreated ( no anisomycin ) cells are only modestly correlated ( 0 . 23 ) with results from short footprints from the anisomycin-treated cells ( data not shown ) . 10 . 7554/eLife . 03735 . 012Figure 7 . RRT analysis of short footprints from anisomycin treatment . The short , seven-codon footprints from anisomycin treatment ( dataset 1b ) from Lareau et al . ( 2014 ) were analyzed for RRT . All 61 sense codons are shown; codons for selected amino acids are color-coded by amino acid . Position along the footprint is shown on the x-axis . DOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 01210 . 7554/eLife . 03735 . 013Table 4 . Top 10 RRTs at positions 3 through 6 of the anisomycin-generated short footprintsDOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 013Pos 3Pos 4Pos 5Pos 6Gly GGG 2 . 64Pro CCC 2 . 36Leu TTA 2 . 75Arg CGA 3 . 72Gly GGC 2 . 52Pro CCA 2 . 34Leu CTC 2 . 73Arg CGG 3 . 50Gly GGT 2 . 36Met ATG 2 . 25Val GTA 2 . 43Pro CCG 2 . 74Gly GGA 2 . 32Pro CCT 2 . 17Leu CTA 2 . 36Lys AAA 2 . 59Asp GAC 1 . 80Ala GCC 2 . 13Leu TTG 2 . 29Lys AAG 2 . 49Ala GCC 1 . 79Phe TTC 2 . 03Val GTG 2 . 21Arg CGC 2 . 46Ala GCA 1 . 70Ala GCA 2 . 01Leu CTT 2 . 16Arg CGT 2 . 34Ala GCT 1 . 65Ala GCT 1 . 98Val GTC 2 . 12Arg AGG 2 . 32Ala GCG 1 . 59Tyr TAC 1 . 98Val GTT 2 . 11Arg AGA 2 . 21Blu GAG 1 . 58Ser TCC 1 . 97Ile ATA 2 . 03Asp GAT 2 . 1210 . 7554/eLife . 03735 . 014Figure 8 . Short footprints are amino acid-specific; long footprints are codon-specific . For the set of codons corresponding to each amino acid ( x-axis ) , a test was done to see if all the codons behaved similarly or not . For the short footprints ( left , panel A ) , p-values ( y-axis ) are generally small , showing that each codon for a particular amino acid behaves similarly ( ‘Materials and methods’ ) . For the long footprints ( right , panel B ) , p-values are generally large , showing that the codons for each particular amino acid behave differently ( ‘Materials and methods’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03735 . 014 It appeared that the RRT values at position 6 for the long footprints were codon-specific ( Figure 4 , Table 2 ) , while the RRT values for the short footprints were amino acid-specific ( Figure 7 , Table 4 ) . To confirm this , we developed a statistical test for the coherence of the results for a particular amino acid ( ‘Materials and methods’ ) . Briefly , this method tests whether every codon for a particular amino acid behaves similarly , and it yields a small p-value if it does . Indeed , this analysis confirms that the short footprints give results specific to the amino acid , while the long footprints generally do not ( i . e . , the long footprints are codon-specific ) ( Figure 8 ) . This suggests that the long footprints are reporting on the process of decoding ( which depends on specific codons ) , while the short footprints are reporting on events after decoding .
To our knowledge , this is the first measurement of the differential rate of translation of all 61 codons in vivo . There is a correlation between a high codon usage and a high rate of decoding . Although this is a correlation that has been widely expected , there has been little evidence for it; indeed , the most recent experiments suggested that all codons were decoded at the same rate ( Qian et al . , 2012; Charneski and Hurst , 2013 ) . Some workers have had other expectations for decoding rates . For instance , an important theory was that the more common codons were common because their translation might be more accurate ( Plotkin and Kudla , 2011 ) ( and this still might be correct ) . Translation is optimized for both speed and accuracy ( Bieling et al . , 2006 ) . During translation , the ribosome must sample many incorrect tRNAs at the A-site before finding a correct tRNA . It must match the anti-codon of that correct tRNA with the codon; after such matching , there is a conformational change around the codon–anticodon interaction at the decoding center ( Demeshkina et al . , 2012; Zeng et al . , 2014 ) . The ribosome must form the peptide bond ( Rodnina , 2013; Polikanov et al . , 2014 ) , translocate ( Semenkov et al . , 2000; Khade and Joseph , 2011; Zhou et al . , 2014 ) , and eject the empty tRNA . The nascent peptide must make its way through the ribosome exit tunnel ( Lu and Deutsch , 2008; Petrone et al . , 2008; Lu et al . , 2011; Wilson and Beckmann , 2011 ) . Depending on the rate of each of these events , the concentration of the various tRNAs might or might not have a detectable effect on the overall rate of translation . Our findings that ( i ) the more frequent codons ( i . e . , the ones with the highest tRNA concentrations ) are decoded rapidly; and ( ii ) GC-rich codons are decoded slowly; and ( iii ) proline is slow in the P-site , suggest that there are at least three processes that happen somewhat slowly and on a similar timescale . The high rate of decoding for high concentration tRNAs may reflect the relatively short time it takes for the ribosome to find a high-concentration correct tRNA among many incorrect tRNAs . The fact that we detect proline-specific delays of a similar magnitude to the rare-codon specific delays suggest that peptide bond formation and identification of the correct tRNA are happening on similar time scales . In general , this is what one might expect from the evolution of such an important process as protein synthesis—if one process was entirely rate-limiting , there would be very strong selection for greater speed in that process , until a point is reached where it ‘catches up’ with other processes , and several processes together are then rate-limiting . Even though these data establish that common codons are translated relatively rapidly , this does not on its own explain the success of codon optimization for increasing protein expression , since the rate of translation is primarily limited by the rate of initiation , not elongation ( Andersson and Kurland , 1990; Plotkin and Kudla , 2011 ) ( although one recent study identifies a mechanism whereby rapid elongation causes rapid initiation [Chu et al . , 2014] ) . Nevertheless , on a genome-wide ( and not gene-specific ) scale , the use of faster codons would mean that a given genomic set of mRNAs would require ( or titrate out ) fewer ribosomes to make a given amount of protein than the same set of mRNAs using slower codons ( Andersson and Kurland , 1990; Plotkin and Kudla , 2011 ) . Based on our RRT measurements , and taking into account the different copy numbers of different mRNAs ( Lipson et al . , 2009 ) , we roughly estimate that yeast requires about 5% fewer ribosomes than if they were to make protein at the same overall rate but using each synonymous codon at an equal frequency ( ‘Materials and methods’ ) . This provides at least a sufficient reason for the bias towards faster synonymous codons . We applied RRT analysis to the short footprints identified by Lareau et al . ( Figure 7 ) . These short footprints seem to report on a different translational process than the long footprints seen in cycloheximide experiments . We see that the basic amino acids Arg and Lys are slow at position 6; small hydrophobic amino acids are slow at position 5; and glycine is slow at position 3 . While we know too little about the nature of the short footprints to reliably interpret these results , one speculative possibility is that the results report on the interaction of amino acids in the nascent peptide chain with the exit tunnel of the ribosome ( Raue et al . , 2007; Petrone et al . , 2008; Berndt et al . , 2009; Bhushan et al . , 2010; Lu et al . , 2011; Wilson and Beckmann , 2011; Gumbart et al . , 2012 ) . We find Arg and Lys slow at position 6 , and this correlates with the fact that these basic amino acids cause a pause by interacting with the exit tunnel ( Lu et al . , 2007; Lu and Deutsch , 2008; Brandman et al . , 2012; Wu et al . , 2012; Charneski and Hurst , 2013 ) . This would then suggest that small hydrophobic amino acids , and then glycine , might similarly cause pauses by interacting with positions one or three amino acids further out in the exit tunnel . In summary , we believe that RRT analysis is a sensitive high-resolution method that can characterize the interaction of codons and amino acids with the ribosome . It can be applied to ribosome profiling data of many types , from many organisms . In this study , we show that frequent codons are decoded more quickly than rare codons; that codons high in AT are decoded somewhat quickly; that proline forms peptide bonds slowly; and that short footprints from anisomycin treated cells have an interesting RRT profile that may reflect interaction of amino acids with the ribosome exit tunnel .
Informatic analysis was conducted on four ribosome profiling experiments ( YPD1 , YPD2 , SC-lys , and SC-his ) done for other reasons in the Futcher lab . The strains and methods used varied slightly from experiment to experiment; nevertheless similar results were obtained for the RRT analysis ( Table 2 ) . The ribosome profiling experiments YPD1 and YPD2 have been reported previously ( Cai and Futcher , 2013 ) as the ‘WT’ and ‘whi3’ experiments , respectively . All experiments used S . cerevisiae strain background BY4741 . Two biologically independent ribosome-profiling libraries and mRNA-seq libraries were obtained from YPD rich media ( the YPD1 and YPD2 experiments ) , and two biologically independent ribosome-profiling libraries and mRNA-seq libraries were prepared in synthetic media ( the SC-lys and SC-his experiments ) . Two methods for harvesting cells were used . After harvesting and footprint size selection , footprints from all four experiments were processed identically into sequencing libraries using the ARTseq Yeast Ribosome Profiling kit , following the manufacture's instructions beginning with step B3 in the protocol . Unless indicated , data processing and analysis were performed using a collection of custom programs written in Perl . Primary data were generated using Illumina HiSeq2000 . Data were processed using Fastq clipper from the FASTX Toolkit 0 . 0 . 13 to remove the adaptor sequence and all reads shorter than 25 nucleotides were discarded . Alignment to the reference was done using bowtie2 2 . 1 . 0 in local alignment mode . Before performing our analysis on the Ingolia et al . ( 2009 ) data , in order to adhere to the processing guidelines of that paper , we used bowtie 0 . 12 . 8 , reporting all alignments with at most three mismatches , and a seed length of 21 . We then processed the multiple alignments , removing the poly-A tails and picking the one with the greatest number of bases matching to the reference . This analysis uses the general idea that many different mRNA sequences should get an independent and equal vote on decoding speed . We opted to analyze select regions where the effects of codon usage become particularly easy to assay . First , we discounted all reads with more than two mismatches or quality less than 10 . We identified the first in-frame codon of each read and discarded those less than 30 nucleotides long to exclude fragments that may have been over digested by RNAase I . We then examined the coding regions of the genome , ignoring those overlapping with other genes , rRNAs , and tRNAs , in order to maximize our confidence in unique mapping . Each of the footprint reads that fully fit into a coding region that it aligned to was considered for further analysis . For each particular codon , we identified all instances in our coding regions where this codon ( say CTC ) occurs uniquely within a window of 10 codons upstream and 10 codons downstream ( i . e . , a window of 19 codons with the target CTC in the center of the window ) . For footprints that are 10 codons long , there will be 10 classes of footprints where this particular CTC can appear—position 1 , position 2 , . . . , position 10 . Thus , all footprints where the first codon of the footprint aligns to this particular CTC will belong to the position 1 class , all footprints where the second codon of the footprint aligns to this particular CTC will belong to position 2 class , etc . In the absence of any codon preference of the ribosome , we would expect to see a uniform distribution of reads across these 10 classes . In general , the codon-positional preference is described by the relative frequency of reads in each of these classes . These relative frequency distributions can be fairly averaged over all target regions over all genes centered on a specific codon . This average we call the ‘Ribosome Residence Time’ ( RRT ) ; it is intended as a statistical estimate of the relative time spent by the ribosome at a particular codon at a particular position . Typically we discuss the RRT at position 6 ( the A-site ) , but we also discuss the RRT at position 5 ( the P-site ) . Regions on highly expressed genes can be fairly compared with similar regions on genes with lower expression , because we are dealing with relative frequency distributions ( i . e . , percentage instead of read counts ) . Each region represents an independent trial of any positional preference of the given central codon . Averaging over the 100s or 1000s of occurrences on the genome provides for a statistically rigorous analysis . Relative frequency distributions will only be representative if the observed number of reads in the window is high enough that no single position dominates the distribution . For this reason , we restricted our analysis to windows with at least 20 total reads with at least 3 non-empty classes . The frequency distributions are not normally distributed; this is in part because the number of reads is limited , so many windows have zero footprints at many positions , so the mode of the distribution is often 0 . Nevertheless we believe that the mean is a good summary statistic . Maximum values are less than 1 , so the mean cannot be skewed by extremely high values . We have also calculated the RRTs using the median of the windows instead of the mean , but the results are almost indistinguishable . The Spearman rank correlation between the RRTs as calculated by the mean , and by the median , is 0 . 97 , while the Kendall Tau correlation is 0 . 89 . For each codon , we obtain the two-tailed p-value by comparing the experimentally determined relative frequency to the distribution of 10 , 000 relative frequencies based on permuted results . For each of the 10 , 000 instances , for each considered window , we permute the footprint counts of the 10 position classes . We performed our RRT analysis on the Ingolia et al . ( 2009 ) data , with small modifications . We did not perform the checks of read quality and the number of mismatches , as this was taken care of in pre-processing steps ( See Sequence Processing and Alignment ) . We also considered all reads with at least 24 nucleotides and performed our relative frequency calculations on the eight codons , because the majority of the reads were shorter than the reported size selection of RNA fragments ∼27–31 nucleotides in length . The statistical significances shown in Table 1 were obtained by constructing 10 , 000 simulated frequency distributions by randomly and independently permuting each region's frequency distribution prior to averaging . The rank of each observed positional peak among these simulated distributions established the p-value . We developed a p-value computation to assess whether the codons for a given amino acid behave similar to one another ( i . e . , are coherent ) or not . Each codon's RRT values along the positions of a footprint may be considered as a k-dimensional vector , where k is the number of positions in the footprint ( 10 for long reads vs 7 for short reads ) . We consider the position in k-dimensional space of the end-point of this vector . For the set of synonymous codons for a particular amino acid , we consider the set of endpoints . For any given set of c such endpoints , we can compute the average pairwise distance d between them over all c ( c-1 ) /2 pairs of points . If all codons for an amino acid behave similarly , then the endpoints are close together , and the distance d is relatively small , indicating codon coherence ( amino-acid specific behavior ) , whereas if the various codons for a given amino acid behave differently ( non-coherence , codon-specific behavior ) , then the distance d is relatively large . To judge the sizes of these distances for a particular set of points , S , containing c codons ( c ranges from 2 to 6 ) for a particular amino acid , we use a p-value . We construct 10 , 000 random samples of c codons drawn from the 61 possible sense codons . For each sample , we compute the average pairwise distance and compare this to the average pair distance of S . The rank of S in this distribution provides a p-value , which is significant if the vast bulk of random samples have greater pairwise distance than S . Results are shown in Figure 8 . An mRNA encoding a given protein could use only the fastest codon for each amino acid or only the slowest or it could use a mixture . In each case , the mRNA would occupy , or titrate out , a different number of ribosomes . A transcriptome of mRNAs using only the slowest codons would require more ribosomes to make a given amount of total protein in a given time than a transcriptome of mRNAs using only the fastest codons . We roughly estimated the size of this effect for the range of codon decoding speeds we observed . We generated in silico a yeast transcriptome using only the fastest codon for each amino acid at position 6 ( from Table 1 ) or only the slowest codon or a random mixture of codons . Furthermore , we weighted the abundance of each mRNA according to its actual abundance as measured by Lipson et al . ( 2009 ) . We then compared the relative time required to translate each of these in silico transcriptomes by a set number of ribosomes based on the RRT values for each codon at position 5 and 6 , and also assuming that the relevant delay is the delay at position 5 plus the delay at position 6 ( since these two reactions must occur sequentially and not simultaneously before the ribosome can shift along the mRNA ) . In doing this , we noted that the RRT values for position 5 are negatively correlated with those at position 6 . Results are as follows: the random encoding requires 1 . 050 as long as WT; the slowest encoding requires 1 . 168 as long as WT; and the fastest encoding requires 0 . 930 as long as WT . Note that this estimate uses the simplification that each species of mRNA will initiate translation at the same rate . A more accurate calculation in which the more abundant mRNAs initiate more rapidly than average would increase the difference between the WT and the random encodings . When the accepted manuscript was published , RRT values from an earlier version of the algorithm were erroneously used for Figure 5 ( but not for other figures ) , giving a correlation of –0 . 7 between RRT and codon usage . The current algorithm , used here , gives a corrected version of Figure 5 , shown here , with a correlation of –0 . 52 . | Genes contain the instructions for making proteins from molecules called amino acids . These instructions are encoded in the order of the four building blocks that make up DNA , which are symbolized by the letters A , T , C , and G . The DNA of a gene is first copied to make a molecule of RNA , and then the letters in the RNA are read in groups of three ( called ‘codons’ ) by a cellular machine called a ribosome . ‘Sense codons’ each specify one amino acid , and the ribosome decodes hundreds or thousands of these codons into a chain of amino acids to form a protein . ‘Stop codons’ do not encode amino acids but instead instruct the ribosome to stop building a protein when the chain is completed . Most proteins are built from 20 different kinds of amino acid , but there are 61 sense codons . As such , up to six codons can code for the same amino acid . The multiple codons for a single amino acid , however , are not used equally in gene sequences—some are used much more often than others . Now , Gardin , Yeasmin et al . have instantly halted the on-going processes of decoding genes and building proteins in yeast cells . Codons being translated into amino acids are trapped inside the ribosome; and codons that take the longest to decode are trapped most often . By using a computer algorithm , Gardin , Yeasmin et al . were able to measure just how often each kind of sense codon was trapped inside the ribosome and use this as a measure of how quickly each codon is decoded . The more often a given codon is used in a gene sequence , the less likely it was found to be trapped inside the ribosome—which suggests that these codons are decoded quicker than other codons and pass through the ribosome more quickly . Put another way , it appears that genes tend to use the codons that can be read the fastest . Certain properties of a codon also affected its decoding speed . Codons with more As and Ts , for example , are decoded faster than codons with more Cs and Gs . Furthermore , whenever a chemically unusual amino acid called proline has to be added to a new protein chain , it slowed down the speed at which the protein was built . The method described by Gardin , Yeasmin et al . for peering into a decoding ribosome may now help future studies that aim to answer other questions about how proteins are built . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2014 | Measurement of average decoding rates of the 61 sense codons in vivo |
Social relationships have profound effects on health in humans and other primates , but the mechanisms that explain this relationship are not well understood . Using shotgun metagenomic data from wild baboons , we found that social group membership and social network relationships predicted both the taxonomic structure of the gut microbiome and the structure of genes encoded by gut microbial species . Rates of interaction directly explained variation in the gut microbiome , even after controlling for diet , kinship , and shared environments . They therefore strongly implicate direct physical contact among social partners in the transmission of gut microbial species . We identified 51 socially structured taxa , which were significantly enriched for anaerobic and non-spore-forming lifestyles . Our results argue that social interactions are an important determinant of gut microbiome composition in natural animal populations—a relationship with important ramifications for understanding how social relationships influence health , as well as the evolution of group living .
Vertebrate intestines are home to thousands of bacterial species that exert profound effects on their hosts: they train the immune system , produce vitamins , help resist pathogens , and contribute substantially to daily energy acquisition ( Bergman , 1990; Turnbaugh et al . , 2006; Hooper et al . , 2012; Bengmark , 2013; Morgan et al . , 2013 ) . In humans , inter-individual variation in gut microbiome composition has repeatedly been linked to major health concerns , including obesity , diabetes , cancer , heart disease , and autoimmune disorders ( e . g . , Turnbaugh et al . , 2009; Hooper et al . , 2012; Bengmark , 2013; Iida et al . , 2013; Koeth et al . , 2013; Viaud et al . , 2013 ) . However , despite its importance , large gaps remain in our understanding of the forces that shape gut microbiome composition . Among the least understood but potentially most significant such forces are the effects of host social interactions . From an evolutionary perspective , social effects on the gut microbiome may be an underappreciated consequence of group living , associated with both fitness costs and benefits ( Lombardo , 2008; Archie and Theis , 2011; Ezenwa et al . , 2012; Montiel-Castro et al . , 2013 ) . For example , co-housing in lab mice promotes the transmission of bacterial communities that contribute to inflammatory bowel disease , implicating social relationships in microbiome-associated disease risk ( Garrett et al . , 2010 ) . In bumblebees , socially transmitted gut bacteria protect against a widespread and virulent gut parasite , suggesting that socially mediated microbial transmission can also confer powerful benefits ( Koch and Schmid-Hempel , 2011 ) . If social interactions predict gut microbiome composition in free-living vertebrates as well , this link could help explain the strong association between social interactions and health in highly social species ( e . g . , Berkman and Syme , 1979; House et al . , 1988; Sapolsky , 2004; Holt-Lunstad et al . , 2010 ) . A handful of recent studies in humans and other primates provide circumstantial evidence for social effects on the gut microbiome ( Degnan et al . , 2012; Kinross and Nicholson , 2012; Yatsunenko et al . , 2012; Song et al . , 2013 ) . For instance , in wild chimpanzees , social group membership predicts the identity and abundance of gut microbes , while kinship , age , and sex do not ( Degnan et al . , 2012 ) . In humans , shared residence predicts gut microbiome similarity ( Kinross and Nicholson , 2012; Yatsunenko et al . , 2012; Song et al . , 2013 ) . To date , these effects have largely been attributed to shared diets , as members of the same household or social group tend to consume similar foods in similar proportions ( Claesson et al . , 2011; Kinross and Nicholson , 2012; Yatsunenko et al . , 2012 ) . However , social relationships could also shape gut microbiomes more directly , via transmission from shared environments ( Lax et al . , 2014 ) or during physical contact . Differentiating between these mechanisms requires fine-grained data on social interactions , shared environments , and diet . Such complementary data sets are rare , but are frequently collected in long-term primate field studies . Here , we leveraged one such study , on the intensively studied Amboseli baboons of Kenya ( Alberts and Altmann , 2012 ) , to test whether social group structure and social interactions within groups predict either the taxonomic or the functional composition of the gut microbiome . Like humans , baboons are highly social , group-living primates . Members of the same social group travel together , consume similar foods , and drink from the same water sources . Within social groups , individuals selectively engage in frequent affiliative grooming interactions , which solidify social bonds and have the potential to mediate bacterial transmission . Within this context , we addressed three central questions . We first asked ( i ) does social group membership predict gut microbiome composition , as shown for humans and chimpanzees ( Degnan et al . , 2012; Kinross and Nicholson , 2012; Yatsunenko et al . , 2012; Song et al . , 2013 ) ? We then asked two novel questions that have not been addressed in prior studies: ( ii ) within social groups , do rates of social interactions ( captured here by grooming-based social networks ) predict gut microbiome similarity after accounting for dietary patterns , shared environments , and kinship ? And ( iii ) which bacterial species , with what lifestyle traits , are most likely to be socially transmitted , both between and within social groups ?
Across all 48 individuals , social group membership explained 18 . 6% of global variation in gut microbial species composition ( as summarized by a Bray–Curtis dissimilarity matrix; PERMANOVA for social group effects: p < 10−4; Figure 1C ) . Social group membership was also the dominant source of variance in the abundance of enzyme gene orthologs encoded by gut microbes , explaining 10 . 8% of global variance in a Bray–Curtis dissimilarity matrix ( PERMANOVA: p = 0 . 003; Figure 1D ) . In contrast , sex , age , and sequencing read depth made comparatively minor or non-significant contributions to gut microbiome composition ( PERMANOVA: sex , age and read depth explained 3 . 6% , p = 0 . 026; 5 . 3% , p = 0 . 052; 6 . 0% , p = 0 . 024 of variance in taxonomic composition , respectively; no significant variation was explained by sex , age , or read depth for enzyme gene orthologs ) . Furthermore , social group remained a strong and significant predictor of taxonomic and enzyme gene ortholog composition even after controlling for genetic relatedness between study subjects ( partial Mantel test for taxonomic composition: r = 0 . 378 , p < 10−5; for enzyme gene orthologs: r = 0 . 140 , p = 1 . 6 × 10−3 ) . Previous associations between social proximity and gut microbial composition in humans and other primates have largely been attributed to diet ( Degnan et al . , 2012; Kinross and Nicholson , 2012; Yatsunenko et al . , 2012 ) . However , the two social groups in our study inhabited a relatively homogeneous savannah environment and exploited very similar resources . During the sample collection period , half of each group's diet was devoted to grass corms , and similar proportions were devoted to other food types , including grass seed heads , Acacia tortilis seed pods , leaves ( primarily grass blades ) , and Acacia xanthophloea gum ( Figure 1B; Supplementary file 7 ) . The only diet component that differed significantly between the two groups was the proportion devoted to fruit ( permutation test: p = 0 . 05 ) . However , we found no differences between the two groups in the abundance of two common fruit-associated bacterial enzymes , pectinesterase ( p-value for social group in a linear mixed effects model: p = 0 . 306 ) and pectate lyase ( p-value for social group in a linear mixed effects model: p = 0 . 869 ) . Furthermore , patterns of differential taxonomic abundance between groups did not recapitulate differences associated with differential consumption of fresh fruits and vegetables described in a human gut microbiome data set ( Davenport et al . , 2014; see ‘Materials and methods’ ) . Despite few detectable differences in diet , unidentified environmental differences between Mica's group and Viola's group could explain the differences in gut microbiome composition we observed . To test whether social contacts per se predicted gut microbiome composition , we turned to fine-grained data on within-group grooming interactions . Grooming is by far the most common form of physical contact in baboons . Importantly , the strength of grooming relationships between pairs of individuals in the same social group varies considerably , despite the fact that all members of a social group travel together and use the same resource base . To test whether physical contact predicted gut microbiome composition , we constructed grooming networks for each social group , using all grooming interactions observed in the year prior to and during microbiome sampling ( Figure 2A , B ) . We found that , in both groups , closer grooming partners harbored more similar communities of gut bacteria ( Mantel test between Bray–Curtis microbiome dissimilarity matrices and social network matrices: Mica's group r = −0 . 257 , p = 3 . 0 × 10−4; Viola's group r = −0 . 173 , p = 8 . 0 × 10−4; Figure 2C , D ) . This pattern was not driven by host genetic effects: although female relatives have stronger grooming bonds , controlling for pairwise relatedness still produced strong support for a relationship between grooming and taxonomic composition for Viola's group ( partial Mantel test controlling for kinship: r = −0 . 148 , p = 2 . 0 × 10−3 ) , and a consistent trend in Mica's group ( partial Mantel test controlling for kinship: r = −0 . 163 , p = 0 . 060 ) . Interestingly , extending this analysis to the level of enzyme gene orthologs suggested that close grooming partners also have functionally more similar gut microbiomes . Grooming networks predicted variation in within-group enzyme gene ortholog abundance for Mica's group ( partial Mantel test controlling for kinship: r = −0 . 22 , p = 0 . 014 ) , but not Viola's group ( partial Mantel test controlling for kinship: r = −0 . 051 , p = 0 . 166 ) . 10 . 7554/eLife . 05224 . 009Figure 2 . Grooming-based social networks predict microbiome composition . Social networks based on grooming interactions in the year prior to and including the month of microbiome sampling in ( A ) Mica's and ( B ) Viola's social groups . Each circle represents an individual ( with the individual's ID listed within the circle ) . Lines represent grooming interactions between individuals , and heavier lines reflect stronger grooming relationships . ( C and D ) Violin plots depicting the relationships between pairwise grooming bond strength vs pairwise Bray–Curtis dissimilarity in taxonomic composition in Mica's and Viola's groups , respectively . White dots represent median values and grey rectangles represent the first and third quartiles of the data . Rotated kernel density plots representing the underlying data are shown on each side . Stronger bonds predict more similar gut microbiotas in both groups ( Mica's group: Mantel test r = −0 . 257 , p = 3 . 0 × 10−4; Viola's group: r = −0 . 173 , p = 8 . 0 × 10−4 ) . Parallel results based on de novo assembly are shown in Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 05224 . 00910 . 7554/eLife . 05224 . 010Figure 2—figure supplement 1 . Evidence for social structuring of the gut microbiome based on de novo assembly . Estimating gut microbiome taxonomic composition by comparison to de novo bacterial genome assemblies also produces congruent evidence for social structuring . ( A ) Proportional representation of common phyla in each sample , grouping phyla not present at >1% in at least one sample together are ‘rare phyla’ . ( B ) Principal coordinates projection for individuals from Mica's group and Viola's group separates samples by social group along the first axis . ( C ) Strength of pairwise grooming relationships , and thus within group social structure , explains levels of similarity and dissimilarity in gut microbiome taxonomic composition . Data are shown for Mica's group . DOI: http://dx . doi . org/10 . 7554/eLife . 05224 . 010 Despite the relative homogeneity of diet within social groups , our results could still be explained by a diet-related mechanism if close grooming partners consumed more similar diets . Alternatively , close social partners might experience similar environmental exposures if they used more similar microenvironments in the group's home range . We tested these possibilities directly , focusing on adult females for whom diet composition and spatial proximity data were routinely collected ( N = 11 females in Mica's group and N = 20 females in Viola's group ) . Grooming network proximity also predicted microbiota composition in this restricted data set ( Mantel tests: Mica's group: r = −0 . 328 , p = 9 . 0 × 10−3; Viola's group: r = −0 . 228 , p = 2 . 6 × 10−3 ) , and remained a significant predictor of microbiota composition after accounting for dietary similarity ( partial Mantel test controlling for dietary similarity: Mica's group p = 0 . 020; Viola's group: p = 0 . 005 ) and spatial proximity ( partial Mantel test controlling for spatial proximity for Mica's group p = 0 . 039; Viola's group: p = 0 . 005 ) . Additionally , we found no evidence that close social partners consumed more similar diets ( Mantel tests: Mica's group: Mantel r = −0 . 200 , p = 0 . 080; Viola's group: Mantel r = 0 . 0942 , p = 0 . 876 ) . We next investigated which bacterial species were associated with the strong signature of social structure in our data set . To identify these ‘socially structured’ species , we focused on the 327 most prevalent species in our data set ( i . e . , those found in ≥50% of samples ) . Using a mixed effects model controlling for age , sex , read depth , and host genetic relatedness , we identified 64 species ( 19 . 6% , using a 10% false discovery rate ) that were significantly differentially abundant in the two social groups . We performed a complementary analysis , using a test of spatial autocorrelation , to investigate whether close grooming partners exhibited similar bacterial abundances within social groups as well ( due to the larger sample size , we performed these tests in Viola's group; see ‘Materials and methods’ ) . Among the same set of 327 prevalent species , we found 51 species ( 15 . 6% , 10% false discovery rate ) for which proximity within the group's grooming network significantly predicted abundance ( Supplementary file 8 ) . Interestingly , 15 species were significantly socially structured both between groups and within social networks—more species than expected by chance ( hypergeometric test , p = 0 . 020 ) . We next conducted an enrichment analysis to test whether the set of significantly socially structured species contained some taxonomic groups more often than by chance . We found that socially structured species were phylogenetically non-random at both between-group and within social network levels of analysis ( Figure 3A , B ) . Moreover , at both levels of analysis , similar taxonomic groups were significantly enriched for socially structured species ( red asterisks on Figure 3 ) , including the phylum Actinobacteria; the families Bifidobacteriaceae , Coriobacteriaceae , and Veillonellaceae; and the genus Bifidobacterium , a group of Gram-negative bacteria that has been linked to beneficial health effects in humans ( Servin , 2004; Gronlund et al . , 2007; Turroni et al . , 2008 ) . The striking similarities between the two levels of analysis suggest that common underlying mechanisms—mediated by direct social contact rather than diet or general physical proximity—account for both between-group differences and grooming network effects within groups . 10 . 7554/eLife . 05224 . 011Figure 3 . Socially structured species are taxonomically and phenotypically nonrandom . Bacterial taxonomic groups significantly enriched ( 10% FDR ) for socially structured species ( A ) between social groups and ( B ) within the grooming network for Viola's group ( Supplementary file 8 ) . Vertical dashed lines depict a fold enrichment of 1 , representing the background level of taxon abundance in the data set . Red asterisks denote taxonomic groups identified as significantly enriched at both levels of analysis . ( C ) Significant enrichment of anaerobic , non-spore-forming bacterial taxa , both between and within groups , at both species and genus levels ( socially structured species between groups , species level traits: p = 0 . 017; socially structured species within group , species level traits: p = 0 . 067; socially structured species between groups , genus level traits: p = 0 . 036; socially structured species within group , genus level traits: p = 0 . 040 ) . See Figure 3—figure supplements 1 , 2 for a comparison of the enrichment of p-values in our data set vs an empirical null distribution . DOI: http://dx . doi . org/10 . 7554/eLife . 05224 . 01110 . 7554/eLife . 05224 . 012Figure 3—figure supplement 1 . Enrichment of low p-values in the data vs an empirical null: between group analyses . To confirm that our modeling approach ( quantile normalization of species relative abundances , followed by mixed effects modeling in GEMMA ) did not bias us towards detecting false positives , we compared the signal in our true data set against an empirically derived null . The histogram distribution of p-values for the true data ( gold ) is plotted against the distribution of p-values from 10 permutations ( blue ) . In each permutation , group membership was scrambled across the data set while keeping the modeling approach , kinship structure , and all other covariates constant . The inset shows a quantile–quantile plot of the same data , with clear enrichment of differentially abundant species in the actual data vs the empirical null . No differentially abundant species are detected at a 10% FDR in the permuted data sets , while 64 are discovered in the true data set . DOI: http://dx . doi . org/10 . 7554/eLife . 05224 . 01210 . 7554/eLife . 05224 . 013Figure 3—figure supplement 2 . Enrichment of low p-values in the data vs an empirical null: within group network analysis . To confirm that our modeling approach ( Moran's I statistic within Viola's group ) did not bias us towards detecting false positives , we compared the signal in our true data set against an empirically derived null . The histogram distribution of p-values for the true data ( gold ) is plotted against the distribution of p-values from 10 permutations ( blue ) . In each permutation , species abundance was scrambled across group members while keeping the modeling approach and social network structure constant . The inset shows a quantile–quantile plot of the same data , with clear enrichment of socially structured species in the actual data vs the empirical null . No socially structured species are detected at a 10% FDR in the permuted data sets , while 51 are discovered in the true data set . DOI: http://dx . doi . org/10 . 7554/eLife . 05224 . 013 Finally , we extended our enrichment analysis to test whether the set of socially structured species was enriched for particular bacterial lifestyles . We reasoned that , if socially structured species depend on direct transmission between baboons , as our data suggest , they should be less likely than other species to persist outside of a host . Thus , we predicted that socially structured species would tend to be anaerobic and unable to produce spores . To test these predictions , we turned to information about bacterial lifestyles available in the Genomes OnLine Database ( Pagani et al . , 2012 ) , using both species-level ( n = 138 ) and genus-level ( n = 299 ) traits ( see ‘Materials and methods’ for trait assignment criteria ) . We found that socially structured species were consistently enriched ( relative to all species or genera tested ) for an anaerobic , non-spore forming lifestyle ( Figure 3C; hypergeometric tests for socially structured species between groups , species level traits: p = 0 . 017; socially structured species within group , species level traits: p = 0 . 067; socially structured species between groups , genus level traits: p = 0 . 036; socially structured species within group , genus level traits: p = 0 . 040 ) . For instance , 17% of the species in this analysis differed significantly in abundance between social groups; however 32% of anaerobic and non-spore forming species were significantly socially structured . Notably , no species that were both aerobic and spore-forming were socially structured at the level of social groups or social networks , except for one case in the genus-level analysis .
Taken together , our results provide strong evidence that social interactions directly affect the composition of the gut microbiome in wild baboons . To our knowledge , this study is the first to test whether rates of interaction within cohabiting groups , as opposed to between groups or households , explain variation in the gut microbiome . Specifically , we found that an individual's contacts in a grooming-based social network , as well as its membership in a given social group , were highly predictive of its gut microbiome composition at both the species and genic levels . Unlike prior studies , we were able to exclude kinship , shared diet , and shared environment as the basis for our observations . Our results are thus unique among studies to date in the degree to which they implicate direct , affiliative physical contact as a determinant of gut microbiome composition in natural populations . Our data also provide the first evidence in vertebrates that social effects on the microbiome extend to its functional composition . These findings lend important support to the hypothesis that social interactions play a role in the health-related consequences of variation in gut microbiome composition ( e . g . , Turnbaugh et al . , 2009; Hooper et al . , 2012; Bengmark , 2013; Iida et al . , 2013; Koeth et al . , 2013; Viaud et al . , 2013 ) , with potentially important consequences for the evolution of sociality ( Lombardo , 2008; Archie and Theis , 2011; Ezenwa et al . , 2012; Montiel-Castro et al . , 2013 ) . Thus , our results highlight the importance of socially mediated transmission in shaping gut microbiomes . However , unlike some prior studies in mice and bumblebees ( Garrett et al . , 2010; Koch and Schmid-Hempel , 2011 ) , baboons are not coprophagic , raising a question about the mechanisms that facilitate gut microbial transfer between social partners . One possibility is that the duration and intimacy of grooming bouts , which include frequent hand-to-mouth contact , may be important in exposing baboons to the gut bacteria of their grooming partners . Furthermore , some grooming bouts , especially those directed from adult males to estrous females , concentrate heavily on the ano-genital region , increasing the probability of fecal-oral transfer . Such close contact may be especially important in the transmission of anaerobic , non-spore-forming species , as these bacteria are not thought to persist for long periods of time outside of a host ( Wilson , 2008 ) . However , some relatively hardy bacterial species may also be transmitted via social contact ( VanderWaal et al . , 2013 ) , and recent modeling efforts suggest that fecal-oral transmission can be highly efficient in socially structured host populations , even when transmission is indirectly mediated through the soil ( Nunn et al . , 2011 ) . Interestingly , our observations suggest that social partners not only share more similar sets of gut microbes , but also similar abundances of individual microbial species . One explanation for this pattern is that when bacteria from a host colonize a social partner , they arrive pre-adapted to occupy the available gut microbial niches in their new host ( Walter and Ley , 2011 ) . Specifically , because members of a single bacterial species can have markedly different gene contents , a given member of a gut microbial species may perform different functions in different hosts ( Walter and Ley , 2011; Costello et al . , 2012 ) . However , if social partners transmit bacteria with similar capabilities to each other , these bacteria may serve similar functions in both hosts and thus be found in similar abundances . This hypothesis could be further tested by assessing if bacterial species isolated from social partners tend to represent shared strains that perform similar biological functions . In humans , affiliative physical contact ( e . g . , hugging , kissing , holding hands ) is common and may provide a similar route through which close social partners transmit gut bacteria . In addition , surfaces in human homes may act as reservoirs for household-specific bacterial communities ( Lax et al . , 2014 ) , possibly facilitating social transmission through intermediate surfaces . Future work , in both humans and animals , will be important to establishing the relative importance and generality of socially mediated transmission . In particular , population genetic studies have the potential to directly map the genetic structure of microbiome-associated species onto the social structure of host populations to test whether close social partners tend to share genetically more similar bacterial populations than non-partners ( e . g . , VanderWaal et al . , 2013 ) . Fine-grained studies of how gut microbial communities change in social species , before and after perturbations to their social networks , will also be important for understanding the time scales on which social transmission of microbes act . Such efforts would also contribute an important longitudinal perspective . Our power to identify associations between social relationships and microbiome composition in this study was probably facilitated by our sampling scheme , which eliminated the contribution of temporal or seasonal effects . More comprehensive long-term studies will be valuable for placing these effects in context , alongside concomitant changes in season , diet , and resource use . In humans , variation in the taxonomic and genic composition of the microbiome is increasingly linked to health issues , such as obesity and autoimmune disorders ( e . g . , Turnbaugh et al . , 2009; Hooper et al . , 2012; Bengmark , 2013; Koeth et al . , 2013 ) . Health and survival in social species ( including humans and baboons ) are also strongly associated with social relationships ( Berkman and Syme , 1979; House et al . , 1988; e . g . , Sapolsky , 2004; Silk et al . , 2009; Holt-Lunstad et al . , 2010; Silk et al . , 2010; Archie et al . , 2014 ) . However , few studies have connected these two observations . By highlighting the strong relationship between microbiome composition and social networks , our findings indicate the importance of further research in this area . One of the most important unanswered questions is whether social network-mediated microbiome sharing produces net fitness benefits or costs for hosts . Previous research on fecal-oral or social network-mediated transmission has focused almost exclusively on pathogens or parasites . Microbiome studies have the potential to broaden this perspective to include species with beneficial effects . Indeed , while we found several socially structured taxa that have been associated with pathogenic effects ( e . g . , Fusobacterium spp , Campylobacter ureolyticus ) , we found several other bacteria thought to be beneficial to hosts . For example , members of the phylum Actinobacteria , especially the genus Bifidobacterium , are commonly thought to have probiotic effects in humans due to their role in complex carbohydrate digestion , pathogen inhibition , and vitamin production ( Servin , 2004; Gronlund et al . , 2007; Turroni et al . , 2008 ) . Understanding the balance between social transmission of pathogenic vs commensal or beneficial bacteria thus promises to provide valuable new insight into the link between disease risk and the evolution of sociality .
Study subjects were 48 wild , adult baboons living in the Amboseli ecosystem , a semi-arid savannah in southern Kenya ( Supplementary file 1 ) . The baboons were studied as part of the Amboseli Baboon Research Project ( ABRP ) , which has been collecting continuous , individual-based data on all the members of several baboon social groups since 1971 ( Alberts and Altmann , 2012 ) . The specific subjects for this project represented near complete sampling ( 92% ) of all the adult members of two social groups , called ‘Mica's group’ and ‘Viola's group’ . The baboons are individually recognized by experienced observers , who visit each group several times per week , year round , for 5-hr monitoring visits ( Alberts and Altmann , 2012 ) . Distal gut microbiome composition was characterized using fecal samples collected opportunistically from known individuals . All fecal samples were collected during a single 1-month span in the dry season ( 7 July 2012 to 8 Aug 2012: Supplementary file 1 ) . Samples were collected within a few minutes of defecation , thoroughly mixed , and then preserved in 95% ethanol ( 2:5 feces to ethanol ) . DNA was extracted from each sample using MO BIO's PowerSoil DNA Isolation kit , according to the manufacturer's instructions ( MO BIO Laboratories , Inc . , Carlsbad , CA ) . For each individual , 200 ng of extracted DNA were prepared for metagenomic sequencing on an Illumina HiSeq 2500 using the Kapa Biosystems Library Preparation Kits ( Kapa Biosystems , Wilmington , MA ) . Specifically , DNA samples were sheared to an average size of 400 base pairs , ligated to barcoded adapters , and subjected to 100 base pair paired end sequencing at the UCLA Neuroscience Genomics Core . In total , we generated 1 . 4 billion raw , paired-end Illumina sequences across all samples ( mean ± SD = 14 . 4 ± 13 . 7 million read pairs per sample ) . All raw reads are deposited in the National Center for Biotechnology Information ( NCBI ) Short Read Archive ( BioProject PRJNA271618 ) . Species-level taxonomic abundances were inferred for all samples using MetaPhlAn 2 . 0 ( Segata et al . , 2012 ) . MetaPhlAn 2 . 0 estimates the relative abundance of bacterial species by mapping reads against a set of clade-specific marker sequences , which unequivocally identify microbial clades at the species level or higher taxonomic levels . Based on 12 , 926 complete bacterial genomes , MetaPhlAn 2 . 0 is able to provide clade-specific markers for a total of 3848 bacterial species , 925 of which were detected in our data set ( Supplementary files 2 , 3 ) . Specifically , we mapped our sequence reads against the clade-specific markers using the ‘very-sensitive-local’ alignment mode implemented in Bowtie 2 ( Langmead et al . , 2009 ) . This mode produces alignments that can be trimmed at one or both extremes in order to optimize the alignment score . Because spurious or poor-quality reads are unlikely to match any of the pre-defined marker sequences , no preprocessing of the metagenomic DNA sequences was performed , as recommended by the authors . However , we tested the robustness of these estimates by re-running MetaPhlAn 2 . 0 on a subset of our data after trimming the reads to eliminate adapter sequences and bases with a quality score <20 . Correlations between the bacterial abundance estimates obtained using unprocessed data and those obtained using the trimmed data were always above 0 . 97 , confirming that MetaPhlAn 2 . 0 is indeed highly robust to potential sequence artifacts . To investigate variation in the genic composition of the gut microbiome , we combined information from the Kyoto Encyclopedia of Genes and Genomes database ( KEGG: Kanehisa and Goto , 2000; Kanehisa et al . , 2014 ) with the HMP Unified Metabolic Analysis Network ( HUMAnN ) v0 . 99 pipeline ( Abubucker et al . , 2012 ) . We first filtered the forward reads for quality using USEARCH v7 . 0 ( Edgar , 2010 ) . Specifically , for each sample , we ( i ) trimmed reads to a length of 99 bases , ( ii ) excluded reads shorter than 99 bases , and ( iii ) excluded reads with expected error ( a measure of read quality in USEARCH based on base call quality and read length ) > 0 . 5 . An average of 87 . 9% of all reads passed quality filtering ( Figure 1—figure supplement 4 ) . Remaining reads were translated in all three possible reading frames and aligned against a reduced KEGG database ( last free version , June 2011 ) using the ublast function of USEARCH v7 . 0 and default parameters . The reduced KEGG database was generated by removing entries for which no KEGG orthology ( KO ) assignments existed and subsequently clustering each KO individually ( uclust v1 . 5 . 579 , using 85% sequence identity as the clustering cutoff ) ( Edgar , 2010; Kanehisa and Goto , 2000; Kanehisa et al . , 2014 ) . This database was converted to a USEARCH-compatible database file prior to running ublast . An average of 23 . 0% of the input reads across all samples were assigned an identity from the KEGG database ( Figure 1—figure supplement 4 ) . Finally , the ublast output was used as input for HUMAnN . HUMAnN was configured to generate KO abundances from BLAST hits of enzymes as well as coverage and abundances for KEGG pathways and modules . To investigate the correlation between social group membership and the composition of baboon gut microbiomes , we constructed separate summaries of the complete taxonomic composition data set from MetaPhlAn 2 . 0 and the complete enzyme gene ortholog abundance data set from HUMAnN . Specifically , for each data set , we used the vegdist function in the R package vegan ( Oksanen et al . , 2013 ) to calculate a 48 × 48 Bray–Curtis dissimilarity matrix , which describes the global dissimilarity in gut microbial composition between each pair of individuals in the data set . To understand sources of variance in these matrices , we performed PERMANOVA analyses ( adonis function in vegan ) with 10 , 000 permutations . In addition to social group , predictor variables in this analysis were age , sex , and total read depth . Sex was known from direct observation of the study subjects . Ages were known to within a few days' error for 39 of the 48 individuals in the data set . The remaining 9 individuals immigrated into the population after birth , and so their ages were estimated using well-defined metrics and comparison to known-age animals ( Alberts and Altmann , 1995 ) . Of these 9 individuals , 6 animals had birth dates estimated to be accurate within 1 year , and 3 animals had birth dates estimated to be accurate within 2 years . All study subjects were adults ( i . e . , all females had attained menarche , and all males had attained adult dominance rank; Onyango et al . , 2013 ) . To assess the possible confounding effects of kinship , we constructed a matrix of pairwise genetic relatedness values from the extensive pedigree data available for the Amboseli population ( e . g . , Buchan et al . , 2003; Alberts et al . , 2006; current pedigree includes 1409 individuals , with 1298 known maternal links and 526 known paternal links ) using the R package pedantics ( Morrissey and Wilson , 2010 ) . We then used partial Mantel tests to assess the correlation between a matrix describing group co-residency ( cells took a value of 1 if two individuals resided in different groups and a value of 0 if they were co-resident ) and the Bray–Curtis dissimilarity matrix for taxonomic composition , controlling for the pairwise genetic relatedness matrix . To assess differences in diet between the two social groups , we used direct observations of the food consumed by adult female baboons in each group during the month in which samples were collected . Diet composition data were collected in the context of random-order focal animal sampling ( Altmann , 1974 ) . Specifically , ABRP observers spent 4 hr of each group visit rotating through the group , conducting focal animal samples on adult females in the order dictated by a randomized list . Each focal animal sample was 10 min long , during which activity ( feeding , walking , resting etc ) was recorded during point samples collected at 1-min intervals . When feeding was observed ( 353 point samples in Mica's group and 731 point samples in Viola's group ) , the observers recorded the type of food consumed . Food types were divided into 7 categories , including: ( 1 ) corms of all grass species , ( 2 ) seed heads of all grass species , ( 3 ) pods from A . tortilis and A . xanthophloea ( 4 ) fruits , including those from Azima tetracantha , Salvadora persica , Solanum dubium , Trianthema ceratosepala , and Tribulus terrestris , ( 5 ) leaves from Lyceum sp . and all grass species , ( 6 ) gum from A . xanthophloea , and ( 7 ) unknown/unidentified diet items ( Supplementary file 7 ) . To calculate the contribution ( including confidence intervals ) of each of the seven major food categories to each group's diet , we conducted 1000 random subsamples of one foraging point sample per focal animal sample . We took this approach to avoid autocorrelation between point samples collected during the same 10-min focal sample . To test for differences in diet between groups , we repeated the same analysis after randomly permuting group membership across the females in our data set . We calculated the proportion of cases in which between-group differences in the proportion of a food consumed exceeded between-group differences in the 1000 permuted data sets . This proportion is equivalent to the p-value for the null hypothesis that the two groups did not differ in diet . Because we detected a nominally significant difference ( p = 0 . 05 ) in the amount of fruit consumed by the members of Mica's group ( 7 . 9% , 95% CI: 0 . 0–8 . 3% ) and the members of Viola's group ( 2 . 2% , 95% CI: 1 . 0–7 . 3% ) during the sampling period , we also compared our results to a published data set of seasonal differences in gut microbiome composition in humans ( Davenport et al . , 2014 ) . These differences are believed to be the result of differences in consumption of fresh fruits and vegetables . Only three genera were detected as both significantly differentially abundant in the diet-related human data set ( FDR = 10% ) and significantly enriched for differential abundance between social groups in our data set , at a conservative ( for comparative purposes ) threshold of p ≤ 0 . 05 . Bifidobacterium was more abundant in humans when they consumed less fresh fruit; Prevotella and Treponema were more abundant when they consumed more fresh fruit . In our data set , however , Bifidobacterium levels were more abundant in Mica's group , which consumed more fruit , and Prevotella and Treponema species were more abundant in Viola's group , which consumed less fruit , suggesting that the patterns we observed are due to other sources of variance . To test whether grooming-based social networks predicted gut microbiome similarity , we constructed grooming networks based on ad lib observations of grooming interactions collected in the year prior to and including the period of microbiome sampling ( 8 August 2011 to 8 August 2012: 1648 total interactions , with 667 in Mica's group and 981 in Viola's group ) . Ad lib grooming interactions were collected throughout the monitoring visit while observers were carrying out focal animal sampling . Ad lib grooming data were used to calculate a count of observed grooming interactions between all adult dyads present in each social group ( range = 0 to 41 interactions per dyad ) . These data were used to construct a matrix of grooming relationship strength by scoring the strongest dyadic grooming relationship in each group as a 1 and weighting all other dyadic relationships relative to this strongest bond . We then used Mantel tests to investigate the strength of the correlation between group-specific grooming networks and group-specific Bray–Curtis dissimilarity matrices , constructed as described above . We used partial Mantel tests to assess whether grooming network-microbiome dissimilarity matrix correlations were driven by kinship ( represented using pedigree-based pairwise relatedness estimates ) . To investigate alternative explanations for social network effects on the microbiome , we collated data on diet and spatial proximity for members of each social group , focusing on adult females only ( comparable data were not available for adult males ) . Parallel to the time span for social network construction , we compiled data for the year prior to and including the month of microbiome sampling . For diet , we extracted all foraging-related point samples from the females in our microbiome data set ( 1380 points in Mica's group; 1989 points in Viola's group ) . We subsampled each data set so that only one point sample was represented per focal sample , which avoids autocorrelation between point samples collected during the same focal . We then constructed a table of the proportion of foods consumed per female , for each group separately , and used this table to calculate group-specific , diet-based Bray–Curtis dissimilarity matrices . For spatial proximity , we calculated the percent of time all adult female dyads spent within 5 m of each other during the same time period . Specifically , during each focal animal sample , the nearest adult female neighbor within 5 m is recorded at each 1-min point sample ( 893 points in Mica's group; 1637 points in Viola's group; range = 0–64 points per dyad ) . The proximity score between each pair of females ( within groups ) was calculated as the total number of point samples in which they were each other's nearest neighbors divided by the total number of point samples collected for each member of the dyad . To identify differentially abundant bacterial taxa by social group membership , we used the linear mixed model approach implemented in the program GEMMA ( Zhou and Stephens , 2012 ) , which allowed us to account for potential kinship effects in our data set . This approach assumes that the response variable ( taxon abundance ) is continuously distributed . To meet this assumption , we used methods established for analyzing high-throughput functional genomic data sets ( Rapaport et al . , 2013 ) . Specifically , we first quantile normalized abundance values across individuals , focusing only on the 327 most prevalent taxa ( i . e . , those found in at least 50% of hosts based on our MetaPhlAn 2 . 0 analysis , regardless of abundance ) , and then transformed the distribution of values for each species to a standard normal . We then fit the following linear mixed model to the data for each species:y=μ+xβx+aβa+sβs+rβr+u+ε , u∼MVN ( 0 , σu2K ) , ∈∼MVN ( 0 , σe2I ) . Here , y is the n by 1 vector of normalized taxon abundances for the n individuals in the sample; μ is the intercept; x is the n by 1 vector denoting social group membership; and βx is the effect size of social group membership . For the other covariates , a is the n by 1 vector denoting age and βa describes its effects on taxon abundance; s is the n by 1 vector denoting sex and βs its effect size; and r is the n by 1 vector denoting read depth and βr its effect size . The n by 1 vector of u is a random effects term to control for relatedness , and the n by n matrix K provides pedigree-based estimates of relatedness . Residual errors are represented by ε , an n by 1 vector , and MVN denotes the multivariate normal distribution . We interpreted significantly non-zero βx values as support for differences in taxon abundance between social groups , using a false discovery rate threshold of 10% ( Storey and Tibshirani , 2003 ) after checking that an empirically derived null distribution of p-values for this analysis was uniform ( Figure 3—figure supplement 1 ) . To identify socially structured bacterial taxa within baboon social groups , we utilized a test of spatial autocorrelation , Moran's I , as implemented in the function Moran . I in the R package ape ( Paradis et al . , 2004 ) . This analysis tests whether individuals with closer social bonds ( as measured by the pairwise matrix of grooming strengths ) tend to have more similar values for taxon abundance than those with weak or absent social bonds . Here , we again investigated the 327 most prevalent species from the MetaPhlAn 2 . 0 analysis . For this analysis , our power was constrained by the number of individuals in the social group . Thus , while we identified a large number of socially structured species within Viola's group ( n = 51 of 327 species tested , at a false discovery rate of 10% ) , we did not observe strong evidence for socially structured species within Mica's group . Further investigation suggests this result is a consequence of sample size , as subsampling Viola's group ( n = 29 individuals ) to the size of Mica's group ( n = 19 individuals ) also resulted in little power to detect socially structured species . More than half of the time ( 58% of 100 random subsamples ) , fewer than 5 such cases were detected in Viola's group after subsampling , and more than a third of the time ( 35% ) no cases could be detected with the smaller sample size . Hence , we focused on results from Viola's group . We again used a 10% FDR threshold to identify significant taxa in this analysis , after ensuring that the empirical null distribution was uniform ( Figure 3—figure supplement 2 ) . For both between-group and within-group analyses , we investigated enrichment of socially structured species in taxonomic units above the level of species ( i . e . , phylum , class , order , family , and genus ) using hypergeometric tests . We required that taxonomic units include at least four species in our analysis to test for significant enrichment , and again employed an FDR threshold of 10% . Descriptive data on bacteria were retrieved from the Genomes OnLine Database ( GOLD; Pagani et al . , 2012 ) . This information included records for 34 , 533 unique entries and was downloaded from the GOLD website using a custom script on 02 June 2014 ( available on GitHub at https://github . com/jklynch/scrape ) . Each record included fields for oxygen requirements and sporulation , as well as taxonomic classifications from the kingdom to species levels . We retained only completely sequenced genomes , and filtered this set to the entry , for any given species , associated with the most information about bacterial lifestyle and phenotype ( n = 3818 unique species in 1280 unique genera ) . To assign ‘genus-level’ traits , we kept only genera in which all species in our filtered database were associated with the same trait value , if assigned ( e . g . , we assigned an anaerobic lifestyle to a genus only when all members of the genus were consistently anaerobic ) . To investigate properties of significantly socially structured species , we merged the set of 327 prevalent species with the set of species with known lifestyle information . 138 species were represented in both sets; the comparable analysis at the genus level yielded n = 299 genera in both sets . We then applied hypergeometric tests to these data sets to ask whether socially structured species or genera , either between or within groups , were enriched for anaerobic , non-spore forming life-styles . Our results were broadly robust to whether anaerobes are distinguished in contrast to aerobes or in contrast to both aerobes and facultatively oxygen tolerant species ( socially structured species between groups , species level traits: p = 0 . 025; socially structured species within group , species level traits: p = 0 . 100; socially structured species between groups , genus level traits: p = 0 . 056; socially structured species within group , genus level traits: p = 0 . 050 ) . As an alternative to taxonomic profiling using MetaPhlAn 2 . 0 , we also performed de novo contig assembly using the complete set of 1 . 4 billion raw reads . This approach allowed us to evaluate whether our results were robust to our methods for estimating species abundance . Reads were assembled de novo using Ray Meta , a short read de Bruijn assembler specifically devised for metagenome data , following the authors' recommendations ( Boisvert et al . , 2012 ) . Bacterial proportions for each sample were then estimated using Ray Communities , utilizing all bacterial genomes available in GenBank and the Greengenes taxonomy as a reference ( DeSantis et al . , 2006 ) . Summary statistics for alpha diversity and bacterial abundances estimated for each sample from the de novo assemblies can be found in Supplementary files 3 , 4 . Across all 48 samples , we identified 1465 taxa that could be identified to the species level . Similar to our results using MetaPhlAn 2 . 0 , we identified substantial representation of phyla typically found in gut microbiomes , including Bacteroidetes , Firmicutes , Proteobacteria , and Actinobacteria ( Figure 1—figure supplement 1 ) . The de novo assembly , however , identified a very large contribution of the phylum Spirochaetes in Viola's group ( mean = 23 . 7% ) , which was primarily driven by the abundance of reads mapping to the bacteria Treponema succinifaciens . Notably , we also identified T . succinifaciens as significantly more abundant in Viola's group members than in Mica's group members using the MetaPhlAn approach ( p = 2 . 46 × 10−10 ) . Thus , while our two approaches differed in the magnitude of this effect , the overall pattern was highly consistent . The relationship between social group membership and gut microbiome composition using the de novo assembly approach broadly recapitulated the results using MetaPhlAn-based estimates ( Figure 2—figure supplement 1 ) . Specifically , social group membership explained 32 . 8% of global variation in gut microbial taxonomic composition , as summarized by a pairwise Bray–Curtis dissimilarity matrix ( PERMANOVA: p < 1 . 0 × 10−4 ) . Kinship did not explain this relationship ( partial Mantel test relating group co-residency to taxonomic composition , controlling for pedigree-based kinship: r = 0 . 434 , p < 1 . 0 × 10−5 ) . Additionally , the de novo assembly approach again revealed that , within groups , closer grooming partners harbored more similar gut microbes ( Mica's group: Mantel test r = −0 . 197 , p = 0 . 016; Viola's group: r = −0 . 147 , p = 1 . 9 × 10−3 ) . However , while this relationship survives correcting for kinship in Viola's group ( r = −0 . 112 , p = 0 . 017 ) , it is not statistically detectable after controlling for kinship in Mica's group ( r = −0 . 091 , p = 0 . 20 ) . This pattern recapitulates our observations in the MetaPhlAn analysis , in which within group structuring of the microbiome tended to be weaker in Mica's group as well . We next restricted the within-group grooming network analysis to adult females only , in order to test for alternative explanations for the grooming-microbiome composition effect . Grooming interactions remained a significant predictor of microbiome composition after accounting for both within-group patterns of dietary similarity ( partial Mantel controlling for dietary similarity: Mica's group p = 0 . 038; Viola's: p = 0 . 006 ) and spatial proximity in Viola's group ( partial Mantel controlling for proximity: Viola's: p = 0 . 009 ) . In Mica's group , controlling for proximity produced a consistent trend with our main analyses , but eliminated the strong statistical signal of grooming on microbiome composition ( Mica's group p = 0 . 124 ) . We surmise that patterns of proximity , kinship , and grooming may be too closely correlated in Mica's group to disentangle in the de novo assembly-based data set , which may produce noisier estimates of taxonomic abundance . Raw metagenomic sequencing data are deposited in NCBI's Short Read Archive ( BioProject PRJNA271618 ) . The custom script to download data from the Genomes Online Database ( GOLD; Pagani et al . , 2012 ) is available on GitHub at https://github . com/jklynch/scrape . Taxonomic and genic abundance data with sample metadata available on Dryad at doi:10 . 5061/dryad . 8gp03 ( Tung et al . , 2015 ) and are uploaded as supplementary files associated with this manuscript . | The digestive system is home to a complex community of microbes—known as the gut microbiome—that contributes to our health and wellbeing by digesting food , producing essential vitamins , and preventing the growth of harmful bacteria . The recent development of rapid genome sequencing techniques has made it much easier to identify the species of microbes found in the gut microbiome , and how this microbiome's composition varies between individuals . Studies in humans and other primates suggest that direct contact during social interactions may alter the composition of the gut microbiome in an individual . This could explain why there is a strong association between social interactions and health in humans and other social animals . However , similarities in the gut microbiomes of individuals within a social group could also be due to a shared diet or a common environment . The information collected during long-term studies of wild primates offers an opportunity to analyze and assess the influence of diet , environment and social interaction on the gut microbiome . Here , Tung et al . studied the gut microbiomes of 48 wild baboons belonging to two different social groups in Amboseli , Kenya . Using a technique called shotgun metagenomic sequencing , they sequenced DNA extracted from samples of feces collected from individual baboons . The sequence data revealed that an individual's social group and social network can predict the species found in its gut microbiome . This remained the case even when other factors—such as diet , kinship , and shared environments—were taken into account . Tung et al . 's findings suggest that direct physical contact during social interactions may be important in transmitting gut microbiomes between members of the same social group . However , scientists still don't know whether this exchange is good or bad for the health of the baboons . Future work will try to understand whether baboons benefit from acquiring gut microbes from their group members , and if the gut microbes of some social groups are better than others . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"genetics",
"and",
"genomics"
] | 2015 | Social networks predict gut microbiome composition in wild baboons |
Neuroimaging has been used to examine developmental changes of the brain . While PET studies revealed maturation-related changes , maturation of metabolic connectivity of the brain is not yet understood . Here , we show that rat brain metabolism is reconfigured to achieve long-distance connections with higher energy efficiency during maturation . Metabolism increased in anterior cerebrum and decreased in thalamus and cerebellum during maturation . When functional covariance patterns of PET images were examined , metabolic networks including default mode network ( DMN ) were extracted . Connectivity increased between the anterior and posterior parts of DMN and sensory-motor cortices during maturation . Energy efficiency , a ratio of connectivity strength to metabolism of a region , increased in medial prefrontal and retrosplenial cortices . Our data revealed that metabolic networks mature to increase metabolic connections and establish its efficiency between large-scale spatial components from childhood to early adulthood . Neurodevelopmental diseases might be understood by abnormal reconfiguration of metabolic connectivity and efficiency .
Recent neuroimaging studies unraveled the developmental changes of the adolescent brains ( Blakemore , 2012 ) . Among those neuroimaging techniques , positron emission tomography ( PET ) provides quantitative information regarding regional metabolism or synaptic neurochemicals with high sensitivity . As 18F-Fluorodeoxyglucose ( FDG ) is taken up in the brain proportional to cerebral energy consumption and neuronal activity , FDG PET studies have been used to investigate regional energy metabolism during physiologic and pathologic processes . Previous brain developmental studies using FDG PET in humans assumed that the regional increase of metabolism was accompanied by regional increase of neuronal activity during brain maturation ( Chugani et al . , 1987; Phelps and Mazziotta , 1985 ) . Despite these previous data , the developmental changes are not understood in terms of dynamic organization of metabolic networks between brain regions . Since a series of regional brain activity is being coherently organized changing throughout time , sets of specific brain regions are activated or deactivated during resting or various cognitive tasks . These specific regional activity patterns , so-called functional brain networks , are identified by neuroimaging studies ( Deco et al . , 2011 ) . Among several imaging approaches , resting state functional magnetic resonance image ( fMRI ) -based connectivity analyses revealed maturation of the functional networks during childhood and adolescence ( Fair et al . , 2008; Fair et al . , 2007; Supekar et al . , 2009 ) . FDG PET can also reveal functional metabolic network and its maturation , exploiting the coupling between neuronal activity and metabolism . While the fMRI-based functional connectivity measures correlation of fast temporal fluctuations , metabolic connectivity measured by FDG PET reflects a cumulative energy consumption in several minutes and provides relatively stable information regarding steady resting state ( Choi et al . , 2014; Di et al . , 2012; Lee et al . , 2008; Lee et al . , 2012; Toussaint et al . , 2012; Yakushev et al . , 2013 ) . Furthermore , FDG PET measures metabolic activity and connection patterns during awaken states even in animals as in humans , because FDG is taken up in the brain mainly during the period after injection and before the imaging under anesthesia . To organize highly connected functional brain networks , which can be measured by fMRI or FDG PET , the brain needs efficient energy metabolism ( Bullmore and Sporns , 2012 ) . As the high energetic costs are required for brain wiring to construct and maintain hub regions with dense connections , hub regions will show high glucose metabolism ( Lord et al . , 2013 ) . Brain regions were found to show different regional energy efficiency so that , for instance , subcortical regions consumed relatively lower metabolic energy per connection ( Tomasi et al . , 2013 ) . Therefore , we can say that metabolism of brain regions could be determined both by the regional complexity of functional networks and their energy efficiency for wiring . Since the developmental changes in functional networks reconfigure the metabolic demands , energy efficiency of the brain networks are going to be changed during maturation . We primarily aimed to find the maturation-related changes in both regional metabolic activity and metabolic connectivity in adolescent period with a longitudinal study in rats . As the animals allow repeated imaging studies , we could study longitudinal maturation of regional metabolism and connectivity in the same animals . As the small animals also have default mode network ( DMN ) ( Lu et al . , 2012 ) defined as brain regions activating during resting state and deactivating during attention-demanding tasks ( Raichle et al . , 2001 ) , specific network components were extracted in the rat brain from minutes-scale metabolic covariance patterns and independent component analysis ( ICA ) . We hypothesized the metabolic networks are dynamically organized during maturation to achieve connection efficiency . We investigated maturation of large-scale metabolic connectivity and analyzed their energy efficiency for wiring in rat brains .
Voxelwise comparisons were firstly performed among rats aged 5 , 10 , and 15 weeks to disclose temporal changes of regional metabolic activity . The comparison using statistical parametric mapping revealed that the metabolism increased in 10-week-old age in bilateral frontal cortices and anterior aspect of striatum compared with 5-week-old age and decreased in bilateral cerebellar cortices , thalamus , parieto-occipital cortices , and retrosplenial cortices ( Figure 1A ) ( representative FDG PET images for each age are shown in Figure 1—figure supplement 1 ) . Again , metabolism increased in 15-week-old age compared with 10-week-old age , in the clusters of both hippocampi and decreased in the small clusters of right striatum and left frontal cortex ( Figure 1B , Figure1—figure supplement 2 ) . Interregional metabolic correlation on FDG PET was supposed to reflect interregional covariance patterns of neuronal activities . An ICA was performed to yield regional contributing components of metabolic networks . A total of 13 metabolic independent components ( ICs ) were identified in rats similar to components found in previous reports in humans on FDG PET ( Di et al . , 2012; Toussaint et al . , 2012; Yakushev et al . , 2013 ) . All ICs were displayed in Figure 2—figure supplement 1 , and a threshold z > 1 . 5 was applied for the voxels for display purposes . Four particular components , IC1 , IC5 , IC8 , and IC9 , were selected ( Figure 2 ) , which anatomically corresponded to previously alleged limbic/anterior DMN ( IC1 ) , posterior DMN ( IC5 ) , motor ( IC8 ) , and somatosensory network ( IC9 ) , respectively ( Lu et al . , 2012 ) . IC1 included the clusters of dorsal hippocampi and medial prefrontal cortex . IC5 included the retrosplenial cortex , known as a hub of posterior DMN in rats ( Lu et al . , 2012 ) . It is anatomically adjacent to the posterior cingulate and precuneus in human as cores of DMN ( Raichle et al . , 2001 ) . These were used further to identify maturing patterns of metabolic networks consisting of IC-derived volume-of-interests ( VOIs ) . Eight VOIs ( 3 for IC1 , 1 for IC5 , 2 for IC8 and IC9 , respectively ) were selected taking anatomical structures into consideration ( Figure 2—figure supplement 2 ) ( stereotaxic coordinates and abbreviations of VOIs are summarized in Table 1 ) . Interregional correlations were calculated between paired VOIs to yield 28 pairs ( 28 edges upon 8 nodes ) . This VOI-based metabolic interregional correlation was assumed to yield functionally coherent covariance patterns among spatial components to represent metabolic connectivity . Metabolic connectivity obtained this way was used to investigate the changes of metabolic networks during maturation . Figure 3A shows interregional correlation representing connection strength of pairs of VOIs in rats aged 5 , 10 , and 15 weeks . Significantly different connections in interregional correlation were examined using nonparametric permutation tests ( Figure 3—figure supplement 1 ) using pseudorandom relabeling 5-week/10-week-old rat images or 10-week/15-week-old ones , followed by FDR correction for multiple comparisons ( Kim et al . , 2015 ) . Connectivity matrices ( Figure 3A ) and p-values for the difference in connection strength ( Figure 3—figure supplement 2 ) between paired groups suggested that the connection between pairs of retrosplenial , medial prefrontal , and sensorimotor cortices was strengthened from 5 to 10 weeks , while connectivity involving limbic regions did not change . According to aging , pairs of anterior-posterior connections were significantly strengthened when comparing 10-week-old rats with 5-week-old rats ( Figure 3B ) . The increase became prominent when we compared 15-week-old rats from 5-week-old rats . Connectivity was not significantly weakened between areas during maturation ( Figure 3—figure supplement 2 ) . We further investigated energy efficiency to configure the metabolic connectivity between spatial components at each period of maturation . The energy efficiency was defined as the ratio of metabolic connection strength , a sum of weights of links connected to each VOI , to normalized FDG uptake of each VOI . Significantly different metabolic energy efficiency was also examined using the permutation test ( Figure 3—figure supplement 1 ) . In 5-week-old rats , energy efficiency was relatively lower in midline structures , that is , medial prefrontal and retrosplenial cortices than sensorimotor cortices . In 10- and 15-week-old age compared with 5-week-old age , energy efficiency of the midline structures significantly increased almost reaching those of the sensorimotor cortices ( Figure 4 ) .
We showed that metabolic activity of rat brain matured to build connectivity between network components accompanied by enhanced energy efficiency from childhood to early adulthood . Metabolism increased during adolescence in frontal cortices compared to childhood in rat brain similarly to that in human brain . Components of metabolic networks were identified , and connectivity analysis showed efficient connection developed between the components particularly pair of VOIs including DMN during adolescent period . Furthermore , the midline structures showed increase in energy efficiency of metabolic connections . This study about metabolic network maturation gave insights into normal brain developments and might elucidate the plausible pathophysiology of neuropsychiatric developmental diseases .
Thirty adult male Sprague-Dawley ( SD ) rats ( Koatech , Seoul , Korea ) were used for FDG PET scans . They were kept at standard laboratory condition ( 22–24°C , 12 hr light and dark cycle ) with free access to water and standard feed . All the experimental procedures were approved by Institutional Animal Care and Use Committee at Seoul National University Hospital ( IACUC Number 13–0224 ) . FDG PET images were acquired at the age of 5 , 10 , and 15 weeks . PET scans were performed on a dedicated small animal PET/CT scanner ( eXplore VISTA , GE Healthcare , WI ) . Rats were fasted for at least 8 hr before the start of the study . Rats were anesthetized 2% isoflurane at 1–1 . 5 L/min oxygen flow for 5–10 min . After an intravenous bolus injection ( 0 . 3–0 . 5 mL/rat ) of FDG ( 100–150 MBq/kg ) , each rat has 45 min period of FDG uptake ( Schiffer et al . , 2007 ) . Static scans at 45 min after the injection best reflects absolute rates of glucose metabolism in rodents ( Schiffer et al . , 2007 ) . During FDG uptake period , rats were awake for 35 min and anesthetized for preparation , 10 min before PET/CT scans . Emission scan data were acquired for 20 min with the energy window 400–700 keV and reconstructed by a three-dimensional ordered-subsets expectation maximum ( OSEM ) algorithm with attenuation , random and scatter correction . The voxel size was 0 . 3875 × 0 . 3875 × 0 . 775 mm . We acquired 88 PET images considering voxelwise statistical difference after multiple comparison correction: 30 scans at 5 weeks , 30 scans at 10 weeks , and 28 scans at 15 weeks . For preprocessing , all voxels were scaled by a factor of 10 in each dimension . All brain PET images were spatially normalized to the FDG rat brain template ( Schiffer et al . , 2006 ) ( PMOD 3 . 4 , PMOD group , Zurich , Switzerland ) using nonlinear registration on Statistical Parametric Mapping ( SPM8 , University College of London , London , UK ) . After spatial normalization , a binary mask for brain was applied . The images were smoothened by Gaussian filter of 12 mm full width at half maximum . For scaling voxel intensities , the voxel counts were normalized to the global brain uptake in each PET image . Testing for age effects on the brain metabolism was performed with SPM . Paired t-test was performed between two groups of PET images acquired at 5 and 10 weeks . Additional paired t-tests were performed between PET images at 10 and 15 weeks and PET images at 5 and 15 weeks . False discovery rate ( FDR ) corrected p < 0 . 05 was set as the significance threshold and an extent threshold of 100 contiguous voxels was applied . We applied a group ICA algorithm to define coherent network components ( GIFT , http://mialab . mrn . org/ , GIFT ver 2 . 0a ) . All preprocessed PET images ( n = 88 ) were included in this spatial ICA to find spatially independent components and coherently activated regions among subjects . The group ICA approaches have been used to obtain multivariate patterns for metabolic networks ( Toussaint et al . , 2012; Yakushev et al . , 2013 ) . Prior to perform ICA , the optimal number of components extracted from PET images was determined . We used the dimensional estimation algorithm implemented in GIFT software based on the assessment of entropy rate of independent and identically distributed ( i . i . d . ) Gaussian random process ( Li et al . , 2007; Xu et al . , 2009 ) . The estimated optimal number was thirteen components . ICA was performed using an infomax neural network algorithm that minimized the mutual information of the outputs . The resulting independent components were z-transformed and visualized using the threshold z > 1 . 5 . To compare the changes of brain metabolic connectivity according to age , we engaged VOI-based correlation analyses . VOIs were placed on four ICs ( IC1 , IC5 , IC8 , and IC9 ) referring to rat brain atlas , which was constructed on 3D digital map based on Paxinos and Watson atlas ( Schiffer et al . , 2006; Toga et al . , 1995 ) . We selected these ICs to concentrate on the relationship between functional components including DMN as intrinsically coherent regions at resting state and sensory-motor networks . The total eight VOIs ( 3 on IC1 , 1 on IC5 , 2 on IC8 and IC9 ) were defined as spheres with a radius of 8 mm , centered at each IC ( the actual size of a radius was 0 . 8 mm as voxels were scaled by a factor of 10 ) . The size of VOI was determined considering initial voxel size ( 0 . 775 mm ) and to a representative small cluster of a network not to overlap other networks . The VOIs were displayed in Figure 2—figure supplement 2 . We compared the connectivity between rats aged 5 , 10 , and 15 weeks . We obtained normalized FDG uptake in the VOIs of each rat and Pearson’s correlation coefficients were calculated between the pairwise VOIs ( 8x7/2= 28 paired VOIs ) using subjects variation . These correlation coefficient matrices were constructed for 5 week- , 10 week- , and 15 week-old rats . For statistical comparison of correlation matrices between these groups of different ages , we performed permutation tests . We tested whether there was significantly different connectivity , represented by Fisher-transformed Z values , between 5 vs 10 weeks and 10 vs 15 weeks of age . At first , PET images of each age group were randomly permuted to make pseudo-random groups reassigned 10 , 000 times and from each paired group of rats , interregional correlation matrices were calculated . Type I errors were determined by the comparison between the observed Z score for each connection of VOI pairs and Z score distribution of VOI pairs from the permuted data ( Figure 3—figure supplement 1 ) . For multiple comparison correction , we applied false-discovery rate ( FDR ) at a threshold of FDR < 0 . 05 . To analyze energy efficiency for metabolic connectivity , undirected networks with the eight nodes were constructed where strength of each connection was simply defined as correlation coefficients . Strength of metabolic connectivity was calculated for each VOI as a sum of weights of positive links ( correlation coefficients ) per VOI ( Kaiser , 2011 ) . The ratio of metabolic connectivity strength to normalized metabolic activity was defined as energy efficiency for each VOI ( Tomasi et al . , 2013 ) . For statistical comparison of energy efficiency of VOIs between 5 , 10 , and 15 week-old rats , we performed permutation tests . Again PET images of each paired group of rats were permuted to make pseudo-random groups reassigned 10 , 000 times and distribution of metabolic energy efficiency of each VOI was drawn . The observed energy efficiency was compared with distribution of energy efficiency of each of VOIs from the permuted data ( Figure 3—figure supplement 1 ) . | The brain consumes a great deal of a sugar called glucose , which is delivered to the brain through blood vessels . Active regions of the brain need more glucose , and so the brain has a metabolic network that controls when and where glucose is metabolized . Yet precisely how this metabolic network changes during brain development is not yet understood . Choi et al . have now monitored the patterns of glucose metabolism in the brains of awake rats as they matured from 'childhood' to early adulthood . The experiments involved injecting the rats with radioactive glucose , and then using a technique called positron emission tomography ( commonly known as 'PET scan' ) to monitor the metabolism of these radioactive sugar molecules in the animals’ brains . Choi et al . showed that the patterns of glucose consumption in the brain shift drastically as the rats mature . Importantly , the findings showed that these shifts in glucose metabolism seem to support the activity of long distance connections that develop as the brain matures . The findings also showed that the increased long distance connections were energy efficient . The results suggest that these metabolic changes are likely a way of maintaining high-energy efficiency that is crucial for the brain to perform normally . Finally , in addition to revealing the changes involved in normal brain development , these findings may have implications in neurological and psychiatric disorders in which the brain fails to achieve efficient metabolic networks as it matures . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Maturation of metabolic connectivity of the adolescent rat brain |
Much of the vertebrate skeleton develops from cartilage templates that are progressively remodeled into bone . Lineage tracing studies in mouse suggest that chondrocytes within these templates persist and become osteoblasts , yet the underlying mechanisms of this process and whether chondrocytes can generate other derivatives remain unclear . We find that zebrafish cartilages undergo extensive remodeling and vascularization during juvenile stages to generate fat-filled bones . Growth plate chondrocytes marked by sox10 and col2a1a contribute to osteoblasts , marrow adipocytes , and mesenchymal cells within adult bones . At the edge of the hypertrophic zone , chondrocytes re-enter the cell cycle and express leptin receptor ( lepr ) , suggesting conversion into progenitors . Further , mutation of matrix metalloproteinase 9 ( mmp9 ) results in delayed growth plate remodeling and fewer marrow adipocytes . Our data support Mmp9-dependent growth plate remodeling and conversion of chondrocytes into osteoblasts and marrow adipocytes as conserved features of bony vertebrates .
Vertebrate bones develop via two largely distinct processes . Intramembranous ( i . e . dermal ) bone , which makes up a large portion of the skull , arises through the direct differentiation of mesenchymal precursors into osteoblasts and then osteocytes . In contrast , endochondral bone , which comprises the majority of the axial and limb skeletons , arises through the progressive remodeling of an embryonic cartilage template . On the outside of developing endochondral bone , perichondral cells mature into periosteal progenitors that contribute to the bone collar . The cartilage templates of endochondral bone are organized into distinct zones of chondrocytes: resting , proliferative , pre-hypertrophic , and hypertrophic . In mammals , chondrocytes at the edge of the developing hypertrophic zone largely disappear as the cartilage matrix is degraded , a process concurrent with the invasion of blood vessels , hematopoietic cells , and progenitors for osteoblasts and marrow adipocytes ( Maes et al . , 2010 ) . This growth plate remodeling contributes to the establishment of trabecular bone , complementing the cortical bone largely derived from the periosteum , and the marrow cavity supports continued hematopoiesis . As with mammals , zebrafish also have intramembranous and endochondral bones . Their endochondral bones are hollow and filled predominantly with fat yet do not support hematopoiesis as in mammals ( Witten and Huysseune , 2009; Weigele and Franz-Odendaal , 2016 ) . It has remained unclear , however , whether zebrafish bone arises solely through osteoblast differentiation in the periosteum , or also through invasion of the vasculature and conversion of growth plate cartilage to bone as in mammals . The source of marrow adipocytes also remains unclear in either fish or mammals . It has long been appreciated that many hypertrophic chondrocytes undergo cell death during endochondral ossification , with osteoblasts forming from periosteal cells brought into the bone along with the vasculature ( Maes et al . , 2010 ) . At the same time , there are numerous studies showing that cultured chondrocytes can dedifferentiate into mesenchymal progenitors and/or transdifferentiate into osteoblasts ( Shimomura et al . , 1975; Mayne et al . , 1976; von der Mark and von der Mark , 1977 ) . Recent lineage tracing studies of hypertrophic chondrocytes , using constitutive and inducible Col10a1-Cre- and Aggrecan-Cre-based transgenes in mice , has revealed that such transdifferentiation may also occur in vivo , with chondrocytes making a major contribution to osteoblasts within trabecular bone and potentially also the bone collar ( Yang et al . , 2014; Kobayashi et al . , 2014; Jing et al . , 2015; Park et al . , 2015 ) . A limitation of these studies is the use of population-based labeling by Cre recombination , which cannot exclude low-level and/or leaky labeling of other cell types . It is also unclear whether hypertrophic chondrocytes can give rise to other cell types , such as marrow adipocytes , and whether hypertrophic chondrocytes directly transform into osteoblasts or do so through a stem cell intermediate . Finally , it is unknown whether the ability of chondrocytes to generate osteoblasts and other cell types is specific to mammals or a more broadly shared feature of vertebrates . In this study , we address the long-term fate of growth plate chondrocytes in zebrafish , as well as potential mechanisms of their fate plasticity . We use the ceratohyal ( Ch ) bone of the lower face as a model . This long bone , which is derived from cranial neural crest cells , exhibits properties in common with the long bones of mammalian limbs , including two prominent growth plates at either end and a marrow cavity ( Paul et al . , 2016 ) . Here , we describe remodeling of the Ch from a cartilage template to a fat-filled bone in juvenile stages , which coincides with extensive vascularization . Using inducible Cre and long-lived histone-mCherry fusion proteins , driven by regulatory regions of the chondrocyte genes sox10 and col2a1a , we reveal contribution of chondrocytes to osteoblasts , adipocytes , and mesenchymal cells within the adult Ch . In mouse , LepR expression marks bone marrow cells that contribute to osteoblasts and adipocytes primarily after birth ( Zhou et al . , 2014b ) . In zebrafish , we find that growth plate chondrocytes express lepr and re-enter the cell cycle during the late hypertrophic phase , raising the possibility that Lepr +skeletal stem cells may derive from growth plate chondrocytes . Further , we find that delayed remodeling of the hypertrophic cartilage zone in zebrafish mmp9 mutants correlates with a paucity of marrow adipocytes . Unlike in mouse where Mmp9 functions in hematopoietic cells for timely growth plate remodeling ( Vu et al . , 1998 ) , we find that Mmp9 is sufficient in neural crest-derived chondrocytes of zebrafish for growth plate remodeling . Our studies reveal that growth plate chondrocytes generate osteocytes and adipocytes in zebrafish bones , potentially by transitioning through a proliferative intermediate .
In order to characterize the progressive remodeling of an endochondral bone in zebrafish , we performed pentachrome staining on sections of the Ch bone from juvenile through adult stages ( Figure 1 ) . The Ch bone is shaped like a flattened barbell , and here we sectioned it to reveal the thin plane of the bone ( see Figure 1—figure supplement 1A ) for a view along the thicker perpendicular plane ) . Unlike the unidirectional growth plates in the mouse limb , the two growth plates of Ch are bidirectional with a central zone of compact , proliferative chondrocytes flanked by hypertrophic chondrocytes on either side ( Paul et al . , 2016 ) . Unlike in many other fish species , the Ch bone , as with other bones in zebrafish , also contains embedded osteocytes ( Witten and Huysseune , 2009 ) . At 11 mm standard length ( SL ) ( approx . 4 . 5 weeks post-fertilization ( wpf ) ) , the Ch contains chondrocytes throughout its length with the exception of a small marrow space at the anterior tip . The Ch is surrounded by a thin layer of cortical bone that has been shown to derive from osteoblasts located on the outside of the cartilage template ( i . e . periosteum ) ( Paul et al . , 2016 ) . By 12 mm SL ( approx . five wpf ) , both tips of the Ch contain marrow spaces , and on the central sides of the growth plates we begin to observe small fissures in the cortical bone and disruption of the hypertrophic zone . By 13 mm SL ( approx . 5 . 5 wpf ) , breaks in the cortical bone become more prominent and are accompanied by further degradation of the cartilage matrix . At later stages ( 16 and 19 mm SL ) ( approx . 7 and 9 wpf ) , cortical bone regains integrity and increases in thickness , and marrow adipocytes containing LipidTOX +lipid vesicles are seen throughout Ch ( Figure 1—figure supplement 1B ) . By adulthood ( one year of age ) , the marrow cavity is filled with large fat cells and the growth plates appear largely mineralized . While we focus on the Ch for this study , a number of other cartilage-derived bones in the face and fins have been reported to have a similar structure in zebrafish , including growth plates and prominent marrow fat ( Weigele and Franz-Odendaal , 2016 ) . Given the transient breakdown of cortical bone , we examined whether this coincides with vascularization of Ch . To do so , we performed confocal imaging of the dissected Ch from fish carrying both a chondrocyte-specific col2a1a:mCherry-NTR transgene and fli1a:GFP ( Figure 2A ) . We used fli1a:GFP to label endothelial cells of the vasculature , but we also noticed that presumptive resting chondrocytes in the middle of the growth plate express this transgene . At 10 and 11 mm SL , vessels expressing fli1a:GFP are found largely on the outside of the Ch bone . By 13 , 16 , and 20 mm SL , we observe increasing numbers of capillaries within the Ch , coinciding with the replacement of col2a1a:mCherry-NTR +chondrocytes with adipocytes . High-magnification confocal sections at 18 mm SL clearly show fli1a:GFP+ vessels in intimate association with adipocytes and within the Calcein Blue+ bone collar ( Figure 1—figure supplement 1C ) . Given the more complicated expression pattern of fli1a:GFP , we also independently confirmed blood vessel identity with kdrl:GFP . At 28 mm SL ( approx . 26 wpf ) , the Ch is heavily supplied with both kdrl:GFP+blood vessels and lyve1:DsRed+lymphatic vessels , which abut each side of the growth plate ( Figure 2B ) . Hence , as in mammalian bones , remodeling of the Ch bone is accompanied by extensive vascular invasion of the cartilage template . Given the extensive remodeling and vascularization of Ch , we next investigated the long-term fate of growth plate chondrocytes by multiple , independent methods . First , we constructed an inducible sox10:CreERT2 line and crossed it to a ubiquitous bactin2:loxP-tagBFP-stop-loxP-DsRed reporter ( Blue to Red conversion: B > R ) . The zebrafish sox10 promoter drives expression in early cranial neural crest cells from 10 to 16 hpf , followed by a second wave of expression in all chondrocytes from two dpf onwards ( Dutton et al . , 2008 ) . Here , we took advantage of this second wave of expression to label developmental chondrocytes . Upon addition of 4-hydroxytamoxifen ( 4-OHT ) at 15 dpf , we observed extensive labeling of chondrocytes within 5 days , as well as some cells in the perichondrium surrounding Ch and other cartilages ( Figure 3A ) . We did not observe leaky conversion in the absence of 4-OHT at either embryonic or adult stages ( Figure 3—figure supplement 1A ) . We then converted sox10/B > R fish by 4-OHT treatment at 14 dpf and raised these to adulthood ( 27 mm SL ) for analysis , with inclusion of an ocn:GFP transgene allowing us to label osteoblasts ( Figure 3B ) . Confocal maximum intensity projections through Ch revealed extensive labeling of growth plate chondrocytes . We also observed numerous cells throughout the Ch , including in and around the marrow cavity . In sections through the middle of Ch , we observed labeling of a number of large diameter cells of adipocyte morphology ( Figure 3C ) . In superficial sections through the cortical bone , we also observed labeled cells that were positive for the osteoblast marker ocn:GFP , as well as some labeled cells negative for ocn:GFP that may represent bone progenitors or other cell types ( Figure 3D ) . As a comparison , we used a constitutive Cre driven by a human SOX10 enhancer that drives expression throughout the neural crest lineage ( note that this human enhancer lacks the second wave of chondrocyte expression seen with the zebrafish regulatory region used for the sox10:CreERT2 line ) ( Kague et al . , 2012 ) . When crossed to the B > R line , this neural crest-specific SOX10:Cre line drives broader conversion in the five dpf head than the later conversion of sox10:CreERT2 ( Figure 3—figure supplement 2A ) . Analysis of the adult Ch in SOX10:Cre fish shows labeling of all growth plate cartilage , as well as most if not all adipocytes and numerous smaller cells throughout the marrow and cortical bone surface ( Figure 3—figure supplement 2B , C ) . This is consistent with previous studies showing that the Ch bone is neural crest-derived ( Schilling and Kimmel , 1994 ) . However , we also detect unconverted cells in the marrow , consistent with contribution of non-neural crest-derived cells such as the mesoderm-derived vasculature ( Figure 2 ) . As sox10:CreERT2-mediated conversion at 14 dpf also labels cells outside the cartilage , which could also contribute to osteoblasts and adipocytes , we next examined animals with lower conversion efficiency to follow discrete growth plate clones . When analyzed at 30 mm SL , growth plate clones could be quite large , consistent with clonal selection as described in the zebrafish heart and skeletal muscle ( Gupta and Poss , 2012; Nguyen et al . , 2017 ) . In one example with three discrete clones , we observed a clone that contributed to a narrow column of growth plate chondrocytes in the middle of Ch that was contiguous with mesenchymal cells and then adipocytes toward the central marrow cavity ( Figure 3E ) . To more definitively follow growth plate clones , we also examined sox10:CreERT2; ubb:Zebrabow fish in which 4OH-T treatment at 14 dpf resulted in conversion of RFP ( red ) to various color combinations of CFP , YFP , and RFP at 23 mm SL . Analysis of uniquely colored clones in the growth plate revealed those that contained both chondrocytes and adjacent adipocytes in the marrow ( Figure 3F ) . We also observed clones that appeared to contain chondrocytes and weakly labeled cells embedded in cortical bone , though the clonal contribution to these and other mesenchymal lineages will require further analysis . Together , our data are consistent with chondrocytes giving rise to adipocytes and mesenchymal cells , and potentially osteoblasts , after growth plate remodeling in zebrafish . As sox10-CreERT2-mediated conversion was broader than just the cartilage , we also generated an inducible col2a1a-CreERT2 line to more precisely trace chondrocytes and their derivatives . In mice , Col2a1:CreERT2-mediated conversion at embryonic and early postnatal stages broadly labels not only chondrocytes but also osteochondroprogenitors , such as those in the perichondrium and periosteum ( Ono et al . , 2014 ) . In zebrafish , col2a1a is similarly expressed at high levels in chondrocytes and in weaker levels in osteoblasts and perichondrium , although direct evidence for col2a1a marking osteochondroprogenitors in zebrafish is lacking ( Eames et al . , 2012 ) . In order to restrict expression to chondrocytes , thus avoiding potential complications of labeling osteoblasts and putative perichondral progenitors , we utilized a chondrocyte-specific ‘R2’ enhancer of the zebrafish col2a1a gene that we had previously characterized ( Dale and Topczewski , 2011; Askary et al . , 2015 ) . Treatment of col2a1a/B > R fish with a single dose of 4-OHT at five dpf resulted in extensive labeling of col2a1a-BAC:GFP+ chondrocytes at 12 dpf , but no labeling of the perichondrium , periosteum , and osteoblasts as marked by the sp7:GFP transgene ( Figure 4A , B; Figure 3—figure supplement 1B ) . We also observed labeling of the notochord in larval fish , but no labeling of the vasculature , blood , or other tissues examined . After conversion of col2a1a/B > R chondrocytes at five dpf and examination at adulthood ( 30 mm SL ) , maximal intensity projections through Ch revealed extensive labeling of growth plate chondrocytes , as well as cells throughout the bone and marrow cavity ( Figure 4C ) . Thus , the majority of adult growth plate chondrocytes in Ch appear to derive from embryonic chondrocytes . Moreover , optical sections revealed cytoplasmic DsRed staining in lipid-filled adipocytes , presumptive osteoblasts lining the inner surface of bone , and mesenchymal cells within the marrow cavity . In some animals displaying lower conversion efficiency , we observed apparent clones of cells containing growth plate chondrocytes , large adipocytes , and osteoblasts embedded in Calcein +mineralized matrix ( Figure 4D ) . Analysis of individual sections at higher magnification revealed contribution of col2a1a-lineage cells to a subset of osteoblasts within both the endosteal and periosteal surfaces of bone , as well as embedded osteocytes with characteristic cellular processes ( Figure 4E–G ) . A limitation of CreER-mediated lineage tracing is that it can be difficult to rule out contributions from rare converted cells outside the population of interest , for example in the perichondrium and periosteum . We therefore employed a Cre-independent approach to independently assess the fate of growth plate chondrocytes . To do so , we expressed a Histone2A-mCherry fusion protein in col2a1a+ cells . An advantage of this type of lineage approach is that the levels of Histone2A-mCherry protein , which is stably incorporated into chromatin , reflect those of endogenous col2a1a in chondrocytes provided continued cell division does not dilute out the fusion protein . This is in contrast to Cre-mediated approaches in which a strong ubiquitous promoter determines the level of a reporter protein; hence , even low levels of Cre recombinase activity outside of chondrocytes ( e . g . in osteochondroprogenitors ) can result in strong reporter expression . In zebrafish , col2a1a is expressed at high levels in the proliferative zone of the Ch growth plate and then downregulated in the hypertrophic zone ( Paul et al . , 2016 ) . In a newly generated col2a1a:Histone2A-mCherry-T2A-GFP-CAAX transgenic line , membrane-localized GFP-CAAX , which is rapidly turned over , is seen primarily in the proliferative zone , whereas Histone2A-mCherry is seen uniformly throughout the Ch cartilage , confirming the long-lived nature of this fusion protein ( Figure 5A and Figure 5—figure supplement 1A , B ) . At 7–8 mm SL ( approx . three wpf ) , which is well before the start of growth plate remodeling at 11–12 mm SL , all Ch chondrocytes are Histone2A-mCherry+ and we do not detect Histone2A-mCherry+ cells associated with Calcein Blue+ bone or co-expressing the osteoblast transgene ocn:GFP ( Figure 5A and Figure 5—figure supplement 1C , D ) . In contrast , at post-remodeling stages ( 12 and 18 mm SL ) , we observe extensive overlap of Histone2A-mCherry with ocn:GFP+ osteoblasts associated with cortical bone ( Figure 5B and Figure 5—figure supplement 1E ) . We also observe numerous Histone2A-mCherry+ cells embedded in the endosteal and periosteal surfaces of the Ch bone ( labeled by Calcein Blue ) , with several of these cells co-expressing the pre-osteoblast transgene RUNX2:GFP or the early osteoblast transgene sp7:GFP ( Figure 5C–E ) . Note that ocn:GFP , RUNX2:GFP , and sp7:GFP can all be readily distinguished from membrane GFP-CAAX by their much stronger and cytoplasmic expression . In addition , we observed that sp7:GFP+ osteoblasts further from the growth plate tended to have weaker Histone2A-mCherry signal than those more closely associated with the edge of the hypertrophic zone , suggesting that hypertrophic chondrocytes and/or their osteoblast derivatives undergo cell division to dilute out the Histone2A-mCherry signal . These findings independently confirm the conclusions of our CreER lineage tracing studies that col2a1a+ chondrocytes generate osteoblasts in zebrafish . The conversion of hypertrophic chondrocytes to osteoblasts and adipocytes could occur in the absence of cell division ( i . e . ‘transdifferentiation’ ) and/or through partial dedifferentiation into a proliferative progenitor . To test these possibilities , we first used incorporation of bromodeoxyuridine ( BrdU ) to detect proliferative cells in the Ch during remodeling stages ( Figure 6A , B ) . At the beginning of remodeling ( 11 mm SL ) , we detected BrdU+ cells in the central zone of chondroblasts in the growth plate , as well as in the perichondrium and periosteum . In addition , we observed BrdU+ cells at the edge of the hypertrophic zone where the cartilage matrix is being actively degraded , similar to what has been reported in mouse ( Park et al . , 2015 ) . We observed BrdU incorporation in similarly positioned hypertrophic chondrocytes at 15 and 19 mm SL , with BrdU+ cells becoming fewer in the perichondrium and periosteum by 19 mm . We next tested whether hypertrophic chondrocytes that re-enter the cell cycle also express known skeletal stem cell markers . In mice , LepR expression marks a heterogeneous population of cells in endochondral bone , including a putative postnatal skeletal stem cell population ( Zhou et al . , 2014b ) . First , we examined expression of lepr and found dynamic expression in the zebrafish Ch endochondral bone from juvenile through adult stages . A comparison with the hypertrophic chondrocyte and early osteoblast marker runx2b shows higher lepr expression in proliferative versus hypertrophic chondrocytes at 8 , 12 , and 20 mm SL stages , with chondrocyte expression decreasing by 27 mm SL ( Figure 6—figure supplement 1A ) . During remodeling of zebrafish Ch ( 15 mm SL ) , we also observe lepr+ cells in the marrow cavity that are derived from chondrocytes , based on labeling by col2a1a/B > R ( Figure 6C ) , and we continue to observe lepr+ cells in the Ch marrow at later stages ( 20 and 27 mm SL ) ( Figure 6—figure supplement 1A ) . These results are consistent with lepr+ cells in the bone marrow deriving from growth plate chondrocytes in zebrafish , although direct evidence will be needed to determine if any of these chondrocyte-derived lepr+ marrow cells behave as skeletal stem cells in zebrafish . In mice , Mmp9 has been reported to function in hematopoietic lineage cells for growth plate remodeling , as bone marrow transplants can rescue the delay in growth plate remodeling seen in Mmp9 mutants ( Vu et al . , 1998 ) . Here , we tested whether mmp9 might have a conserved requirement for growth plate remodeling in zebrafish , including the generation of marrow adipocytes . At the beginning of remodeling ( 11 mm SL ) , we observe expression of mmp9 at the edge of the hypertrophic zone , with this restricted expression in late-stage hypertrophic chondrocytes continuing through 17 mm SL stages ( Figure 7A ) . Higher magnification views show that mmp9 expression is prominent in hypertrophic chondrocytes that appear to be exiting their lacunae . Next , we used CRISPR/Cas9 mutagenesis to create an early frame-shift mutation in the mmp9 gene that is predicted to abolish most if not all protein function ( Figure 7B ) . mmp9 homozygous mutants are adult viable and do not display obvious larval craniofacial defects . Whereas trichrome staining revealed no significant differences in the mutant Ch growth plates at 17 mm SL , by 21 mm we observed that the hypertrophic zone was significantly larger , compared to the proliferative zone , in mmp9 mutants versus controls , indicating a delay in growth plate remodeling ( Figure 7C , D ) . The defect in growth plate remodeling was still evident at 27 mm SL , with mutants displaying a wider Ch growth plate ( Figure 7E , G ) . Strikingly , mmp9 mutants also had fewer adipocytes in the central marrow cavity compared to stage-matched controls ( Figure 7E , G ) , consistent with adipocytes deriving from hypertrophic chondrocytes that are released from the cartilage matrix by Mmp9 activity . Given mmp9 expression in late-stage hypertrophic chondrocytes , we next tested whether Mmp9 might function in chondrocytes as opposed to hematopoietic lineage cells for growth plate remodeling and marrow adipocyte generation in zebrafish . As the Ch bone is generated from neural crest-derived cells , we used transplantation of ubiquitously labeled wild-type ectodermal cells into the neural crest precursor domain of unlabeled mmp9-/- shield-stage hosts to generate wild-type Ch bones in otherwise mutant hosts . At adult stages , we were able to recover eight mutant recipients with contribution of wild-type cells to chondrocytes of the Ch growth plate . In these animals , we observed a rescue of growth plate width , with a trend toward better rescue in wild-type versus mutant regions of the growth plate ( p = 0 . 06 ) , as well as a trend toward rescue of adipocyte number ( p = 0 . 06 ) ( Figure 7E–G ) . As a control , transplantation of wild-type neural crest cells into wild-type animals had no effect on Ch growth plate width and adipocyte number . These results indicate that mmp9 is required in the neural crest lineage , and potentially chondrocytes themselves , for efficient remodeling of the growth plate and the generation of marrow adipocytes from chondrocytes .
Despite anatomical differences between zebrafish and mammalian bones , we find that growth plate remodeling is remarkably well conserved and thus likely ancestral to bony vertebrates . The zebrafish Ch undergoes a transient breakdown of cortical bone near the growth plates , which coincides with extensive vascularization and an Mmp9-dependent replacement of hypertrophic chondrocytes with fat and bone . Using multiple methods of lineage tracing , including a Cre-independent technique , we show that late-stage hypertrophic chondrocytes generate not only osteoblasts but also marrow adipocytes in zebrafish . Further support for the ability of chondrocytes to generate adipocytes is that delayed growth plate remodeling in mmp9 mutants results in a paucity of marrow adipocytes in adults . Lastly , we show that hypertrophic chondrocytes re-enter the cell cycle and contribute to lepr+ mesenchymal cells , raising the possibility that partial dedifferentiation into proliferative progenitors underlies chondrocyte fate transitions inside endochondral bones . As zebrafish have hollow bones that lack hematopoiesis , one possibility was that most if not all of the cortical bone of adult zebrafish would be simply derived from the periosteum . However , we find significant contribution of chondrocyte-derived cells to both the endosteal and periosteal surfaces of the Ch bone , similar to what has been described in the mouse ( Yang et al . , 2014; Zhou et al . , 2014a; Jing et al . , 2015 ) . We also reveal chondrocytes to be a significant source of marrow adipocytes in zebrafish . Although the incomplete conversion efficiency of the CreER lines , as well as the expression of sox10:CreER outside of chondrocytes , made it difficult to precisely quantify what proportion of osteoblasts and marrow adipocytes derive from chondrocytes , there are likely other sources for these cells , in particular the periosteum which houses several types of skeletal stem cells ( Debnath et al . , 2018 ) . Our findings raise the question of whether marrow adipocytes also derive in part from chondrocytes in mammals . Indeed , older studies have demonstrated the ability of murine chondrocytes to differentiate into adipocytes in vitro ( Heermeier et al . , 1994; Hegert et al . , 2002 ) . Further , Col2a1:CreER cells converted at postnatal day three in mouse were found to give rise to adipocytes in the metaphyseal bone marrow , although a caveat is that Col2a1:CreER marks both chondrocytes and progenitors at early postnatal stages ( Ono et al . , 2014 ) . The finding that late-stage hypertrophic chondrocytes can re-enter the cell cycle and express lepr suggests that at least some of these cells may dedifferentiate into stem-like cells , which subsequently expand in number and differentiate into osteoblasts and adipocytes . Future molecular profiling of these chondrocyte-derived marrow mesenchymal cells will be needed to better characterize their relationship to previously identified skeletal stem cells . We also cannot rule out that some hypertrophic chondrocytes directly change into adipocytes and/or osteoblasts in the absence of cell division ( i . e . ‘transdifferentiation’ ) . Indeed , the detection of osteoblasts with strong col2a1a:Histone2A-mCherry signal suggests that in some cases chondrocytes can form osteoblasts with little to no cell division , as otherwise the Histone2A-mCherry signal would have been diluted out with successive cell divisions . On the other hand , osteoblasts farther from the growth plate tended to have weaker Histone2A-mCherry signal , suggesting proliferation of a progenitor intermediate , and we directly observed late-stage hypertrophic chondrocytes undergoing cell division . Whereas it has been suggested in mouse that only chondrocytes near the periosteal surface , that is ‘borderline chondrocytes’ , may be capable of lineage plasticity ( Bianco et al . , 1998; Maes et al . , 2010 ) , we detected lineage clones containing growth plate chondrocytes , mesenchymal cells , and adipocytes in the central part of the growth plate ( Figure 3E , F ) , arguing against such a model . A notable feature of LepR-lineage cells in mice is that they contribute to osteoblasts and adipocytes primarily in postnatal phases , that is when growth plate remodeling is already underway ( Yang et al . , 2014 ) . It would therefore be interesting to test whether LepR-expressing skeletal stem cells have a similar origin from hypertrophic chondrocytes in mammals . One caveat is that we detect endogenous lepr expression in both marrow cells and growth plate chondrocytes in zebrafish . However , the more specific labeling of marrow cells by the LepR-Cre in mouse likely reflects the Cre insertion being in the long LepR isoform ( containing exon 18b ) that displays more restricted expression than the short isoform ( Zhou et al . , 2014b ) . Indeed , LepR mRNA and protein has also been reported in chondrocytes of mouse ( Hoggard et al . , 1997 ) , rat , and human ( Morroni et al . , 2004 ) , a finding we confirmed in the postnatal mouse femur , including the same higher expression in immature versus hypertrophic chondrocytes that we observe in zebrafish ( Figure 6—figure supplement 1B ) . Without a comparable long-isoform lepr-Cre line in zebrafish , we cannot therefore conclude whether zebrafish lepr+ marrow cells are comparable to those described in mouse . The future generation of Cre lines to specifically mark lepr+ marrow cells in zebrafish will be needed to determine whether these chondrocyte-derived marrow cells also act as stem cells for osteoblasts and adipocytes in post-embryonic fish . A similar delay in the remodeling of the hypertrophic zone in mouse Mmp9 and zebrafish mmp9 mutants reveals genetic conservation of growth plate remodeling from fish to mammals . In contrast to mice , we find that Mmp9 in zebrafish appears to function primarily in chondrocytes for growth plate remodeling . We also observed rescue of marrow adipocyte number by wild-type chondrocytes , though this had moderate statistical significance ( p = 0 . 06 ) , potentially owing to low sample size ( n = 8 ) , the mosaic contribution of wild-type cells to the growth plate , and/or roles of cells outside the growth plate to marrow adipocyte generation . In mice , loss of Mmp13 in chondrocytes does result in a growth plate remodeling delay , with global Mmp13 deletion enhancing the remodeling defects of Mmp9 mutants ( Inada et al . , 2004; Stickens et al . , 2004 ) . It may be that MMPs are derived from both chondrocytes and invading hematopoietic cells ( e . g . osteoclasts ) , with the relative importance of each cell source varying between zebrafish and mammals . Compensation by mmp13 might also explain why we detected growth remodeling defects in mmp9 zebrafish mutants at 21 and 27–31 mm SL stages , but not at 17 mm SL . Mmp9 and Mmp13 have known roles in degrading components of the cartilage extracellular matrix ( Page-McCaw et al . , 2007 ) , consistent with our observed loss of collagen-rich matrix in hypertrophic chondrocytes at the edges of the zebrafish growth plate . Secreted Mmp9 may therefore function simply to degrade cartilage matrix and facilitate release of dedifferentiating chondrocytes into the marrow . Intriguingly , Mmp9 has recently been shown to have an additional , non-canonical function in the nucleus for histone H3 tail cleavage . Whereas this has only been demonstrated so far for osteoclasts ( Kim et al . , 2016 ) , it remains possible that Mmp9 could have a similar non-canonical function in altering the chromatin structure of hypertrophic chondrocytes to allow them to acquire new potential . In conclusion , our data support conservation of mammalian-like growth plate remodeling in zebrafish , which provides new opportunities for better understanding the molecular and cellular mechanisms by which hypertrophic chondrocytes transform into osteoblasts , marrow adipocytes , and potentially adult skeletal stem cells within endochondral bones .
All procedures were approved by the University of Southern California Institutional Animal Care and Use Committee . Published Danio rerio lines include Tg ( Has . RUNX2:EGFP ) zf259 and Tg ( Mmu . Sox10-Mmu . Fos:Cre ) zf384 ( Kague et al . , 2012 ) , Tg ( sp7:EGFP ) b1212 ( DeLaurier et al . , 2010 ) , Tg ( Ola . Osteocalcin . 1:EGFP ) hu4008 ( Knopf et al . , 2011 ) , Tg ( col2a1aBAC:GFP ) el483 and Tg ( col2a1aBAC:mCherry-NTR ) el559 ( Askary et al . , 2015 ) , Tg ( bactin2:loxP-BFP-loxP-DsRed ) sd27 ( Kobayashi et al . , 2014 ) , Tg ( fli1a:eGFP ) y1 ( Lawson and Weinstein , 2002 ) , Tg ( kdrl:eGFP ) s843 ( Jin et al . , 2005 ) , Tg ( −5 . 2lyve1b:DsRed ) nz101 ( Okuda et al . , 2012 ) , and Zebrabow - Tg ( ubb:LOX2272-LOXP-RFP-LOX2272-CFP-LOXP-YFP ) a131 ( Pan et al . , 2013 ) . The sox10:CreERT2 transgene was generated with Gateway Cloning ( Invitrogen ) and the Tol2 kit ( Kwan et al . , 2007 ) by combining p5E-sox10 ( Das and Crump , 2012 ) , pME-CreERT2 , p3E-pA , and pDestTol2pA2 . The col2a1a-R2-E1b:CreERT2 and col2a1a-R2-E1b:H2A . F/Z-mCherry-P2A-GFPCAAX transgenes utilize a zebrafish col2a1a R2 enhancer element with a minimal E1B promoter sequence ( Dale and Topczewski , 2011 ) . For col2a1a-R2-E1b:CreERT2 , CreERT2 was amplified using pENTR/D-CreERT2 as template and primers 5’- TTCTTGTACAAAGTGGCCACCGGCCACCATGTCCAATTTACTGACCGTACAC-3’ and 5’-TAGAGGCTCGAGAGGCCTTGTCAAGCTGTGGCAGGGAAACCCTC-3’ . The amplified PCR product was combined with an NcoI/EcoRI fragment of pDestTol2-col2a1aR2-E1B-eGFPpA ( Askary et al . , 2015 ) using Gibson Assembly ( New England Biolabs ) . For col2a1a-R2-E1b:H2A . F/Z-mCherry-P2A-GFPCAAX , H2A . F/Z-mCherry was amplified using template pME-H2A . F/Z-mCherry and primers 5’-TTTCTTGTACAAAGTGGCCAAAGCTTGGATCCCGGCCACCATGGCAGGTGGAAAAGCAGG-3’ and 5’-AAGTTGGTTGCTCCCGACCCCTTGTACAGCTCGTCCATGCCGCCGGTG-3’ . P2A-GFPCAAX was synthesized as a gBlock ( IDT ) . H2A . F/Z-mCherry and P2A-GFPCAAX were combined with an NcoI/EcoRI fragment of pDestTol2-col2a1aR2-E1B-eGFPpA using Gibson Assembly . All transgenes were injected at 30 ng/μL along with 50 ng/μL Tol2 mRNA into one-cell-stage embryos , with these animals raised and outcrossed to identify stable germline founders . CreERT2 transgenes were injected into Tg ( bactin2:loxP-BFP-loxP-DsRed ) sd27 embryos , followed by overnight treatment with 10 μM ( Z ) −4-Hydroxytamoxifen ( 4-OHT ) ( Sigma-Aldrich H7904 ) at two dpf and screening for DsRed+ chondrocytes at six dpf using a Leica fluorescent stereomicroscope . Embryos with fluorescent cartilages were raised to adulthood and outcrossed to identify founders . We used Tg ( sox10:CreERT2 ) el777 , Tg ( col2a1a-R2-E1b:CreERT2 ) el712 , and Tg ( col2a1a-R2-E1b:H2A . F/Z-mCherry-P2A-GFPCAAX ) el695 lines for our experiments . Two additional alleles of Tg ( col2a1a-R2-E1b:CreERT2 ) and one additional allele of Tg ( col2a1a-R2-E1b:H2A . F/Z-mCherry-P2A-GFPCAAX ) gave similar cartilage-specific expression . To generate mmp9el734 we used CRISPR/Cas9 mutagenesis to target the second exon . gRNA targeting the sequence 5’-TTGATGCCATGAAGCAGCCC-3’ was injected at 25 ng/μl with 50 ng/μl Cas9 RNA into one-cell-stage embryos as described ( Hwang et al . , 2013 ) . The mmp9el734 allele is an 8 bp deletion that results in the incorporation of 13 additional amino acids after amino acid 98 ( P ) , followed by a premature stop codon that is predicted to completely abolish the catalytic metalloproteinase domain . Live bone staining of dissected ceratohyal bones was performed by treating with 50 µg/ml Alizarin Red ( Sigma Aldrich , cat . no . A5533 ) , Calcein Green ( Thermofisher Scientific , cat . no . C481 ) , or Calcein Blue , AM ( Thermofisher Scientific , cat . no . C1429 ) for 5 min and repeatedly rinsing in embryo medium as described ( Paul et al . , 2016 ) . We performed adipocyte labeling of dissected ceratohyal bones by incubating in a 1:200 solution of HCS LipidTOX Deep Red ( Life Technologies , cat . no . H34477 ) for 15 min and rinsing in embryo medium as described ( Minchin and Rawls , 2017 ) . Paraffin embedding and histology were performed as described ( Paul et al . , 2016 ) . We cut blocks into 5 μm sections on a Shandon Finesse Me+ microtome ( cat . no . 77500102 ) and collected sections on Apex Superior Adhesive slides ( Leica Microsystems , cat . no . 3800080 ) . Pentachrome and Trichrome staining were performed according to manufacturer's instructions ( Movat-Russell modified pentachrome stain kit , Newcomer Supply cat . no . 9150A; Gomori One-Step , Aniline Blue , trichrome stain kit , Newcomer Supply cat . no . 9176A ) . For cell proliferations assays , fish were immersed in system water containing 4 . 5 mg/ml BrdU ( Sigma Aldrich , cat . no . B5002 ) for 1 hr , followed by two washes in system water , fixation in 4% paraformaldehyde , and paraffin embedding . Immunohistochemistry was performed as described except that we blocked with 2% normal goat serum ( Jackson ImmunoResearch , cat . no . 005-000-121 ) . After cutting thin sections , we performed antigen retrieval by treating slides with citrate buffer ( pH 6 . 0 ) in a steamer set ( IHC World , cat . no . IW-1102 ) for 35 min . Primary antibodies include rat anti-BrdU ( 1:100 , Bio-Rad , cat . no . MCA2060GA ) and rabbit anti-mCherry ( 1:250 , Novus Biologicals , cat . no . NBP2-25157 ) . We used AlexaFluor secondary antibodies and Hoechst to visualize nuclei . Colorimetric in situ hybridization was performed as described ( Paul et al . , 2016 ) . The mmp9 riboprobe was generated by PCR amplification of zebrafish genomic DNA with primers 5’-GTCTCCAATACTAAAGCTCTGAAGAAG-3’ and 5-‘TAGGATGTCGAAGGTCTATAGAGAATG-3’ and cloning into pCR-BluntII-TOPO ( Life Technologies ) . RNA probe was synthesized with T7 polymerase ( Roche ) after linearizing the plasmid with BamHI restriction . RNAscope probes for leptin receptor ( lepr ) and runx2b were synthesized by Advanced Cell Diagnostics in Channel 1 and Channel 2 , respectively , and for mouse Lepr in channel 1 . Paraformaldehyde-fixed paraffin-embedded sections were deparaffinized , and the RNAscope fluorescent multiplex v2 assay combined with immunofluorescence was performed according to manufacturer’s protocols and with the ACD HybEZ Hybridization oven . At six hpf , donor ectoderm from the animal cap of bactin2:loxP-tagBFP-loxP-DsRed embryos was transplanted into the neural crest precursor domain of mmp9-/- hosts as described ( Crump et al . , 2004 ) . Embryos displaying unilateral tagBFP fluorescence in the face at three dpf were raised in the nursery and then genotyped at 14 dpf for the mmp9 mutant allele . Homozygous mutant and wild-type siblings were raised at similar density until they reached the indicated sizes for analysis . Brightfield images of pentachrome and trichrome stains , and colorimetric in situ hybridizations , were acquired with a Zeiss AxioImager . A1 microscope and a Zeiss slide scanner AxioScan . Z1 . Focus stacking of multiple images was done in Adobe Photoshop CS5 . Fluorescence images were captured on a Zeiss LSM800 confocal microscope , with representative sections or maximum intensity projections shown as specified . Tiling was performed using ZEN software or manually stitched together using Fiji . Brightness and contrast were adjusted in Adobe Photoshop CS5 with similar settings for experimental and control samples . We stage-matched control and experimental zebrafish by measuring standard body length ( SL ) from the tip of the snout to the edge of the hypuralia . Adipocyte counts and area/width measurements of growth plates were calculated using Fiji . The proliferative zone of the growth plate was defined as the central region of compact chondrocytes . All measurements were performed blinded to genotype . Statistical significance was determined by a student’s t-test for pair-wise comparisons or Tukey-Kramer HSD tests for multiple comparisons , using GraphPad’s Prism software . | Our adult bones are made of a fatty tissue , called marrow , wrapped inside a hard outer layer produced by bone cells . They may appear stiff and unyielding , but our bones are actually dynamic structures . Early in life , most bones start as small ‘templates’ made of another , flexible tissue called cartilage . As the templates grow into adult bones , the cartilage is gradually replaced by bone and fat , but this process is still poorly understood . For example , it is not clear whether cartilage cells simply die and make way for new cells , or instead if they turn into bone and fat cells . To investigate this question , Giovannone , Paul et al . set out to follow the fate of early cartilage cells in zebrafish , and to compare this with what happens in mammals . Zebrafish were chosen because their skeleton and ours develop in similar ways; yet , these animals are much easier to study , in particular because their embryos are transparent . Young cartilage cells were ‘tagged’ with a long-lasting fluorescent protein in genetically engineered zebrafish embryos , and then followed over time . As the embryos started to form bones , the cartilage cells gave rise to bone cells , fat cells , and also potentially adult stem cells within the marrow , which can become other types of cells . This process required a protein called Mmp9 , which also helps shape bone development in other organisms , including humans . Defects in how early cartilage templates morph into bone and fat may contribute to dwarfism and other severe conditions . Fully grasping the molecular mechanisms that preside over this complex transition may one day help design drugs to treat skeletal disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"developmental",
"biology"
] | 2019 | Programmed conversion of hypertrophic chondrocytes into osteoblasts and marrow adipocytes within zebrafish bones |
Interoception , the sensitivity to visceral sensations , plays an important role in homeostasis and guiding motivated behaviour . It is also considered to be fundamental to self-awareness . Despite its importance , the developmental origins of interoceptive sensitivity remain unexplored . We here provide the first evidence for implicit , flexible interoceptive sensitivity in 5 month old infants using a novel behavioural measure , coupled with an established cortical index of interoceptive processing . These findings have important implications for the understanding of the early developmental stages of self-awareness , self-regulation and socio-emotional abilities .
Forty-one healthy , full-term infants were tested in total , at 5 months of age ( 19 males , mean age = 5 . 10 months , SD = 0 . 29 ) . The expected effect sizes were not known in advance , so samples were selected according to similar adult literature ( e . g . [Fukushima et al . , 2011] ) assuming an approximate 50% attrition rate , which is usual in infant EEG studies with this age range . Infants were recruited using a marketing company database , which provides data information from consenting mothers to be . Recruitment leaflets were sent to each household . Parents were able to participate by signing up to our online database or by contacting us via email . The study was completed in one session , and conducted according to the Declaration of Helsinki and all methods were approved by the Royal Holloway University of London Departmental Ethics Committee . Mothers and infants familiarised themselves with the testing room . Infants first performed the iBEAT task . After a short break for feeding and changing if needed , they were then placed on their mother’s lap for the EEG cap and electrodes to be fitted . Once the EEG signal was clear and the infant was ready , they were returned to the high-chair and the Emotion Observation Task was run . The cap and ECG electrodes were then removed , and mother and infant were taken to a comfortable rest area to take a break , feed and change if necessary . The mother then completed a brief questionnaire and further behavioural task ( the results of which will be reported in a separate publication ) before being thanked , debriefed and given a small gift and monetary compensation for travel costs . All tests were evaluated against a two-tailed p<0 . 05 level of significance . For the HEP analysis , a Monte-Carlo random cluster-permutation method was implemented in FieldTrip . This method corrects for multiple comparisons in space and time ( Maris and Oostenveld , 2007 ) . Using this method , all samples that showed a significant ( p< . 05 ) relationship with our independent variable were clustered according to spatiotemporal adjacencies , and cluster-level statistics were calculated by taking a sum of the t-values for each cluster . A Monte-Carlo permutation method then generated a p-value by calculating the probability that this cluster-level statistic could be achieved by chance , by randomly shuffling and resampling the independent variable structure a large number of times ( 2000 repetitions ) ( Maris and Oostenveld , 2007 ) . Spatiotemporal clusters that had a resulting Monte-Carlo corrected p-value of less than the critical alpha level of . 05 were interpreted as ‘significant’ . For both the iBEAT task and the HEP measurement , data collection was not performed blind to the experimental condition to which each trial belonged , due to the requirements for stimulus and cardiac monitoring during the task . However , for HEP analysis , experimental condition was removed from the data after data collection and all EEG pre-processing was performed blind to the conditions of the experiment . Condition was revealed at final statistical analysis so that specific emotions could be compared to the neutral condition . Both reported tasks had within-subjects designs involving no group allocation; therefore , blinding to any between-subject conditions and randomization to such conditions was not applicable . | From the beginning till the end of a person’s life , parts of the body continuously send signals to the brain . Most of this happens without the person even being aware of it , yet people can become aware of the signals under certain circumstances . For example , we can feel our racing heart rate or the “butterflies in our stomach” when we are anxious or excited . This ability to consciously sense signals from the body is called interoception , and some people are more aware of these signals than others . These differences between people can influence a wide range of psychological processes , including how strongly they feel emotions , how they make decisions , and their mental health . Despite the crucial role that interoception plays in thought processes in adults , scientists know practically nothing about how it first develops . Progress in this field has been hindered largely because there was no way to measure sensitivity to interoceptive signals in infants . Now , Maister et al . have developed a new task called iBEATS that can measure how sensitive an infant is to their own heartbeat . During the task , five-month old infants were shown an animated character that either moved in synchrony with their own heartbeat or out of synchrony with their heartbeat . The infants spent longer looking at the character that was moving out of synchrony than the one moving in synchrony , suggesting that even at this early age , infants can sense their own interoceptive signals . As with adults , some of the infants were more sensitive to their heartbeats than others , and Maister et al . could see these differences played out in the infant’s brain activity via electrodes placed on the infant’s head . Infants who had shown a strong preference in the iBEATS task also showed a larger brain signal known as the Heart-Evoked Potential ( or HEP ) . Furthermore , this brain signal got larger when infants viewed a video clip of an angry or fearful face . This suggests that the infants’ brains were monitoring their hearts more closely when they were confronted with negative emotions . This study provides a validated measure of interoception for very young participants . Using this task , researchers can now investigate which factors affect how awareness to interoceptive signals develops , including social interactions and the infant’s temperament . Maister et al . also plan to carry out longer-term experiments to learn exactly how interoception may influence the development of emotional abilities , and also what role it might play in disorders such as anxiety and depression . The findings of these future experiments may eventually guide interventions to treat these conditions . | [
"Abstract",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2017 | Neurobehavioral evidence of interoceptive sensitivity in early infancy |
Assembly , maintenance and function of synaptic junctions depend on extracellular matrix ( ECM ) proteins and their receptors . Here we report that Tenectin ( Tnc ) , a Mucin-type protein with RGD motifs , is an ECM component required for the structural and functional integrity of synaptic specializations at the neuromuscular junction ( NMJ ) in Drosophila . Using genetics , biochemistry , electrophysiology , histology and electron microscopy , we show that Tnc is secreted from motor neurons and striated muscles and accumulates in the synaptic cleft . Tnc selectively recruits αPS2/βPS integrin at synaptic terminals , but only the cis Tnc/integrin complexes appear to be biologically active . These complexes have distinct pre- and postsynaptic functions , mediated at least in part through the local engagement of the spectrin-based membrane skeleton: the presynaptic complexes control neurotransmitter release , while postsynaptic complexes ensure the size and architectural integrity of synaptic boutons . Our study reveals an unprecedented role for integrin in the synaptic recruitment of spectrin-based membrane skeleton .
The extracellular matrix ( ECM ) and its receptors impact every aspect of neuronal development , from axon guidance and migration to formation of dendritic spines and neuromuscular junction synaptic junctions and function . The heavily glycosylated ECM proteins provide anchorage and structural support for cells , regulate the availability of extracellular signals , and mediate intercellular communications ( Reichardt and Prokop , 2011 ) . Transmembrane ECM receptors include integrins , syndecans and the dystrophin-associated glycoprotein complex ( Bökel and Brown , 2002; Häcker et al . , 2005; Waite et al . , 2009 ) . Integrins in particular are differentially expressed and have an extensive repertoire , controlling multiple processes during neural development . In adults , integrins regulate synaptic stability and plasticity ( Morini and Becchetti , 2010; McGeachie et al . , 2011 ) . However , integrin roles in synapse development have been obscured by their essential functions throughout development . How integrins are selectively recruited at synaptic junctions and how they engage in specific functions during synapse development and homeostasis remain unclear . One way to confer specificity to ECM/integrin activities is to deploy specialized ECM ligands for the synaptic recruitment and stabilization of selective heterodimeric integrin complexes ( Reichardt and Tomaselli , 1991 ) . For example , at the vertebrate NMJ , three laminins containing the β2 subunit ( laminin 221 , 421 and 521 , that are heterotrimers of α2/4/5 , β2 and γ1 subunits ) are deposited into the synaptic cleft and basal lamina by skeletal muscle fibers and promote synaptic differentiation . However , only laminin 421 interacts directly with presynaptic integrins containing the α3 subunit and anchors a complex containing the presynaptic Cavα and cytoskeletal and active zone-associated proteins ( Carlson et al . , 2010 ) . Studies with peptides containing the RGD sequence , recognized by many integrin subtypes , have implicated integrin in the morphological changes and reassembly after induction of long-term potentiation ( LTP ) ( reviewed in [McGeachie et al . , 2011] ) . Several integrin subunits ( α3 , α5 , α8 , β1 and β2 ) with distinct roles in the consolidation of LTP have been identified , but the relevant ligands remain unknown . Drosophila neuromuscular junction ( NMJ ) is a powerful genetic system to examine the synaptic functions of ECM components and their receptors . In flies , a basal membrane surrounds the synaptic terminals only in late embryos; during development , the boutons ‘sink’ into the striated muscle , away from the basal membrane ( Prokop et al . , 1998 ) . The synaptic cleft relies on ECM to withstand the mechanical tensions produced by the muscle contractions . The ECM proteins , including laminins , tenascins/teneurins ( Ten-a and -m ) and Mind-the-gap ( Mtg ) , interact with complexes of five integrin subunits ( αPS1 , αPS2 , αPS3 , βPS , and βν ) ( Broadie et al . , 2011 ) . The αPS1 , αPS2 and βPS subunits localize to pre- and post-synaptic compartments and have been implicated in NMJ growth ( Beumer et al . , 1999; Beumer et al . , 2002 ) . The αPS3 and βν are primarily presynaptic and control activity-dependent plasticity ( Rohrbough et al . , 2000 ) . The only known integrin ligand at the fly NMJ is Laminin A , which is secreted from the muscle and signals through presynaptic αPS3/βν and Focal adhesion kinase 56 ( Fak56 ) to negatively regulate the activity-dependent NMJ growth ( Tsai et al . , 2012 ) . Teneurins have RGD motifs , but their receptor specificities remain unknown ( Mosca et al . , 2012 ) . Mtg secreted from the motor neurons influences postsynaptic βPS accumulation ( Rushton et al . , 2009 ) , but that may be indirectly due to an essential role for Mtg in the organization of the synaptic cleft and the formation of the postsynaptic fields ( Rohrbough et al . , 2007; Rushton et al . , 2012 ) . The large size of these proteins and the complexity of ECM-integrin interactions made it difficult to recognize relevant ligand-receptor units and genetically dissect their roles in synapse development . Here , we report the functional analysis of Tenectin ( Tnc ) , an integrin ligand secreted from both motor neurons and muscles; Tnc accumulates at synaptic terminals and functions in cis to differentially engage presynaptic and postsynaptic integrin . We uncovered tnc , which encodes a developmentally regulated RGD-containing integrin ligand ( Fraichard et al . , 2006; Fraichard et al . , 2010 ) , in a screen for ECM candidates that interact genetically with neto , a gene essential for NMJ assembly and function ( Kim et al . , 2012 ) . We found that Tnc selectively recruits the αPS2/βPS integrin at synaptic locations , without affecting integrin anchoring at muscle attachment sites . Dissection of Tnc functions revealed pre- and postsynaptic biologically active cis Tnc/integrin complexes that function to regulate neurotransmitter release and postsynaptic architecture . Finally , we exploited the remarkable features of this selective integrin ligand to uncover a novel synaptic function for integrin , in engaging the spectrin-based membrane skeleton .
To search for novel ECM proteins important for NMJ development we set up a synthetic lethality screen that took advantage of the 50% lethality of an allele with suboptimal levels of Neto , neto109 ( Kim et al . , 2012 ) . Neto , an obligatory subunit of ionotropic glutamate receptor ( iGluR ) complexes , controls the distribution and function of iGluRs as well as the assembly and organization of postsynaptic structures ( Han et al . , 2015; Kim et al . , 2015; Ramos et al . , 2015 ) . Using this lethality screen we have previously uncovered genetic interactions between neto and several BMP pathway components ( Sulkowski et al . , 2014; Sulkowski et al . , 2016 ) . Lowering the dose of Mtg , an ECM protein known to organize the synaptic cleft ( Rohrbough et al . , 2007 ) , induced 95% lethality ( n = 286 ) in neto109/Y;; mtg1/+ animals , further validating our strategy . We focused on ECM candidates ( Broadie et al . , 2011 ) and identified a set of overlapping deficiencies ( Df ( 3R ) BSC-318 , , –492 , −494 , and −655 ) that drastically increased the lethality of neto109 hemizygotes ( from 50% for neto109/Y up to 82% for neto109/Y;; Df/+ ) . Among the common loci disrupted by these deficiencies was tnc , a gene coding for a large mucin-type protein conserved in many insects but with no obvious mammalian homologue ( Figure 1A–B ) ( Fraichard et al . , 2006; Syed et al . , 2008; Fraichard et al . , 2010; Syed et al . , 2012 ) . Tnc is a secreted molecule with five vWFC ( von Willebrand factor type-C ) protein interaction domains separated by two PTS-rich regions . In mucins , the PTS domains are highly O-glycosylated and form gel-like structures . Tnc has one RGD and several more RGD-like motifs that have been implicated in interaction with integrin ( Fraichard et al . , 2010 ) . During development , Tnc is secreted in the lumen of several epithelial organs , including foregut , hindgut and trachea . Tnc is also expressed in the embryonic CNS . Using a polyclonal anti-Tnc antibody ( Materials and methods ) we found that Tnc signals are strongly enriched in the neuropile , in the proximity of the anti-Fasciclin II ( FasII ) positive axons ( Figure 1—figure supplement 1A , B and [Fraichard et al . , 2006] ) . The Tnc signals were absent in the CNS of a tnc mutant ( tncEP- P[EPgy2]EY03355 ) , predicted to disrupt both known tnc transcripts ( Syed et al . , 2012 ) . These mutant embryos had a normal FasII pattern indicating no obvious CNS defects during late embryogenesis ( Figure 1—figure supplement 1C–D ) . The tncEP mutant showed partial lethality ( 11 . 6% of expected homozygous progenies were viable , n = 303 ) , which was not enhanced in heteroallelic combinations ( 12 . 2% viability for tncEP/Df , n = 392 ) suggesting that this mutant is equivalent to a genetic null . We also generated a small deletion mutant ( tnc82 ) by FRT-induced recombination . These animals die as homozygous pharate adults ( 100% lethality , n = 221 ) but produce some heteroallelic escapers ( 17 . 5% viability for tnc82/Df , n = 389 ) . Western blot analysis revealed a Tnc-positive band of ~300 kD ( the calculated MW for Tnc is 299 kD ) in extracts from brains and body-wall muscles of control larvae ( Figure 1C ) . This band was undetectable in tnc hetero-allelic combinations , tncEP/Df and tnc82/Df . A band of similar size was found in S2 cells transfected with a Tnc expression construct and was enriched in the conditioned media , indicating that Tnc is efficiently secreted in cell culture ( Figure 1D , Material and methods ) . Neuron specific RNAi knockdown reduced Tnc levels in larval brains to 43% of the control group; this generated very strong phenotypes ( below ) suggesting that the residual band could reflect additional Tnc-expressing cells in the larval brain . The muscle-specific knockdown reduced the muscle Tnc levels to 19% of the control ( Figure 1E ) . During larval stages , we found Tnc positive signals throughout the muscles with weak accumulation at the NMJ ( Figure 1F and Figure 1—figure supplement 1F ) . Under the same imaging conditions , the signals were significantly reduced in tncEP/Df mutants ( Figure 1G and Figure 1—figure supplement 1G , quantified in Figure 1—figure supplement 1E ) . Such weak NMJ immunoreactivities were previously reported for proteins secreted in the synaptic cleft ( Rushton et al . , 2009 ) . To further confirm the specificity of Tnc signals , we overexpressed Tnc in motor neurons ( elav-Gal4 ) or muscles ( BG487-Gal4 ) and examined the NMJs ( Figure 1H–J and Figure 1—figure supplement 1H–I ) . Paneuronal expression of Tnc induced strong accumulation of Tnc-positive signals at synaptic terminals , as well as along the motor neuron axons . At these NMJs , Tnc-labeled puncta were concentrated at the edge of anti-horseradish peroxidase ( HRP ) -stained boutons ( Jan and Jan , 1982 ) . Most of these signals were also observed in the absence of detergents , suggesting that Tnc is secreted in the synaptic terminal . In contrast , muscle overexpression of tnc showed increased Tnc-positive signals throughout the muscle; at synaptic terminals these signals appeared diffuse and farther away from the neuronal membrane . Thus , excess muscle Tnc may not be properly targeted and/or stabilized at synaptic terminals and may have detrimental effects on Tnc-mediated functions . We repeated these results using independent HA-tagged tnc transgenes and staining with anti-HA antibodies ( Figure 1—figure supplement 2 ) . Both tagged and untagged transgenes rescued the viability of tnc mutants ( see below ) , indicating that Tnc functions are unaffected by addition of the tag . As above , expression of tnc-HA in neurons but not in muscles induced high accumulation of HA-positive puncta , accessible without detergent , around the synaptic boutons , indicating extracellular distribution ( Figure 1—figure supplement 2A–D ) . Thus , Tnc is expressed in both motor neurons and muscles and appears to accumulate at the ECM surrounding synaptic terminals . To investigate a possible role for Tnc in the function of the nervous system and/or the musculature , we examined the morphology and physiology of tnc mutants . During larval stages , both tnc mutants ( tncEP/Df and tnc82/Df ) had largely normal NMJ , with minimally increased bouton numbers ( Figure 2—figure supplement 1A–B ) . A closer examination revealed smaller boutons , with less clear bouton-interbouton delimitations ( details below ) . The few adult escapers did not fly and exhibited climbing defects ( Figure 2—figure supplement 1C ) . Such phenotypes are consistent with previously reported flightless adults generated by RNAi-mediated Tnc knockdown ( Fraichard et al . , 2010 ) . We next recorded the evoked excitatory junction potentials ( EJPs ) and spontaneous miniature excitatory junction potentials ( mEJPs ) from muscle 6 of third instar larvae ( Figure 2A–E ) . The mEJPs amplitude was normal in tnc mutants . However , the mean frequency of mEJPs was significantly reduced in tnc mutants compared with the control ( w1118 , 2 . 67 ± 0 . 14 Hz vs . tncEP/Df , 1 . 68 ± 0 . 13 Hz , p=0 . 0001 , and tnc82/Df , 2 . 11 ± 0 . 14 Hz , p=0 . 0322 , Figure 2D ) . tnc mutations caused 23% and 18% reduction in evoked EJPs amplitude of tncEP/Df and tnc82/Df animals , respectively ( w1118 , 26 . 30 ± 0 . 99 mV vs . tncEP/Df , 20 . 15 ± 1 . 26 mV , p=0 . 0042 , and tnc82/Df , 21 . 53 ± 1 . 38 mV , p=0 . 0375 , Figure 2E ) . Moreover , tnc mutants showed a significant decrease in quantal content ( w1118 , 29 . 05 ± 1 . 81 vs . tncEP/Df , 21 . 44 ± 1 . 60 and tnc82/Df , 21 . 53 ± 1 . 38 , Figure 2F ) . Since we found no change in the resting potential and input resistance in mutant animals , the decrease in EJPs amplitude and quantal content was probably not caused by abnormal passive membrane properties in the muscle . Instead , the reduction of quantal content could be due to a decreased number of vesicle release sites or reduced probability of release . The reduced mEJP frequency is in agreement with reduced quantal content due to fewer release sites . To evaluate the vesicle release probability , we measured the paired-pulse ratio ( PPR ) using the EJP amplitudes evoked by two stimuli separated by duration of 50 ms ( Wong et al . , 2014 ) . At 0 . 5 mM extracellular Ca2+ concentration , the control larvae showed mild short-term depression following paired-pulse stimulation ( PPR < 1 , Figure 2G–H ) , indicating a relatively high initial probability of vesicle release; the second stimulus , provided before the resting Ca2+ returned to baseline , lead to the exocytosis of fewer synaptic vesicles than the first stimulus . In contrast , the tnc mutant NMJs showed elevated facilitation and significantly increased PPR ( w1118 , 0 . 80 ± 0 . 03 vs . tncEP/Df , 1 . 21 ± 0 . 08 , p=0 . 0231 , and tnc82/Df , 1 . 16 ± 0 . 09 , p=0 . 0295 ) . The higher the ratio of EJP amplitudes following the first and second pulses , the lower is the probability of release . Thus , tnc mutants have significantly decreased probability of vesicle release . Since Tnc is expressed in both pre- and post-synaptic compartments , we next examined which Tnc pool ( s ) is required for NMJ function . We found that tnc knockdown in motor neurons resulted in mEJPs with significantly reduced frequency as compared to the controls ( tncRNAi transgene with no driver , or driver alone ) ( control , 2 . 69 ± 0 . 2 Hz vs . N > tncRNAi , 1 . 31 ± 0 . 10 Hz p=0 . 0001 , and M > tncRNAi , 2 . 4 ± 0 . 17 Hz p=0 . 4434 ) ( Figure 3A–C , and Figure 3—figure supplement 1 ) . Similar results were obtained with a second RNAi line ( GD14952 ) confirming that these phenotypes are specific to tnc depletion ( Figure 3—figure supplement 1 ) . In contrast , tnc knockdown in muscles had no effect on mEJPs frequency or amplitude . Thus , neuronal but not muscle Tnc is required for normal neurotransmitter release . Expression of Tnc in neurons but not in muscles also rescued the mEJPs frequency and EJPs amplitude defects observed in tnc mutants ( Figure 3D–F ) . Similar to tnc mutants , mutants carrying only a tnc transgene but no driver showed reduced mEJP frequency and EJP amplitude ( compare Figure 3D–F and Figure 2 ) . Expression of the tnc transgene in neurons , but not in muscles of tnc mutants restored the mEJPs frequency ( w1118 , 1 . 97 ± 0 . 27 Hz vs . N > tnc; tncEP/Df , 1 . 56 ± 0 . 18 Hz p=0 . 07 and M > tnc; tncEP/Df , 1 . 20 ± 0 . 16 Hz p=0 . 0019 ) and EJPs amplitude ( w1118 , 71 . 51 ± 2 . 39 mV vs . N > tnc; tncEP/Df , 70 . 1 ± 2 . 08 mV p=0 . 198 and M > tnc; tncEP/Df , 45 . 03 ± 3 . 67 mV p<0 . 0001 ) to levels that are no longer significantly different from the w1118 control . Together , these data indicate that neuron-derived Tnc regulates normal neurotransmitter release at the NMJ . Previous studies suggest that Tnc functions as a ligand for αPS2/βPS integrin during wing morphogenesis ( Fraichard et al . , 2010 ) . Drosophila integrins have been implicated in NMJ growth and synaptic function ( Keshishian et al . , 1996 ) , with the βPS-containing complexes primarily in the postsynaptic compartment ( Prokop et al . , 1998; Beumer et al . , 1999; Koper et al . , 2012 ) . If Tnc functions by recruiting integrin at the NMJ , then presynaptic but not postsynaptic βPS or αPS2 should similarly modulate the neurotransmitter release . Indeed , neuronal knockdown of myospheroid ( mys ) , which encodes the βPS integrin subunit , significantly reduced the mEJPs frequency compared to control ( mysRNAi transgene with no driver ) ( Figure 3G–H ) . In contrast , muscle knockdown of mys had no detectable effect on mEJPs frequency ( control , 0 . 95 ± 0 . 14 Hz vs . N > mysRNAi , 0 . 47 ± 0 . 05 Hz p=0 . 0044 and M > mysRNAi , 1 . 00 ± 0 . 07 Hz p=0 . 9142 ) or amplitude . Similarly , neuronal but not muscle knockdown of inflated ( if ) , which codes for αPS2 integrin , significantly reduced the mEJPs frequency compared to control ( ifRNAi transgene with no driver ) ( Figure 3J–L ) ( control , 1 . 62 ± 0 . 51 Hz vs . N > ifRNAi , 0 . 99 ± 0 . 23 Hz p=0 . 0041 and M > ifRNAi , 1 . 67 ± 0 . 42 Hz p=0 . 9662 ) . The ifRNAi transgene appeared stronger than mysRNAi and generated larval lethality and , occasionally , muscle attachment defects; using these transgenes we observed diminished mEJP amplitude compared to control ( Figure 3L ) . Overall , the reduction of presynaptic αPS2/βPS mirrored the mEJP frequency deficits observed for tnc neuronal knockdown suggesting that neuron-derived Tnc functions as a ligand for presynaptic αPS2/βPS to modulate neurotransmitter release . Postsynaptic αPS2/βPS integrin does not appear to influence neurotransmitter release . Alternatively , a role for postsynaptic αPS2/βPS may be obscured by partial knockdowns and essential functions for these genes in the muscle . If Tnc recruits integrin to modulate neurotransmitter release , then tnc and mys ( or if ) should interact genetically . We tested this prediction by examining the trans-heterozygote animals ( mys/+;; tnc/+ ) . Indeed , these trans-heterozygotes exhibited severe mEJP deficits that resembled tnc mutants , whereas individual heterozygote larvae ( mys/+ or tnc/+ ) showed normal mEJPs frequency and amplitude ( Figure 3M–O ) . This indicates that tnc and mys function together to modulate neurotransmitter release . If Tnc recruits and/or stabilizes βPS integrin at synaptic terminals , then Tnc should co-localize with βPS and form Tnc/integrin complexes at synaptic locations and perturbations of Tnc should alter the recruitment of βPS integrin at larval NMJ . Indeed , the βPS signals were dramatically reduced at tnc mutant NMJs ( Figure 4A–D ) . In these analyses the βPS immunoreactivities concentrated at perisynaptic locations , surrounding the control boutons , consistent with previous observations that the muscle pool constitutes the major fraction of βPS at synaptic terminals ( Beumer et al . , 1999 ) . Interestingly , the βPS levels remained unchanged at the muscle attachment sites; we also observed no detachment of the muscle fibers or defects in costamere organization ( Maartens and Brown , 2015 ) . This indicates that loss of Tnc selectively impairs the recruitment of βPS integrin at synaptic terminals . The anti-Tnc antibodies marked discrete puncta at the edge of the HRP- labeled boutons in a region strongly stained by anti-βPS antibodies , but Tnc was not detectable at the muscle attachment sites ( Figure 4E–F ) . This suggests that Tnc and βPS may directly associate at synaptic terminals . Our attempts to co-immunoprecipitate Tnc/integrin complexes from larval carcasses failed; instead , we tested for their close juxtaposition at the NMJ using a proximity ligation assay ( PLA ) , which indicates a less than 40 nm distance between two proteins ( Wang et al . , 2015 ) . As shown in Figure 4G–H , PLA signals between Tnc and βPS were detected at control but not at tnc mutant NMJs . These PLA signals were tightly packed around the synaptic boutons , unlike the Tnc or βPS immunoreactivities , which spread into the postsynaptic specializations , suggesting that Tnc and βPS form complexes in the close proximity of the synaptic terminal . Such complexes could influence the presynaptic neurotransmitter release , as described above , but could also function in the postsynaptic compartment , where most of the perisynaptic βPS resides ( see below ) . Since different integrin heterodimers provide spatial and temporal specificities , we next examined the distribution of integrin α-subunit ( s ) at tnc boutons . We found that αPS2 but not αPS1 levels were selectively reduced at tnc mutant NMJs ( Figure 4I–K and Figure 4—figure supplement 1A–B ) , consistent with the RGD-containing Tnc being a ligand for αPS2/βPS ( Fraichard et al . , 2010 ) . In addition , the levels of phosphorylated Fak were normal at tnc mutant NMJs ( Figure 4—figure supplement 1C–D ) suggesting that Tnc does not influence the LanA-induced αPS3/βν activation of the Fak signaling pathway ( Tsai et al . , 2008; Tsai et al . , 2012 ) . The synaptic abundance of βPS integrin subunit was inversely correlated with the synaptic accumulation of FasII , a homophilic adhesion molecule required for synapse stabilization and growth ( Schuster et al . , 1996a; Schuster et al . , 1996b; Beumer et al . , 2002 ) . We found that FasII synaptic levels were increased by 40% ( p<0 . 05 , n = 28 ) in tnc mutants compared with the controls ( Figure 4—figure supplement 1E–F ) . This increase resembles the elevated levels of FasII reported at mys mutant NMJs , and indicates that loss of Tnc recapitulates some of the phenotypes reported for selective mys mutants ( Beumer et al . , 2002 ) . Since secreted Tnc accumulates at synaptic terminals , both neuron- and muscle-derived Tnc could potentially recruit βPS . However , we found that tnc knockdown in muscles , but not in neurons , reduced the synaptic βPS levels ( Figure 4M–O , quantified in P ) . In fact , tnc knockdown in neurons induced a significant increase of βPS synaptic levels , suggesting that neuron-derived Tnc limits the accumulation of predominantly postsynaptic βPS . This unexpected result prompted us to examine the distribution of Tnc itself in RNAi experiments ( Figure 4—figure supplement 2 ) . Compared to the control ( tncRNAi transgene with no driver ) , tnc knockdown in motor neurons produced a significant increase ( by 28% , p=0 . 0044 , n = 27 ) of synaptic Tnc levels; this result is consistent with the apparent increase in Tnc net levels in muscle extracts from N > tncRNAi larvae ( Figure 1E , right panel ) . Thus , neuron-derived Tnc limits the accumulation of muscle-derived Tnc at synaptic terminals . In contrast , tnc knockdown in muscle diminished the synaptic Tnc levels by 37% ( p<0 . 0001 , n = 30 ) . This partial reduction may reflect an inefficient RNAi treatment and/or a complementary increase in the neuron-derived Tnc . Nonetheless our data indicate that muscle Tnc is required in cis for the postsynaptic recruitment of βPS and that neuron-derived Tnc limits the accumulation of both Tnc and βPS postsynaptically ( see below ) . To determine the function of Tnc in the muscle we first tested whether Tnc influences the assembly and organization of postsynaptic iGluR fields by examining the levels and distribution of various postsynaptic components . Drosophila NMJ utilizes two types of iGluRs , type-A and -B , which require the essential auxiliary protein Neto for their distribution and function . Lack of Tnc did not alter the intensities of GluRIIA and GluRIIB synaptic signals or the IIA/IIB ratio ( Figure 5—figure supplement 1 ) . This result is consistent with the normal mini amplitude observed at tnc mutant NMJs ( Figure 2 ) . Neto itself appeared properly recruited at Tnc-depleted synapses ( Figure 5—figure supplement 1 ) . In addition to iGluRs , Neto is critical for the recruitment of p21-activated kinase , PAK , a postsynaptic protein that stabilizes type-A receptors at PSDs ( Ramos et al . , 2015 ) . We found that PAK signals are normal at Tnc-deprived NMJs , even though the βPS levels are reduced ( Figure 5—figure supplement 1 ) . This suggests that Neto controls PAK recruitment at synaptic terminals; alternatively , a very low level of βPS may suffice in recruiting/stabilizing PAK at synaptic terminals . In the course of these experiments , we noted that tnc mutant NMJs have smaller boutons and often poorly defined bouton/interbouton boundaries . To characterize these defects , we first examined the distribution of HRP-marked neuronal membranes and Discs large ( Dlg ) , a PDZ ( PSD-95/Dlg/Zona occludens-1 ) domain-containing scaffolding protein ( Budnik et al . , 1996 ) . Dlg localizes perisynaptically to the subsynaptic reticulum ( SSR ) , a stack of membrane folds that surrounds the type I boutons ( Guan et al . , 1996 ) . In the absence of Tnc , the type Ib boutons appeared significantly smaller and had diminished Dlg signals ( Figure 5A–B , quantified in F-G ) . tnc knockdown in neurons induced a slight increase in the Dlg signals and no change in the bouton area , whereas tnc knockdown in muscles significantly decreased both the Dlg synaptic levels and the size of the type Ib boutons ( Figure 5C–E ) . Similar reduction in bouton size has been reported for selective mys mutants ( Beumer et al . , 1999 ) , suggesting that postsynaptic Tnc/integrin complexes control bouton size . In electron micrographs , type Ib control boutons are surrounded by a thick SSR ( Figure 5H , quantified in 5I ) ; presynaptic T-bars and electron-dense membranes mark individual synapses . The synapses appeared to have normal organization at tnc mutant boutons , with electron-dense synaptic membranes separated by a dense synaptic cleft ( Figure 5J ) . However , the mutant boutons had irregular shapes and were surrounded by sparse SSR , with wider spaces between the membrane layers . These ultrastructural defects are consistent with the observed morphological phenotypes and indicate that muscle Tnc regulates the SSR thickness and bouton architecture . The tnc phenotypes may reflect increased adhesion due to elevated FasII levels ( Figure 4—figure supplement 1 ) . Alternatively , bouton architecture could be disrupted when muscles contract in the absence of a properly reinforced synaptic skeleton , including the microtubule-based and the cortical membrane skeleton . We found no microtubule defects at tnc mutant NMJs ( Figure 6A–B ) . At the presynaptic arbor , the mature microtubule bundles that traverse the NMJ branches can be visualized with antibodies against the microtubule-associated protein Futsch . These microtubules remained organized in smooth sheaths and loops in tnc mutant larvae , similar to the control . In contrast , loss of tnc severely disrupted the α-Spectrin accumulation at the NMJs ( Figure 6C–F ) . The net levels of α-Spectrin were normal in extracts from tnc larval muscles , as determined by Western blot analysis , but the synaptic abundance of α-Spectrin at tnc NMJs was reduced to 37% from control levels ( p<0 . 001 , n = 21 ) . Since α-Spectrin is essential for the integrity of the SSR ( Pielage et al . , 2006 ) , loss of synaptic α-Spectrin is consistent with the sparse SSR observed at tnc mutant NMJs . Moreover , α- and β-spectrin mutant embryos have reduced neurotransmitter release and diminished EJPs amplitudes without any apparent defects in postsynaptic receptor fields ( Featherstone et al . , 2001 ) ; these defects are reminiscent of Tnc-deprived NMJs . Interestingly , the distorted shapes of tnc mutant boutons resemble the less individuated boutons seen in α-spectrinR22S larvae , which are impaired for spectrin tetramerization ( Khanna et al . , 2015 ) . Tetramerization is required for formation of the spectrin-based membrane skeleton ( SBMS ) but is not required for viability in Drosophila , whereas spectrins are essential genes ( Lee et al . , 1993; Pielage et al . , 2005; Pielage et al . , 2006 ) . A key protein involved in the organization of SBMS is Adducin ( Bennett and Baines , 2001 ) . Drosophila adducin gene encodes several isoforms , all but one detectable with the anti-Adducin antibody 1B1 ( Wang et al . , 2014 ) . Using this antibody , we found that Adducin is also diminished at tnc mutant terminals ( Figure 6G–I ) . These data suggest that Tnc is required for the proper recruitment of SBMS at the NMJ . Similar to βPS , the accumulation of α-Spectrin at synaptic terminals appeared to be limited by neuronal Tnc and promoted by muscle-derived Tnc , as indicated by knockdown analyses ( Figure 6—figure supplement 1A–D ) . Moreover , the changes in synaptic α-Spectrin levels followed the variations in integrin levels at synaptic terminals , since manipulations of mys/ ( βPS ) elicited a similar profile for the synaptic α-Spectrin ( Figure 6—figure supplement 1E–H ) . Together our data suggest that the postsynaptic Tnc/integrin complexes function to recruit the SBMS at synaptic terminals; this activity seems restricted by neuron-derived Tnc , which appears to limit the accumulation and function of postsynaptic complexes . Complete loss of Tnc triggers severe reduction of integrin and α-Spectrin at synaptic terminals , consistent with the observed disruption of bouton architecture . Since neuron-derived Tnc recruits βPS integrin in the motor neurons to modulate neurotransmitter release ( Figure 3 ) , we next examined whether muscle Tnc functions similarly in cis to recruit postsynaptic integrin and SBMS and ensure bouton integrity . For this rescue experiment we used a wide range of Tnc levels and monitored both βPS accumulation and postsynaptic function ( Ib bouton area ) ( Figure 7A–G ) . Paneuronal expression of Tnc rescued the postsynaptic βPS accumulation at tnc mutant NMJs; however , these animals had small , tnc-like boutons . Thus , neuron-derived Tnc can engage postsynaptic βPS to form non-productive complexes that cannot restore the bouton architecture . When low levels of Tnc were provided in the muscle using a weak promoter and low rearing temperatures ( Figure 7D ) , βPS accumulation was rescued to levels exceeding the control and bouton size was fully restored . In contrast , high levels of Tnc induced massive lethality and exacerbated the loss of βPS at tnc mutant NMJs; these larvae had poorly defined boutons , with almost tubular NMJ branches ( Figure 7E ) . Together these results indicate that only low levels of muscle Tnc could rescue the distribution and function of postsynaptic Tnc/integrin complexes , while excess muscle Tnc is toxic . On the other hand , neuron-derived Tnc can recruit and/or stabilize integrin but cannot form fully functional complexes . In support of this interpretation , we found that only low levels of muscle Tnc could fully restore the accumulation of α-Spectrin at tnc mutant NMJs ( Figure 7H–M ) ; neuronal Tnc induced only a modest increase in synaptic α-Spectrin levels . The fact that the trans Tnc/integrin complexes cannot rescue the bouton size suggests that the cis and trans complexes have different activities . This could be due to different processing or post-translational modifications of Tnc in motor neurons vs . muscles , which have been reported to modulate the activity of ligand-integrin complexes ( Reichardt and Tomaselli , 1991 ) . Our Western blot analyses did not provide clear evidence for different post-translational modifications for neuron- and muscle-derived Tnc , although the large Tnc-specific band appeared to include multiple species only when Tnc was overexpressed in the muscle ( Figure 7—figure supplement 1 ) . Alternatively , neuron- and muscle-derived Tnc may be packaged differently and/or associate with molecules that influence their activities . Indeed , Tnc has multiple vWFC domains and RGD-like motifs that could enable a large repertoire of protein interactions . Our rescue results may also reflect different distributions for neuron- and muscle-derived Tnc: While neuron-secreted Tnc accumulates at synaptic terminals , muscle-secreted Tnc likely distributes throughout the muscle membrane and may sequester integrin away from NMJ locations , further reducing the βPS accumulation at synaptic terminals . To test this possibility , we overexpressed Tnc in an otherwise wild-type genetic background and examined the βPS signals at perisynaptic vs . muscle attachment sites . In these experiments we observed no changes in βPS levels and distribution at muscle attachment sites or costameres ( not shown ) . However , βPS recruitment at larval NMJ was drastically reduced when Tnc was overexpressed in either neurons or muscles ( Figure 8A–E ) . Neuronal excess of Tnc produced a reduction of βPS synaptic levels ( to 54% from control , p=0 . 0002 , n = 23 ) , and activities , as reflected by the reduction of bouton size compared with the control ( transgene only ) ( Figure 8F ) . Excess Tnc in the muscles practically abolished the perisynaptic βPS signals . With strong muscle drivers ( 24B-Gal4 in Figure 8D , or G14-Gal4 in Figure 8—figure supplement 1 ) , these larvae showed ribbon-like NMJs with no interbouton/bouton delimitations exceeding the severity of defects observed at tnc mutant NMJs ( not shown ) . A significant number of these animals died during larval and pupal stages and only 46% ( n = 211 ) of third instar larvae developed into adult flies . The toxicity of excess Tnc during development prompted us to examine the distribution of βPS after a short ( 8 hr ) pulse of Tnc expression in the muscle , using BG487-Gal4 to induce moderate , gradient muscle expression ( Figure 8G–H’ ) . This pulse triggered an increase in Tnc immunoreactivities , which distributed diffusely around the synaptic terminals . Importantly the βPS signals were also drastically diminished , particularly at postsynaptic locations . The βPS signals were no longer concentrated around the synaptic boutons , and instead appeared as thin lines inside the boutons , along the HRP-marked neuronal membrane . Thus , it appears that excess Tnc in the muscle gradually disrupted the postsynaptic βPS accumulation , revealing a small but clear pool of presynaptic βPS; further Tnc overexpression in the muscle completely disrupted both the Tnc and βPS synaptic accumulation and altered the boutons morphology . This dose dependent depletion of Tnc and βPS synaptic signals indicates that excess Tnc may form large aggregates that may be physically excluded from the intercellular space and dispersed away from the synaptic terminals . Alternatively , excess Tnc may trap integrin in the secretory compartment and/or overload a limiting step for the synaptic targeting and recruitment of Tnc/integrin complexes . We favor the former possibility because ( i ) Tnc itself diffused away from the synaptic terminal when in mild excess , ( ii ) Tnc has been previously implicated in the formation of large aggregates that fill the lumen of epithelial organs ( Syed et al . , 2012 ) , and ( iii ) simple disruption of Tnc trafficking and targeting within the muscle cannot explain the loss of presynaptic integrin accumulation and function ( Figure 8D–E and not shown ) . To our knowledge this is the first example where genetic manipulations completely abolished the synaptic accumulation of βPS integrin . These larvae also exhibited drastically reduced levels of synaptic α-Spectrin and Adducin ( Figure 8—figure supplement 1 ) . Together these results indicate that optimal levels of secreted Tnc are required for proper βPS accumulation at the NMJ , which ensures normal NMJ morphology and function . These experiments also uncovered a novel function for βPS integrin in anchoring α-Spectrin at synaptic locations and coupling the ECM of the synaptic cleft ( Tnc ) with the spectrin-based membrane skeleton . To examine whether Tnc directed the recruitment of integrin and spectrin complexes at the cell membrane , we took advantage of our ability to produce full-length Tnc and Tnc-HA proteins in S2 and S2R + insect cells . Secreted Tnc-HA was concentrated from S2 conditioned media , affinity coupled to Neutravidin beads of 1 μm diameter , then presented to S2R + cells , which express relatively high levels of integrin and spectrin . Unlike control beads ( coated with an unrelated HA-tagged protein ) , Tnc-HA-coupled beads induced a local accumulation of βPS at the periphery of S2R + cells , in the close proximity of the beads ( Figure 9A–B ) . This local recruitment of βPS was not caused by mechanical stress , since control beads , either alone or in clusters , did not trigger βPS accumulation . The βPS recruitment was dose-dependent , as beads with variable Tnc-HA levels elicited proportional βPS accumulation ( not shown ) . The Tnc-coupled beads triggered local recruitment of αPS2 but not αPS1 integrin subunit at the surface of S2R+ cells ( Figure 9C–D , and not shown ) . This is consistent with our NMJ observations ( Figure 4 ) and with previous reports on small Tnc fragments mediating αPS2/βPS-dependent spreading of S2 cells via RGD and RGD-like motifs ( Fraichard et al . , 2010 ) . Moreover , the Tnc-HA-coupled beads induced similar accumulation of α-Spectrin and Adducin at the bead-cell membrane interfaces ( Figure 9E–J ) . Thus Tnc triggers the local recruitment of α-Spectrin and Adducin at the cell membrane . Whether this recruitment is mediated either directly by integrin or indirectly , via other integrin activated scaffolds , remains to be determined . Our attempts to knockdown βPS in S2R + cells , and establish a requirement for integrin for Tnc-induced α-Spectrin recruitment , were hampered by an incomplete RNAi knockdown . Nonetheless , our data establish that Tnc provides a powerful means to recruit and/or stabilize integrin and cortical skeleton components at synaptic terminals . Although extracellular Tnc binds integrin on cell surfaces , our genetic analyses indicate that in vivo Tnc functions in cis . This suggests that Tnc positively affects the surface delivery and/or stabilization of βPS integrin . We tested this possibility by following Tnc and βPS levels and distribution in S2R + cells transiently transfected with Tnc ( Figure 9—figure supplement 1 ) . In this system , Tnc or Tnc-HA were efficiently secreted and accumulated in the media . Inside the cells , Tnc marked large aggregates , likely corresponding to secretory compartments . We found that βPS levels were significantly elevated in cells expressing Tnc; also , βPS co-localized with Tnc in large intracellular aggregates , suggesting that ( a ) Tnc promotes βPS secretion , and that ( b ) βPS and Tnc likely associate during intracellular trafficking . It is also possible that excess Tnc may trap βPS inside the secretory compartment , although Tnc appeared to be efficiently secreted and βPS levels were elevated throughout the Tnc-positive cells . Binding to Tnc may stabilize βPS at the cell surface through clustering of integrin complexes and/or by conformational changes and activation of integrins ( Liddington and Ginsberg , 2002 ) . In both cases , Tnc binding could reduce the rate of βPS integrin endocytosis as well as overall turnover ( López-Ceballos et al . , 2016 ) , which may explain the elevated levels of βPS observed in Tnc-expressing cells .
Tnc appears to fulfill unique , complementary functions with the other known synaptic ECM proteins at the Drosophila NMJ . Unlike Mtg , which organizes the active zone matrix and the postsynaptic domains ( Rohrbough et al . , 2007; Rushton et al . , 2009 ) , Tnc does not influence the recruitment of iGluRs and other PSD components . LanA ensures a proper adhesion between the motor neuron terminal and muscle ( Koper et al . , 2012 ) and also acts retrogradely to suppress the crawling activity-dependent NMJ growth ( Tsai et al . , 2012 ) . The latter function requires the presynaptic βν integrin subunit and phosphorylation of Fak56 via a pathway that appears to be completely independent of Tnc . Several more classes of trans-synaptic adhesion molecules have been implicated in either the formation of normal size synapses , for example Neurexin/Neuroligin , or in bridging the pre- and post-synaptic microtubule-based cytoskeleton , such as Teneurins ( Banovic et al . , 2010; Mosca et al . , 2012 ) . However , genetic manipulation of Tnc did not perturb synapse assembly or microtubule organization , indicating that Tnc functions independently from these adhesion molecules . Instead , Tnc appears to promote expression and stabilization of αPS2/βPS complexes , which in turn engage the spectrin-based membrane skeleton ( SBMS ) at synaptic terminals . On the presynaptic side these complexes modulate neurotransmitter release . On the postsynaptic side , the Tnc-mediated integrin and spectrin recruitment modulates bouton morphology . A similar role for integrin and spectrin in maintaining tissue architecture has been reported during oogenesis; egg chambers with follicle cells mutant for either integrin or spectrin produce rounder eggs ( Bateman et al . , 2001; Ng et al . , 2016 ) . Our data are consistent with a local function for the Tnc/βPS-recruited SBMS at synaptic terminals; this is distinct from the role of spectrin in endomembrane trafficking and synapse organization ( Kizhatil et al . , 2007; Lorenzo et al . , 2010; Tjota et al . , 2011 ) . Embryos mutant for spectrins have reduced neurotransmitter release ( Featherstone et al . , 2001 ) , a phenotype shared by larvae lacking presynaptic Tnc or βPS integrin ( Figures 3 and 4 ) . However , Tnc perturbations did not induce synapse retraction and axonal transport defects as seen in larvae with paneuronal α- or β- spectrin knockdown ( not shown ) ( Pielage et al . , 2005 ) . Spectrins interact with ankyrins and form a lattice-like structure lining neuronal membranes in axonal and interbouton regions ( Koch et al . , 2008; Pielage et al . , 2008; Goellner and Aberle , 2012 ) . We found that Tnc manipulations did not affect the distribution of Ankyrin two isoforms ( Ank2-L and Ank2-XL ) in axons or at the NMJ ( not shown ) ; also loss of ankyrins generally induces boutons swelling , whereas Tnc perturbations shrink the boutons and erode bouton-interbouton boundaries . Like tnc , loss of spectrins in the striated muscle shows severe defects in SSR structure ( [Pielage et al . , 2006] and Figure 5 ) . Lack of spectrins also disrupts synapse assembly and the recruitment of glutamate receptors ( Pielage et al . , 2005; Pielage et al . , 2006 ) . In contrast , manipulations of tnc had no effect on PSD size and composition ( Figure 5—figure supplement 1 ) . Instead , tnc perturbations in the muscle led to boutons with altered size and individualization and resembled the morphological defects seen in spectrin tetramerization mutants , spectrinR22S ( Khanna et al . , 2015 ) . spectrinR22S mutants have more subtle defects than tnc , probably because spectrin is properly recruited at NMJs but fails to crosslink and form a cortical network . Spectrins are also recruited to synaptic locations by Teneurins , a pair of transmembrane molecule that form trans-synaptic bridges and influence NMJ organization and function ( Mosca et al . , 2012 ) . Drosophila Ten-m has an RGD motif; we found that βPS levels were decreased by 35% at ten-mMB mutant NMJs ( not shown ) . Thus , Ten-m may also contribute to the recruitment of integrin and SBMS at the NMJ , a function likely obscured by the predominant role both play in cytoskeleton organization . We have previously reported that α-Spectrin is severely disrupted at NMJs with suboptimal levels of Neto , such as neto109- a hypomorph with 50% lethality ( Kim et al . , 2012 ) . These mutants also had sparse SSR , reduced neurotransmitter release , as well as reduced levels of synaptic βPS ( [Kim et al . , 2012] and not shown ) . In this genetic background , lowering the dose of tnc should further decrease the capacity to accumulate integrin and spectrin at synaptic terminals and enhance the lethality . This may explain the increased synthetic lethality detected in our genetic screen . In flies or vertebrates , the ECM proteins that comprise the synaptic cleft at the NMJ are not fully present when motor neurons first arrive at target muscles . Shortly thereafter , the neurons , muscles and glia begin to synthesize , secrete and deposit ECM proteins . At the vertebrate NMJ , deposition of the ECM proteins forms a synaptic basal lamina that surrounds each skeletal myofiber and creates a ~ 50 nm synaptic cleft . In flies , basal membrane contacts the motor terminal in late embryos , but is some distance away from the synaptic boutons during larval stages ( Prokop et al . , 1998 ) . Nonetheless , the NMJ must withstand the mechanical tensions produced by muscle contractions . Our data suggest that Tnc is an ideal candidate to perform the space filling , pressure inducing functions required to engage integrin ( Pines et al . , 2012 ) and establish a dynamic ECM-cell membrane network at synaptic terminals . First , Tnc is a large mucin with extended PTS domains that become highly O-glycosylated , bind water and form gel-like complexes that can extend and induce effects similar to hydrostatic pressure ( Syed et al . , 2008 ) . In fact , Tnc fills the lumen of several epithelial tubes and forms a dense matrix that acts in a dose-dependent manner to drive diameter growth . Second , the RGD and RGD-like motifs of Tnc have been directly implicated in αPS2/βPS-dependent spreading of S2 cells ( Fraichard et al . , 2010 ) . Third , secreted Tnc appears to act close to the source ( Syed et al . , 2012 ) , presumably because of its size and multiple interactions . In addition to the RGD motifs , Tnc also contains five complete and one partial vWFC domains , that mediate protein interactions and oligomerization in several ECM proteins including mucins , collagens , and thrombospondins ( Bork , 1991 ) . The vWFC domains are also found in growth factor binding proteins and signaling modulators such as Crossveinless-2 and Kielin/Chordin ( Lin et al . , 2005; Wharton and Serpe , 2013 ) suggesting that Tnc could also influence the availability of extracellular signals . Importantly , Tnc expression is hormonally regulated during development by ecdysone ( Fraichard et al . , 2010 ) . Tnc does not influence integrin responsiveness to axon guidance cues during late embryogenesis; unlike integrins , the tnc mutant embryos have normal longitudinal axon tracks ( Figure 1—figure supplement 1 and [Stevens and Jacobs , 2002] ) . Instead , Tnc synthesis and secretion coincide with the NMJ expansion and formation of new bouton structures during larval stages . Recent studies have reported several mucin-type O-glycosyltransferases that modulate integrin signaling and intercellular adhesion in neuronal and non-neuronal tissues , including the Drosophila NMJ ( Tran and Ten Hagen , 2013; Dani et al . , 2014 ) . Tnc is likely a substrate for these enzymes that may further regulate Tnc activities . In flies as in vertebrates , integrins play essential roles in almost all aspects of synaptic development . Early in development , integrins have been implicated in axonal outgrowth , pathfinding and growth cone target selection ( Myers et al . , 2011 ) . In adult flies , loss of αPS3 integrin activity is associated with the impairment of short-term olfactory memory ( Grotewiel et al . , 1998 ) . In vertebrates , integrin mediates structural changes involving actin polymerization and spine enlargement to accommodate new AMPAR during LTP , and ‘lock in’ these morphological changes conferring longevity for LTP ( McGeachie et al . , 2011 ) . Thus far , integrin functions at synapses have been derived from compound phenotypes elicited by use of integrin mutants , RGD peptides , or enzymes that modify multiple ECM molecules ( Rohrbough et al . , 2000; McGeachie et al . , 2011; Dani et al . , 2014 ) . Such studies have been complicated by multiple targets for modifying enzymes and RGD peptides and by the essential functions of integrin in cell adhesion and tissue development . In contrast , manipulations of Tnc , which affects the selective recruitment of αPS2/βPS integrin at synaptic terminals , have uncovered novel functions for integrin and clarified previous proposals . We demonstrate that βPS integrin is dispensable for the recruitment of iGluRs at synaptic sites and for PSD maintenance . We reveal an unprecedented role for integrin in connecting the ECM of the synaptic cleft with spectrin , in particular to the spectrin-based membrane skeleton . These Tnc/integrin/spectrin complexes are crucial for the integrity and function of synaptic structures . Our studies uncover the ECM component Tnc as a novel modulator for NMJ development and function; these studies also illustrate how manipulation of a selective integrin ligand could be utilized to reveal novel integrin functions and parse the many roles of integrins at synaptic junctions .
To generate the tnc82 allele , transposons PBac[RB]e01245 and PBac[WH]f00323 were mobilized by FRT-induced recombination as previously described ( Parks et al . , 2004; Thibault et al . , 2004 ) . The UAS-tnc lines were generated by insertion of the tnc cDNA in pUAST vector and germline transformation ( BestGene , Inc . ) . The following fly lines were obtained from the Bloomington Stock Center: the P-element lines tncEP ( P[EPgy2]EY03355 ) and ten-m mutant Mi[ET1]Ten-mMB10734 ( Bellen et al . , 2004 ) ; the deficiency lines Df ( 3R ) BSC-318 , , –492 , −494 , and −655 ( Parks et al . , 2004 ) ; and the TRIP lines UAS-tncRNAi ( P[TRIP . HMC05051] attP40 ) , and UAS-mysRNAi ( P[TRIP . JF02819] attP2 ) , and UAS-if RNAi ( P[TRiP . HMS01872]attP40 ) . Additional tnc , mys and if RNAi lines showed similar phenotypes , but were not reported here because of mild effects ( such as for UAS-tncRNAi line from Vienna Drosophila RNAi Center , P[GD14952]v42326 ) , or aberrant muscle development ( in the case of mys and if knockdown ) . Other fly stocks used in this study were previously described: neto109 ( Kim et al . , 2012 ) ; UAS-Dlg ( Budnik et al . , 1996 ) ; 24B-Gal4 , elav-Gal4 , G14-Gal4 , BG487-Gal4 . Full-length cDNA for tnc was assembled in pRM-Tlr plasmid ( Serpe and O'Connor , 2006 ) from the following elements: ( 1 ) a synthetic PinAI/SalI fragment joining the Tolloid-related signal peptide with the 5’ of tnc CDS; ( 2 ) SalI 7 kb fragment isolated from BACPAC CH322-177A22 ( Venken et al . , 2009 ) ; ( 3 ) SalI/XhoI PCR product covering the 3’ of tnc CDS . We used this plasmid to introduce two HA tags in two steps: ( 1 ) replace the SalI 7 kb fragment with a pair of primers that introduced 2xHA tags separated by a SalI site; ( 2 ) reintroduce SalI kb fragment . The primers used here were: For-5’-Phos- TCGAGCTATCCCTATGACGTCCCGGACTATGCACAGTCGACTACCCGTACGATGTGCCCGATTACGCAC and Rev: 5’-Phos- TCGAGTGCGTAATCGGGCACATCGTACGGGTAGTCGACTGTGCATAGTCCGGGACGTCATAGGGATAGC . All constructs were verified by DNA sequencing . UAS-tnc lines were generated by insertion of the tnc cDNA in pUAST and germline transformation ( BestGene , Inc . ) . To measure the climbing ability , a custom device to fractionate populations based on negative geotaxis was used ( Benzer , 1967 ) . The flies were placed in the first tube ( numbered 0 ) and left to recover for 2 hr . The fractionation consisted of sequential cycles of tapping down the flies , and moving the flies that climbed above the threshold in 15 s to the next tube . The climbing index was calculated by the formula ( 1xN1 + 2xN2 + 3xN3 + 4xN4 + 5xN5 ) /N , where Nr is the number of flies in fraction r , and N is the total number of flies . Drosophila S2 and S2R+ cells were used for expression of recombinant proteins and immunohistochemistry as described previously ( Serpe et al . , 2008 ) . For protein analysis , wandering third instar larvae were dissected , and brains or the body wall ( muscle and cuticle ) were isolated . The tissues were mechanically disrupted and lysed in lysis buffer ( 50 mM Tris-HCl , 150 mM NaCl , 1% Triton X-100 , 1% deoxycholate , protease inhibitor cocktail ( Roche ) for 30 min on ice . The lysates were separated by SDS-PAGE on 4–12% NuPAGE gels ( Invitrogen ) and transferred onto PVDF membranes ( Millipore ) . The rabbit polyclonal anti-Tnc antibodies were generated as previously described ( Fraichard et al . , 2006 ) against a synthetic peptide ( APVQEYTEIQQYSEGC ) ( Pacific Immunology Corp . ) and affinity purified . Primary antibodies were used at the following dilutions: rabbit anti-Tnc 1:1 , 000; anti-Tubulin ( Sigma-Aldrich ) , 1:1000; mouse anti-α-Spectrin ( 3A9 ) , 1:50 . Immune complexes were visualized using secondary antibodies coupled with IR-Dye 700 or IR-Dye 800 followed by scanning with the Odyssey infrared imaging system ( LI-COR® Biosciences ) . Neutravidin beads of 1 μm diameter ( Life Technologies ) were washed in PBS and preloaded for 30 min at room temperature ( RT ) with biotin conjugated anti-HA antibody ( clone 3F10 , Roche ) . An unrelated HA-tagged protein , Tolloid ( Tld ) , was used as a control ( Serpe and O'Connor , 2006 ) . Tld-HA and Tnc-HA were concentrated from S2 conditioned media using Amicon Ultra filters ( 10 kDa ) and bound to the preloaded beads overnight . The beads were washed with PBS and presented for 1 . 5 hr to S2R + cells cultured on chambered coverglass ( Thermo Fisher ) at 26°C . Cells were next fixed with 4% PFA for 20 min and stained using standard procedures as described below . Wandering third instar larvae were dissected as described previously in ice-cooled Ca2+-free HL-3 solution ( Stewart et al . , 1994; Budnik et al . , 2006 ) . Embryos were collected and fixed using standard procedures ( Patel , 1994 ) . Primary antibodies from Developmental Studies Hybridoma Bank were used at the following dilutions: mouse anti-GluRIIA ( 8B4D2 ) , 1:100; mouse anti-Dlg ( 4F3 ) , 1:1000; mouse anti-Brp ( Nc82 ) , 1:200; mouse anti-α-Spectrin ( 3A9 ) , 1:50; mouse anti-FasII ( 1D4 ) , 1:10; mouse anti-Futsch ( 22C10 ) , 1:100; mouse anti-Adducin ( 1B1 ) , 1:50; mouse anti-βPS integrin ( CF . 6G11 ) 1:10; mouse anti-αPS1 integrin ( DK . 1A4 ) 1:10; mouse anti-αPS2 integrin ( CF . 2C7 ) 1:10 . Other primary antibodies were utilized as follow: rabbit anti-Tnc 1:100; rat anti-HA ( 3F10 ) ( Lu and Roche , 2012 ) 1:500; rabbit anti-FAK ( phospho Y397 ) ( Abcam , ab39967 ) , 1:100; rabbit anti-GluRIIB , 1:2000; rabbit anti-GluRIIC , 1:1000 ( Ramos et al . , 2015 ) ; rabbit anti-PAK , 1:5000 ( a gift from Nicholas Harden ) ( Conder et al . , 2004 ) ; 1:1000; rat anti-Neto , 1:1000 ( Kim et al . , 2012 ) ; and Cy5- conjugated goat anti-HRP , 1:1000 ( Jackson ImmunoResearch Laboratories , Inc . ) . Alexa Fluor ( AF ) 405- , AF488- , AF568- , and AF647- conjugated secondary antibodies ( Molecular Probes ) were used at 1:200 . All samples were mounted in ProLong Gold ( Invitrogen ) . Samples of different genotypes were processed simultaneously and imaged under identical confocal settings in the same imaging session with a laser scanning confocal microscope ( CarlZeiss LSM780 ) . All images were collected as 0 . 2 μm optical sections and the z-stacks were analyzed with Imaris software ( Bitplane ) . To quantify fluorescence intensities synaptic ROI areas surrounding anti-HRP immunoreactivities were selected and the signals measured individually at NMJs ( muscle 4 , segment A3 ) from ten or more different larvae for each genotype ( number of samples is indicated in the graph bar ) . The signal intensities were calculated relative to HRP volume and subsequently normalized to control . Boutons were counted in preparations double labeled with anti-HRP and anti-Dlg . Bouton size was estimated by using the ImageJ software . All quantifications were performed while blinded to genotype . The numbers of samples analyzed are indicated inside the bars . Statistical analyses were performed using the Student t-test with a two-tailed distribution and a two-sample unequal variance . Error bars in all graphs indicate standard deviation ±SEM . ***p<0 . 001 , **p<0 . 005 , *p<0 . 05 , ns- p>0 . 05 . PLA was performed following published protocols ( Wang et al . , 2015 ) . In brief , wandering third instar larvae were dissected , fixed and incubated with primary antibodies ( anti-Tnc and anti-βPS integrin ) overnight at 4°C . To detect the markers , the samples were incubated with AF-conjugated secondary antibodies ( anti-rabbit- AF488 , anti-mouse-AF405 and anti-HRP-AF647 ) for 1 hr in the dark . For PLA , the samples were washed with Wash Buffer A ( DUO 92101 Kit , Sigma ) , incubated first with PLA probe anti-mouse MINUS and PLA probe anti-rabbit PLUS ( 1:5 dilution ) for 2 hr at 37°C , then with 200 μl Ligation solution for 1 hr at 37°C , and finally with Amplification solution for 2 hr at 37°C . After washes in Wash Buffer B , the samples were mounted in ProLong Gold ( Invitrogen ) and examined by confocal imaging . The standard larval body wall muscle preparation first developed by Jan and Jan ( 1976 ) ( Jan and Jan , 1976 ) was used for electrophysiological recordings ( Zhang and Stewart , 2010 ) . Wandering third instar larvae were dissected in physiological saline HL-3 saline ( Stewart et al . , 1994 ) , washed , and immersed in HL-3 containing 0 . 5 or 0 . 8 mM Ca2+ using a custom microscope stage system ( Ide , 2013 ) . The nerve roots were cut near the exiting site of the ventral nerve cord so that the motor nerve could be picked up by a suction electrode . Intracellular recordings were made from muscle 6 , abdominal segment 3 and 4 . Data were used when the input resistance of the muscle was >5 MΩ and the resting membrane potential was between −60 mV and −70 mV . The input resistance of the recording microelectrode ( backfilled with 3 M KCl ) ranged from 20 to 25 MΩ . Muscle synaptic potentials were recorded using Multiclamp 700B amplifiers ( Molecular Devices ) and Axon Clamp 2B amplifier ( Axon Instruments ) and analyzed using pClamp 10 software . Following motor nerve stimulation with a suction electrode ( 200 μsec , 1 . 9 V ) , evoked EJPs were recorded . Four to six EJPs evoked by low frequency of stimulation ( 0 . 1 Hz ) were averaged . For mini recordings , TTX ( 1 μM ) was added to prevent evoked release ( Stewart et al . , 1994 ) . To calculate mEJP mean amplitudes , 100 events from each 10 or more NMJs ( only one NMJ per animal was used ) were measured and averaged using the Mini Analysis program ( Synaptosoft ) . Minis with a slow rise and falling time arising from neighboring electrically coupled muscle cells were excluded from analysis ( Gho , 1994; Zhang et al . , 1998 ) . Paired stimuli ( 200 μsec , 1 . 9 V , 0 . 05 Hz ) were applied with a suction electrode at 50 ms inter-stimulus intervals . The amplitude of the eEJP was determined as an average from 5 to 8 steady consecutive sweeps . The paired-pulse ratio ( PPR ) was expressed as the amplitude ratio of the second synaptic response to the first synaptic response ( Zhang and Stewart , 2010 ) . Quantal content was calculated by dividing the mean EJP by the mean mEJP after correction of EJP amplitude for nonlinear summation according to previously described methods ( Stevens , 1976; Feeney et al . , 1998 ) . Corrected EJP amplitude = E[Ln[E/ ( E - recorded EJP ) ]] , where E is the difference between reversal potential and resting potential . The reversal potential used in this correction was 0 mV ( Feeney et al . , 1998; Lagow et al . , 2007 ) . Statistical analysis was performed with KaleidaGraph 4 . 5 ( Synergy Software ) using ANOVA followed by a Tukey post hoc test . Differences were considered significant at p<0 . 05 . Data are presented as mean ±SEM . Wandering third instar larvae were dissected in physiological saline HL-3 saline and fixed for 30 min on dissection plate in fixation buffer ( 0 . 1 M Sodium Cacodylate buffer , pH7 . 2; 2 mM MgCl2; 1% glutaraldehyde; 4% paraformaldehyde ) . The samples were trimmed to include only the abdominal segments A2 and A3 , transferred in a 1 . 5 mL Eppendorf tube , fixed overnight at 4 ˚C , then washed extensively with 0 . 1 M Sodium Cacodylate buffer with 132 mM Sucrose , pH 7 . 2 . The samples were further processed and analyzed according to published protocols ( Ramachandran and Budnik , 2010 ) at the Microscopy and Imaging Facility , NICHD . | Nerve cells or neurons can communicate with each other by releasing chemical messengers into the gap between them , the synapse . Both neurons and synapses are surrounded by a network of proteins called the extracellular matrix , which anchors , protects and supports the synapse . The matrix also helps to regulate the dynamic communication across the synapses and consequently neurons . Little is known about the proteins of the extracellular matrix , in particular about the ones involved in structural support . This is especially important for the so-called neuromuscular junctions , where neurons stimulate muscle contraction and trigger vigorous movement . Receptor proteins on cell surfaces , such as integrins , can bind to the extracellular matrix proteins to anchor the cells and are important for all cell junctions , including synaptic junctions . But because of their many essential roles during development , it was unclear how integrins modulate the activity of the synapse . To investigate this further , Wang et al . studied the neuromuscular junctions of fruit flies . The experiments revealed that both muscle and neurons secrete a large protein called Tenectin , which accumulates into the small space between the neuron and the muscle , the synaptic cleft . This protein can bind to integrin and is necessary to support the neuromuscular junction structurally and functionally . Wang et al . discovered that Tenectin works by gathering integrins on the surface of the neuron and the muscle . In the neuron , Tenectin forms complexes with integrin to regulate the release of neurotransmitters . In the muscle , the complexes provide support to the synaptic structures . However , when Tenectin was experimentally removed , it only disrupted the integrins at the neuromuscular junction , without affecting integrins in other regions of the cells , such as the site where the muscle uses integrins to attach to the tendon . Moreover , without Tenectin an important intracellular scaffolding meshwork that lines up and reinforces cell membranes was no longer organized properly at the synapse . A next step will be to identify the missing components between Tenectin/integrin complexes on the surface of neurons and the neurotransmitter release machinery inside the cells . The extracellular matrix and its receptors play fundamental roles in the development and function of the nervous system . A better knowledge of the underlying mechanisms will help us to better understand the complex interplay between the synapse and the extracellular matrix . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2018 | Tenectin recruits integrin to stabilize bouton architecture and regulate vesicle release at the Drosophila neuromuscular junction |
SNARE proteins play a crucial role in intracellular trafficking by catalyzing membrane fusion , but assigning SNAREs to specific intracellular transport routes is challenging with current techniques . We developed a novel Förster resonance energy transfer-fluorescence lifetime imaging microscopy ( FRET-FLIM ) -based technique allowing visualization of real-time local interactions of fluorescently tagged SNARE proteins in live cells . We used FRET-FLIM to delineate the trafficking steps underlying the release of the inflammatory cytokine interleukin-6 ( IL-6 ) from human blood-derived dendritic cells . We found that activation of dendritic cells by bacterial lipopolysaccharide leads to increased FRET of fluorescently labeled syntaxin 4 with VAMP3 specifically at the plasma membrane , indicating increased SNARE complex formation , whereas FRET with other tested SNAREs was unaltered . Our results revealed that SNARE complexing is a key regulatory step for cytokine production by immune cells and prove the applicability of FRET-FLIM for visualizing SNARE complexes in live cells with subcellular spatial resolution .
One of the central paradigms in cell biology is that all intracellular membrane fusion , except for mitochondrial fusion , is catalyzed by soluble NSF ( N-ethylmaleimide-sensitive fusion protein ) attachment protein receptor ( SNARE ) proteins ( Hong , 2005; Jahn and Scheller , 2006 ) . There are about 36 SNARE proteins identified in mammals which can be classified in R-SNAREs or Q-SNAREs depending on the central residue of their SNARE motif being either arginine ( R ) or glutamine ( Q ) . The interaction of one R motif and three Q motifs ( termed Qa , Qb and Qc ) of cognate SNARE proteins located in opposing membranes leads to the formation of a 4-helix coiled-coil bundle . This trans-SNARE complex formation brings the membranes in close proximity and catalyzes their fusion . After membrane fusion is complete , the SNAREs are in a so-called cis-conformation and need to be disassembled by the AAA-ATPase NSF in conjunction with α-SNAP ( Söllner et al . , 1993; Hong , 2005; Jahn and Scheller , 2006 ) . From studies in yeast and many mammalian cell types , including neurons , neuroendocrine cells , adipocytes and epithelial cells , it is clear that different intracellular trafficking steps are catalyzed by specific sets of SNARE proteins ( Pelham , 2001; Hong , 2005; Jahn and Scheller , 2006 ) . Also in immune cells , such as granulocytes , platelets , mast cells , macrophages and T cells , specific combinations of SNARE proteins mediate release of cytokines , chemokines and delivery of surface receptors via specific secretory pathways ( Collins et al . , 2015; Stow et al . , 2009; Stanley and Lacy , 2010; Lacy and Stow , 2011; Murray and Stow , 2014 ) . In order to identify the combinations of SNARE proteins responsible for a particular intracellular trafficking step , most studies rely on localization microscopy experiments either with endogenous or overexpressed SNAREs . However , these approaches suffer from the problems that many SNAREs locate to multiple organelles , as SNAREs are promiscuous and can be involved in different trafficking steps , and that mere co-localization of SNAREs does not prove their actual interaction ( Pelham , 2001; Hong , 2005 ) . Interactions can be shown with co-immunoprecipitation experiments , but these cannot resolve in which organelle the interactions take place . Perturbation experiments , such as genetic ablation of SNAREs , overexpression of soluble ‘dominant negative’ SNARE fragments , or microinjection of antibodies directed against SNAREs , are also difficult to interpret because SNAREs are functionally redundant and defects in a specific trafficking route cannot be discerned from upstream trafficking steps . Because of this , knockdown or knockout of many SNARE proteins often does not show a clear phenotype ( Hong , 2005; Bethani et al . , 2009; Nair-Gupta et al . , 2014b ) . Thus , a method that allows quantitative visualization of SNARE complexes with subcellular resolution is highly desirable . In this study , we aimed to fill this gap in methodology by developing an assay to locally detect SNARE interactions with Förster resonance energy transfer ( FRET ) based fluorescence lifetime imaging microscopy ( FLIM ) ( Jares-Erijman and Jovin , 2003; Wallrabe and Periasamy , 2005 ) . FRET is based on the energy transfer from a donor to an acceptor fluorophore within the Förster distance range . FRET leads to quenching of the donor fluorophore and sensitization of the acceptor fluorophore and can either be measured from fluorescence intensities ( ratiometric FRET ) or from a decrease in fluorescence lifetime of the donor fluorophore ( FRET-FLIM ) . Ratiometric FRET and FRET-FLIM have been previously used to characterize cytosolic interactions of fungal SNAREs ( Valkonen et al . , 2007 ) and of the neuronal Qa-SNARE syntaxin 1 ( Stx1 ) and the Qb/c-SNARE SNAP25 ( Rickman et al . , 2007; Medine et al . , 2007; Rickman et al . , 2010; Xia et al . , 2001; Kavanagh et al . , 2014 ) . We hypothesized that FRET-FLIM could also be used to measure full ternary cis-SNARE complex formation , as the crystal structure of the neuronal SNARE complex revealed that the luminal/extracellular C-termini of the R-SNARE vesicle-associated membrane protein 2 ( VAMP2 ) and of Stx1 are in immediate proximity ( <1 nm ) after membrane fusion ( i . e . , in the cis-SNARE complex ) ( Stein et al . , 2009 ) . In vitro , FRET between these SNAREs can be measured by labeling their C-termini by means of site-specific labeling with organic fluorophores ( Xia et al . , 2001 ) . In vivo , Degtyar et al . ( 2013 ) showed interactions between VAMP2 and Stx1 C-terminally fused to fluorescent proteins by total internal reflection fluorescence ( TIRF ) microscopy combined with ratiometric FRET ( Degtyar et al . , 2013 ) . However , ratiometric FRET is technically challenging as extensive controls are required to correct for local concentration differences of the donor and acceptor fluorophores ( Jares-Erijman and Jovin , 2003; Wallrabe and Periasamy , 2005 ) . FLIM does not suffer from this limitation , as the lifetime ( τ ) is an intrinsic characteristic of a fluorophore which is not influenced by the probe concentration nor by the excitation intensity ( Wallrabe and Periasamy , 2005; Jares-Erijman and Jovin , 2003 ) . Occurrence of FRET leads to quenching of the donor signal , which in turn shortens its fluorescence lifetime . Therefore , reductions in the donor’s lifetime reflect interactions between the proteins conjugated to the donor and acceptor fluorophores . In this study , we demonstrate the applicability of our FRET-FLIM assay to visualize the complexes of several SNAREs in dendritic cells of the immune system . Dendritic cells are leukocytes essential for the activation of T cells ( Banchereau and Steinman , 1998 ) . Activation of dendritic cells by inflammatory or pathogenic stimuli triggers the production and release of cytokines and chemokines that orchestrate the immune response ( Collins et al . , 2015; Stanley and Lacy , 2010 ) . Despite many studies characterizing the cytokines that are released by dendritic cells and their mechanisms of action , not much is known about the trafficking pathways involved . Using FRET-FLIM , we measured how activation of dendritic cells affects the interactions between the R-SNAREs vesicle-associated membrane protein 3 ( VAMP3 ) and VAMP8 with the Qa-SNAREs syntaxin 3 ( Stx3 ) and syntaxin 4 ( Stx4 ) . While both VAMP3 and VAMP8 are associated with cytokine release from immune cells ( Collins et al . , 2015; Stow et al . , 2009; Stanley and Lacy , 2010; Lacy and Stow , 2011; Murray and Stow , 2014 ) , VAMP3 is mostly associated with early and recycling endosomes and VAMP8 with late endosomes ( Bajno et al . , 2000; Manderson et al . , 2007; Hong , 2005; Murray et al . , 2005; Antonin et al . , 2000 ) . VAMP8 has also been reported to inhibit phagocytosis by dendritic cells ( Ho et al . , 2008 ) . Stx3 and Stx4 are mainly plasma membrane localized SNAREs mediating exocytosis , but also have intracellular roles in immune cells ( Naegelen et al . , 2015; Collins et al . , 2014; Stow et al . , 2009; Stanley and Lacy , 2010; Lacy and Stow , 2011; Murray and Stow , 2014; Gómez-Jaramillo et al . , 2014; Frank et al . , 2011 ) . Our FRET-FLIM experiments revealed FRET of fluorescently labeled Stx3 with VAMP3 in live dendritic cells and this was predominantly present at the plasma membrane , indicative of increased Stx3-VAMP3 complexing at this site . In contrast , FRET of fluorescently labeled Stx3 with VAMP8 was higher at intracellular compartments . Our results also showed an increase in FRET of fluorescently labeled Stx4 , but not of Stx3 , with VAMP3 upon cellular activation with the Toll-like receptor 4 ( TLR4 ) agonist lipopolysaccharide ( LPS ) , indicating that LPS promotes the complex formation of Stx4 with VAMP3 . LPS activation of dendritic cells induces an inflammatory response , which is characterized by the secretion of inflammatory cytokines such as interleukin-6 ( IL-6 ) ( Stow et al . , 2009 ) . VAMP3 plays a key role in this cytokine secretion , as siRNA knockdown of VAMP3 resulted in impaired release of IL-6 from dendritic cells . Our study demonstrates that the secretion of IL-6 is regulated at the level of SNARE complex formation and proves the use of FRET-FLIM for quantitative visualization of SNARE interactions in live cells .
We generated variants of VAMP3 and Stx3 with their C-termini conjugated to mCitrine ( Griesbeck et al . , 2001 ) and mCherry ( Shaner et al . , 2004 ) , respectively . The emission spectrum of mCitrine overlaps with the excitation spectrum of mCherry which makes them a suitable donor-acceptor pair for FRET ( Figure 1—figure supplement 1A ) . Both mCitrine and mCherry are pH-insensitive and therefore their fluorescence lifetimes are not affected by low pH within the lumen of intracellular compartments ( Griesbeck et al . , 2001; Shaner et al . , 2004 ) . In addition , they are monomeric which is important to prevent oligomerization artifacts . Modeling the crystal structure of the neuronal cis-SNARE complex ( Stein et al . , 2009 ) with mCitrine ( Ho et al . , 2008 ) and mCherry ( Shu et al . , 2006 ) fused to C-termini of VAMP2 and Stx1 showed that these fluorophores are immediately juxtaposed to each other ( Figure 1A ) . Based on the high structural homology of SNARE proteins , we expected that mCitrine and mCherry attached to the C-termini of VAMP3 or VAMP8 and Stx3 or Stx4 would also be in close proximity in a cis-SNARE complex and that this would result in FRET and a decrease of the donor ( mCitrine ) fluorescence lifetime ( Figure 1B ) . 10 . 7554/eLife . 23525 . 003Figure 1 . SNARE complex formation by FRET-FLIM . ( A ) Model of the neuronal SNAREs ( crystal structure; protein database 3HD7 [Stein et al . , 2009] ) with the C-termini of syntaxin 1 ( red ) conjugated to mCitrine ( 3DQ1 [Ho et al . , 2008]; yellow ) and VAMP2 ( blue ) conjugated to mCherry ( 2H5Q [Shu et al . , 2006]; magenta ) . mCitrine ( donor fluorophore ) and mCherry ( acceptor ) are within 3 nm proximity resulting in FRET . Green: SNAP25 . ( B ) Scheme of membrane fusion resulting in FRET . ( C ) Representative confocal microscopy ( left ) and FLIM ( right ) images of dendritic cells expressing Stx3-mCitrine ( green in merge; upper panels ) , Stx3-mCitrine with VAMP3-mCherry ( magenta; middle panels ) , or Stx3 conjugated to both mCitrine and mCherry ( Stx3-mCitrine-mCherry; lower panels ) . Apparent fluorescence lifetimes of Stx3-mCitrine with VAMP3-mCherry were lowest at the cell membrane ( yellow arrowheads ) . Scale bars , 10 µm . Full lifetime/intensity lookup table and lifetime images are in Figure 1—figure supplement 1D–E . DOI: http://dx . doi . org/10 . 7554/eLife . 23525 . 00310 . 7554/eLife . 23525 . 004Figure 1—figure supplement 1 . Lifetime images , confocal images and phasor analysis belonging to main Figure 1 . ( A ) The excitation ( open curves ) and emission ( filled curves ) spectra of mCitrine ( yellow ) and mCherry ( pink ) ( data from Chroma Technology ) . ( B ) Representative confocal microscopy images of antibody staining of endogenous VAMP3 ( top; magenta in merge ) and VAMP8 ( bottom ) in dendritic cells . Blue: DAPI . BF: bright-field . ( C ) Distribution of total photon counts for all fluorescence lifetime recordings . ( D ) Full lookup table belonging to the FLIM images . Plotted is the fluorescence intensity ( y-axis ) versus the fluorescence lifetime ( x-axis ) . ( E ) Fluorescence lifetime images belonging to main Figure 1C . FLIM images were generated by convolution of these lifetime images with the fluorescence intensities ( i . e . , the mCitrine images shown in the main figure ) . ( F ) Phasor plots for the cells shown in main Figure 1C . Cyan circles: center of the phasor location for Stx3-mCitrine in absence of acceptor fluorophore ( upper graph ) ; purple circles: center for Stx3-mCitrine with VAMP3-mCherry ( middle graph ) ; pink circle: center for Stx3-mCitrine-mCherry ( lower graph ) . ( G ) Representative confocal microscopy images of live ( top ) and fixed ( bottom ) dendritic cells expressing the Stx3-mCitrine-mCherry tandem construct ( mCitrine: green in merge; mCherry: magenta ) . ( H ) Pearson correlation coefficients for colocalization of the fluorophores of the Stx3-mCitrine-mCherry tandem construct for 3 donors ( average ± SEM; paired two-sided Student’s t-test ) . ( I–J ) Same as panel G , but now for live ( I ) and fixed ( J ) cells co-expressing VAMP3-mCitrine with VAMP3-mCherry ( left ) or VAMP8-mCitrine with VAMP8-mCherry ( right ) . ( K ) Pearson correlation coefficients for panels I–J ( average ± SEM; VAMP3: p=0 . 046 , paired two-sided Student’s t-test ) . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 23525 . 004 Human blood-isolated monocyte-derived dendritic cells were transfected with either Stx3-mCitrine alone or in combination with VAMP3-mCherry ( Figure 1C ) . Stx3-mCitrine localized to the plasma membrane and to intracellular compartments . VAMP3-mCherry was predominantly present in intracellular compartments with only a small fraction located at the plasma membrane similar to the localization of endogenous VAMP3 ( Figure 1—figure supplement 1B ) . We recorded fluorescent lifetime images of live ( unfixed ) cells at an image plane locating at the height of the nucleus and accumulated on average ~750 , 000 photons per cell ( Figure 1—figure supplement 1C ) . The fluorescence life time histograms of single pixels were fitted with a mono-exponential decay function to generate FLIM images showing the apparent lifetime of each pixel ( Figure 1C; Figure 1—figure supplement 1D–E ) . Compared to cells expressing only Stx3-mCitrine , the FLIM images of cells co-transfected with Stx3-mCitrine and VAMP3-mCherry showed a reduction of the donor’s apparent lifetime . The lowest apparent lifetime was present at the plasma membrane ( quantification below ) , indicating that the complexing of VAMP3 with Stx3 mainly took place at the cell membrane and less at intracellular compartments . FRET was also clear from Phasor analysis ( Hinde et al . , 2012 ) , as co-expression of Stx3-mCitrine with VAMP3-mCherry resulted in a shift of the phasor location compared to the condition with the donor fluorophore only ( Stx3-mCitrine ) ( Figure 1—figure supplement 1F ) . Thus , FRET-FLIM allows visualization of SNARE complexes in living cells . In order to obtain an estimate of the lowest lifetime observable , we recorded FLIM images of Stx3 conjugated to both mCitrine and mCherry in tandem ( Stx3-mCitrine-mCherry ) ( Figure 1C; Figure 1—figure supplement 1E–F ) . Although the colocalization of the mCitrine and mCherry signals was apparent , the overlap was not complete and the mCherry signal was more present in a cellular area juxtaposed to the nucleus ( Figure 1C; Figure 1—figure supplement 1G–H ) . To investigate this phenomenon further , we co-expressed VAMP3-mCitrine with VAMP3-mCherry or VAMP8-mCitrine with VAMP8-mCherry and compared the localization of these constructs ( Figure 1—figure supplement 1I ) . Also in this case , the mCherry signals were more prevalent than mCitrine in a juxtanuclear area . This juxtanuclear accumulation of mCherry was also observed upon dissipating the proton gradients by paraformaldehyde fixation ( Figure 1—figure supplement 1G and J ) , indicating that the differential localization of mCherry and mCitrine was not caused by pH-quenching of mCitrine , but possibly due to differences in maturation speed or stability of the used fluorophores ( Griesbeck et al . , 2001; Shaner et al . , 2004 ) . Consequently , our FRET-FLIM approach likely underestimates the amount of FRET in juxtanuclear regions and more stable or rapidly maturating YFP analogs need to be developed for this . However , we observed clear overlap of mCherry and mCitrine in more peripheral cellular regions ( Pearson correlation coefficients between 0 . 6–0 . 8 for the whole cell; Figure 1—figure supplement 1H and K ) , supporting the conclusion that our FRET-FLIM method can report on SNARE interactions in those regions . We performed whole-cell fluorescence lifetime analysis to directly compare between cells and conditions . We fitted all pooled photons collected from each individual cell with single exponential decay functions convoluted with the instrument response function ( IRF ) ( Figure 2A–B; Figure 2—figure supplement 1A–B ) . We obtained reasonable fitting accuracy ( within ~2% deviation ) , although these deviations were larger at very short time intervals ( <2 . 5 ns; Figure 2—figure supplement 1A ) . These deviations at short time intervals are likely caused by imperfect fitting of the IRF due to drift of the laser pulsing or timing of the detectors , reflections and/or ( auto ) fluorescence with fast kinetics , but do not cause major deviations of the resulting apparent lifetimes as our experiments with the reference dye rhodamine B show ( Figure 2—figure supplement 1C ) . In Stx3-mCitrine expressing cells , co-transfection with VAMP3-mCherry significantly reduced the apparent lifetime of mCitrine from 2 . 79 ± 0 . 02 ns to 2 . 49 ± 0 . 03 ns ( mean ± SEM ) ( Figure 2C ) . Cells from at least four different donors ( >8 cells/donor ) were measured for each condition ( Figure 2—figure supplement 1D ) . The spread of apparent lifetimes for cells co-expressing Stx3-mCitrine with VAMP3-mCherry was quite large , ranging from 2 to 3 ns ( Figure 2C ) . This large spread was at least partly caused by the availability of VAMP3-mCherry and competition with endogenous SNAREs , because the fluorescence lifetimes inversely correlated with the expression levels of VAMP3-mCherry ( Figure 2D; Figure 2—figure supplement 1E ) . To account for the whole cell population with varying levels of acceptor fluorophore-labeled SNAREs , we randomly selected cells with visible expression of both donor and acceptor fluorophores for our FRET-FLIM experiments . In order to quantify the difference in apparent lifetimes between the plasma membrane and intracellular compartments , we manually selected peripheral and intracellular regions of the image cell areas and fitted the fluorescence lifetime histograms of these two areas with mono-exponential decay functions ( Figure 2—figure supplement 1F–G ) . Most cells showed a lower apparent lifetime at the cellular periphery compared to the intracellular region ( Figure 2E ) . This reduction was on average 55 ps ( paired two-sided Student’s t-test; p=0 . 0078 ) , but was larger for cells with higher FRET ( linear regression: β = 1 . 18 , R2 = 0 . 82 ) . 10 . 7554/eLife . 23525 . 005Figure 2 . SNARE complex formation by whole-cell fluorescence lifetime measurements . ( A ) Representative whole-cell fluorescence lifetime histograms of dendritic cells expressing Stx3-mCitrine ( red curves; left graph ) , Stx3-mCitrine with VAMP3-mCherry ( green; middle graph ) , or Stx3 conjugated to both mCitrine and mCherry ( Stx3-mCitrine-mCherry; cyan; right graph ) . Dashed lines: fits with mono-exponential decay functions convoluted with the instrument response function ( IRF; gray ) . Graphs are normalized to the maximum photon counts ( depicted in each graph ) . Apparent fluorescence lifetimes for Stx3-mCitrine: 2 . 90 ns; for Stx3-mCitrine with VAMP3-mCherry: 2 . 29 ns; for Stx3-mCitrine-mCherry: 2 . 05 ns . Shown with logarithmic scaling in Figure 2—figure supplement 1A . ( B ) Overlap of the fluorescence lifetime decay curves from panel A ( logarithmic scaling in Figure 2—figure supplement 1B ) . ( C ) Whole-cell apparent fluorescence lifetimes for the conditions from panel A . Shown are individual cells pooled from at least 4 donors ( mean ± SEM shown; one-way ANOVA with Bonferroni correction; n: number of cells; individual donors in Figure 2—figure supplement 1D ) . ( D ) Whole-cell apparent fluorescence lifetimes of Stx3-mCitrine as a function of the expression level of VAMP3-mCherry ( by fluorescence intensities ) of a representative donor ( more donors in Figure 2—figure supplement 1E ) . Dashed line: linear regression ( β = −0 . 008; R2 = 0 . 800 ) . ( E ) Apparent fluorescence lifetimes of dendritic cells expressing Stx3-mCitrine with VAMP3-mCherry at the peripheral region ( P ) vs . the internal region ( I ) of the imaged cell areas ( Figure 2—figure supplement 1F–G ) . Individual cells from five donors are shown ( grey curve: line of equality; black dashed curve: linear regression ( β = 1 . 174; R2 = 0 . 821 ) ) . Note that for most cells , the lifetime at the periphery is lower than at intracellular regions ( Paired two-sided Student’s t-test; p=0 . 0078 ) . ( F ) Fluorescence lifetime histogram from panel A for a dendritic cell co-expressing Stx3-mCitrine with VAMP3-mCherry , but now fitted with a double-exponential decay function with the lifetimes of the slow ( 2 . 8 ns ) and fast ( 2 . 0 ns ) components fixed and convoluted with the IRF ( gray curve ) . The percentage FRET ( % FRET ) was calculated as the amplitude of the fast component over the total amplitude and was 81% ( logarithmic scaling in Figure 2—figure supplement 2A ) . ( G ) Same as panel C , but now fitted with double-exponential decay functions and % FRET shown . ( H ) Same as panel D , but now fitted with double-exponential decay functions and % FRET shown ( more donors in Figure 2—figure supplement 2B ) . Dashed line: linear regression ( β = 0 . 927; R2 = 0 . 771 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23525 . 00510 . 7554/eLife . 23525 . 006Figure 2—figure supplement 1 . Fluorescence lifetime histograms fitted with mono-exponential decay functions and calibration of FLIM setup . ( A ) Same as main Figure 2A , but now with logarithmic scaling . Shown are representative whole-cell fluorescence lifetime decay curves of dendritic cells expressing Stx3-mCitrine ( red curves; left graphs ) , Stx3-mCitrine with VAMP3-mCherry ( green; middle graphs ) , or Stx3 conjugated to both mCitrine and mCherry ( Stx3-mCitrine-mCherry; cyan; right graphs ) . Dashed lines: fits with mono-exponential decay functions convoluted with the instrument response function ( IRF; gray; residuals from the fits shown ) . Graphs are normalized to the maximum photon counts ( depicted in each graph ) . ( B ) Same as main Figure 2B , but now with logarithmic scaling . Shown is the overlap of the fluorescence lifetime decay curves from panel A . ( C ) Representative fluorescence lifetime decay curve of rhodamine B in methanol ( about 1 µM concentration; magenta curves ) . Dashed line: fit with a mono-exponential decay function . The fit is shown with a linear ( left ) and a logarithmic ( right ) scaling . A fluorescence lifetime was obtained of 2 . 19 ± 0 . 04 ns ( mean ± SEM from four independent measurements ) , close to the reported lifetime of 2 . 32 ns ( Kristoffersen et al . , 2014 ) . ( D ) Same as main Figure 2C , but now the average whole-cell apparent fluorescence lifetimes of individual donors are shown ( n: number of donors; mean ± SEM shown; one-way ANOVA with Bonferroni correction ) . ( E ) Same as main Figure 2D for three more donors . Shown are whole-cell apparent fluorescence lifetimes of Stx3-mCitrine from fits with mono-exponential decay functions for individual cells co-expressing Stx3-mCitrine with VAMP3-mCherry . Dashed lines: linear regression ( top: β = −0 . 001 , R2 = 0 . 756; middle: β = −0 . 001 , R2 = 0 . 702; bottom: β = −0 . 001 , R2 = 0 . 869 ) . ( F ) Representative confocal microscopy and FLIM images of a dendritic cell co-expressing Stx3-mCitrine ( green in merge ) and VAMP3-mCherry ( magenta ) . The intracellular ( I; light blue in lower right panel ) and peripheral ( P; white ) regions of the imaged cell area were manually selected . BF: bright-field . Scale bar , 10 µm . ( G ) Whole-cell fluorescence lifetime decay curves of the intracellular ( I ) and peripheral ( P ) cell regions for the cell from panel F . The apparent fluorescence lifetimes obtained from mono-exponential fits were 2 . 41 ns ( I ) and 2 . 24 ns ( P ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23525 . 00610 . 7554/eLife . 23525 . 007Figure 2—figure supplement 2 . Fluorescence lifetime histograms fitted with double-exponential decay functions . ( A ) Same as main Figure 2F , but now with logarithmic scaling . Shown is a representative whole-cell fluorescence lifetime histogram of a dendritic cell co-expressing Stx3-mCitrine with VAMP3-mCherry . Dashed line: fit with double-exponential decay function convoluted with the instrument response function ( IRF; gray; residuals from the fit shown ) . Graphs are normalized to the maximum photon counts ( depicted in graph ) . ( B ) Same as main Figure 2H for three more donors . Shown are % FRET from fits with double-exponential decay functions for individual cells co-expressing Stx3-mCitrine with VAMP3-mCherry . Dashed lines: linear regression ( top: β = 0 . 055 , R2 = 0 . 704; middle: β = 0 . 065 , R2 = 0 . 674; bottom: β = 0 . 147 , R2 = 0 . 867 ) . ( C ) The apparent fluorescence lifetime from mono-exponential fits as a function of the % FRET from double exponential fits for individual cells co-expressing Stx3-mCitrine with VAMP3-mCherry . Dashed line: linear regression ( β = −0 . 007 , R2 = 0 . 812 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23525 . 007 The co-expression of Stx3-mCitrine and VAMP3-mCherry will result in a mixed population , with part of Stx3-mCitrine interacting with VAMP3-mCherry and the remainder free or engaged with endogenous VAMP3 or other R SNAREs . In order to estimate the fraction of FRET , we fitted the whole-cell lifetime histograms of cells co-expressing Stx3-mCitrine and VAMP3-mCherry with double exponential decay functions ( Figure 2F; Figure 2—figure supplement 2A ) . In our biexponential fitting , we had to fix the lifetime components , as we did not have sufficient photon counts for extra free fit parameters since small errors in the lifetimes will influence the amplitudes of the two components and vice versa . The time constants of the slow and fast components of the double exponential decay function were fixed to the values of the donor only ( i . e . , no FRET; 2 . 79 ns ) and of the tandem Stx3-mCitrine-mCherry construct ( 100% FRET; 2 . 0 ns ) . This allowed estimating the percentage of FRET from the amplitudes of the fast and slow components , which provides a measure for the fraction of Stx3-mCitrine in complex with VAMP3-mCherry ( Figure 2G ) . Using this analysis , the percentage of complexed Stx3-mCitrine again correlated with the expression level of VAMP3-mCherry , and reached up to almost 100% at high expression levels ( Figure 2H; Figure 2—figure supplement 2B ) . These data support our conclusion that complex formation of the donor fluorophore-labeled SNARE depends on the availability of acceptor fluorophore-labeled SNAREs . The observed fluorescence lifetimes of mCitrine not only depend on FRET , but will also be sensitive to other factors including the microenvironment , dipole orientation , and self-quenching of mCitrine within SNARE domains ( Zhu et al . , 2015 ) . As a consequence , the observed fluorescence lifetimes in our samples may not be the same as in our control samples ( i . e . , donor only and tandem Stx3-mCitrine-mCherry ) . Fitting with mono-exponential decay functions does not require any a-priori knowledge of the fluorescence lifetimes . Moreover , correlation of the apparent lifetimes from mono-exponential fits with the amplitudes from the bi-exponential fits resulted in a clear linear correlation ( Figure 2—figure supplement 2C; linear regression: R2 = 0 . 812 ) , demonstrating that these apparent lifetimes provide a reliable measure of the percentage FRET . We therefore analyzed the remainder of our FLIM data with mono-exponential decay functions . We performed two sets of control experiments to confirm that FRET-FLIM is a reliable method to measure interactions between SNARE proteins in live cells . First , we generated fusion constructs of SNARE proteins with either the FK506 binding protein 12 ( FKBP ) or FKBP rapamycin binding ( FRB ) domain of mTOR fused to their N-termini . Dendritic cells were transfected with FRB-VAMP3-mCherry in combination with FKBP-Stx3-mCitrine or FKBP-Stx4-mCitrine ( Figure 3A; Figure 3—figure supplement 1A ) . FKBP and FRB are rapamycin binding domains and the presence of this compound induces their dimerization ( Putyrski and Schultz , 2012 ) . Thus , we reasoned that the rapamycin-induced heteromerization of FKBP and FRB would force or stabilize SNARE association . Indeed , addition of rapamycin or the rapamycin analogue A/C Heterodimerizer resulted in a decrease of the apparent fluorescence lifetimes ( Figure 3A ) . 10 . 7554/eLife . 23525 . 008Figure 3 . Forced interactions between SNAREs increases FRET while FRET is reduced with a fusion incompetent VAMP3 mutant . ( A ) Whole-cell apparent fluorescence lifetimes for dendritic cells expressing FKBP-Stx3-mCitrine or FKBP-Stx4-mCitrine together with FRB-VAMP3-mCherry and incubated in absence or presence of rapamycin or a rapamycin analogue . ( B ) Alignment of VAMP2 and VAMP3 ( mouse sequences ) showing 100% identity of the region containing leucines 84 ( VAMP2 ) and 71 ( VAMP3 ) . ( C ) Whole-cell apparent fluorescence lifetimes for dendritic cells expressing Stx3-mCitrine with wild-type VAMP3-mCherry or a mutant lacking leucine 71 ( VAMP3 ( Δ71 ) -mCherry ) . Shown in panels A and C are individual cells pooled from at least 4 donors ( mean ± SEM shown; one-way ANOVA with Bonferroni correction; n: number of cells ) . Representative confocal and FLIM images are in Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 23525 . 00810 . 7554/eLife . 23525 . 009Figure 3—figure supplement 1 . FLIM images belonging to main Figure 3 . ( A ) Representative confocal microscopy , convoluted FLIM and fluorescence lifetime images of dendritic cells expressing FKBP-Stx3-mCitrine ( upper panel ) or FKBP-Stx4-mCitrine ( lower panel; green in merge ) together with FRB-VAMP3-mCherry ( magenta ) and incubated in absence or presence of a rapamycin analogue . ( B ) Representative confocal microscopy , convoluted FLIM and lifetime images of dendritic cells expressing Stx3-mCitrine ( green ) with mutant VAMP3-mCherry lacking leucine 71 ( VAMP3 ( Δ71 ) -mCherry; magenta ) . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 23525 . 009 As a second approach to validate our FLIM method , we generated a mutant form of VAMP3-mCherry lacking leucine 71 ( VAMP3 ( Δ71 ) ) ( Figure 3B–C; Figure 3—figure supplement 1B ) . This residue is located at the C-terminal end of the SNARE region which is identical to VAMP2 ( Figure 3B ) . For VAMP2 , deletion of leucine 84 , homologous to leucine 71 of VAMP3 , allows formation of a trans-SNARE complex but impairs fusion of membranes as progression to the cis-conformation cannot take place ( Hernandez et al . , 2012 ) . For VAMP3 , deletion of leucine 71 also impairs membrane fusion , as cells co-expressing Stx3-mCitrine with VAMP3 ( Δ71 ) -mCherry showed a significantly higher apparent fluorescence lifetime compared to cells expressing non-mutant VAMP3-mCherry ( Figure 3C ) . These data confirm that FRET-FLIM allows detection of SNARE complexes . Next , we used our FRET-FLIM assay to compare complex formation between different SNARE proteins . Dendritic cells were transfected with Stx3-mCitrine or Stx4-mCitrine alone or in combination with VAMP3-mCherry or VAMP8-mCherry ( Figure 4A; Figure 4—figure supplement 1A ) . Stx4-mCitrine mainly localized to the plasma membrane and only somewhat to intracellular compartments . Similar to VAMP3-mCherry , the majority of VAMP8-mCherry was present in intracellular compartments ( Figure 4A ) and this corresponded to the localization of endogenous VAMP8 ( Figure 1—figure supplement 1B ) . Whole-cell fluorescence lifetime analysis revealed interactions between all SNARE pairs tested , as reflected by reductions of the apparent fluorescence lifetime of mCitrine in the presence of mCherry-fused VAMP3 or VAMP8 , although for Stx4 these reductions were relatively small ( <100 ps ) and not significant when analyzing donor-averaged lifetimes ( Figure 4B; Figure 4—figure supplement 1B ) . FLIM imaging showed that whereas interactions of Stx3-mCitrine with VAMP3-mCherry mainly occurred at the plasma membrane , the interactions with VAMP8-mCherry predominantly occurred at intracellular compartments ( Figure 1C , Figure 4A; Figure 4—figure supplement 1A ) . As a positive control , we treated our cells with N-ethylmaleimide ( NEM ) , which impairs the AAA-ATPase NSF responsible for dissociation of SNARE complexes . Addition of this compound traps all SNARE complexes in their cis-conformation ( Söllner et al . , 1993 ) and resulted in a trend showing further reductions of the apparent fluorescence lifetimes for all combinations of donor and acceptor SNAREs tested ( Figure 4B; Figure 4—figure supplement 1B ) . Thus , our FLIM data resulted in two new observations of SNARE complex formation in unstimulated dendritic cells: ( i ) FRET at the plasma membrane of fluorescently-labeled VAMP3 with Stx3 was higher than with Stx4 , suggesting more interactions of VAMP3 with Stx3 even though most Stx4 is present at the plasma membrane . ( ii ) FRET of fluorescently-labeled Stx3 , but not Stx4 , with VAMP3 was observed mainly at the plasma membrane while FRET with VAMP8 was found mainly at intracellular compartments , indicating that complex formation of these SNAREs occurred in different subcellular compartments . 10 . 7554/eLife . 23525 . 010Figure 4 . FRET-FLIM of Stx3 and Stx4 with VAMP3 and VAMP8 . ( A ) Representative confocal microscopy ( left ) and convoluted FLIM ( right ) images of dendritic cells expressing Stx3-mCitrine or Stx4-mCitrine ( green in merge ) with VAMP3-mCherry or VAMP8-mCherry ( magenta ) . The apparent fluorescence lifetimes of Stx3-mCitrine with VAMP3-mCherry were lowest at the cell membrane , whereas the lifetimes with VAMP8-mCherry were lowest at intracellular compartments ( yellow arrowheads ) . Scale bars , 10 µm . Lifetime images are in Figure 4—figure supplement 1A . ( B ) Whole-cell apparent fluorescence lifetimes for the conditions from panel A , both in presence or absence of NEM . Shown are individual cells pooled from at least 3 donors ( mean ± SEM shown; one-way ANOVA with Bonferroni correction; n: number of cells; individual donors in Figure 4—figure supplement 1B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23525 . 01010 . 7554/eLife . 23525 . 011Figure 4—figure supplement 1 . Fluorescence lifetime images belonging to main Figure 4 . ( A ) Fluorescence lifetime images belonging to main Figure 4A . FLIM images were generated by convolution of these lifetime images with the fluorescence intensities ( i . e . , the mCitrine images shown in the main figure ) . Scale bars , 10 µm . ( B ) Same as main Figure 4B , but now with the averages for individual donors . Shown are donor-averaged whole-cell apparent fluorescence lifetimes of dendritic cells expressing Stx3-mCitrine or Stx4-mCitrine with or without VAMP3-mCherry or VAMP8-mCherry and in absence or presence of NEM treatment ( mean ± SEM shown; one-way ANOVA with Bonferroni correction; n: number of donors ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23525 . 011 Activation of dendritic cells by inflammatory stimuli triggers the production and secretion of immunomodulatory cytokines and alters the surface receptors displayed on the cell membrane and this is accompanied by marked changes of the membrane trafficking machinery ( Collins et al . , 2015; Stow et al . , 2009; Stanley and Lacy , 2010; Lacy and Stow , 2011; Murray and Stow , 2014; Murray et al . , 2005; Pagan et al . , 2003 ) . We therefore wondered whether the SNARE interactions that we observed in unstimulated dendritic cells by FRET-FLIM would be altered by activation of the cells . We transfected dendritic cells with combinations of SNAREs and stimulated the cells overnight with the bacterial antigen LPS , a component from the outer cell membrane of gram-negative bacteria and a strong agonist of TLR4 ( Figure 5A; Figure 5—figure supplement 1A ) . TLR4 stimulation triggers danger signaling in dendritic cells leading to cell maturation characterized by a strong increase in cytokine production . Although LPS stimulation can upregulate the expression of several SNAREs in immune cells ( Chiaruttini et al . , 2016; Murray et al . , 2005; Collins et al . , 2014; Ho et al , 2009; Pagan et al . , 2003 ) , the protein levels of both endogenous and overexpressed Stx3 , Stx4 , VAMP3 and VAMP8 in monocyte-derived dendritic cells were not altered by LPS ( Figure 5—figure supplement 1B–C ) . Whole-cell FLIM analysis revealed that LPS activation of dendritic cells expressing Stx4-mCitrine in combination with VAMP3-mCherry led to a significant reduction of the apparent lifetime compared to the condition without LPS ( Figure 5B; Figure 5—figure supplement 1D ) . In contrast , no significant changes were observed for Stx3-mCitrine or for VAMP8-mCherry . FLIM imaging demonstrated that the interactions of VAMP3-mCherry with Stx4-mCitrine mainly occurred at the plasma membrane ( Figure 5A; Figure 5—figure supplement 1A ) . Because of the photobleaching we were unable to measure FLIM of the same cell pre- and post-LPS treatment . VAMP3 is known to mediate release of inflammatory cytokines by the murine macrophage cell line RAW264 . 7 ( Manderson et al . , 2007 ) and the human synovial sarcoma cell line SW982 ( Boddul et al . , 2014 ) . VAMP3 is also required for the secretion of cytokines by LPS activated monocyte-derived dendritic cells , as siRNA knockdown of this SNARE impaired release of the cytokine IL-6 , as shown by ELISA ( Figure 5C–D; Figure 5—figure supplement 1E ) . These FLIM data support the conclusion that LPS activation of dendritic cells results in an increased complex formation of Stx4 with VAMP3 at the plasma membrane and that this promotes IL-6 secretion . 10 . 7554/eLife . 23525 . 012Figure 5 . LPS activation of dendritic cells increases complex formation of Stx4 with VAMP3 at the plasma membrane . ( A ) Representative confocal microscopy ( left ) and convoluted FLIM ( right ) images of unstimulated or LPS-activated dendritic cells expressing Stx3-mCitrine or Stx4-mCitrine ( green in merge ) with VAMP3-mCherry or VAMP8-mCherry ( magenta ) . Apparent fluorescence lifetimes of Stx3-mCitrine and Stx4-mCitrine with VAMP3-mCherry were lowest at the cell membrane ( yellow arrowheads ) , whereas the lifetimes of Stx3-mCitrine with VAMP8-mCherry were lowest at intracellular compartments ( yellow arrowheads ) . Scale bars , 10 µm . Lifetime images are in Figure 5—figure supplement 1A . ( B ) Whole-cell apparent fluorescence lifetimes for the conditions from panel A in presence or absence of LPS . Shown are individual cells pooled from at least 4 donors ( mean ± SEM shown; one-way ANOVA with Bonferroni correction; n: number of cells; individual donors in Figure 5—figure supplement 1D ) . ( C ) Representative Western blot and quantification of siRNA knockdown of VAMP3 ( siV3; p=0 . 0001 , paired two-sided Student’s t-test ) . siCntrl: non-targeting siRNA control . GAPDH: loading control . ( D ) IL-6 production after 24 hr LPS exposure by dendritic cells with VAMP3 knockdown ( siV3 ) or siCntrl ( individual donors shown; p=0 . 0199 , paired two-sided Student’s t-test; time traces shown in Figure 5—figure supplement 1E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23525 . 01210 . 7554/eLife . 23525 . 013Figure 5—figure supplement 1 . Fluorescence lifetime images belonging to main Figure 5 , expression levels of endogenous and overexpressed SNAREs upon LPS treatment and time traces of IL-6 secretion . ( A ) Fluorescence lifetime images belonging to main Figure 5A . FLIM images were generated by convolution of these lifetime images with the fluorescence intensities ( i . e . , the mCitrine images shown in the main figure ) . Scale bars , 10 µm . ( B ) Expression levels of overexpressed Stx3 , Stx4 , VAMP3 , and VAMP8 by dendritic cells with and without overnight LPS-activation . Shown are representative Western blots and quantification ( mean ± SEM from 3 donors ) . GAPDH: loading control . VAMP8-mCherry appears as a double band because of incomplete stripping of the GAPDH antibody ( lower band ) . LPS treatment did not result in significant changes in SNARE expression ( paired two-sided Student’s t-tests ) . ( C ) Same as panel B , but now for endogenous Stx3 , Stx4 , VAMP3 , and VAMP8 . ( D ) Same as main Figure 5B , but now with the averages for individual donors . Shown donor-averaged whole-cell apparent fluorescence lifetimes of dendritic cells expressing Stx3-mCitrine or Stx4-mCitrine with or without VAMP3-mCherry or VAMP8-mCherry and in absence or presence of LPS treatment ( mean ± SEM shown; one-way ANOVA with Bonferroni correction; n: number of donors ) . ( E ) Time course of IL-6 secretion with siCntrl or siV3 knock-down for four individual donors . DOI: http://dx . doi . org/10 . 7554/eLife . 23525 . 013
SNARE proteins are pivotal for membrane fusion and coordinate intracellular organellar trafficking , exocytosis and endocytosis ( Hong , 2005; Jahn and Scheller , 2006 ) . Although it is generally accepted that different intracellular trafficking routes are catalyzed by specific sets of SNARE proteins ( Collins et al . , 2015; Stow et al . , 2009; Stanley and Lacy , 2010; Lacy and Stow , 2011; Murray and Stow , 2014 ) , current microscopy techniques often cannot resolve distinct SNARE pairs for specific trafficking routes . In this study , we demonstrate that FRET-FLIM enables the assignment of SNARE partners to specific trafficking routes , as it allows visualization of SNARE pairing at subcellular resolution . Our data show that FLIM-FRET is a quantitative method that in principle allows to determine the fraction of SNAREs engaged in complex formation by fitting the photon counts with bi-exponential decay functions with fixed lifetimes ( i . e . , for no FRET ( donor only ) and 100% FRET ( all SNAREs interacting ) ) . While the transient overexpression system used for our primary cells does not allow meaningful quantification of SNARE complexing , we expect that fusing chromosomal SNARE-encoding genes with fluorescent proteins using CRISPR/CAS9 in cell lines will enable to quantify endogenous SNARE interactions by FRET-FLIM . We used FRET-FLIM to measure SNARE complex formation in human blood-derived dendritic cells . In unstimulated dendritic cells , the apparent lifetime of fluorescently labeled Stx3 with VAMP3 was lower at the periphery of the imaged cell areas , whereas the apparent lifetime with VAMP8 was lower at intracellular areas . This indicates that complex formation of Stx3 with VAMP3 mainly occurs at the plasma membrane , while interactions with VAMP8 predominantly occur at intracellular compartments . VAMP3 is primarily localized in early and recycling endosomes and VAMP8 locates predominately at late endosomal compartments ( Bajno et al . , 2000; Manderson et al . , 2007; Hong , 2005; Antonin et al . , 2000 ) . Our data are therefore perfectly in line with the widely accepted notion that fusion of compartments of early/recycling endosomal nature with the plasma membrane is more abundant than of late endosomal compartments . Based on differences in apparent fluorescence lifetimes , we also conclude that the interactions of VAMP3 at the plasma membrane with Stx4 are lower than with Stx3 in unstimulated dendritic cells , even though the vast majority of Stx4 is present at the plasma membrane . This changes upon activation of the dendritic cells by the pathogenic stimulus LPS , which results in a decreased apparent lifetime of fluorescently labeled Stx4 with VAMP3 at the plasma membrane , whereas the apparent lifetimes of Stx3 remain unaltered . This result indicates that LPS triggers increased exocytosis by VAMP3 and Stx4 . VAMP3 is a recycling endosomal SNARE ( Bajno et al . , 2000; Manderson et al . , 2007; Hong , 2005; Murray et al . , 2005 ) and many cytokines traffic via recycling endosomes to the plasma membrane , including IL-6 and tumor necrosis factor alpha ( TNFα ) ( Manderson et al . , 2007; Murray et al . , 2005 ) . Indeed , our data show that siRNA knockdown of VAMP3 impairs IL-6 secretion by monocyte-derived dendritic cells as previously observed for RAW264 . 7 murine macrophages ( Manderson et al . , 2007 ) and SW982 human synovial sarcoma cells ( Boddul et al . , 2014 ) . As supported by our FRET-FLIM data the increased complexing of VAMP3 with Stx4 is thus likely responsible for the increased secretion of IL-6 and other cytokines by activated dendritic cells . Interestingly , we observed that LPS increased FRET only of VAMP3 with Stx4 and not with Stx3 . Stx3 and Stx4 have been well described in epithelial cells , where they catalyze apical and basolateral exocytosis respectively ( Low et al . , 1996; ter Beest et al . , 2005; Procino et al . , 2008 ) , but why non-polarized cells such as dendritic cells express both syntaxins is not well understood . Our study provides a possible clue why this could be the case , as our data indicate that activation of the dendritic cells leads to a rerouting of intracellular trafficking with more VAMP3-containing compartments fusing with the plasma membrane by interacting with Stx4 . Perhaps Stx3 catalyzes the constitutive cycling of early/recycling endosomes in resting cells and Stx4 offers the spare capacity to fulfil the additional secretory requirements upon dendritic cell activation needed for the secretion of massive amounts of cytokines . What could cause the increased Stx4 complex formation with VAMP3 upon dendritic cell activation ? Our data show that such increased SNARE complexing is not caused by increased levels of Stx4 or VAMP3 . Possibly , the increased complexing might be regulated by direct phosphorylation of Stx4 , VAMP3 and/or the Qb/c-SNARE SNAP23 ( Malmersjö et al . , 2016 ) . SNAP23 was recently shown to be phosphorylated in an IκB-kinase two dependent fashion upon stimulation of mouse dendritic cells with LPS and this mediates the delivery of major histocompatibility complex ( MHC ) class I from recycling endosomes to phagosomes , presumably via Stx4 and VAMP3 or VAMP8 ( Nair-Gupta et al . , 2014a ) . Alternatively , or additionally , the increased complex formation could be regulated by the Sec1/Munc18-protein Munc18c ( STXBP3 ) . Munc18c and Stx4 are well known to regulate GLUT4 trafficking in adipocytes and skeletal muscle cells ( Tellam et al . , 1997; Hong , 2005 ) , and are also implied in insulin secretion from pancreatic beta cells ( Zhu et al . , 2015 ) , secretion of dense-core granules , α-granules and lysosomes from human platelets ( Schraw et al . , 2003 ) and TNFα secretion from macrophages ( Pagan et al . , 2003 ) . In any case , the mechanism must be specific for Stx4 and VAMP3 , as our data show that FRET of Stx3 with VAMP3 and of Stx4 with VAMP8 are not altered by LPS-activation of dendritic cells . Overall , our study demonstrates the use of FRET-FLIM for the identification of the SNARE partners orchestrating specific organellar membrane trafficking steps . We expect this technique will allow deciphering the precise intracellular trafficking pathways of molecules important in health and disease , such as cytokines , receptors , hormones , metabolic enzymes and metabolites .
Dendritic cells were derived from monocytes by culturing in the presence of IL-4 and GM-CSF for 6 days as described ( Dingjan et al . , 2016 ) . Monocytes were isolated from blood of healthy individuals ( informed consent and consent to publish obtained , approved by Sanquin ethical committee and according to Radboudumc institutional guidelines ) . Dendritic cells were transfected with plasmid DNA ( overnight ) or siRNA ( for 48 hr; VAMP3 Stealth; HSS113848 , HSS113849 , HSS113850; Thermo Scientific , Waltham , MA ) using a Neon Transfection system ( Invitrogen , Carlsbad , CA ) as described ( Dingjan et al . , 2016 ) . To induce SNARE complex formation , NEM ( 400 µM; 10 min before imaging ) , rapamycin ( Selleck Chemicals , Houston , TX ) or the rapamycin analog A/C Heterodimerizer ( Clontech , Mountain View , CA; #635057; 0 . 5 µM; 90 min before imaging ) were added to the samples . LPS was used at 1 µg ml−1 and incubated for 16 hr for FLIM measurements and for 4 , 8 or 24 hr after which the supernatant was collected for IL-6 determination . IL-6 concentration was measured by ELISA ( 88-7066-88; Thermo Scientific ) . DNA coding for mCitrine was ordered as a synthetic gene and inserted in the BamHI/NotI sites of pEGFP-N1 ( Clontech ) yielding pmCitrine-N1 . Stx3 and Stx4 ( ter Beest et al . , 2005 ) were inserted in the EcoRI/BamHI sites of pmCitrine-N1 . VAMP3 and VAMP8 were inserted in the EcoRI/BamHI site of pmCherry-N1 . The VAMP3 ( Δ71 ) -mCitrine mutant was generated by site directed mutagenesis . For FKBP-Stx4-mCitrine , synthetic DNA coding for human FKBP1A residues 1–108 with a serine-glycine linker ( SGGGGSGGGGSGGGG ) ( Genscript , Piscataway , NJ ) was subcloned in the restriction sites XhoI/EcoRI of Stx4-mCitrine . Stx4 was then replaced with Stx3 for FKBP-Stx3-mCitrine . For FRB-VAMP3-mCherry , residues 2 , 015–2114 of human mTOR with T2098L and a serine-glycine linker ( Genscript ) was subcloned in the XhoI/EcoRI restriction sites of the VAMP3-mCherry construct . For Stx3-mCitrine-mCherry , mCitrine with BamHI restriction sites on both sides was subcloned in the Stx3-mCherry construct . FLIM was recorded with live ( unfixed ) dendritic cells on a Leica ( Wetzlar , Germany ) SP8 confocal microscope equipped with a 63 × 1 . 20 NA water immersion objective . Confocal images of the cells were recorded prior to each FLIM measurement . The image plane was selected at the height of the nucleus and was between 2–5 µm above the surface of the cover slips . Excitation was done with a pulsed white light laser ( Leica; 80 , 000 MHz pulsing ) operating at 516 nm . Emission from 521 to 565 nm was collected with a photomultiplier tube and processed by a PicoHarp 300 Time-Correlated Single Photon Counting ( TCSPC ) system ( PicoQuant , Berlin , Germany ) . At least 50 , 000 photons with on average ~750 , 000 photons were recorded for each individual cell ( Figure 1—figure supplement 1C ) . Samples were imaged in Live Cell Imaging Solution ( Thermo Scientific ) at 37°C . Photon traces in Picoquant PT3 format were used to construct FLIM images in Image Cytometry Standard ( ICS ) format using in-house developed PT32ICS conversion software ( Membrane Trafficking Group , Radboudumc , Nijmegen , The Netherlands ) . For single cell FLIM , all photons were pooled for each individual cell and fitted with exponential decay functions convoluted with the IRF using OriginPro2016 ( Originlab , Northampton , MA ) . Single pixel fitted FLIM images were generated for individual cells ( with at least 1 , 000 , 000 photons per cell ) using the TRI2 software ( version 2 . 8 . 6 . 2; Gray institute , Oxford , UK ) ( Barber et al . , 2005 , 2009 ) with monoexponential ( Marquardt ) fitting algorithm , 7 × 7 pixel circular binning and thresholding from 15% to 100% intensity . Phasor plots were generated using FIJI ImageJ ( Schneider et al . , 2012; Schindelin et al . , 2012 ) with the Time Gated Phasor plugin from Spechron ( developed by F . Fereidoni , UC Davis Medical Center ) . Cells were plated and stained as described previously ( Dingjan et al . , 2017 ) . For immunofluorescence labeling , primary antibodies rabbit anti-VAMP3 ( Abcam , Cambridge , United Kingdom; #5789; 1:200 dilution ( v/v ) ) or rabbit anti-VAMP8 ( Synaptic Systems , Goettingen , Germany; #104–303; 1:200 dilution ( v/v ) ) were used in combination with secondary antibody goat anti-rabbit conjugated with Alexa fluor 647 ( Thermo Scientific; 1;400 dilution ( v/v ) ) . The cells were imaged on a Leica SP8 confocal microscope equipped with a 63 × 1 . 20 NA water immersion objective . Dendritic cells were incubated for 16 hr in complete medium in the absence or presence of LPS ( 1 µg ml−1 ) . Cells were harvested and analyzed by Western blot . Primary antibodies used were a rabbit monoclonal against GAPDH ( clone 14C10; Cell Signaling , Danvers , MA; #2118; 1:1000 dilution ( v/v ) ) , a rabbit polyclonal against Stx3 ( ter Beest et al . , 2005 ) ( 1:500 dilution ( v/v ) ) , a mouse monoclonal against Stx3 ( clone 1–146; Millipore , Billerica , MA; MAB2258; 1:500 dilution ( v/v ) ) , a rabbit polyclonal against VAMP3 ( Abcam #5789; 1:1000 dilution ( v/v ) ) , a mouse monoclonal against VAMP8 ( Santa Cruz , Dallas , TX; sc-166820; 1:500 dilution ( v/v ) ) , a mouse IgG1 against Stx4 ( Abcam #77037; 1:1000 dilution ( v/v ) ) and a rabbit polyclonal against mCherry ( Abcam #167453; 1:500 dilution ( v/v ) ) . Secondary antibodies were goat anti-rabbit or goat anti-mouse conjugated with IRDye 800 ( LI-COR , Lincoln , NE; 1:5000 dilution ( v/v ) ) or goat anti-rabbit conjugated with Alexa fluor 647 ( Thermo Scientific; 1:500 dilution ( v/v ) ) . Due to experimental restraints ( time duration of the experiments and the viability of the cells ) , a maximum of 6 conditions a day could be measured . During single days , all relevant comparisons and appropriate controls were included . Cells from the same human subjects were used across the data displayed in Figures 2–5 . For all the FLIM experiments , significant difference between two samples was assessed using one-way ANOVA with post-hoc Bonferroni correction for relevant pairs . Error bars indicate the SEM for at least >30 cells or for at least 3 donors . Western blots and ELISA results were analyzed by paired two-sided Student’s t-test and error bars indicate the SEM for at least 3 donors . The level of statistical significance is represented by *p≤0 . 05 , **p≤0 . 01 , ***p≤0 . 001 . | Many processes in living cells involve membranes coming together and fusing . For example , white blood cells known as dendritic cells rely on membrane fusion to fight off infections . When a dendritic cell detects a bacterial infection , it releases signaling molecules called cytokines to recruit other immune cells that help to eliminate the bacteria . The cytokines are contained in membrane-bound packages inside the cell , called vesicles , and are transported outside when these vesicles fuse with the membrane that surrounds the dendritic cell . Proteins called SNAREs drive the fusion of a cell’s membranes . These proteins , which are found on both membranes that will fuse , entwine to form a tight complex that pulls the membranes together . Mammals have over 30 different SNARE proteins , and many scientists believe that specific transport routes within cells use distinct pairs of SNAREs . However , to date , it has been difficult to assign specific pairs of SNAREs to specific transport routes with existing techniques . Verboogen et al . have now engineered human dendritic cells to add labels onto their SNAREs that fluoresce if the proteins interact . This approach meant that the interactions could be tracked via a microscope . The experiments showed that exposing dendritic cells to a bacterial compound that stimulates the release of cytokines caused two SNARE proteins called syntaxin 4 and VAMP3 to interact more at the cell membrane . This indicates that syntaxin 4 and VAMP3 are important for the release of cytokines from these cells . This finding was supported by an additional experiment in which Verboogen et al . switched off the gene for VAMP3 in the dendritic cells and found that this reduced the amount of cytokines that were released . This new microscope-based approach will be useful for identifying the specific pairs of SNARE proteins that are needed for the release and transport of molecules – like hormones and enzymes – that are important in health and disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"immunology",
"and",
"inflammation"
] | 2017 | Fluorescence Lifetime Imaging Microscopy reveals rerouting of SNARE trafficking driving dendritic cell activation |
The evolutionarily conserved Crumbs protein is required for epithelial polarity and morphogenesis . Here we identify a novel role of Crumbs as a negative regulator of actomyosin dynamics during dorsal closure in the Drosophila embryo . Embryos carrying a mutation in the FERM ( protein 4 . 1/ezrin/radixin/moesin ) domain-binding motif of Crumbs die due to an overactive actomyosin network associated with disrupted adherens junctions . This phenotype is restricted to the amnioserosa and does not affect other embryonic epithelia . This function of Crumbs requires DMoesin , the Rho1-GTPase , class-I p21-activated kinases and the Arp2/3 complex . Data presented here point to a critical role of Crumbs in regulating actomyosin dynamics , cell junctions and morphogenesis .
Dorsal closure ( DC ) in the Drosophila embryo is an established model for epithelial morphogenesis . The power of Drosophila genetics and cell biological tools have contributed to understand how signalling pathways , cell polarity and cell adhesion regulate the coordinated movements of two epithelial sheets , the epidermis and the amnioserosa ( AS ) , a transient extraembryonic tissue [reviewed in ( Ríos-Barrera and Riesgo-Escovar , 2013 ) ] . More recently , elaborate biophysical techniques combined with high resolution imaging have elucidated how contractile forces are coordinated between cells in order to drive coherent changes in tissue morphology ( Sokolow et al . , 2012; Jayasinghe et al . , 2013; Fischer et al . , 2014; Wells et al . , 2014; Eltsov et al . , 2015; Saias et al . , 2015 ) . DC is a complex morphogenetic process taking about 2 hr , during which the epidermis expands dorsally to encompass the embryo . The process can be subdivided into three phases: i ) elongation of the dorsal-most epidermal cells ( DME ) along the dorso-ventral axis; ii ) contraction of AS cells and migration of the lateral epidermal cells towards the dorsal midline; iii ) “zippering” , i . e . adhesion of the epidermal cells from both sides on the dorsal midline [reviewed in ( Gorfinkiel et al . , 2011 ) ] . Several forces contribute to these processes . First , pulsed contraction of AS cells produces a pulling force . These pulsed contractions are correlated with dynamic apical actomyosin foci , which transiently form in the apical medial cytocortex ( Kiehart et al . , 2000; Hutson et al . , 2003; Solon et al . , 2009; Gorfinkiel et al . , 2009; Blanchard et al . , 2010; Heisenberg and Bellaiche , 2013 ) . Cells delaminating from the AS contribute additional pulling forces ( Muliyil et al . , 2011; Sokolow et al . , 2012; Toyama et al . , 2008 ) . Second , a supracellular actomyosin cable , formed in the DME cells , surrounds the opening and provides contractile forces ( Hutson et al . , 2003; Rodriguez-Diaz et al . , 2008 ) . Finally , “zippering” of the two lateral epithelial sheets occurs , mediated by dynamic filopodia and lamellipodia ( Eltsov et al . , 2015; Jacinto et al . , 2000 ) . A plethora of proteins contribute to coordinate this highly dynamic morphogenetic process . Beside transcription factors , these include adhesion molecules and signalling pathways , a variety of cytoskeletal proteins and their regulators . Non-muscle myosin-II heavy chain ( MHC ) and the non-muscle myosin regulatory light chain ( MRLC ) , encoded by zipper ( zip ) and spaghetti-squash ( sqh ) , respectively , are , together with the essential light chain , part of a force-producing molecular motor during DC [reviewed in ( Vicente-Manzanares et al . , 2009; Liu and Cheney , 2012 ) ] . The small G-proteins of the Rho family , namely Rho1 , Rac1 , Rac2 , Mtl , and Cdc42 , regulate actomyosin activity and cell-cell adhesion ( Abreu-Blanco et al . , 2014; Magie et al . , 1999; 2002 ) . These GTPases stimulate myosin contraction through Rho-kinase ( Rok ) ( Mizuno et al . , 1999; Harden et al . , 1999 ) or p21-activated kinase ( DPak ) ( Harden et al . , 1996; Conder et al . , 2004; Hofmann et al . , 2004 ) . They also modulate the Arp2/3 complex , which consists of seven subunits conserved in almost all eukaryotes ( Rotty et al . , 2013; Veltman and Insall , 2010 ) . The Arp2/3 complex promotes the formation of densely branched , rapidly treadmilling actin filament arrays that , together with the Wiskott-Aldrich syndrome protein ( WASP ) and the WASP-family verprolin-homologous protein ( WAVE ) , coordinate membrane-cytoskeleton dynamics ( Lecuit et al . , 2011; Kurisu and Takenawa , 2009; Blanchoin et al . , 2014 ) . The Arp2/3 complex also regulates endocytosis of DE-cadherin ( Georgiou et al . , 2008; Leibfried et al . , 2008 ) and thus contributes to the regulation of the zonula adherens ( ZA ) , an adhesion belt encircling the apex of epithelial cells ( Tepass et al . , 1996; McEwen et al . , 2000; Sarpal et al . , 2012 ) . Moreover , the Drosophila WAVE homolog SCAR , the main activator of Arp2/3 in fly embryos ( Zallen et al . , 2002 ) , is a downstream effector of Rac , Cdc42 and DPak ( Lecuit et al . , 2011; Kurisu and Takenawa , 2009 ) . DPak , in turn , can also activate the Arp2/3 complex independently of SCAR ( Lecuit et al . , 2011; Kurisu and Takenawa , 2009; Zallen et al . , 2002 ) . Thus , the regulation of cell-cell adhesion and cytoskeleton activity is closely linked to each other . During epithelial morphogenesis , mechanisms controlling cell polarity have to be set in place to ensure tissue integrity . One of the key regulators of epithelial cell polarity in the Drosophila embryo is the Crumbs protein complex . Its core components are the type I transmembrane protein Crumbs ( Crb ) and the scaffolding proteins Stardust ( Sdt ) , DLin-7 and DPATJ , which are conserved from flies to mammals [reviewed in ( Bulgakova and Knust , 2009; Tepass , 2012 ) ] . Drosophila embryos mutant for crb or sdt are unable to maintain apico-basal polarity in most of their epithelia ( Tepass and Knust , 1990; 1993; Bachmann et al . , 2001; Hong et al . , 2001 ) . This leads to a complete breakdown of tissue integrity due to failure in positioning and maintaining the ZA , followed by apoptosis in many tissues , e . g . the epidermis and the AS ( Grawe et al . , 1996; Tepass and Knust , 1990; 1993; Tepass , 1996 ) . Comparable defects in epithelial integrity are observed in mice lacking Crb2 or Crb3 ( Whiteman et al . , 2014; Xiao et al . , 2011; Szymaniak et al . , 2015 ) . Conversely , over-expression of Drosophila Crb can lead to an expansion of the apical membrane domain , both in embryonic epithelial cells ( Wodarz et al . , 1995 ) and in photoreceptor cells ( Muschalik and Knust , 2011; Pellikka et al . , 2002; Richard et al . , 2009 ) . These results define Crb as an important apical determinant of epithelial cells . Besides a role in epithelial cell polarity , Drosophila crb controls tissue size in imaginal discs by acting upstream of the Hippo pathway [reviewed in ( Boggiano and Fehon , 2012; Genevet and Tapon , 2011 ) ] , regulates morphogenesis of photoreceptor cells and prevents light-dependent retinal degeneration [reviewed in ( Bazellières et al . , 2009; Bulgakova and Knust , 2009 ) ] . Crb contains in its extracellular domain an array of epidermal growth factor-like repeats , interspersed by four laminin A globular domain-like repeats . Its small cytoplasmic portion of only 37 amino acids contains two highly conserved motifs , a C-terminal PDZ ( Postsynaptic density/Discs large/ZO-1 ) domain-binding motif ( PBM ) , -ERLI , which can bind the PDZ-domain of Sdt and DPar-6 ( Li et al . , 2014; Roh et al . , 2002; Bulgakova et al . , 2008; Bachmann et al . , 2001; Hong et al . , 2001; Kempkens et al . , 2006; Ivanova et al . , 2015 ) , and a FERM ( protein 4 . 1/ezrin/radixin/moesin ) domain-binding motif ( FBM ) ( Klebes and Knust , 2000 ) , which can directly interact with the FERM-domain of Yurt ( Yrt ) , Expanded ( Ex ) and Moesin ( Moe ) ( Laprise et al . , 2006; Ling et al . , 2010; Wei et al . , 2015 ) . Our previous structure-function analysis of Crb using a fosmid-based approach revealed that the PBM is essential for the maintenance of cell polarity in embryonic epithelia ( Klose et al . , 2013 ) . In contrast , the FBM is non-essential for normal development of most embryonic epithelia . At later stages of development , however , embryos with a mutation in the FBM fail to undergo DC ( Klose et al . , 2013 ) . This phenotype now provides access to unravel additional functions of this highly conserved polarity regulator . Using live imaging and genetic analysis we elucidate a novel function of Crb as a key negative regulator of actomyosin dynamics during DC . Our results also further our understanding on the mechanisms that couple the regulation of the cytoskeleton and cell-cell adhesion with the control of embryonic morphogenesis .
We previously showed ( Klose et al . , 2013 ) that a fosmid covering the entire crb locus , named foscrb , completely rescues the lethality caused by the lack of endogenous crb . We also showed that a variant , in which the conserved tyrosine10 in the FERM-domain binding motif ( FBM ) is replaced by an alanine ( foscrbY10A variant ) does not rescue embryonic lethality . Interestingly , the fosCrbY10A variant properly localises at the apical domain in most embryonic epithelia , which undergo normal morphogenesis ( i . e . germ band elongation , salivary gland invagination ) . But later in development , germ band ( GB ) retraction , dorsal closure ( DC ) and head involution fail to occur properly ( Klose et al . , 2013 ) . This indicated that the FBM of Crb fulfils a tissue- and stage-specific morphogenetic function in the embryo . Moreover , these defects appear to be independent of a putative Tyr phosphorylation , because another variant , in which the Y10 is replaced by a phenylalanine ( foscrbY10F ) , completely rescues the embryonic lethality of crb mutants ( Klose et al . , 2013 ) . To get a better understanding of the mechanisms by which Crb regulates these morphogenetic processes , we performed detailed in vivo analyses of embryos expressing the different fosmid variants together with a DE-cad::GFP or a DE-cad::mTomato knock-in allele ( Huang et al . , 2009 ) in a crb null background ( crbGX24 or crb11A22 ) ( for simplicity , these are called foscrb , foscrbY10A and foscrbY10F from now on ) . Because staging of embryos depends on morphological criteria , and foscrbY10A mutant embryos show morphological defects , we imaged control and mutant embryos always in parallel , and stages were classified according to elapsed time after egg collection , i . e . , after equal developmental times ( see Materials and methods for details about staging and imaging ) . By the time foscrb embryos finish GB retraction ( Figure 1A , Video 1 ) , foscrbY10A embryos ( Figure 1B , Video 2 ) exhibit major defects in GB retraction , as revealed by a highly disorganised amnioserosa ( AS ) in which individual AS cells could hardly be followed . While foscrb embryos proceed through DC ( Figure 1C , E , Video 1 ) , those expressing the foscrbY10A variant progressively lose the AS ( Figure 1D , F ) and ultimately fail to complete DC ( Video 2 ) . Embryos expressing the foscrbY10F variant complete DC similar as foscrb embryos ( Figure 1—figure supplement 1 ) , indicating that the Y10A mutation specifically affects the progress of DC . 10 . 7554/eLife . 07398 . 003Figure 1 . The FERM-binding domain motif ( FBM ) of Crb is essential for dorsal closure ( DC ) . ( A-F ) Stills from dorsal views of live imaging of embryos expressing DE-cad::GFP . In all images the anterior part is towards the left . A , C and E , w;foscrb , DE-cad::GFP;crbGX24 ( Video 1 ) . B , D and F , w;foscrbY10A , DE-cad::GFP;crbGX24 ( Video 2 ) . All embryos were collected at the same time ( 1 hr collection ) , incubated at 28ºC for 7 hr and imaged together . Numbers in ( B , D and F ) indicate the time in minutes for the corresponding row . While DC is completed in foscrb embryos ( E ) , in foscrbY10A embryos , the amnioserosa ( AS ) is disorganised and progressively lost ( F ) . Scale bar: 100 μm . ( G-J’ ) Localisation of phosphotyrosine ( PY ) , Crb and DPatj in the dorsal epidermis at the beginning of DC . In all images the AS is at the top ( see reference axis in G and in the scheme K ) . ( G , I , I’ ) w;foscrb;crbGX24 . ( H , J , J’ ) w;foscrbY10A;crbGX24 . ( K ) Schematic representation of the dorsal epidermis at the beginning of DC indicating that the leading edge ( LE ) of the dorsal most epidermal ( DME ) cells is in contact with the AS . Arrows in ( G , H ) indicate LE of the DME ( row of cells marked by brackets ) . The arrowheads indicate where the corresponding protein is absent from the LE ( I-J’ ) . The asterisks mark LE membranes positive for Crb ( J ) and DPatj ( J’ ) in foscrbY10A mutant . Scale bar: 10 μm . Representative images from 8–12 different embryos for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 00310 . 7554/eLife . 07398 . 004Figure 1—figure supplement 1 . DC in foscrbY10F embryos . ( A-C ) Stills from dorsal views of live imaging of embryos expressing DE-cad::GFP in w;foscrbY10F , DE-cad::GFP;crbGX24 . Embryos collected and imaged as described in Figure 1 . Numbers indicate the time in minutes for the corresponding row . DC proceeds as in foscrb embryos . Scale bar: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 00410 . 7554/eLife . 07398 . 005Video 1 . Dorsal closure ( DC ) in a w;foscrb , DE-cad::GFP;crbGX24 embryo . Note that the granules from the yolk are visible because of their strong auto-fluorescence in the green part of the spectrum . Time-lapse: 3 . 5 min; 12 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 00510 . 7554/eLife . 07398 . 006Video 2 . Defective germ band ( GB ) retraction and DC phenotype in a w;foscrbY10A , DE-cad::GFP;crbGX24 embryo . Time-lapse: 3 . 5 min; 12 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 006 Various mechanisms have been documented to contribute to DC , including elongation of the dorsal most epidermal ( DME ) cells ( Riesgo-Escovar et al . , 1996 ) . This elongation occurs normally in foscrb embryos , as revealed by phosphotyrosine ( PY ) staining associated with the ZA ( Figure 1G ) . In contrast the DME cells of foscrbY10A embryos do not elongate co-ordinately ( Figure 1H ) . We analysed the localisation of Crb and DPatj at this stage . Both proteins are expressed at higher levels in the epidermis compared to the AS ( Figure 1I–J' ) . In foscrb embryos , Crb ( Figure 1I ) and DPatj ( Figure 1I' ) are mostly absent from the leading edge ( LE –Figure 1I–I' arrowheads ) of the DME cells . In contrast , in foscrbY10A embryos both CrbY10A ( Figure 1J , asterisks ) and DPatj ( Figure 1J' , asterisks ) are detected at the LE , particularly in those cells that remain short , while both are removed in cells that elongate properly ( Figure 1J , J' , arrowheads ) . Thus proper elongation of the DME cells fails in foscrbY10A embryos . Besides elongation of the DME cells , a complex actomyosin machinery is established at their LE . The DME cells extend filopodia and lamellipodia that are essential for correct ‘zippering’ ( Young et al . , 1993; Edwards et al . , 1997; Jacinto et al . , 2000; Eltsov et al . , 2015 ) . These filopodia , revealed by staining with an antibody against Stranded at Second [Sas ( Denholm et al . , 2005 ) ] , extend dorsally in foscrb embryos ( Figure 2A arrow ) . In contrast , filopodia in foscrbY10A embryos are disorganised and often absent ( Figure 2B , empty arrowhead and arrowhead , respectively ) . This is confirmed by live imaging of embryos expressing a Venus-tagged Sas protein ( Video 3 ) . Filopodia of foscrbY10A embryos are erratic , and some even appear to move out of the plane ( Video 3 , arrow in foscrbY10A embryo ) , probably because of the loss of contact with the AS . 10 . 7554/eLife . 07398 . 007Figure 2 . The FBM of Crb is important the establishment of the supracellular actomyosin cable at the LE of the DME cells during DC . ( A-L ) Localisation of Stranded at second ( Sas , A , B ) , Enabled ( Ena , C , D ) , Actin ( E , F ) , Zipper ( Zip , E’ , F’ ) , Echinoid ( Ed , G , H ) , phosphotyrosine ( PY , G’ , H’ ) , Bazooka ( Baz , I , J ) , and DE-cadherin ( DE-cad , K , L ) at the beginning of stage 14 . In all images the AS is at the top half , for the genotypes w;foscrb;crbGX24 and w;foscrbY10A;crbGX24 . Filopodia extend dorsally in foscrb embryos ( A , arrow ) , but in foscrbY10A embryos filopodia are absent ( B , arrowhead ) or disorganised ( B , empty arrowhead ) . Ena , Actin and Zip concentrate at the LE in foscrb embryos ( C , E and E’ , arrows ) , but these proteins are almost absent from the LE in foscrbY10A embryos ( D , F and F’ , arrowheads ) . Ed is absent from the LE of foscrb embryos ( G , arrowhead ) , but the DME cells of foscrbY10A embryos show an important decrease of the protein ( H , magenta overlay ) though the PY staining is still clearly associated with the ZA in the same cells ( H’ , magenta overlay ) . Similarly , Baz decreases at the LE of foscrb embryos ( I , arrowhead ) , but in foscrbY10A embryos , the cells that do not elongate keep Baz at the LE ( J , arrow ) , while other DME cells show a reduction of Baz ( J , and Figure 2—figure supplement 3 ) . DE-cad ( mTomato signal ) localises at all cell-cell contacts in foscrb embryos ( K ) . However , in foscrbY10A , the DE-cad localisation is affected in both the dorsal epidermis ( L , solid arrowhead ) and the AS ( L , empty arrowheads ) . Scale bar: 10 μm . ( M ) Schematic representation of the changes in DME cells at the beginning of DC in embryos expressing either fosCrb or fosCrbY10F . The elongation of the DME cells is accompanied by the removal of the Crb protein complex , Ed , Baz and the septate junction components from the LE . At the LE a supracellular actomyosin cable is established and filopodia extend dorsally and attach to the AS cells . Representative images from 8–12 different embryos for each genotype . ( N ) Schematic representation of the defects in the DME cells of embryos expressing the fosCrbY10A variant . At the beginning of DC , the DME cells do not elongate uniformly . In the cells that do not elongate , the Crb protein complex and Baz remain at the LE . Reduced DE-cad suggest defects in the ZA function . Ed is dramatically reduced in DME cells , probably contributing to the absence of the supracellular actomyosin cable . Also , the DME cells exhibit disorganised filopodia . Nevertheless , the septate junction components are properly removed from the LE . The Crb protein complex is apical to the ZA , but Ed and the actomyosin cable are associated with the ZA . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 00710 . 7554/eLife . 07398 . 008Figure 2—figure supplement 1 . Localisation of Pyd , Dia and DAAM in foscrb and foscrbY10F embryos . Localisation of Polychaetoid ( Pyd , A , B ) , Phosphotyrosine ( PY , A’ , B’ ) , Diaphanous ( Dia , C , D ) , and Dishevelled Associated Activator of Morphogenesis ( DAAM , E , F ) in embryos at the beginning of stage 14 . In all images the AS is at the top , for the genotypes w;foscrb;crbGX24 , and w;foscrbY10A;crbGX24 . The localisation of Pyd ( A , B’’ ) is comparable between the different genotypes , despite the irregularly extended DME cells in w;foscrbY10A;crbGX24 embryos ( B , B’ , B’’ ) . The PY staining ( A’ , B’ ) marks the ZA . The localisation of Dia ( C , D ) and DAAM ( E , F ) is similar in the different genotypes . Scale bar: 10 μm . Representative images from 8–12 different embryos for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 00810 . 7554/eLife . 07398 . 009Figure 2—figure supplement 2 . The FBM of Crb is important for the establishment of the supracellular actomyosin cable . Stills from live imaging of embryos expressing Zip::GFP . In all images the anterior part is to the left . ( A-C ) w;foscrb/Zip::GFP;crbGX24 and ( D-F ) w;foscrbY10A/Zip::GFP;crbGX24 embryos were followed during GB retraction . Numbers in ( D-F ) indicate the time in minutes for the corresponding row . Arrow in ( B ) marks the incipient formation of the supracellular actomyosin cable in a foscrb embryo . The supracellular actomyosin cable is continuous at later time points ( C , arrow ) . In foscrbY10Aembryos , some segments of the DME cells concentrate Zip::GFP at the LE ( E , arrow ) . At the time when GB retraction should be completed and thereafter , the actomyosin cable forms randomly at the LE ( F , arrows ) , and several discontinuities are present ( F , arrowheads ) . Scale bar: 100 μm . Representative images from 6–8 different embryos for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 00910 . 7554/eLife . 07398 . 010Figure 2—figure supplement 3 . Reduction of Baz in DME cells of foscrbY10A embryos . Localisation of Bazooka ( Baz , A , B ) , and phosphotyrosine ( PY , A’ , B’ ) at the beginning of stage 14 in w;foscrb;crbGX24 and w;foscrbY10A;crbGX24 embryos . The black lines in A-B’ mark the position for the plot profile ( C , D ) of the Baz signal ( C , D , black line ) and the PY signal ( C , D , magenta line ) in the DME cells . Maxima intensities overlap for both markers , but note that the intensity of Baz in foscrbY10A embryos is lower than in foscrb embryos . The arrows indicate where Baz is preserved at the LE of those cells that do not elongate properly , while the asterisks mark the DME cells that extend normally , and have a reduction of Baz signal in the junctions . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 01010 . 7554/eLife . 07398 . 011Figure 2—figure supplement 4 . Distribution of septate junction components in DME cells . Localisation of Coracle ( Cora , A , B ) , DE-cad ( A’ , B’ ) , Disc large ( Dlg , C , D ) and Yurt ( Yrt , E , F ) in embryos at the beginning of stage 14 . In all images the AS is at the top , for w;foscrb;crbGX24 and w;foscrbY10A;crbGX24 embryos . The septate junction proteins Cora ( A , B ) , Dlg ( C , D ) and Yrt ( E , F ) are absent from the LE in all genotypes ( arrowheads ) . Bracket in ( B ) marks bunching of dorsal epidermis observed in foscrbY10Aembryos . The DE-cad staining ( A’ , B’ ) , is a maximal projection of the first ∼1 . 5 μm from the surface of the embryo , while the Cora staining is a maximal projection of the whole Z-stack . The merge of these projections ( A’’ , B’’ ) shows that Cora is mainly present in the epidermis . Scale bar: 10 μm . Representative images from 8–12 different embryos for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 01110 . 7554/eLife . 07398 . 012Figure 2—figure supplement 5 . Distribution of actomyosin and junctional components in DME cells of foscrbY10F embryos . ( A-K ) Localisation of Sas at the filopodia ( A , arrow ) . Ena ( B ) , Actin ( C ) , and Zip ( C’ ) concentrate at the LE ( arrows ) . Ed ( D , and PY , D’ ) , and Baz ( E ) are absent from the LE ( arrowheads ) . DE-cad::mTomato ( F ) and Pyd ( G , and PY , G’ ) localise at all cell-cell contacts . Localisation of Dia ( H ) and DAAM ( I ) . The septate junction components Cora ( J , the corresponding DE-cad , J’ and the merge , J’’ ) , and Dlg ( K ) are absent from the LE ( J , K , arrowheads ) . The localisation of all these proteins is similar to the one observed in foscrb embryos . Scale bar: 10 μm . Representative images from 8–12 different embryos for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 01210 . 7554/eLife . 07398 . 013Video 3 . Filopodia movement at the leading edge ( LE ) of the dorsal most epidermal ( DME ) cells in w;foscrb;crbGX24 , Sas::Venus ( top ) and w;foscrbY10A;crbGX24 , Sas::Venus ( bottom ) embryos . The filopodia at the DME cells were followed for 5 min and the movie loops 6 times . Note that the filopodia in the foscrbY10A embryo move randomly and some filopodia , like the one label with the arrow ( bottom embryo ) , appear to detach and move out of the plane . Time-lapse: 10 sec; 8 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 013 A key regulator of the number and length of filopodia during DC is the actin-elongation promoting protein Enabled ( Ena ) ( Gates et al . , 2007; Nowotarski et al . , 2014; Bilancia et al . , 2014; Homem and Peifer , 2009 ) . Ena concentrates at the LE of DME cells in foscrb embryos ( Figure 2C , arrows ) . In contrast , Ena is strongly reduced at the LE of foscrbY10A embryos ( Figure 2D , arrowhead ) . Localisation of Ena at the LE depends on the ZA–associated protein Polychaetoid ( Pyd ) ( Choi et al . , 2011 ) . However , Pyd localisation at the ZA shows no major difference in foscrb and foscrbY10A embryos ( Figure 2—figure supplement 1A–B''' ) . The localisation of the formins Dia and DAAM , both involved in the growth of actin-based protrusions ( Matusek et al . , 2006; Homem and Peifer , 2008; Liu et al . , 2010 ) , is also similar in foscrb and foscrbY10A embryos ( Figure 2—figure supplement 1C–F ) . This suggests that different regulators of Ena are affected in foscrbY10A mutant embryos . In addition to filopodia , forces produced by a supracellular actomyosin cable at the LE contribute to DC ( Franke et al . , 2005; Hutson et al . , 2003; Kiehart et al . , 2000; Jacinto et al . , 2002; Young et al . , 1993 ) . This supracellular cable , which contains actin ( Figure 2E ) and the non-muscle myosin II Zipper ( Zip , Figure 2E' ) , is correctly formed in foscrb embryos ( Figure 2E , E' arrows ) . However , it is virtually absent in foscrbY10A embryos ( Figure 2F , F' , arrowheads ) . Live imaging experiments using a zipper::GFP protein trap line ( Buszczak et al . , 2007; Morin et al . , 2001 ) reveal that Zip::GFP appears homogenously along the LE in foscrb embryos . In contrast , it randomly concentrates in some segments along the LE of foscrbY10A embryos ( Figure 2—figure supplement 2 ) . Together , these results show that the FBM of Crb is important for the generation and maintenance of actin-based protrusions and the correct organisation of the supracellular actomyosin cable at the LE . The formation of the actomyosin cable at the LE depends on the removal of the adhesion protein Echinoid ( Ed ) from the LE and the AS cells ( Laplante and Nilson , 2011; Lin et al . , 2007 ) . As expected , Ed in foscrb embryos is distributed as in wild type embryos ( Figure 2G , arrowheads mark Ed absence at the LE ) . However , in foscrbY10A embryos , Ed levels are strongly reduced in the DME cells ( Figure 2H , magenta overlay ) , even though the DME cells are still in contact with the AS , as revealed by PY staining ( Figure 2H' ) . It has been suggested that the asymmetric distribution of Ed is essential to exclude the polarity protein Bazooka ( Baz ) away from the LE ( Laplante and Nilson , 2011; Pickering et al . , 2013 ) . We found that , in contrast to foscrb embryos ( Figure 2I , arrowhead ) , foscrbY10A embryos preserve Baz at the LE of those cells that fail to elongate ( Figure 2J , arrow ) . In addition , there is a general reduction of Baz at the junctions of the DME cells of foscrbY10A embryos ( Figure 2—figure supplement 3 ) . Together , these results suggest that the FBM of Crb is important for Ed stability and hence Baz redistribution and amount in DME cells . The asymmetric distribution of different proteins in the DME cells reflects the planar cell polarity of these cells , a feature that also includes the removal of septate junction ( SJ ) components from the LE ( Kaltschmidt et al . , 2002 ) . We found that removal of Coracle ( Cora ) , Discs Large ( Dlg ) and Yurt ( Yrt ) from the LE appears normal in the different fosmid variants ( Figure 2—figure supplement 4 ) , suggesting that not all aspects of the planar polarisation of the DME cells are affected in embryos expressing the foscrbY10A variant . Ed , Baz and DE-cadherin ( DE-cad ) are all proteins associated with the ZA , which is essential in maintaining adhesion between the dorsal epidermis and the AS and for transmitting the forces generated during DC ( Gorfinkiel and Arias , 2007; Heisenberg and Bellaiche , 2013; Lecuit et al . , 2011 ) . In foscrb , DE-cad localises at all cell-cell contacts , including the LE ( Figure 2K , arrow ) . In foscrbY10A embryos , however , the DE-cad signal is strongly reduced at the LE ( Figure 2L , solid arrowhead ) . Moreover , disruption of DE-cad suggests a discontinuous adhesion belt in the AS cells of these embryos ( Figure 2L , empty arrowheads ) . The loss of DE-cad from the LE in the foscrbY10A embryos at this early stage is different from the normal redistribution of DE-cad that occurs at late stages during the zippering phase ( Gorfinkiel and Arias , 2007 ) . As expected , in foscrbY10F embryos , all proteins mentioned above localise as in foscrb embryos ( Figure 2—figure supplement 5 ) . Taken together , these results show that the DC phenotype in foscrbY10A embryos is accompanied by defects in the establishment of the complex actomyosin apparatus at the LE of the DME cells and by the disturbance or even loss of different components of the ZA ( schematised in Figure 2M , N ) . As described above , GB retraction is defective and the AS is strongly disorganised in foscrbY10A embryos ( Figure 1F ) . Because the AS is required during GB retraction ( Lamka and Lipshitz , 1999; Lynch et al . , 2013; Scuderi and Letsou , 2005 ) , we analysed by live imaging whether the AS is affected before GB retraction . In foscrb and foscrbY10A embryos , at the beginning of stage 11 , AS cells are elongated along the antero-posterior axis ( Figure 3A , D ) , highlighted by DE-cad::mTomato along the ZA ( Figure 3B , E , arrows ) . In foscrbY10A embryos , however , the continuity of DE-cad::mTomato is frequently disrupted ( Figure 3E , arrowhead ) and DE-cad::mTomato additionally appears in large intracellular clusters of unknown identity ( Figure 3E , concave arrowheads ) , which are never observed in foscrb embryos . As GB retraction proceeds , fragmentation of the ZA continues in the AS of foscrbY10A embryos and the tissue disintegrates ( Figure 3F arrowheads and Video 4; and for a dorsal view of a different set of embryos see Video 5 ) , while the dorsal aspect of foscrb embryos is covered by a continuous epithelial sheet ( Figure 3C ) . 10 . 7554/eLife . 07398 . 014Figure 3 . The FBM of Crb is important for the maintenance of the AS . ( A-F ) Stills from lateral views of live imaging of DE-cad::mTomato knock-in at the beginning of germ band ( GB ) retraction ( Video 4 ) . In all images the anterior part is towards the left , for the genotypes w;foscrb , DE-cad::mTomato;crbGX24 and w;foscrbY10A , DE-cad::mTomato;crbGX24 . All embryos were collected at the same time ( 1 hr collection ) , incubated at 28ºC for 5 hr and imaged together . The numbers in ( D , F ) indicate the time in min . for the corresponding row . At stage 11 ( A , B , D , E ) , the AS cells are elongated along the AP-axis , and DE-cad::mTomato localises along the ZA ( B , E , arrows ) ; in foscrbY10A mutant , the continuity of DE-cad::mTomato along the ZA is lost ( E , arrowhead ) and DE-cad::mTomato is also found in large clusters ( E , white concave arrowhead ) . At the end of GB retraction the AS covers the dorsal aspect of foscrb embryos ( E ) , but in foscrbY10A ( F ) , GB retraction is impaired and DE-cad::mTomato signal is fragmented in the AS ( F , arrowheads ) . Scale bar: 100 μm , except for ( B , E ) 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 01410 . 7554/eLife . 07398 . 015Figure 3—figure supplement 1 . The FBM of Crb is important for the integrity of the AS . ( A-B’ ) Scanning electron micrographs of dorsal views of embryos incubated for 8 hr at 28ºC after egg collection ( 1 hr collection ) for the genotypes w;foscrb;crbGX24 and w;foscrbY10A;crbGX24 . The boxed area in ( A , B ) is shown in ( A’ , B’ ) respectively . In foscrb embryos ( A’ ) the AS appears as a flat continuous monolayer , while in foscrbY10Aembryos ( B’ ) , the AS is disorganised and some cells exhibit large filopodia ( B , B’ , arrow ) . Other cells are completely detached and may be AS cells or haemocytes ( B , B’ , arrowheads ) , and some cells have the appearance of apoptotic cells ( B’ , concave arrowhead ) . Scale bars: 100 μm ( A , B ) and 10 μm ( A’ , B’ ) . Representative images from 17–37 embryos for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 01510 . 7554/eLife . 07398 . 016Video 4 . Lateral views during germ band ( GB ) retraction in w;foscrb , DE-cad::mTomato;crbGX24 ( top ) and w;foscrbY10A , DE-cad::mTomato;crbGX24 ( bottom ) embryos . Time-lapse: 10 min; 8 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 01610 . 7554/eLife . 07398 . 017Video 5 . Dorsal views during GB retraction and the beginning of DC in w;foscrb , DE-cad::GFP;crbGX24 ( top ) and w;foscrbY10A , DE-cad::GFP;crbGX24 ( bottom ) embryos . Note that the yolk aggregates are clearly visible because they have an intense autofluorescence in the green part of the spectrum . Time-lapse: 10 min; 8 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 017 The defects of the AS in foscrbY10A embryos become very obvious in scanning electron micrographs ( Figure 3—figure supplement 1 ) . At stage 14 , the AS forms a flat monolayer of epithelial cells in foscrb embryos ( Figure 3—figure supplement 1A , A' ) . In contrast , in foscrbY10A embryos developed for the same period of time , the AS is completely disorganised . Large processes form , some of which extend over the caudal end of the embryos ( Figure 3—figure supplement 1B , B' , arrow ) . Some isolated cells are visible over the epidermis ( whether these are detached AS cells or migrating haemocytes was not determined –Figure 3—figure supplement 1B , arrowhead ) , while others have the appearance of apoptotic cells ( Figure 3—figure supplement 1B' , concave arrowhead ) . Together , these observations suggest that cell-cell adhesion in the AS is strongly disrupted in foscrbY10A embryos , and define the FBM of Crb as an important regulator of cytoskeletal organisation and cell-cell adhesion of the AS . Our scanning electron microscopy analyses suggest that the AS of foscrbY10A embryos undergo apoptosis . In order to determine whether apoptosis contributes to the disruption of the AS , we used the apoptotic reporter Apoliner , an RFP-GFP fusion protein localising at cell membranes of live cells . Caspase activation releases the GFP moiety , which is relatively unstable after cleavage , so dying cells have a stronger red appearance ( Bardet et al . , 2008; Kolahgar et al . , 2011 ) . Apoliner expression in the AS ( specifically driven by the line GAL4332 . 3 ) of foscrb embryos ( Video 6 ) revealed some apoptotic cells at the posterior canthus at the end of GB retraction ( Figure 4A , arrow ) . In foscrbY10A embryos developed for the same period of time , more apoptotic cells are visible , some of which detach ( Figure 4B , arrowheads ) , while others remain attached to the posterior edge of the remaining AS ( Figure 4B , arrow ) . As DC progresses in foscrb embryos , some apoptotic cells delaminate from the AS and are easily distinguished ( Video 6 , blinking arrows –some of these cells could be hemocytes with engulfed apoptotic debris , as reported previously [Bardet et al . , 2008] ) . At this stage , almost all AS cells in foscrbY10A embryos are apoptotic ( Video 6 , compare embryos at 210 min ) . Finally , at the end of DC , the internalised AS cells are localised in a central rod-like structure in foscrb embryos and subsequently die by apoptosis ( Figure 4C ) [as has been reported for wild type embryos ( Reed et al . , 2004; Shen et al . , 2013 ) ] , while in foscrbY10A embryos at this time point the remaining AS cells are completely disaggregated ( Figure 4D ) . To summarise , the AS in foscrbY10A embryos breaks apart and undergoes premature apoptosis ( Video 6 ) , supporting the conclusion that an intact FBM is required for maintaining the integrity of the AS . 10 . 7554/eLife . 07398 . 018Video 6 . Dorsal views during DC in w;foscrb , GAL4332 . 3/foscrb , UAS-Apoliner;crbGX24 ( first row ) , and two examples of w;foscrbY10A , GAL4332 . 3/foscrbY10A , UAS-Apoliner;crbGX24 ( second and third rows ) embryos . Apoliner GFP signal is on the left ( green ) , the RFP signal on the middle ( magenta ) , and the merge on the right . At the time 210 min , the blinking arrows in the merge of the foscrb embryo indicate some apoptotic AS cells separated clearly . Time-lapse: 10 min; 8 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 01810 . 7554/eLife . 07398 . 019Figure 4 . AS detachment in foscrbY10A embryos is accompanied by premature apoptosis . ( A-D ) Stills from dorsal views of live imaging of embryos in which the apoptotic reporter Apoliner is driven in the AS with the line GAL4332 . 3 ( Video 6 ) . Apoptotic cells in magenta appear more intense than their neighbours . In all images the anterior part is towards the left for the genotypes w;foscrb , GAL4332 . 3/foscrb , UAS-Apoliner;crbGX24 , and w;foscrbY10A , GAL4332 . 3/foscrbY10A , UAS-Apoliner;crbGX24 . All embryos were collected at the same time ( 1 hr collection ) , incubated at 28ºC for 7 hr and imaged together . The numbers in ( B , D ) indicate the time in minutes for the corresponding row . After GB retraction in foscrb embryos ( A ) , some apoptotic cells are found mainly at the posterior canthus ( A , arrow ) . In comparison , in foscrbY10A embryos , some of the cells that have detached from the AS ( B , arrowheads ) , as well as those in the posterior edge of the AS ( B , arrow ) , are apoptotic . As DC is completed in foscrb embryos ( C ) , a significant portion of the internalised AS cells are apoptotic , while the remaining internalised cells are still localised in a rod-like structure along the dorsal part of the embryo . In contrast , in foscrbY10A embryos ( D ) all the remaining AS cells are apoptotic cells ( the GFP signal in ( D ) does not belong to the AS ) . Scale bar: 100 μm . Representative images from 8–12 different embryos for each genotype . ( E-K ) Activation of the JNK pathway in the DME cells analysed with the enhancer trap pucE69 ( β–galactosidase staining ) . DE-cad staining is in green . In all images anterior is to the left for the genotypes w;foscrb/+;crbGX24/pucE69 , crbGX24 and w;foscrbY10A/+;crbGX24/pucE69 , crbGX24 . From the beginning to the end of DC , Puc expression is normally induced on each side of the embryo in the single row of DME cells in both genotypes , and few positive β–gal nuclei appear below the row of DME cells ( E , F , arrowheads ) . In foscrbY10A embryos at middle DC some β–gal positive cells appear below the DME cells ( H , arrowheads ) . When DC is completed in foscrb embryos ( I ) , a single row of cells on each side of the embryo is β–gal positive , even in foscrbY10A embryos , independently of whether the epidermis contacted the corresponding segment of the epidermis on the dorsal midline ( J , dashed line ) , bunched on the same side of the embryo ( J , dotted line ) or fail to touch the complementing segment ( J , arrow ) . Scale bar: 10 μm . ( K ) No significant difference in the number of β–gal positive nuclei at middle DC along 50 μm at the dorsal epidermis ( indicated by the brackets in G , H ) , mean ± SD , n= 17 embryos per genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 01910 . 7554/eLife . 07398 . 020Figure 4—figure supplement 1 . Hindsight expression in foscrb and foscrbY10A embryos . ( A-D ) Expression of Hindsight ( Hnt ) at stage 12 ( A , C , lateral view ) and stage 14 ( B , D , dorsal view ) . In all images the AS is inside the green dotted line . Note that the AS is properly specified in foscrb and foscrbY10A embryos , and at stage 14 , Hnt staining is comparable between the two genotypes ( B , D ) , and Hnt is present even in the cells that have detached from the AS in the foscrbY10A embryos ( D , arrowhead ) . Scale bar: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 02010 . 7554/eLife . 07398 . 021Figure 4—figure supplement 2 . Localisation of integrin βPS in the AS of foscrb and foscrbY10A embryos . ( A , B ) The localisation of the integrin-βPS is similar in foscrb and foscrbY10Aembryos . The images are projections of ∼1 μm thickness; thus , in some cells it is possible to see the localisation of the integrin-βPS at the basal membrane ( arrows ) , while in other cells it is possible to see the protein localisation at the lateral membrane ( arrowheads ) . The inserts are magnification of a single confocal plane ( 0 . 45 μm ) through the middle part of the AS cells in the respective genotypes . Scale bars: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 02110 . 7554/eLife . 07398 . 022Figure 4—figure supplement 3 . Localisation of DPatj and Yrt in the dorsal epidermis . ( A-C’’ ) Cross section ( ZX view –see reference axis in Figure 1K ) of the dorsal epidermis of embryos at stage 14 stained for DPatj ( green ) and Yrt ( fire LUT-pseudocolor ) . In all images the apical aspect of the cells is at the top and the dotted line marks the basal aspect . ( A-A’’ ) w;foscrb;crbGX24 . ( B-B’’ ) w;foscrbY10F;crbGX24 . ( C-C’’ ) w;foscrbY10A;crbGX24 . Note that Yrt is concentrated toward the apical aspect of the cells in all genotypes . Scale bar: 5 μm . Representative images from 8–12 different embryos for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 02210 . 7554/eLife . 07398 . 023Figure 4—figure supplement 4 . JNK signalling is normal in foscrbY10F embryos . ( A-C ) Activation of the JNK pathway in the DME cells analysed with the enhancer trap pucE69 ( β–galactosidase staining ) . DE-cad staining is in green . In all images anterior is to the left . From the beginning to the end of DC , Puc expression is normally induced on each side of the embryo in the single row of DME cells . When DC is completed , a single row of cells on each side of the embryo is β–gal positive ( C ) . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 023 Several other processes are required for proper DC and integrity of the AS . At early stages , specification of the AS requires the U-shaped-group of genes ( hindsight –hnt , tail-up –tup , u-shaped –ush , and serpent –srp ) , mutations in which produce phenotypes similar to those observed in foscrbY10A embryos ( Frank and Rushlow , 1996; Lamka and Lipshitz , 1999; Yip et al . , 1997; Scuderi and Letsou , 2005; Lynch et al . , 2013 ) . Hnt shows a strong and comparable expression pattern in the AS of foscrb and foscrbY10A embryos at early and late stages ( Figure 4—figure supplement 1 ) , even in the detached AS cells of foscrbY10A embryos ( Figure 4—figure supplement 1D , arrowhead ) . This indicates that fate specification is not affected in foscrbY10A embryos . AS integrity also requires integrin-mediated attachment to the yolk sac membrane ( Reed et al . , 2004 ) . Therefore , we analysed the localisation of integrin-βPS , and found no major differences between foscrb and foscrbY10A embryos ( Figure 4—figure supplement 2A , B ) . Yrt function is also important during DC , and zygotic yrt mutants have DC defects ( Hoover and Bryant , 2002 ) , similar to the ones observed upon Crb over-expression in the AS ( Harden et al . , 2002; Wodarz et al . , 1995 ) . Because Yrt is a FERM protein that negatively regulates Crb by directly interacting with its FBM ( Laprise et al . , 2006 ) , Yrt appeared as a likely candidate in mediating the foscrbY10A mutant phenotype . Yrt localises at the lateral domain and concentrates towards the apical aspect in a Crb-dependent manner from stage 13 onwards ( Laprise et al . , 2006 ) . We found that independently of the fosmid genotype , Yrt concentrates correctly towards the apical aspect of the cells ( Figure 4—figure supplement 3 ) . Moreover , embryos expressing foscrb and lacking zygotic yrt show defects in DC mainly after GB retraction , when a failure in the zippering at the posterior canthus is patent ( Video 7 , arrow in the upper embryo ) . Despite this , the overall AS integrity is preserved during DC and most of the zippering is completed , leaving a hole only at the posterior canthus . This phenotype is completely different from the phenotype of foscrbY10A embryos described above ( Video 2 ) . Significantly , embryos with both the zygotic yrt mutant allele and the foscrbY10A variant do not show amelioration of the foscrbY10A phenotype ( Video 7 , bottom embryo ) . These embryos show strong defects in GB retraction , and the integrity of the AS is lost as development progresses . These results show that the DC phenotype of foscrbY10A embryos starts earlier in development and is more complex than that in yrt mutants , as the former fail in germ band retraction , lose the AS and do not progress on the zippering process . Thus , Yrt seems not to be involved in the phenotype of foscrbY10A embryos . 10 . 7554/eLife . 07398 . 024Video 7 . DC in yrt∆75a zygotic mutants expressing the different fosmids . w;foscrb , DE-cad::GFP;yrt∆75acrb11A22 ( top ) and w;foscrbY10A , DE-cad::GFP;yrt∆75acrb11A22 ( bottom ) embryos . The arrow in the top embryo marks the characteristic defects in the posterior canthus observed during DC in yrt∆75a zygotic mutants . In the w;foscrbY10A , DE-cad::GFP;yrt∆75acrb11A22 embryo the GB retraction and the DC phenotypes are comparable to the ones in the w;foscrbY10A , DE-cad::GFP;crbGX24 ( Video 2 ) . Time-lapse: 6 min; 12 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 024 The AS regulates aspects of DME differentiation ( Stronach and Perrimon , 2001 ) and embryos carrying mutations in components of the JNK signalling pathway show defective elongation of DME cells and fail to establish the supracellular actomyosin cable at the LE ( Riesgo-Escovar et al . , 1996; Martín-Blanco et al . , 1998; Ricos et al . , 1999; Glise et al . , 1995; Hou et al . , 1997; Kockel et al . , 1997; Reed et al . , 2001; Ríos-Barrera and Riesgo-Escovar , 2013 ) . The mutant phenotype described here is characterised by defects in both the AS and the DME cells . To assess whether defects in the DME observed in foscrbY10A embryos are the result of impaired JNK signalling , we used the reporter line puc-lacZ ( Martín-Blanco et al . , 1998; Ring and Martinez Arias , 1993 ) . At the beginning of DC , the DME cells of foscrb and foscrbY10A embryos are β-gal positive ( Figure 4E , F ) , with few lacZ-positive nuclei in the row of cells ventral to DME cells ( Figure 4E , F , arrowheads ) . At advanced DC , foscrb embryos still show a single row of β-gal positive cells ( Figure 4G ) , while in foscrbY10A embryos β-gal positive nuclei can also be found at positions more ventral to the DME cells ( Figure 4H , arrowheads ) . However , given that there is no significant difference in the number of β-gal positive nuclei along 50 μm of the dorsal epidermis between these genotypes ( Figure 4G , H , brackets and 4K ) , we suggest that this phenotype is the result of aberrant elongation of the DME cells in foscrbY10A embryos ( see for example Figure 1H ) . Accordingly , at the time when foscrb embryos complete DC , these embryos ( Figure 4I ) and foscrbY10A embryos exhibit a single row of β-gal positive cells on each side of the dorsal epidermis ( Figure 4J ) . This is independent of whether the epidermis fuses on the dorsal midline ( Figure 4J , encircled by dashed line ) , closes on the same side of the epidermis , thus causing bunching of the tissue ( Figure 4J , encircled by dotted line ) or does not touch any contra-lateral epidermis ( Figure 4J , arrow ) . A normal activation of JNK signalling is also observed in foscrbY10F embryos ( Figure 4—figure supplement 4 ) , showing that JNK signalling appears to be normal in the DME cells of foscrbY10A embryos . Taken together , these results support the conclusion that the FBM of Crb is an important regulator of the integrity and morphogenesis of the AS without affecting its specification during development . It has been previously shown that perturbing actomyosin dynamics of the AS cells interferes with normal DC ( Solon et al . , 2009; Gorfinkiel et al . , 2009; Fischer et al . , 2014 ) . These dynamics , which are evident in stage 13 foscrb embryos ( Video 8 ) similar as in wild-type embryos , is characterised by pulsed contractions of the AS cells . In foscrbY10A embryos , however , the pulsed contraction are difficult to follow , since individual cells can hardly be distinguished due to the highly disrupted ZA ( Video 8 , compare Figure 5A and 5B ) . Pulsed-contraction of wild-type AS cells has been correlated with a regular appearance and disappearance of medial actomyosin foci ( Blanchard et al . , 2010; David et al . , 2010; Solon et al . , 2009 ) . These actomyosin foci are observed in foscrb embryos as revealed by Zip::GFP ( Video 9 and Figure 5C ) . Kymographs show that these foci are transient and disassemble after contraction ( Figure 5C’ , D’ ) . In contrast , the AS of foscrbY10A embryos shows more Zip::GFP foci ( Figure 5D ) , some of which are more prominent ( Figure 5D’ , and Figure 5—figure supplement 1 and Video 10 ) . A similar behaviour was observed for F-actin ( labelled with Utrophin::GFP ( Rauzi et al . , 2010 ) -data not shown ) . Importantly , analysis of the periodicity of foci formation shows that foscrb and foscrbY10F embryos have similar pulsed contractions , while foscrbY10A embryos have aberrant contractions , in that foci are more persistent ( Figure 5E ) . These observations support the hypothesis that the AS of embryos expressing the CrbY10A variant is under both constant and uncoordinated contraction . 10 . 7554/eLife . 07398 . 025Video 8 . Dorsal views during the pulsed contractions of AS cells in w;foscrb , DE-cad::GFP;crbGX24 ( left ) and w;foscrbY10A , DE-cad::GFP;crbGX24 ( right ) . Time-lapse: 10 sec; 15 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 02510 . 7554/eLife . 07398 . 026Figure 5 . The FBM of Crb is essential for the regulation of actomyosin activity in the AS . Stills from views of the AS in live imaging of embryos expressing DE-cad::GFP knock-in ( A , B , Video 8 ) or Zip::GFP ( C-D’ , Video 9 ) . In all images the anterior part is towards the left . Scale bar: 10 μm . ( A ) w;foscrb , DE-cad::GFP;crbGX24 . ( B ) w;foscrbY10A , DE-cad::GFP;crbGX24 . ( C ) w;foscrb/Zip::GFP;crbGX24 . ( D ) w;foscrbY10A/Zip::GFP;crbGX24 . The embryos were collected during 30 min , incubated at 28ºC for 7 hr and imaged under the same conditions . The numbers in ( C , D ) indicate the time in seconds for the corresponding frame in Video 9 . In foscrb embryos ( A ) , DE-cad::GFP is localised at cell-cell junctions; but in foscrbY10A ( B ) embryos DE-cad::GFP continuity is strongly disturbed . ( C’ , D’ ) Kymographs of the Zip::GFP foci in the magenta box in ( C , D ) . Scale bar in ( C’ ) 10 sec . ( E ) Histogram of the relative frequency of Zip::GFP foci duration during the pulsed contractions of the AS in w;foscrb/Zip::GFP;crbGX24 , w;foscrbY10F/Zip::GFP;crbGX24 and w;foscrbY10A/Zip::GFP;crbGX24 embryos . The graph in the insert shows all data points collected , and indicates the mean ± SD . ANOVA test followed by a Dunnett’s multiple-comparison test; ns-not significant difference . n = 150 foci collected from each of the three different embryos . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 02610 . 7554/eLife . 07398 . 027Figure 5—figure supplement 1 . The FBM of Crb regulates the actomyosin activity in the AS . Stills from Video 10 where a Zip::GFP cluster forms and disappears ( followed by the bracket ) in an AS cell during the pulsed contraction in a w;foscrb/Zip::GFP;crbGX24 embryo ( A ) . In contrast , several Zip::GFP foci are present and do not disappear in the w;foscrbY10A/Zip::GFP;crbGX24 embryo ( B ) . Scale bar: 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 02710 . 7554/eLife . 07398 . 028Video 9 . Dorsal views during the pulsed contractions of AS cells in w;foscrb/Zip::GFP;crbGX24 ( left ) and w;foscrbY10A/Zip::GFP;crbGX24 ( right ) . Time-lapse: 10 sec; 15 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 02810 . 7554/eLife . 07398 . 029Video 10 . Magnifications of a small group of cells shown in the Video 11 to see in more detail the medial foci accumulation of Zip::GFP during the cell contraction . These magnifications ( 2X from original ) were created using a bicubic algorithm in Fiji . w;foscrb/Zip::GFP;crbGX24 ( left ) and w;foscrbY10A/Zip::GFP;crbGX24 ( right ) . Time-lapse: 10 sec; 15 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 02910 . 7554/eLife . 07398 . 030Video 11 . Dorsal views during DC in embryos expressing the phosphatase Flw in the AS cells under the control of the GAL4332 . 3 driver . The signal from the UAS-Actin::RFP is not shown . w;foscrb , GAL4332 . 3/UAS-flw-HA , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFP ( top ) and w;foscrbY10A , GAL4332 . 3/UAS-flw-HA , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFP ( bottom ) . Time-lapse: 5 min; 12 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 030 The activity of non-muscle myosin-II ( Zip ) is mainly regulated by the phosphorylation state of the myosin-regulatory light chain [reviewed in ( Tan et al . , 1992 ) ] , encoded by the gene spaghetti squash ( sqh ) . Thus , if over-active actomyosin is responsible for the DC defects of foscrbY10A embryos , we expect that expressing Flapwing ( flw ) , the major Drosophila Sqh phosphatase ( Vereshchagina et al . , 2004 ) , may suppress the DC defects . In fact , UAS-driven expression of Flw in the AS of foscrbY10A embryos leads to a suppression of the DC phenotype ( Figure 6D–F , Video 11 ) , while it does not produce any evident dominant phenotype in foscrb or foscrbY10F embryos ( Figure 6A–C , and Figure 6—figure supplement 1 ) . Interestingly , Flw over-expression also suppresses the disruption of the ZA in the AS ( Video 12 , compare B vs . D ) . This result supports our hypothesis that the FBM of Crb negatively regulates actomyosin activity in the AS . 10 . 7554/eLife . 07398 . 031Figure 6 . Expression of the myosin phosphatase Flapwing in the AS of foscrbY10A embryos suppresses the DC defects . ( A-F ) Stills from dorsal views of live imaging of embryos expressing DE-cad::GFP knock-in and Flw-HA in the AS cells under the control of the GAL4332 . 3 driver ( Video 11 ) , for the genotypes w;foscrb , GAL4332 . 3/UAS-flw-HA , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFP and w;foscrbY10A , GAL4332 . 3/UAS-flw-HA , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFP . All embryos were collected at the same time ( 1 hr collection ) , incubated at 28ºC for 7 hr and imaged together . The numbers on ( D-F ) indicate the time in minutes for the corresponding row . The over-expression of Flw-HA in the AS cells does not produce any obvious phenotype in foscrb ( A-C ) embryos , and it suppresses the DC defects in foscrbY10A ( D-F ) embryos; some defects found include an irregular zippering at the posterior canthus ( E , arrow ) as well as bunching of the dorsal epidermal ( F , bracket ) . Scale bar: 100 μm . Representative images from 6–9 different embryos for each genotype . ( G ) Scheme of the possible pathways regulated by the FBM of Crb in the AS . Crb: Crumbs; Rok: Rho-kinase; Dpak: Drosophila p21-activated kinase; Flw: Flapwing; DMBS: Drosophila myosin-binding-subunit; Sqh: spaghetti-squash; Mlck: myosin-light chain kinase . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 03110 . 7554/eLife . 07398 . 032Figure 6—figure supplement 1 . Normal DC after Flapwing expression in the AS of foscrbY10F embryos . ( A-C ) Stills from dorsal views of live imaging of embryos expressing DE-cad::GFP knock-in and Flw-HA in the AS cells under the control of the GAL4332 . 3 driver , for the genotype w;foscrbY10F , GAL4332 . 3/UAS-flw-HA , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFP . Embryo collection , incubation and imaging as described in Figure 6 . The numbers on ( A-C ) indicate the time in minutes for the corresponding row . The over-expression of Flw-HA in the AS cells does not produce any obvious phenotype . Scale bar: 100 μm . Representative images from 7 different embryos . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 03210 . 7554/eLife . 07398 . 033Video 12 . Flw expression in the AS of foscrbY10A embryos suppresses the disruption of the ZA . Dorsal views during the pulsed contractions of AS cells . The signal from the UAS-Actin::RFP is not shown . ( A , B ) Embryos that do not express the Flw and are trans-heterozygous for DE-cad::GFP; ( A ) w;foscrb/UAS-flw-HA , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFP and ( B ) w;foscrbY10A/UAS-flw-HA , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFP . ( C , D ) Embryos that express Flw in the AS cells under the control of the GAL4332 . 3 driver; ( C ) w;foscrb , GAL4332 . 3/UAS-flw-HA , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFP and ( D ) w;foscrbY10A , GAL4332 . 3/UAS-flw-HA , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFP . Time-lapse: 10 sec; 15 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 033 Rho GTPases have been shown to stimulate myosin contraction by activating Rho-kinase ( Rok ) or the p21-activated kinase ( DPak ) , and are required for proper DC ( Mizuno et al . , 1999; Harden et al . , 1999; 1996; Conder et al . , 2004; Magie et al . , 1999; 2002 ) . To test whether Rho-GTPases are involved in the Crb-mediated DC phenotype , we expressed different versions of established Rho family effectors ( see working model in Figure 6G ) and examined their effects on DC in the embryonic cuticle , a suitable read-out of DC . We grouped the embryos according to their cuticle phenotype into two major categories ( Figure 7A ) : 1 ) embryos with “DC-defect” , which exhibit a range of defects from extensive dorsal opening ( in which the mouthparts are exposed ) , to embryos with complete DC , which , however , still failed to hatch; and 2 ) embryos with “WT-like” cuticle , which includes all those that hatch ( for more details about the different categories and phenotypes see Figure 7—figure supplement 1 ) . Depending on the crb allelic combination , 89–98% of embryos expressing the foscrbY10A variant fall into the “DC-defect” category ( Figure 7A , 1st-6th black bars ) . 10 . 7554/eLife . 07398 . 034Figure 7 . Reduction in actomyosin activity suppresses the DC defects in embryos expressing the foscrbY10A variant . ( A ) Quantification of the defects observed in cuticle preparations from the genotypes indicated in the graph . For the complete genotype see Figure 7—figure supplement 1 . The category “DC defect” includes a range of defects ranging from cuticles of embryos that completed DC but do not hatch , to cuticles with large DC openings . The category “WT-like” includes all larvae that hatch . For details about the classifications see Figure 7—figure supplement 1 . Note that all the genotypes have the foscrbY10A background , except the ones highlighted in magenta , numbers 18 and 19 , that have the foscrb background . mean ± SD from 2–4 independent crosses . n = total number of cuticles counted for the indicated genotype . Note that suppression of the DC phenotype in foscrbY10A embryos is particularly evident upon expression of Flw-HA ( 10 ) , Pak-AID ( 17 ) , and DE-cad ( 22 ) . ( B-F ) Adult flies of the indicated genotypes . In ( F ) , the arrowhead marks the defects in the dorsal abdomen . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 03410 . 7554/eLife . 07398 . 035Figure 7—figure supplement 1 . Reduction in the actomyosin activity suppresses the DC defects in embryos expressing the foscrbY10A variant . Quantification of the defects observed in cuticle preparations from the genotypes indicated in the graph . In the category “Open cuticle” , the dorsal opening is so prominent that in some cases the mouthparts are exposed ( arrowhead ) . Category “Dorsal hole” corresponds to those cuticles in which a medium ( left picture ) or small ( right picture ) dorsal hole is present , but the anterior part is closed . In the category “Closed but not hatched” , the closure is complete , the puckering of the epidermis is noticeable ( arrowhead ) , but the larvae fail to hatch . In the category “Kinked larvae” , the puckering of the epidermis ( arrowhead ) results in larvae with the tail pointing upwards , so the larvae seem to have a kink . In the category “WT-like” , no defects are evident so the larvae are alike to wild type . mean ± SD from 2–4 independent crosses . n = total number of cuticles counted for the indicated genotype . For the statistical analysis see Table 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 03510 . 7554/eLife . 07398 . 036Figure 7—figure supplement 2 . Phosphorylated DMoesin levels are reduced in embryos expressing the foscrbY10A variant . Localisation of phospho-DMoesin ( P-DMoe , A , B ) in embryos at the beginning of stage 14 . In all images the AS is at the top , for the genotypes w;foscrb;crbGX24 and w;foscrbY10A;crbGX24 . The LE of foscrbY10Aembryo is marked with a magenta line ( B ) . Scale bar: 10 μm . Representative images from 9 different embryos for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 03610 . 7554/eLife . 07398 . 037Figure 7—figure supplement 3 . Weak head phenotype of embryos expressing the foscrbY10A variant . Examples of cuticles with a weak head phenotype: the arrows mark an opening in the anterior part . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 03710 . 7554/eLife . 07398 . 038Table 1 . Statistical analyses of the results shown in the Figure 7—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 038Open cuticleDorsal holeClosed but not hatchedKinked larvaeWT-like1foscrbY10A;crbGX24vs14foscrbY10A/Rho11B;crbGX24∗∗nsns∗∗∗∗2foscrbY10A , DE-cad::GFP;crbGX24vs23foscrbY10A , DE-cad::GFP/SCAR∆37 , DE-cad::GFP;crbGX24∗∗∗∗nsns∗∗∗∗∗∗∗∗vs25foscrbY10A , DE-cad::GFP/Arpc1Q25st , DE-cad::GFP;crbGX24nsnsnsnsns3foscrbY10A , DE-cad::GFP/+;crb11A22/crbGX24vs11flw6/Y/w*;foscrbY10A , DE-cad::GFP/+;crb11A22/crbGX24∗∗∗∗∗∗∗∗nsnsvs15rok2/Y/w*;foscrbY10A , DE-cad::GFP/+;crb11A22/crbGX24∗nsnsnsnsvs24foscrbY10A , DE-cad::GFP/+;crb11A22 , Arp3EP3640/crbGX24nsnsnsns∗∗5foscrbY10A , GAL4332 . 3/DE-cad::GFP;crbGX24/crb11A22vs729ºC foscrbY10A , GAL4332 . 3/UAS-Crbfull length;crbGX24/crb11A22nsnsnsnsnsvs825ºC foscrbY10A , GAL4332 . 3/UAS-Crbfull length;crbGX24/crb11A23nsnsnsns∗∗∗∗vs918ºC foscrbY10A , GAL4332 . 3/UAS-Crbfull length;crbGX24/crb11A24nsnsnsns∗∗∗∗vs12foscrbY10A , GAL4332 . 3/DE-cad::GFP;crbGX24/crb11A22 , UAS-RhoN19∗∗nsnsns∗∗vs13foscrbY10A , GAL4332 . 3/DE-cad::GFP;crbGX24/crb11A22 , UAS-rok . CAT-KGnsnsnsns∗vs16foscrbY10A , GAL4332 . 3/DE-cad::GFP;crbGX24/crb11A22 , UAS-RacN17nsnsnsnsnsvs21foscrbY10A , GAL4332 . 3/DE-cad::GFP;crbGX24/crb11A22 , UAS-Dmoe-mycnsnsnsnsnsvs22foscrbY10A , GAL4332 . 3/UAS-DE-cad , DE-cad::GFP;crbGX24/crb11A22∗∗∗∗∗∗∗nsns∗∗∗∗6foscrbY10A , GAL4332 . 3/DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFPvs10foscrbY10A , GAL4332 . 3/UAS-flw-HA , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFPns∗∗∗∗ns∗∗∗∗∗vs17foscrbY10A , GAL4332 . 3/UAS-DPak-AID , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFPns∗∗∗∗nsns∗∗∗∗vs20foscrbY10A , GAL4332 . 3/UAS-DmoeT559D , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFPns∗nsns∗∗∗∗18foscrb , GAL4332 . 3/UAS-Dpak-myr , DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFPvs19foscrb , GAL4332 . 3/DE-cad::GFP;crbGX24/crb11A22 , UAS-Act::RFP∗∗∗∗∗∗∗∗∗ns∗∗∗∗One-way-ANOVA analysis followed by a Dunnet’s multiple comparisons test between the indicated categories of the different genotypes . Statistical significant difference indicated as follows: ns p>0 . 05; *p≤0 . 05; **p≤0 . 01; ***p≤0 . 001; ****p≤0 . 0001 . Using this read-out , we confirm that over-expression of the myosin phosphatase Flw in the AS strongly suppresses the DC defects of foscrbY10A embryos . In fact , >75% hatch ( Figure 7A , 10th vs . 6th bars ) and even some foscrbY10A adults eclose with no obvious defect ( Figure 7C ) . Interestingly , cuticles from foscrbY10A and hemi- or homozygous for the flw6 allele show an enhanced DC phenotype in comparison with the foscrbY10A with a wild type flw allele ( Figure 7A , 3rd vs . 11th black bars: 91 . 2% to 97 . 1%; and Figure 7—figure supplement 1 , 3rd vs . 11th black bars , completely open cuticle from 27 . 7% to 73 . 5% ) . These results support the conclusion that the FBM of Crb regulates the AS actomyosin dynamics by regulating myosin activity . In line with this conclusion we found that over-expression of dominant-negative Rho ( RhoN19 ) or a kinase-dead Rok ( Rok-CAT-KG ) in the AS of foscrbY10A increases the number of hatched larvae ( Figure 7A , 5th vs . 12th and 13th gray bars: from 2 . 9% to 13 . 4% and 10 . 0% , respectively ) , and the proportion of embryos with open cuticles is reduced ( Figure 7—figure supplement 1 , 5th vs 12th and 13th black bars , from 52 . 7% to 23 . 6% and 28 . 5% , respectively ) . Moreover , Rho11B hemizygosity effectively suppresses the DC defects of foscrbY10A embryos ( Figure 7A , 14th bar vs . 1st black bars , 79 . 2 vs . 98 . 3% ) . In contrast , foscrbY10A embryos hemi- or homozygous for rok2 show no suppression of the DC phenotypes ( Figure 7A , 15th vs . 3rd bars ) , which suggests that rok deficiency may be deleterious in the foscrbY10A background and that other morphological processes dependent on Rok could be affected ( Simões et al . , 2010; Krajcovic and Minden , 2012; Mason et al . , 2013; Bertet et al . , 2004 ) . Similarly , over-expression of dominant-negative Rac1 ( Rac1N17 ) in the AS of foscrbY10A embryos does not suppress the DC phenotype ( Figure 7A , 16th vs . 5th bars ) and even appears to increase the proportion of embryos with open cuticles ( Figure 7—figure supplement 1 , 5th vs . 16th black bars , from 52 . 7% to 72 . 9% ) . We assume that the phenotypic enhancement is due to an additive effect , since over-expression of Rac1N17 in wild-type embryos results in DC defects ( Harden et al . , 2002 ) . An important regulator of cytoskeleton activity downstream of Rho GTPases is DPak ( Hofmann et al . , 2004 ) . Interestingly , over-expression of the auto-inhibitory domain of DPak [DPak-AID - ( Conder et al . , 2004 ) ] in the AS of foscrbY10A embryos leads to a very strong suppression of the DC phenotype , as 59% of those embryos hatch ( Figure 7A , 17th vs . 6th bars ) , and even adult flies eclose ( Figure 7D ) . Accordingly , over-expression of constitutive active DPak ( DPak-myr ) in the AS of otherwise viable foscrb embryos leads to embryonic lethality with >90% of embryos with a DC-defect ( Figure 7A , 18th vs . 19th bars ) . These results indicate that unregulated activation of DPak in the AS is sufficient to produce defects in DC , and that this kinase plays a major role in the defects observed in the foscrbY10A embryos . DMoe has been shown to antagonise the activity of the Rho pathway ( Speck et al . , 2003; Neisch et al . , 2010; Hipfner et al . , 2004 ) . The participation of DMoe in the process under discussion here is supported by the fact that the FBM of Crb can recruit DMoesin ( DMoe ) to the membrane ( Médina et al . , 2002 ) and physically interacts with it ( Wei et al . , 2015 ) , and that phosphorylated-DMoe ( P-DMoe ) is reduced in stage 11 foscrbY10A embryos ( Klose et al . , 2013 ) . This reduction in P-DMoe persists during DC ( Figure 7—figure supplement 2 ) . In line with this , over-expression of the phosphomimetic form DMoeT559D in the AS of foscrbY10A embryos notably increases the number of larvae that hatch ( Figure 7A , 6th vs . 20th gray bars , from 10 . 8% to 30 . 9% ) , while over-expression of DMoe does not ameliorate the DC defects in those embryos ( Figure 7A , 21st bar ) . This suggests that the regulation of the cytoskeleton dynamics by Crb is mediated in part by the active form of DMoe . Together these results let us to conclude , that the FBM of Crb regulates actomyosin dynamics in the AS during DC by down-regulating the activity of the Rho1 pathway . We wanted to exclude the possibility that the phenotypes observed are due to a dominant effect of the Y10A mutation . In fact , over-expression of full-length CrbWT in the AS of wild-type embryos leads to premature contraction of the AS and a DC phenotype ( Harden et al . , 2002; Wodarz et al . , 1995 ) . Driving the expression of UAS-CrbWT in the AS of foscrbY10A embryos leads to a suppression of the DC phenotype , as >36% hatch at 18ºC ( Figure 7A , 8th and 9th bars vs . 5th gray bars ) , while inducing a stronger over-expression by maintaining embryos at 29°C does not ameliorate the foscrbY10A phenotype ( Figure 7A , 5th vs . 7th bars ) . These results show that the DC phenotype of foscrbY10A embryos is due to loss of Crb function . Besides an over-active actomyosin network , foscrbY10A embryos exhibit interruptions in DE-cad distribution ( Figures 2L , 3F and 5B ) . In addition some embryos show weak head-involution defects ( Figure 7—figure supplement 3 ) , a phenotype reminiscent to that of weak alleles of shotgun ( shg ) ( the gene encoding DE-cad ) ( Tepass et al . , 1996 ) , armadillo ( arm ) ( the gene encoding β-catenin ) ( McEwen et al . , 2000 ) or α-Cat ( Sarpal et al . , 2012 ) . Therefore we asked whether the DC phenotype of foscrbY10A embryos could be rescued by restoring a functional adhesion belt . Over-expression of DE-cad in the AS of these embryos indeed can suppress the DC phenotype , as 70% of the larvae hatched ( Figure 7A , 22nd vs . 6th bars ) , and even adult animals are obtained ( Figure 7E ) . A likely candidate of DE-cad regulation is the Arp2/3 complex , which has been shown to regulate endocytosis of DE-cad ( Georgiou et al . , 2008; Leibfried et al . , 2008 ) . In addition , reducing the activity of the Arp2/3 complex suppresses the DC phenotype of α-Cat mutants ( Sarpal et al . , 2012 ) . Therefore , we tested the effects of removing one copy of SCAR , Arp3 or Arpc1 on the DC phenotype of foscrbY10A embryos . Strikingly , foscrbY10A embryos that are heterozygous for SCAR∆37 exhibit only minor defects in GB retraction ( Figure 8B ) , partially restore DE-cad::GFP localisation in the AS ( compare Figure 8H with Figure 5B ) and completed DC ( Figure 8F , Video 13 ) . In fact , ~28% of these larvae hatch , as revealed by the cuticle phenotype ( Figure 7A , 23rd vs . 2nd bar ) , and even some of the w;foscrbY10A , DE-cad::GFP/SCAR∆37 , DE-cad::GFP;crbGX24 develop into adult flies that exhibit defects in abdominal development ( Figure 7F , arrowhead ) . A similar suppression was obtained in foscrbY10A embryos heterozygous for Arp3EP3640 ( Video 14 ) ( Figure 7A , 24th vs . 3rd bar ) . foscrb embryos heterozygous for SCAR∆37 or Arp3EP3640 show normal DC ( Figure 8E and Video 14 ) . 10 . 7554/eLife . 07398 . 039Figure 8 . Reduction of the SCAR-Arp complex activity suppresses the DC defects and ameliorates the loss of DE-cadherin in the AS of embryos expressing the foscrbY10A variant . ( A-F ) Stills from dorsal views of live imaging of embryos expressing DE-cad::GFP knock-in and heterozygous for the SCAR∆37 loss of function allele ( Video 13 ) . In all images the anterior is to the left , for the genotypes w;foscrb , DE-cad::GFP/SCAR∆37 , DE-cad::GFP;crbGX24 and w;foscrbY10A , DE-cad::GFP/SCAR∆37 , DE-cad::GFP;crbGX24 . All embryos were collected at the same time ( 1 hr collection ) , incubated at 28ºC for 7 hr and imaged together . The numbers in ( B , D , F ) indicate the time in minutes for the corresponding row . DC occurs normally in foscrb ( A , C , D ) embryos heterozygous for the SCAR∆37 allele , and DC defects are suppressed in foscrbY10A ( B , D , F ) embryos; some defects still visible include the impaired GB retraction ( compare B with A ) , asymmetric position of the posterior spiracles ( D , arrows ) , and bunching of the dorsal epidermis ( D , bracket ) . Scale bar: 100 μm . ( G , H ) Magnified views of AS from ( A , B , respectively ) . Note that , in order to make the localisation of DE-cad::GFP more perceptible , the autofluorescence of the yolk ( visible in A , B ) was removed from the original stack by hand using Fiji . Scale bar: 100 μm . Representative images from 6–9 different embryos for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 03910 . 7554/eLife . 07398 . 040Video 13 . Dorsal views during DC in embryos heterozygous for the SCAR∆37 allele . w;foscrb , DE-cad::GFP/SCAR∆37 , DE-cad::GFP;crbGX24 ( top ) and w;foscrbY10A , DE-cad::GFP/SCAR∆37 , DE-cad::GFP;crbGX24 ( bottom ) . Time-lapse: 10 min; 8 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 04010 . 7554/eLife . 07398 . 041Video 14 . Dorsal views during DC in embryos heterozygous for the Arp3EP3640 allele . w;foscrb , DE-cad::GFP/+;crb11A22 , Arp3EP3640/crbGX24 ( top ) and w;foscrbY10A , DE-cad::GFP/+;crb11A22 , Arp3EP3640/crbGX24 ( bottom ) . Time-lapse: 10 min; 8 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 041 In summary we could demonstrate that the DC phenotype of embryos expressing CrbY10A is due to enhanced Rho-mediated actomyosin activity and reduced adhesion . Whether these two processes are linked or independent functions downstream of Crb remains to be discussed .
The most prominent phenotype of foscrbY10A embryos is the over-contraction of AS cells , most likely mediated by DPak . In fact , cortical localisation of DPak in the AS of foscrbY10A embryos appears to be increased in some cells ( data not shown ) . In addition , over-expression of Pak-AID in the AS of foscrbY10A suppresses the GB retraction and DC phenotypes . A similar degree of suppression was observed upon over-expression of Flw , a negative regulator of Sqh . Members of the Rho GTPase family are well-established upstream regulators of actomyosin dynamics . Our data suggest that Rho1 plays a crucial role downstream of Crb , since heterozygosity of Rho11B partially suppresses the DC phenotype of foscrbY10A embryos . Previous data showed that over-expression of the constitutively active or dominant-negative form of Rac1 in the AS of wild-type embryos results in AS disruption ( Harden et al . , 2002 ) . Our observation that the phenotype of foscrbY10A embryos is enhanced upon expression of a dominant negative form of Rac1 in the AS of foscrbY10A embryos suggests that Rac1 may act upstream of Crb or in a parallel pathway . Since the effects of dominant negative Cdc42N17 could not be studied due to technical difficulties ( see Materials and methods ) , we cannot exclude any contribution of Cdc42 in this process . Therefore , our data so far support a role of Rho1 in the Crb-mediated control of actomyosin dynamics in the AS ( Figure 6G ) . The FERM protein DMoe is a likely candidate to link the FBM of Crb to Rho1 activity . Dmoe mutant imaginal epithelial cells lose epithelial markers and intercellular adhesion , become motile and show invasive behaviour ( Speck et al . , 2003 ) . In addition , lack of DMoe activates the Rho1-Rok-myosin cascade and JNK-mediated apoptosis in imaginal discs ( Warner et al . , 2010; Neisch et al . , 2010 ) . In fact , the FBM of Crb can recruit Moe to the cell membrane , a process that fails upon replacement of Tyr10 or Arg7 by Ala in the FBM of Crb ( Neisch et al . , 2010; Médina et al . , 2002 ) . Similarly , mutating Tyr10 in the FBM of the intercellular adhesion molecule ( ICAM ) -2 or the equivalent Tyr residue in the FBM of the neural cell adhesion molecule L1 impairs interaction with the FERM proteins radixin and ezrin , respectively ( Hamada et al . , 2003; Cheng et al . , 2005 ) . Moreover , it has been shown recently that the FBM of Crb is necessary for organising DMoe , aPKC and the actin cytoskeleton at the marginal zone in the developing follicular epithelium ( Sherrard and Fehon , 2015 ) . And in cervical carcinoma cells , over-expression of the mammalian CRB3 protein restores an epithelial-like morphology by organising a cortical actomyosin network through the regulation of the p114RhoGEF-RhoA-ROCK1/2 pathway via the FERM protein Ehm2 ( Loie et al . , 2015 ) . Finally , recent works documented direct binding between Moesin and Crb , which was abolished upon Y10A substitution ( Wei et al . , 2015 ) . It is unlikely that one of the other two established binding partners of the FBM of Crb , Ex and Yrt ( Ling et al . , 2010; Robinson et al . , 2010; Laprise et al . , 2006 ) , mediates the Crb function in the AS . So far , no role of Ex during DC has been reported , and ex mutant embryos reach stage 16 of development without showing major morphogenetic defects ( Marcinkevicius and Zallen , 2013 ) . Yrt is expressed in the AS and the epidermis , but this is not affected in foscrbY10A embryos . In addition , the DC phenotype of zygotic yrt∆75a mutants is less severe than the one observed in foscrbY10A embryos . Finally , we do not observe increased Crb protein levels in foscrbY10A embryos , which would be expected if the interaction between Yrt and Crb is impaired ( Laprise et al . , 2006 ) . Further support for a more direct role of Crb in regulating the actomyosin network comes from the observation that Crb co-localises with DPar-6 , aPKC and Baz at the medial actomyosin foci in the AS ( David et al . , 2010; 2013 ) . Given the known interactions between members of the Crb complex with members of the Par complex [reviewed in ( Bulgakova and Knust , 2009; Tepass , 2012; Rodriguez-Boulan and Macara , 2014 ) ] , David et al . ( David et al . , 2010 ) suggest that Crb in apical medial foci provides an anchor for PAR proteins . They go on to show that Baz and Par6-aPKC have opposite effects on foci duration , in that Baz promotes and Par6-aPKC complex inhibits the duration of foci . The interplay between these polarity complexes and the actomyosin system seems to establish a delayed negative feedback that promotes the cyclic contractions in the AS ( David et al . , 2010; 2013 ) . In fact , Crb::GFP also exhibits a similar pulsation as Zip::GFP in the AS ( own unpublished observations ) , so it will be important to analyse whether CrbY10A::GFP mutant proteins have different dynamics in comparison to the wild type Crb . Given the observation that at early stages of embryonic development the PBM is required for ZA stability , and that the CrbY10A mutant protein has an intact PBM , it is possible that during DC , Crb-mediated regulation of actomyosin dynamics impacts on ZA stability . Interestingly , DPak is not only a regulator of actomyosin dynamics , but is also involved in supporting ZA stability , both in Drosophila and in mammalian cells ( Lozano et al . , 2008; Braga et al . , 2000; Akhtar and Hotchin , 2001; Pirraglia et al . , 2010; Menzel et al . , 2007; 2008 ) . The role of DPak itself in DC morphogenesis is still controversial . Previous work showed that cell shape changes in the AS occur normally in embryos lacking maternal and zygotic Dpak and that inhibition of DPak in the AS does not prevent apical constriction of amnioserosa cells ( Conder et al . , 2004 ) . However , wild-type embryos expressing Pak-AID in the AS show defects in head involution and DC , which are stronger than those of embryos devoid of maternal and zygotic DPak . This led the authors to suggest that Pak-AID may also affect the activity of a second kinase , Pak3 , in the AS ( Conder et al . , 2004 ) . Thus , whether inhibition of DPak , Pak3 or both upon expression of Pak-AID in foscrbY10A embryos accounts for the rescuing effect of the DC phenotype , including rescue of the ZA , remains to be clarified . How can DPak regulate ZA integrity ? ZA remodelling is essential for morphogenesis , and this remodelling is driven by the endocytosis and recycling of junctional components ( Harris , 2012; Matsubayashi et al . , 2015 ) . DPak can activate the Arp2/3 complex directly or via the Drosophila WAVE homolog SCAR ( Lecuit et al . , 2011; Kurisu and Takenawa , 2009; Zallen et al . , 2002 ) . Arp2/3 , in turn , has been implicated in the regulation of ZA stability , e . g . in the Drosophila notum , where it maintains ZA stability by regulating the endocytosis of junctional components ( Watanabe et al . , 2009; Quiros and Nusrat , 2014; Lecuit et al . , 2011; Georgiou et al . , 2008; Leibfried et al . , 2008 ) . Moreover , reducing the activity of the Arp2/3-complex suppresses the DC phenotype of α-Cat mutants ( Sarpal et al . , 2012 ) , and the Arp2/3–WAVE/SCAR complexes associate with E-cad clusters and regulate their endocytosis ( Verma et al . , 2012; Kovacs et al . , 2002; Lecuit and Yap , 2015 ) . In fact , DE-cad endocytosis is enhanced in a Rho1-dependent manner when junctions are under stress and DE-cad clusters are also down-regulated via inhibition of Par3 by Rok ( Levayer et al . , 2011; Lecuit and Yap , 2015 ) . Our results are in agreement with a role of Arp2/3 in regulating ZA stability in the AS . Heterozygosity of SCAR∆37 , Arp1Q25st or Arp3EP3640 not only partially restored DE-cad::GFP localisation at the ZA in the AS of foscrbY10A embryos and suppressed DC defects , but even rescued the lethality of foscrbY10A flies . Fusion of abdominal segments in adult escapers suggest that Crb may also be involved in histoblast fusion during metamorphosis ( Madhavan and Madhavan , 1980; Ninov et al . , 2007 ) . Myosin-II activity itself has also been shown to be essential for the maintenance of AJs in some cases . Mice ablated for NMHC II-A die by E7 . 5 due to massive defects in cell-cell contacts and epithelial multi-layering accompanied by loss of E-cad and β-catenin from adhesion sites ( Conti et al . , 2004 ) . Similarly , ZA stability in the Drosophila embryonic ectoderm depends on myosin-II contractility and requires interactions with actin ( Engl et al . , 2014; Truong Quang et al . , 2013 ) . Finally , Rok and myosin-II activities participate in ZA remodelling in the Drosophila pupal eye by regulating the formation of DE-cad recycling endosomes ( Yashiro et al . , 2014 ) . Because the SCAR-Arp2/3 complex is an important enhancer of actin protrusions ( Wood et al . , 2002; Abreu-Blanco et al . , 2012; Georgiou and Baum , 2010 ) , it is also plausible that reducing its activity in foscrbY10A embryos stabilises the ZA indirectly . On the other hand , misregulation of actomyosin activity is not always associated with defects in ZA stability and integrity of the AS . Expressing a constitutively active form of MLCK to increase myosin II activity or over-expression of RhoGEF2 , an activator of Rho1 , results in an increase in the number and density of actin foci without affecting the integrity of the AS ( Azevedo et al . , 2011; Fischer et al . , 2014 ) , which could be due to the use of a weak GAL4 driver . Alternatively , the difference to our results could be explained by the fact that these authors performed the over-expression in a background with more than two copies of E-cad ( using a ubi-DE-cad::GFP line ) , while we performed the experiments in a knock-in DE-cad::GFP line ( Huang et al . , 2009; 2011 ) , which thus may represent a more sensitive background . Another possibility to interpret our results is that Crb , or an interacting protein , couples the actomyosin network and the ZA . During gastrulation in C . elegans a molecular clutch has been postulated to connect the myosin network with the adhesion sites to transmit the force generated by the actomyosin contractions ( Roh-Johnson et al . , 2012 ) . In Drosophila , the actomyosin contractions in the AS are initially uncoupled from apical contractions and hence the ZA ( Solon et al . , 2009; Gorfinkiel et al . , 2009; Blanchard et al . , 2010 ) . Successive rows of amnioserosa cells are then sequentially stabilised in a contracted state , driving further contraction of the tissue . The surface stabilization mechanism is not known , but is likely to involve an increase in cellular stiffness [reviewed in ( Paluch and Heisenberg , 2009 ) ] . In foscrbY10A embryos the actomyosin foci in the AS emerge prematurely before the onset of germ band retraction , whereas in wild-type these foci are more abundant after the end of germ band retraction ( Figure 2—figure supplement 2 and data not shown ) . Thus , the early over-contraction of the actomyosin in foscrbY10A embryos may induce a premature coupling to the ZA , thus disrupting germ band retraction and DC . An interesting candidate for this coupling is the protein Canoe , which binds to α-catenin ( Sawyer et al . , 2009; Pokutta et al . , 2002 ) , and whose absence results in a DC phenotype ( Jürgens et al . , 1984; Takahashi et al . , 1998; Boettner et al . , 2003; Choi et al . , 2011 ) . Absence of Canoe induces the detachment of the actomyosin apparatus from cell-cell junctions during Drosophila mesoderm invagination ( Sawyer et al . , 2009; 2011 ) . In conclusion , we show a novel function of the FBM of Crb as an essential regulator of cytoskeleton dynamics and tissue integrity during DC . Different lines of evidence show that Crb regulation of AS morphogenesis involves DMoesin , Rho-GTPases , class-I Pak , and the SCAR-Arp2/3 complex . Further work will determine at which level Crb regulates actomyosin dynamics and why it is just the morphogenesis of the AS that depends on the FBM of Crb , while all other embryonic epithelia are not affected .
Flies were maintained at 25ºC on standard food . All the mutant alleles where balanced over fluorescent balancers to identify the homozygous mutants in fixed embryos or live imaging microscopy ( see below ) . All crosses and analyses were carried in a crb null background ( crbGX24 or crb11A22 , homozygous or trans-heterozygous ) , so the expression of the different variants of Crb is exclusively provided by the fosmid ( Klose et al . , 2013 ) . The different UAS-lines where recombined with the DE-cad::GFP knock-in allele or the null crb11A22 allele . The driver line GAL4332 . 3 was recombined with each of the different fosmid alleles . Embryo stage refers to the foscrb;crbGX24 genotype morphology accordingly to ( Campos-Ortega and Hartenstein , 1985 ) . All genotypes ( foscrb;crbGX24 , foscrbY10F;crbGX24 and foscrbY10A;crbGX24 ) were collected under the same conditions , at the same time and during the same period ( indicated in the respective figure legend ) . In this way , the comparison between foscrb or foscrbY10F and foscrbY10A mutant phenotypes show the differences observed at a specific time after egg laying . Embryos were collected on apple juice plates at 25ºC and then incubated for the appropriate times at 25ºC or 28ºC , dechorionated in 3% sodium hypochlorite for 3 min , fixed for 20 min in 4% formaldehyde in phosphate-buffered saline ( PBS ) solution/heptane V/V 1:1 . Vitelline membrane was removed by strong shaking in heptane/methanol v/v 1:1 , except for the staining of actin in which the vitelline membrane was removed by strong shaking in 80% ethanol . Embryos were blocked for 2 hr at room temperature in PBT ( PBS + 0 . 1% Triton X-100 ) + 5% normal horse serum ( Sigma-Aldrich H1270 , St . Louis , Missouri , USA ) . Embryos were incubated for 2 hr at room temperature or overnight at 4ºC with primary antibodies ( see Table 3 ) . For analysis of Zipper localisation , we used the protein trap line Zipper::GFP ( see Table 2 ) and the staining was done using the anti-GFP antibody . Incubations with the appropriate secondary antibodies were performed for 1 hr at room temperature . Stained embryos were mounted in glycerin propyl gallate ( 75% glycerol , 50 mg/mL propyl gallate ) and visualized using a Zeiss LSM 780 NLO confocal microscope ( ZEISS Microscopy , Jena , Germany ) with a C-Apochromat 40x/1 . 2W Corr objective with the correction collar at 0 . 18 ( at this position the brightness and contrast was enhanced ) . To distinguish homozygous embryos , in all the stainings an anti-GFP antibody was included to stain for the balancer-provided GFP . All images for a given marker in different genotypes were taken under the same settings for laser power , PMT gain and offset . Maximal projections , merging and LUT-pseudocolor assignment was performed using Fiji ( Schindelin et al . , 2012 ) . For the FIRE-LUT pseudocolor 0 is black and 255 is white . Mounting was done in Adobe Photoshop CC 2015 . 0 . 1 and when brightness and contrast was adjusted , the modifications were equally applied to all the set of images for a given marker . 10 . 7554/eLife . 07398 . 043Table 3 . Antibodies and probes employed . DOI: http://dx . doi . org/10 . 7554/eLife . 07398 . 043DilutionSourcePhalloidin Alexa Fluor 5551:500InvitrogenAlexa Fluor 488- , 568- , and 647-conjugated1:500InvitrogenRat antibodiesanti-Crb2 . 81:500 ( Richard et al . , 2006 ) anti-DE-cadherin1:20DSHB DCAD2anti-Yurt1:100 ( Laprise et al . , 2006 ) Mouse antibodiesanti-α-Spectrin1:25DSHB 3A9anti-β-galactosidase1:200DSHB 40-1aanti-Coracle1:25DSHB C566 . 9anti-Crb-Cq41:300DSHB Cq4anti-Disc large1:100DSHB 4F3anti-Enabled1:100DSHB 5G2anti-GFP1:500Roche 11814460001 ( Mannheim , Germany ) anti-Hindsight1:100DSHB 1G9anti-Integrin βPS1:2DSHB CF . 6G11anti-Phosphotyrosine1:100BD Transduction Laboratories cat . no . 610000anti-SCAR1:25DSHB P1C1Rabbit antibodiesanti-Bazooka1:500kindly provided by A . Wodarzanti-DAAM1:3000kindly provided by József Mihály ( unpublished ) anti-Diaphanous1:5000kindly provided by Steven A . Wasserman ( Afshar et al . , 2000 ) anti-DPatj1:1000 ( Richard et al . , 2006 ) anti-Echinoid1:5000kindly provided by Laura Nilson ( Laplante and Nilson , 2006 ) anti-Expanded1:300 ( Boedigheimer and Laughon , 1993 ) anti-GFP1:500Invitrogenanti-DPak1:8000kindly provided by Nicholas Harden ( Harden et al . , 1996 ) anti-Polychaetoid1:5000kindly provided by Sarah Bray ( Djiane et al . , 2011 ) anti-Phospho-Moesin1:100Cell Signaling Technology 3150 ( Danvers , Massachusetts , USA ) anti-Stranded at second1:500kindly provided by E . Organ and D . CavenerInvitrogen , Molecular Probes ( Eugene , Oregon , USA ) ; DSHB - Developmental Studies Hybridoma Bank ( Iowa city , Iowa , USA ) Embryos were collected overnight on apple juice plates at 25ºC and then incubated for > 6 hr at 28ºC . All the GFP or YFP positive eggs ( the GFP or YFP is provided by the balancer ) were removed and the remaining eggs where maintained at 25ºC . The next day , the plates were screened again to remove remaining GFP/YFP positive eggs/larvae . Thus , all the remaining eggs/larvae had a crb null background ( crbGX24 or crb11A22 , homozygous or trans-heterozygous ) . These eggs/larvae were collected , dechorionated in 3% sodium hypochlorite for 3 min , mounted on Hoyer’s medium ( gum arabic 30 g , chloral hydrate 200 g , glycerol 20 g , H2O 50 ml ) , and the slide was incubated overnight at 60ºC . In this way , all the eggs laid in the plate were at least >28 hr at 25º , enough time to let the larvae hatch when they are viable . The preparations were analysed by phase contrast with a Zeiss Axio Imager . Z1 microscope with an EC Plan-NEOFLUAR 10X/0 . 3 objective . Embryos were collected on apple juice plates for 1 hr at 25ºC and then incubated for 8 hr at 28ºC , dechorionated in 3% sodium hypochlorite for 2 min 30 sec , and fixed for 30 min in 25% glutaraldehyde/heptane v/v 1:1 . Devitellinization was done by hand in 25% glutaraldehyde . Then , the embryos were postfixed in modified Karnovsky ( 2% paraformaldehyde/2% glutaraldehyde in 50 mM HEPES ) followed by 1% osmium tetroxide in PBS , dehydrated in a graded series of ethanol , transferred to microporous capsules ( 78 μm pore size , Plano Cat . 4614 ) and critical point dried using the Leica CPD 300 ( Leica Microsystems GmbH , Wetzlar , Germany ) . Embryos were mounted on 12 mm aluminium stubs and sputter coated with gold using a Leica Baltec SCD 050 . Samples were analysed with a Jeol JSM 7500F cold field emission SEM ( JEOL Ltd , Tokyo , Japan ) at 10 kV acceleration voltage . Embryos were collected and incubated as describe above ( see 'Embryo collection and antibody staining' ) . In the analysis of pulsed contractions in the AS , sequential collections of 30 min interspaced by 1 hr between each genotype allowed us to analyse 2–3 embryos of each genotype on the same session , so the acquisition conditions for all the genotypes were identical . To eliminate crbGX24 or crb11A22 heterozygous embryos , all GFP or YFP positive embryos were removed . The remaining eggs were dechorionated by hand or in 3% sodium hypochlorite for 2 min , mounted and oriented in a bottom glass Petri dish ( MatTek P35G-1 . 5 . 14-C , Ashland , Massachusetts , USA ) . Previously , the glass was cover with a thin layer of glue ( adhesive dissolved from double sided tape in heptane ) . The embryos were covered with water and visualized by multi-position scanning using a Zeiss LSM 780 NLO confocal microscope with a W Plan-Apochromat 40x/1 . 0 objective . Excitation was performed with 488 nm for GFP or YFP , and 561 nm for RFP or mTomato from an Argon Multiline Laser . The pinhole was adjusted for faster acquisition , so the step sizes correspond to 2 . 01 μm ( Videos 1 , 2 , 7 , 11 , 13 , 14 ) , 2 . 3 μm ( Videos 4 , 5 , 6 ) , 1 . 2 μm ( Video 8 , 12 ) , 1 . 46 μm ( Videos 3 and 9 ) . 4D-Hyperstacks were processed with Fiji ( Schindelin et al . , 2012 ) and the movies were rendered with Adobe Photoshop CC 2015 . 0 . 1 . Under these conditions we observed that w;foscrb , DE-cad::GFP;crbGX24 embryos imaged for >7 hr at 5 min time lapse hatched and survived without showing any obvious damage ( data not shown ) . Statistical analyses were performed with GraphPad Prism 6 . Results are expressed as means ± SD . Statistical significance was evaluated in a one-way analysis of variance ( ANOVA ) followed by a Dunnett’s multiple-comparison test . In the analysis of the statistical significance of the data presented in the Figure 7—figure supplement 1 , the percentages were first converted to arcsin values and then analysed by a one-way-ANOVA followed by a Dunnet’s multiple comparisons test . | A layer of epithelial cells covers the body surface of animals . Epithelial cells have a property known as polarity; this means that they have two different poles , one of which is in contact with the environment . Midway through embryonic development , the Drosophila embryo is covered by two kinds of epithelial sheets; the epidermis on the front , the belly and the sides of the embryo , and the amnioserosa on the back . In the second half of embryonic development , the amnioserosa is brought into the embryo in a process called dorsal closure , while the epidermis expands around the back of the embryo to encompass it . One of the major activities driving dorsal closure is the contraction of amnioserosa cells . This contraction depends on the highly dynamic activity of the protein network that helps give cells their shape , known as the actomyosin cytoskeleton . One major question in the field is how changes in the actomyosin cytoskeleton are controlled as tissues take shape ( a process known as “morphogenesis” ) and how the integrity of epithelial tissues is maintained during these processes . A key regulator of epidermal and amnioserosa polarity is an evolutionarily conserved protein called Crumbs . The epithelial tissues of mutant embryos that do not produce Crumbs lose polarity and integrity , and the embryos fail to develop properly . Flores-Benitez and Knust have now studied the role of Crumbs in the morphogenesis of the amnioserosa during dorsal closure . This revealed that fly embryos that produce a mutant Crumbs protein that cannot interact with a protein called Moesin ( which links the cell membrane and the actomyosin cytoskeleton ) are unable to complete dorsal closure . Detailed analyses showed that this failure of dorsal closure is due to the over-activity of the actomyosin cytoskeleton in the amnioserosa . This results in increased and uncoordinated contractions of the cells , and is accompanied by defects in cell-cell adhesion that ultimately cause the amnioserosa to lose integrity . Flores-Benitez and Knust’s genetic analyses further showed that several different signalling systems participate in this process . Flores-Benitez and Knust’s results reveal an unexpected role of Crumbs in coordinating polarity , actomyosin activity and cell-cell adhesion . Further work is now needed to understand the molecular mechanisms and interactions that enable Crumbs to coordinate these processes; in particular , to unravel how Crumbs influences the periodic contractions that drive changes in cell shape . It will also be important to investigate whether Crumbs is involved in similar mechanisms that operate in other developmental events in which actomyosin oscillations have been linked to tissue morphogenesis . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] | [
"cell",
"biology"
] | 2015 | Crumbs is an essential regulator of cytoskeletal dynamics and cell-cell adhesion during dorsal closure in Drosophila |
Zona pellucida ( ZP ) , the extracellular matrix sheltering mammalian oocytes and embryos , is composed by 3 to 4 proteins . The roles of the three proteins present in mice have been elucidated by KO models , but the function of the fourth component ( ZP4 ) , present in all other eutherian mammals studied so far , has remained elusive . Herein , we report that ZP4 ablation impairs fertility in female rabbits . Ovulation , fertilization and in vitro development to blastocyst were not affected by ZP4 ablation . However , in vivo development is severely impaired in embryos covered by a ZP4-devoided zona , suggesting a defective ZP protective capacity in the absence of ZP4 . ZP4-null ZP was significantly thinner , more permeable , and exhibited a more disorganized and fenestrated structure . The evolutionary conservation of ZP4 in other mammals , including humans , suggests that the structural properties conferred by this protein are required to ensure proper embryo sheltering during in vivo preimplantation development .
Mammalian oocytes and early preimplantation embryos are surrounded by a glycoprotein shell termed zona pellucida ( ZP ) that serves important functions during folliculogenesis , species-specific fertilization and early development . In eutherian mammals , this glycoprotein shell is solely composed of 3 to 4 proteins depending on the species . The differences in ZP composition between eutherian mammals is the consequence of duplication and/or pseudogenization events during evolution . In all eutherian mammals studied so far ZP always contain ZP2 and ZP3 proteins , but ZP1 or ZP4 may be absent , as both are paralogous genes formed by duplication of a common ancestry gene ( reviewed by Goudet et al . , 2008 ) . The functions of the three proteins ( ZP1 , ZP2 and ZP3 ) present in mouse ZP have been elucidated by gene ablation experiments . ZP1 serves a structural function , being dispensable for sperm binding or fertilization ( Rankin et al . , 1999 ) . ZP1 null females display perturbed folliculogenesis and , although ovulation rates are not affected , both cleavage rate and litter size are reduced , likely due to ZP1-null zona being unable to protect the developing embryo . ZP2 ablation also results in structural defects , which are more severe , as the number of antral follicles is reduced and only few ZP-free oocytes and no two-cell embryos are recovered from Zp2 knock-out ( KO ) females after mating ( Rankin et al . , 2001 ) . Finally , Zp3 KO mice are unable to form ZP , showing a drastic reduction in ovulation rates and also being unable to produce cleaved embryos after mating ( Liu et al . , 1996; Rankin et al . , 1996 ) . Given the lack of ZP on ovulated Zp2- or Zp3-null oocytes , these KO models could not assess the role of these proteins on sperm binding . To study that process , more complex gain-of-function transgenic mouse models have been generated . In particular , the generation of ZP composed of different combinations of human ZP proteins ( Baibakov et al . , 2012 ) or an heterologous ZP lacking mouse ZP2 protein ( Avella et al . , 2014 ) have uncovered a crucial role of ZP2 on mouse and human sperm recognition . The fourth ZP protein component , ZP4 , is absent in mice , which seems to be an exception for mammals , as ZP4 is present in multiple mammalian species including humans ( Hughes and Barratt , 1999; Lefièvre et al . , 2004 ) , rabbits ( Stetson et al . , 2012 ) , ungulates ( Hedrick and Wardrip , 1987; Noguchi et al . , 1994; Topper et al . , 1997 ) , carnivores ( Moros-Nicolás et al . , 2018 ) and other rodents ( Hoodbhoy et al . , 2005; Izquierdo-Rico et al . , 2009 ) . The lack of ZP4 in laboratory mouse ( Mus musculus ) , the only species where KO models were readily available , has precluded the study of the role of this protein by of loss-of-function experiments . Gene ablation experiments are essential to unequivocally know the function of a gene , but were largely restricted to mice , given the technical difficulties to perform site-specific genome modification in other mammalian species ( Lamas-Toranzo et al . , 2017 ) . Luckily , the advent of site-specific endonucleases , and particularly CRISPR , the last technology to be developed and the easiest to tailor to its genomic target , allows direct KO generation in one step . Using this technology , we have generated ZP4-null rabbits uncovering that ZP4 confers structural properties to the ZP that are essential to protect the developing preimplantation embryo .
ZP4 gene ablation was achieved by CRISPR-mediated mutagenesis at the zygote stage . For this aim , a sgRNA targeting the beginning of ZP4 coding region was designed ( Figure 1A ) . Cas9-coding capped polyadenylated mRNA and ZP4-targeting sgRNA were injected into the ooplasm at 100 and 25 ng/µl , respectively . Following microinjection , 50 embryos were transferred to the oviduct of two pseudopregnant recipients ( 20 and 30 embryos/female ) , resulting in the delivery of 12 pups ( 5 and 7 , respectively ) . Genotyping was performed on ear biopsies and blood samples from each pup . Genome edition rates were assessed by sequencing a PCR product containing the target site . This initial screening revealed that all 12 pups were edited at the target site , but it did not allow to identify each mutated allele harbored by each individual . For this aim , clonal sequencing was performed , revealing that all pups but one male were mosaic , that is contained more than two alleles . This phenomenon appears when DNA replication precedes genome edition and has been commonly observed following zygote microinjection in different mammalian species ( Lamas-Toranzo et al . , 2017 ) , including rabbits ( Guo et al . , 2016; Honda et al . , 2015; Lv et al . , 2016; Song et al . , 2016a; Song et al . , 2016b; Sui et al . , 2016; Yan et al . , 2014; Yang et al . , 2016; Yuan et al . , 2016 ) , where genome replication takes place shortly after fertilization ( Oprescu et al . , 1965; Szollosi , 1966 ) . In the absence of an homologous recombination template , CRISPR-mediated gene ablation relies on the generation of frame-disrupting insertion-deletions ( indels ) by non-homologous end joining ( NHEJ ) ( Lamas-Toranzo et al . , 2017 ) . As some of the indels generated do not alter the Open Reading Frame of the gene , only those individuals harboring frame-disrupting mutations in all alleles can be considered KO . None of the seven females generated in F0 were KO , as all contained at least one in-frame indel ( Figure 1A ) . This situation precluded the use of F0 generation for direct experimentation . Consequently , founders were breed to obtain wild-type ( WT ) , heterozygous ( Hz ) or knock-out ( KO ) females in subsequent generations . A non-mosaic male containing two frame-disrupting indels was able to breed normally , as expected , given that ZP4 protein is exclusively expressed in ovaries . This male was crossed to mosaic females to generate KO animals ( i . e . , harboring two frame-disrupting indels ) that were used for an initial fertility ( pregnancy ) screening . Subsequently , one of the KO alleles carried by that male and formed by a G insertion was selected in Hz ( WT/KO ) animals to maintain the line and generate WT , Hz and KO individuals for experiments . Off-target analysis revealed no genome edition in any of the five most probable off-target sites ( Figure 1—figure supplement 1 ) . Protein ablation in KO female individuals was confirmed by IHC in ovaries ( Figure 1B ) . Proteomics analysis of solubilized extracts of ZP from oocytes obtained from KO and WT individuals by tandem mass spectrometry further confirmed the absence of ZP4 in KO individuals . ZP4 ablation did not prevent the synthesis of all other three rabbit ZP proteins ( ZP1-3 ) , which were also detected in the solubilized extract of ZP4 KO ZP ( Supplementary file 1 ) . This analysis also provides a semi-quantitative estimation of the amount of ZP1-3 proteins based on the number of peptides detected for each protein , which were similar for KO and WT oocytes , thereby suggesting that ZP4 ablation does not induce major alterations in ZP1-3 synthesis . The effect of ZP4 ablation on female fertility was initially tested by crossing WT , Hz or KO females with fertile WT males . WT and Hz females delivered normal litter sizes , whereas only one out of 10 KO females , delivered a single litter of reduced size ( four pups , Figure 2A ) . In order to elucidate the root for the reproductive failure caused by ZP4 ablation , ovulation and embryonic cleavage rates were assessed in WT , Hz or KO females . For this aim , five females per group were crossed with WT males and sacrificed 15 hr after mating . Ovulation rates following natural mating were similar between all groups , suggesting that ZP4 ablation did not impair folliculogenesis ( Figure 2B ) . To further test the normalcy of ZP4-null follicles , a histological analysis was performed on KO or WT ovaries . ZP4 disrupted ovaries were grossly indistinguishable from WT or Hz and no obvious abnormalities were noted in follicular population ( Figure 2E–H and Figure 2—figure supplement 1 ) . To analyse embryonic cleavage rates , presumptive zygotes were obtained by oviductal flushing and subsequently cultured in vitro . Cleavage rates did not differ between the different groups , suggesting that ZP4 does not play an essential role in fertilization in vivo ( Figure 2C ) . Cleaved embryos were allowed to develop to the blastocyst stage in vitro . Again , no differences were noted in blastocyst rates or morphology , showing that embryos derived from ZP4-null oocytes ( Hz embryos enclosed by a ZP4-null ZP , MATKO embryos ) are able to reach the blastocyst stage in vitro ( Figure 2D ) . To further confirm that spermatozoa were able to penetrate through the ZP4-disrupted zona , pronuclei formation was assessed in zygotes collected from KO and WT ( three animals per group ) . Normal pronuclear formation was noted for embryos derived from ZP4-null oocytes , showing no effect of ZP4 ablation on the in vivo fertilization process ( Figure 2G ) . As expected , given that rabbit ZP is not involved in polyspermy blockage ( Pincus and Enzmann , 1932 ) , no polyspermy was found on embryos produced from ZP4-null oocytes . As ovulation and fertilization were not affected in ZP4-null females , the negative effect of ZP4 ablation can only be attributed to a developmental failure of the embryos surrounded by a ZP lacking ZP4 . However , given that these Hz embryos derived from ZP4-disrupted oocytes were able to develop in vitro to the blastocyst stage , this developmental failure must occur in vivo . ZP protection is essential for in vivo embryo development , as rabbit embryos without zona are unable to establish pregnancy following embryo transfer ( Moore et al . , 1968; Rottmann and Lampeter , 1981 ) . In vivo development was analysed in KO and WT females ( three per group ) crossed with fertile males and sacrificed on Day 4 post-mating . At this developmental time , rabbit embryos have developed to the blastocyst stage and are surrounded by a mucin coat , but lack ZP , which disappears between Days 3 and 4 ( Denker and Gerdes , 1979; Fischer et al . , 1991 ) . ZP dissolution does not occur in vitro as it requires both from embryo development and uterine secretions ( Fischer et al . , 1991 ) . All embryos recovered from WT females at Day 4 post-mating had reached the blastocyst stage , however , less than half of the MATKO embryos recovered from KO females were able to develop to the blastocyst stage ( Figure 3A ) . Moreover , those embryos reaching the blastocyst stage in the KO group showed a significant reduction in blastocyst expansion ( ~500 vs . ~200 µm diameter for WT vs . KO , respectively , Figure 3B ) . ZP was absent in all WT blastocysts ( Figure 3C ) and in most expanded or collapsed blastocysts from the KO group , but was present in all degenerated embryos ( Figure 3D ) , in agreement with the requirement of embryo development for ZP dissolution previously reported ( Fischer et al . , 1991 ) . Evident signs of mechanical damage ( non-spherical , crushed ZP ) could be observed in degenerated embryos in the KO group ( Figure 3D ) , suggesting that mechanical pressure may be at least partially responsible for the developmental failure . Beyond Day 4 and before implantation , the rabbit embryo undergo a relevant embryonic growth that has not been recapitulated yet on an in vitro setting . This growth , converting a ~ 500 µm blastocyst to a ~ 3 mm expanded blastocysts from Days 4 to 6 , is concomitant to the formation of a new glycoprotein matrix termed neozona and the replacement of the mucin coat by the gloiolemma ( Fischer et al . , 1991 ) . To determine the carry over effects of the developmental delay observed on Day 4 , in vivo development was analysed in KO and WT females ( four per group ) crossed with fertile males and sacrificed on Day 6 post-mating . Following post-mortem recovery , we observed that , based on the number of corpora lutea of each individual , more than half of the MATKO embryos ( Hz embryos lacking ZP4 in their ZP ) had been already lost by Day 6 after mating ( Figure 4A ) , which roughly coincide with the percentage of degenerated embryos on Day 4 . Furthermore , the few embryos surrounded by ZP4-null ZP present in the uterus that were not degenerated by Day 6 were developmentally arrested compared to those protected by WT ZP , as blastocyst growth was severely impaired ( Figure 4B–E ) . These results suggest that ZP4 confers structural properties to the ZP that are required for embryo protection during preimplantation development in vivo . Morphological differences between KO and WT ZP were clearly noticeable under light microscopy: ZP lacking ZP4 appeared thinner and more irregular ( less spherical ) compared to WT ZP ( Figure 5A ) . Furthermore , although a quantifiable mechanical analysis was not performed , ZP4-devoided ZP appeared easily deformable under the mechanical pressure exerted by a blunt micromanipulation needle compared to WT ZP ( Video 1 ) . In regard to ZP thickness , measurements by contrast light microscopy revealed a significant reduction ( ~4 µm ) in the absence of ZP4 to the ZP of WT or Hz females ( Figure 5A ) . No differences in ZP thickness were noticed between WT or Hz females , suggesting that ZP4 haploinsufficiency does not produce any obvious alteration in ZP formation , in agreement with the lack of differences in fertility between WT and Hz individuals . To further characterize the effect of ZP4 ablation on ZP texture , we analysed WT or KO ZP from in vivo derived zygotes ( ~14–15 hr post-mating ) by Scanning Electron Microscopy ( SEM ) . SEM images evidenced notable differences in textural properties between WT and ZP4 lacking ZP: while the former exhibited a smooth and compact texture , the latter displayed an uneven , rougher and porous aspect ( Figure 5B ) . This less compact structural organization can explain its higher deformability and the larger fenestrations could lead to increased permeability . ZP permeability was tested by incubating zygotes obtained from WT or ZP4-null females with green fluorescent nanospheres of different diameters , ranging from 20 to 40 and 100 nm . This analysis revealed that , while 100 nm-sized nanospheres were successfully blocked by both ZPs , all 9 ZPs lacking ZP4 analyzed were permeable to 20 to 40 nm nanospheres , which are efficiently blocked by all 9 WT ZPs ( Figure 5C ) .
Despite its essential functions during folliculogenesis , fertilization and preimplantation embryo development , ZP is composed of just 3 to 4 proteins , depending on the species . Gene ablation experiments have deciphered the functions of the three proteins present in mouse ZP , but the role of the fourth component ( ZP4 ) has remained controversial . Previous studies have suggested that ZP4 may play a role during fertilization . Rabbit ZP4 was found to bind to rabbit sperm acrosome ( Prasad et al . , 1996 ) , although specific blocking by ZP4-antisera fragments could not be demonstrated . Similarly , other studies based on in vitro protein-binding assays have attributed sperm attachment to the zona matrix to ZP4 in humans ( Chiu et al . , 2008 ) or to heterocomplexes of ZP4 and ZP3 in porcine and bovine ( Kanai et al . , 2007; Yurewicz et al . , 1998 ) . In contrast with this notion , the expression of human ZP4 in transgenic mouse ZP was not sufficient to support human sperm binding ( Yauger et al . , 2011 ) . In this article , we have tested ZP4 role by gene ablation in a species where it is naturally present . Although we cannot assure that ZP4 ablation might reduce sperm binding , acrosome reaction or sperm penetration in vitro , our results unequivocally show that ZP4 does not play an essential role during the fertilization process in vivo , but serves a structural and mechanical function which is fundamental to protect the developing embryo prior to implantation . In agreement with our findings , a recent study on ZP characterization hypothesized a structural role of ZP4 on increasing CP thickness in humans ( Nishimura et al . , 2019 ) . ZP plays an essential protective role during preimplantation embryo development , as despite the exception of a single case report where zona-free embryos were developed in vitro up to blastocyst before transfer ( Shu et al . , 2010 ) , zona denuded embryos have been largely considered unable to complete in vivo development ( Bronson and McLAREN , 1970; Modliński , 1970 ) . This protective function was severely impaired following ZP4 ablation , ultimately causing embryonic death and infertility . ZP4 ablation in rabbits lead to similar structural defects in ZP to those obtained when ZP1 is ablated in mice ( Rankin et al . , 1999 ) : reduction in ZP thickness and increased ZP porosity . However , the effects on fertility of Zp1 KO were less severe than those caused by ZP4 ablation , as litter size was halved in Zp1 KO female mice compared to WT mice , whereas ZP4 ablation abolished almost completely embryo development . A plausible explanation for this difference is that mouse embryos may demand a lesser degree of ZP protection during preimplantation development than other domestic mammals , including rabbits . The early blastocyst hatching occurring in mouse ( 3 . 5 days compared to more than 6 days in humans and most domestic mammalian species ) and a reduced hydrostatic pressure in female reproductive tract motivated by its small body size may be partly responsible for this difference . Remarkably , the laboratory mouse remain the only mammalian species studied so far where ZP4 is not present in ZP ( Moros-Nicolás et al . , 2018 ) , and mouse ZP is noticeably softer and more elastic compared to other mammalian ZPs containing ZP4 with or without ZP1 ( Yu Sun et al . , 2003 ) . Further evidence for the fundamental structural role of ZP4 comes from transgenic mouse models producing humanized ZP: the expression of human ZP4 in mouse ZP was found to be able to recover the thickness and robustness lost following mouse ZP2 ablation ( Avella et al . , 2014 ) . In order to determine whether ZP4 deficiency could lead to infertility in women , we searched for loss-of-function mutations in the Genome Aggregation Database , which includes data from 141456 individuals ( Lek et al . , 2016 ) . This search revealed 48 polymorphisms in ZP4 coding region leading to premature stop codons ( PTCs ) , which are the most evident loss-of-function mutations ( Supplementary file 2 and Figure 6 ) . The frequency of individuals carrying at least one of these PTC mutations in ZP4 was 0 . 3% in the global population analysed , raising to 1% in the African dataset . Hence , in this population , roughly 1 into 40000 women would carry a PTC in both alleles , being thereby potentially infertile due to ZP4 deficiency . Mutations in ZP1-3 genes have been recently associated with women infertility ( Zhou et al . , 2019; Nishimura et al . , 2019 ) but the association between these PTC mutations in ZP4 and woman infertility remains to be explored . The evolutionary conservation of ZP4 in humans ( Hughes and Barratt , 1999; Lefièvre et al . , 2004 ) and many other mammals ( Goudet et al . , 2008; Hedrick and Wardrip , 1987; Hoodbhoy et al . , 2005; Izquierdo-Rico et al . , 2009; Moros-Nicolás et al . , 2018; Noguchi et al . , 1994; Stetson et al . , 2012 ) points towards a conserved role of this protein . In this sense , these results highlight ZP4 mutations as a possible cause for female infertility in both humans and livestock or wild species , and pave the way for the development of contraceptive methods based on ZP4 disruption .
All animal protocols were approved by INIA Animal Welfare Committee and Madrid Region authorities ( authorization PROEX040/17 ) . Rabbits ( Oryctolagus cuniculus , farm commercial hybrids of New Zealand White ) were housed individually maintaining visual and olfactory contact with others to allow social interactions . Temperature was controlled to 20–24°C and light cycle was 14:10 . The generation of Type I genetically modified O . cuniculus was approved by Spanish National Biosafety Agency ( authorization A/ES/17/03 ) . ZP4 KO were generated by CRISPR/Cas system . A single guide RNA ( sgRNA ) was designed at the beginning of the coding region of rabbit ZP4 gene using MIT CRISPR design tool ( Yang et al . , 2014 ) , which also provided the five most probable off-targets . sgRNA and capped polyadenylated Cas9 mRNA were synthesized in vitro as described previously ( Bermejo-Álvarez et al . , 2015 ) . Briefly , capped polyadenylated Cas9 mRNA was produced by in vitro transcription ( mMESSAGE mMACHINE T7 ULTRA kit , Life Technologies ) from the plasmid pMJ920 ( Addgene 42234 ) linearized with BstBI . sgRNA was produced by in vitro transcription ( MEGAshortscript T7 kit , Life Technologies ) from a PCR-amplified template using px330 vector ( Addgene 42230 ) as previously described ( Yang et al . , 2014 ) . Cas9 mRNA and sgRNA were co-injected at 100 ng/µl and 25 ng/µl , respectively , into rabbit zygotes collected by oviductal flushing 14 hr after natural mating . Microinjected embryos were transferred one day after microinjection to the oviduct of a pseudopregnant recipient stimulated by a 0 . 02 mg gonadorelin IM injection ( Inducel , Laboratorios Ovejero ) on the previous day . Resulting pups were genotyped by clonal sequencing as previously described ( Bermejo-Álvarez et al . , 2015 ) . Genotyping was performed on both ear biopsies and blood samples in F0 generation . Following DNA purification , the genomic sequence surrounding CRISPR target site was amplified by PCR and the purified PCR product was cloned into pMD20 vector ( Takara ) and transformed into competent cells . 15 clones were sequenced to detect the unknown alleles ( indels ) generated by NHEJ-repair of CRISPR-induced DSBs on each individual . In subsequent generations , Sanger PCR sequencing was performed , as all parental alleles were already known , allowing its identification in a mixed sequencing reaction . Details of primers used for sgRNA synthesis , genotyping and off-target analysis are provided in Table 1 . Immunohistochemistry was performed on ovaries ( 3/group ) collected 14 hr after mating , fixed in 10% formaldehyde for 24 hr , processed and paraffin embedded . Rehydrated sections of 5 µm were incubated in 0 . 3% hydrogen peroxide in TBS to block endogenous peroxidase activity and then in primary antibody solution ( goat polyclonal anti-human ZP4 ( Santa-Cruz sc-49586 ) dilution 1:2000 in TBS-1% BSA solution ) for 1 hr at room temperature in a moist chamber . After a brief wash in TBS , sections were incubated in secondary antibody solution ( ImmPRESS horse anti-goat IgG HRP , Vector Labs MP-7405 ) for 30 min at 37°C and immunoreaction was revealed with 3–3´diaminobencidine . Finally , sections were counterstained with Harry´s hematoxylin ( Thermo Scientific ) , dehydrated , cleared and mounted . Positive immunoreaction was identified as a dark-brown precipitated . The images were collected with a Leica DM 6000 microscope and Software Leica Application Suite . Proteomics was performed on 50 grown oocytes/group recovered by slicing ovaries in PBS . ZP was solubilized by incubation at 75°C for 45 min ( Izquierdo-Rico et al . , 2009 ) and analyses were carried out on one technical replicate with a HPLC/MS system consisting of an Agilent 1290 Infinity II Series HPLC ( Agilent Technologies , Santa Clara , CA , USA ) equipped with an Automated Multisampler module and a High Speed Binary Pump , and connected to an Agilent 6550 Q-TOF Mass Spectrometer ( Agilent Technologies , Santa Clara , CA , USA ) using an Agilent Jet Stream Dual electrospray ( AJS-Dual ESI ) interface . Experimental parameters for HPLC and Q-TOF were set in MassHunter Workstation Data Acquisition software ( Agilent Technologies , Rev . B . 08 . 00 ) . Data processing and analysis was performed on Spectrum Mill MS Proteomics Workbench ( Rev B . 06 . 00 . 201 , Agilent Technologies , Santa Clara , CA , USA ) . All experiments were performed following natural mating with WT males of proven fertility . Ovulation in rabbits is induced by mating , allowing precise determination of embryo developmental timing . Zygotes were collected post-mortem 14 hr after mating by oviductal flushing with Dulbecco´s PBS media supplemented with 1% BSA ( DPBS ) in five independent replicates . Ovulation rates were assessed by counting corpora lutea . Embryo culture took place on 25 µl drops of TCM-199 media ( Sigma ) supplemented with 5% FCS at 38 . 5°C in a 5% CO2 and 5% O2 water saturated atmosphere . Embryonic cleavage and development to blastocysts were assessed on Days 1 and 5 of culture , respectively . Follicular counts were performed in histological sections of ovaries collected 14 hr after mating , fixed and stained with Haematoxylin-Eosin as detailed above ( four sections per ovary , four animals per group ) . Pronuclear formation rates were analysed in the zygotes obtained from three females/group fixed 20 hr following mating as described previously ( Bermejo-Alvarez et al . , 2012 ) . Pronuclei were stained by incubation in 1 µM SYTOX Green ( Thermofisher ) for 30 min and observed under fluorescent inverted microscope ( Nikon Eclipse ) . To analyse in vivo development to Day 4 blastocyst , three females/group were sacrificed 4 days after mating and oviducts ( where no structures were recovered ) and uteri were independently flushed with DPBS . The number of structures recovered were counted and blastocysts were observed and measured under light estereomicroscopy ( Zeiss V20 coupled with Hammamatsu Orca Flash 4 . 0 camera ) . A similar approach was followed to analyse in vivo development to Day 6 blastocysts , but in this case four females/group were used and only uteri were flushed . Corpora lutea and the number of structures were counted to determine recovery rates . The structures recovered were observed and measured under light estereomicroscopy ( Zeiss Stemi 305 coupled with Izasa CMOS 1080P camera ) . Zona pellucida thickness was analysed in 30 fresh embryos/group 40–43 hr post-mating by inverted light microscopy ( Nikon Eclipse ) using the measurement tool of the software NIS ( Nikon ) . Scanning Electron Microscopy ( SEM ) was performed on five zygotes/group fixed in 2% glutaraldehyde for 2 hr at 4°C , washed in PBS and postfixed in 1% osmium tetroxide for 1 hr . After washing , the zygotes were dehydrated in increasing concentration of acetone and air dried . Finally , samples were sputter coated with gold and studied in Jeol-6100 scanning electron microscope . ZP permeability test was performed by incubating fresh zygotes ( 3 replicates of 3 embryos/group ) in 0 . 005% ( 1:1000 dilution ) solutions of carboxylate-modified 505/515 Fluospheres ( Invitrogen ) of different sizes ( 20 , 40 and 100 nm ) for 30 min , using the culture media and atmosphere conditions detailed above . Immediately following incubation and washing , zygotes were observed under an inverted fluorescent microscope to assess the penetration of fluorescent particles . This test yields a binary result: green fluorescence can be detected inside ZP ( ZP is permeable to that size of particle ) or fluorescence cannot be detected inside ZP ( particles do pass through ZP ) . Loss of function ZP4 polymorphisms passing QC filters were queried online at ( http://gnomad . broadinstitute . org/gene/ENSG00000116996 ) , and a cvs file was downloaded and manipulated with R v3 . 5 . 1 . Figure 6 was plotted taking as reference the protein sequence UniProt Q12836 and using the packages ggplot2 and drawproteins ( Brennan , 2018 ) , which was customized to depict frameshift mutations . Biosequence analysis using profile hidden Markov models was performed online with HMMER v1 . 32 ( Potter et al . , 2018 ) . The differences between groups in litter size , ovulation and developmental rates , follicle number , in vivo embryo survival and ZP and embryo size were analysed by One-way ANOVA following Tukey´s post-hoc test or by t-test to compare 3 or two groups , respectively , except for permeability test ( binary variable ) , where Chi-square test were used . All statistical tests were performed using the software package SigmaStat . Differences were considered statistically significant at p<0 . 05 . | The egg cells of mammals , called oocytes , are encased in a protective layer called the zona pellucida . This layer is made from proteins called ZP1 to 4 . Most studies of the zona pellucida use mice , which do not have ZP4 . This means that the research community have limited knowledge of what ZP4 does in humans and other mammals . Scientists can now use a technique called CRISPR to selectively modify the genetics of living things to help us to understand what specific genes and proteins do . The ZP4 protein can be eliminated from rabbit oocytes using CRISPR to help understand its role in egg cell fertilization and development . Lamas-Toranzo et al . examined the effect of losing ZP4 from rabbit oocytes . Without ZP4 the zona pellucida becomes thinner , irregular and more flexible . However , the loss of ZP4 did not affect ovulation ( i . e . the release of egg cells from an ovary ) , fertilization , or the early stages of development of embryos when studied in the laboratory . However , rabbits without ZP4 were much less fertile . Indeed , only one out of 10 female rabbits without ZP4 was able to deliver pups because in most cases the development of embryos in the womb failed . These findings show that ZP4 has a structural role in the zona pellucida . Without ZP4 fertility is reduced . This work lays the ground for further investigation of the role of ZP4 . It could also offer new insights into the causes of infertility . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2019 | ZP4 confers structural properties to the zona pellucida essential for embryo development |
Microtubule asters - radial arrays of microtubules organized by centrosomes - play a fundamental role in the spatial coordination of animal cells . The standard model of aster growth assumes a fixed number of microtubules originating from the centrosomes . However , aster morphology in this model does not scale with cell size , and we recently found evidence for non-centrosomal microtubule nucleation . Here , we combine autocatalytic nucleation and polymerization dynamics to develop a biophysical model of aster growth . Our model predicts that asters expand as traveling waves and recapitulates all major aspects of aster growth . With increasing nucleation rate , the model predicts an explosive transition from stationary to growing asters with a discontinuous jump of the aster velocity to a nonzero value . Experiments in frog egg extract confirm the main theoretical predictions . Our results suggest that asters observed in large fish and amphibian eggs are a meshwork of short , unstable microtubules maintained by autocatalytic nucleation and provide a paradigm for the assembly of robust and evolvable polymer networks .
Animal cells use asters , radial arrays of microtubules , to spatially organize their cytoplasm ( Wilson , 1896 ) . Specifically , astral microtubules transport organelles ( Grigoriev et al . , 2008; Wang et al . , 2013; Waterman-Storer and Salmon , 1998 ) , support cell motility by mediating mechanical and biochemical signals ( Etienne-Manneville , 2013 ) , and are required for proper positioning of the nucleus , the mitotic spindle , and the cleavage furrow ( Field et al . , 2015; Grill and Hyman , 2005; Neumüller and Knoblich , 2009; Tanimoto et al . , 2016; Wilson , 1896 ) . Within asters , individual microtubules undergo dynamic instability ( Mitchison and Kirschner , 1984 ) : They either grow ( polymerize ) or shrink ( depolymerize ) at their plus ends and stochastically transition between these two states . The collective behavior of microtubules is less well understood , and it is not clear how dynamic instability of individual microtubules controls aster growth and function . The standard model of aster growth posits that centrosomes nucleate and anchor all microtubules at their minus ends while the plus ends polymerize outward via dynamic instability ( Brinkley , 1985 ) . As a result , aster growth is completely determined by the dynamics of individual microtubules averaged over the growing and shrinking phases . In particular , the aster either expands at a velocity given by the net growth rate of microtubules or remains stationary if microtubules are unstable and tend to depolymerize ( Belmont et al . , 1990; Dogterom and Leibler , 1993; Verde et al . , 1992 ) . The standard model of aster growth is being increasingly challenged by reports of microtubules with their minus ends located far away from centrosomes ( Akhmanova and Steinmetz , 2015; Keating and Borisy , 1999 ) . Some of these microtubules may arise simply by detachment from centrosomes ( Keating et al . , 1997; Waterman-Storer et al . , 2000 ) or severing of pre-existing microtubules ( Roll-Mecak and McNally , 2010 ) . However , new microtubules could also arise due to a nucleation process independent of centrosomes ( Clausen and Ribbeck , 2007; Efimov et al . , 2007; Petry et al . , 2013 ) and contribute to both aster growth and its mechanical properties . Indeed , we recently observed that centrosomal nucleation is insufficient to explain the large number of growing plus ends found in asters ( Ishihara et al . , 2014 ) . Moreover , the standard model demands a decrease in microtubule density at aster periphery , which is inconsistent with aster morphology in frog and fish embryos ( Wühr et al . , 2008 , 2010 ) . To resolve these inconsistencies , we proposed an autocatalytic nucleation model , where microtubules or microtubule plus ends stimulate the nucleation of new microtubules at the aster periphery ( Ishihara et al . , 2014a , 2014b; Wühr et al . , 2009 ) . This mechanism generates new microtubules necessary to maintain a constant density as the aster expands . We also hypothesized that autocatalytic nucleation could effectively overcome extinction of individual microtubules , and allow rapid growth of large asters made of short , unstable microtubules . However , we did not provide a quantitative model that can be compared to the experiments or even show that the proposed mechanism is feasible . Here , we develop a quantitative biophysical model of aster growth with autocatalytic nucleation . It predicts that asters can indeed expand even when individual microtubules turn over and disappear by depolymerization . In this regime , aster expansion is driven by the increase in the total number of microtubules , and the resulting aster is a network of short interconnected microtubules . The transition from stationary to growing asters depends on the balance between polymerization dynamics and nucleation . At this transition , our theory predicts a minimum rate at which asters grow , which we define as the gap velocity . This gap velocity arises due to the dynamic instability of microtubule polymerization and excludes a wide class of alternative models . More importantly , this mode of aster growth allows the cell to assemble asters with varying polymer densities at consistently large speeds . Using a cell-free reconstitution approach ( Field et al . , 2014; Nguyen et al . , 2014 ) , we perform biochemical perturbations and observe the slowing down and eventual arrest of aster growth with a substantial gap velocity at the transition . By combining theory and experiments , we provide a quantitative framework for how the cell cycle may regulate the balance between polymerization dynamics and nucleation to control aster growth . We propose that the growth of large interphase asters is an emergent property of short microtubules that constantly turnover and self-amplify .
Asters are large structures comprised of thousands of microtubules . How do the microscopic dynamics of individual microtubules determine the collective properties of asters such as their morphology and growth rate ? Can asters sustain growth when individual microtubules are unstable ? To address these questions , we develop a theoretical framework that integrates polymerization dynamics and autocatalytic nucleation ( Figure 1A ) . Our main goal is to determine the distribution of microtubules within asters and the velocity at which asters grow: ( 1 ) V=dRadiusdt . 10 . 7554/eLife . 19145 . 003Figure 1 . A biophysical model for the collective growth of microtubule asters . ( A ) We propose that asters grow via two microscopic processes: polymerization and nucleation . Individual microtubules follow the standard dynamic instability with a growing state with polymerization rate vg and a shrinking state with depolymerization rate vs . Transitions between the states occur at rates fcat and fres , which model catastrophe and rescue events , respectively . New microtubules are added at a rate r via a nucleation at pre-existing plus ends in the growing state . ( B ) Individual vs . collective growth of asters . In the standard model of ‘individual growth’ , asters increase their radius at rate V=dRadiusdt only via a net polymerization from the centrosome ( yellow ) . Thus , this model predicts that the rate of aster growth equals the mean polymerization rate V=J , the number of microtubules is constant , and their density decreases away from the centrosomes . In the collective growth model , the microtubule density is constant and the number of microtubules increases . Autocatalytic nucleation makes asters grow faster than the net polymerization rate J and can sustain growth even when individual microtubules are unstable J<0 . DOI: http://dx . doi . org/10 . 7554/eLife . 19145 . 003 Beyond being the main experimental readout , the aster velocity is crucial for cell physiology because it allows large egg cells to divide its cytoplasm rapidly . Polymerization dynamics of plus ends is an individual property of microtubules . To describe plus end dynamics , we adopt the two-state model of microtubule dynamic instability ( Figure 1A , left ) . In this model , a single microtubule is in one of the two states: ( i ) the growing state , where plus ends polymerize at rate vg and ( ii ) the shrinking state , where plus ends depolymerize at rate vs . A growing microtubule may transition to a shrinking state ( catastrophe event ) with rate fcat . Similarly , the shrinking to growing transition ( rescue event ) occurs at rate fres . For large asters growing in Xenopus egg cytoplasm , we provide estimates of these parameters in Table 1 . 10 . 7554/eLife . 19145 . 004Table 1 . Model parameters used to describe large aster growth reconstituted in interphase Xenopus egg extract . DOI: http://dx . doi . org/10 . 7554/eLife . 19145 . 004QuantitySymbolValueCommentPolymerization ratevg30 μm/minMeasured from growing plus ends and EB1 cometsDepolymerization ratevs42 μm/minMeasured from shrinking plus ends ( Ishihara et al . , 2014a ) Catastrophe ratefcat3 . 3 min−1Measured from EB1 comet lifetimes ( see Materials and methods ) Rescue ratefres2 . 0±0 . 3 min−1Estimated from Equations ( 4 ) and ( 6 ) Autocatalytic nucleation rater2 . 1±0 . 2 min−1Estimated from Equations ( 4 ) and ( 6 ) Carrying capacity of growing endsK0 . 4 μm−2Estimated from comparing Cgbulk to predicted ( see SI ) Mean microtubule length⟨l⟩16 ± 2 μmEstimated from from dynamics parameters ( see SI ) Aster velocityV22 . 3±2 . 6 μm/minMeasured from rate of aster radius increaseGap velocityVgap12 . 8±1 . 7 μm/minMeasured from aster growth at 320 nM MCAK-Q710Bulk growing plus end densityCgbulk0 . 053±0 . 030 μm−2Measured from EB1 comet density ( Ishihara et al . , 2014a ) Plus end dynamics can be conveniently summarized by the time-weighted average of the polymerization and depolymerization rates ( Dogterom and Leibler , 1993; Verde et al . , 1992 ) : ( 2 ) J=vgfres-vsfcatfres+fcat . This parameter describes the tendency of microtubules to grow or shrink . When J<0 , microtubules are said to be in the bounded regime because their length inevitably shrinks to zero , i . e . microtubule disappears . When J>0 , microtubules are said to be in the unbounded regime , because they have a nonzero probability to become infinitely long . Parameter J also determines the mean elongation rate of a very long microtubule that persists over many cycles of catastrophe and rescue . The dynamics of short microtubules , however , depends on their length and initial state ( growing vs . shrinking ) and should be analyzed carefully . The standard model posits that asters are produced by the expansion of individual microtubules , so the transition from small mitotic asters to large interphase asters is driven by a change in the sign of J ( Dogterom and Leibler , 1993; Verde et al . , 1992 ) ( Figure 1B left , ‘individual growth’ ) . With bounded dynamics J<0 , the standard model predicts that every microtubule shrinks to zero length and disappears . This microtubule loss is balanced by nucleation of new microtubules at the centrosomes , the only place where nucleation is allowed in the standard model . As a result , asters remain in the stationary state and are composed of a few short microtubules , and the aster velocity is thus V=0 . With unbounded dynamics J>0 , the standard model predicts an aster that has a constant number of microtubules and increases its radius at a rate equal to the elongation rate of microtubules ( i . e . V=J ) . Below , we add autocatalytic microtubule nucleation ( Figure 1A , right ) to the standard model and propose the regime of ‘collective growth’ ( Figure 1B , right ) . In this regime , asters grow ( V>0 ) although individual microtubules are bounded ( J<0 ) and are , therefore , destined to shrink and disappear . The growth occurs because more microtubules are nucleated than lost , and new microtubules are typically nucleated further along the expansion direction . Indeed , when a new microtubule is nucleated , it is in a growing state and starts expanding outward before its inevitable collapse . During its lifetime , this microtubule can nucleate a few more microtubules all of which are located further along the expansion direction . As we show below , this self-amplifying propagation of microtubules is possible only for sufficiently high nucleation rates necessary to overcome microtubule loss and sustain collective growth . Specifically , we assume that new microtubules nucleate at locations away from centrosomes at rate Q . This rate could depend on the local density of growing plus ends if they serve as nucleation sites or the local polymer density if nucleation occurs throughout a microtubule . The new microtubules have zero length and tend to grow radially due to mechanical interactions with the existing microtubule network . These non-centrosomal microtubules disappear when they shrink back to their minus ends . Our assumptions are broadly consistent with known microtubule physiology ( Clausen and Ribbeck , 2007; Petry et al . , 2013 ) , and we found strong evidence for nucleation away from centrosomes in egg extract by microtubule counting in growing asters ( Ishihara et al . , 2014a ) . Below , we consider plus-end-stimulated nucleation and the analysis for the polymer-stimulated nucleation is summarized in the SI . Without negative feedback , autocatalytic processes lead to exponential growth , but there are several lines of evidence for an apparent ‘carrying capacity’ of microtubules in a given cytoplasmic volume ( Clausen and Ribbeck , 2007; Ishihara et al . , 2014a; Petry et al . , 2013 ) . Saturation is inevitable since the building blocks of microtubules are present at a fixed concentration . In our model , we impose a carrying capacity by expressing autocatalytic nucleation as a logistic function of the local density of growing plus ends , which is qualitatively consistent with local depletion of nucleation factors such as the gamma-tubulin ring complex . Other forms of negative feedback ( e . g . at the level of polymerization dynamics ) are possible as well . In SI , we show that the type of negative feedback does not affect the rate of aster growth , which is determined entirely by the dynamics at the leading edge of a growing aster where the microtubule density is small and negative feedback can be neglected . Assuming a large number of microtubules , we focus on the mean-field or deterministic dynamics ( SI ) and formalize our model as a set of partial differential equations . Specifically , we let ρg ( t , x , l ) and ρs ( t , x , l ) denote respectively the number of growing and shrinking microtubules of length l with their minus ends at distance x>0 from the centrosome . The number of newly nucleated microtubules is given by Q ( x ) =rCg ( t , x ) ( 1-Cg ( t , x ) /K ) , where r is the nucleation rate , K is the carrying capacity controlling the maximal plus end density , and Cg ( t , x ) is the local density of the growing plus ends at point x . The polymerization dynamics and nucleation are then described by , ( 3 ) {∂ρg∂t=−vg∂ρg∂l−fcatρg+fresρs+Q ( x ) ⋅δ ( l ) ∂ρs∂t=+vs∂ρs∂l+fcatρg−fresρs . Note that polymerization and depolymerization changes the microtubule length l , but not the minus end position x . Equations at different x are nevertheless coupled due to the nucleation term , which depends on x through Cg . To understand this system of equations , consider the limit of no nucleation ( r=0 ) . Then , the equations at different x become independent and we recover the standard model that reduces aster growth to the growth of individual microtubules ( Dogterom and Leibler , 1993; Verde et al . , 1992 ) . With nucleation , aster growth is a collective phenomenon because microtubules of varying length and minus end positions contribute to Cg ( t , x ) , which can be expressed as a convolution of ρg ( see SI ) . The delta-function δ ( l ) ensures that newly nucleated microtubules have zero length . Finally , we need to specify what happens when microtubules shrink to zero length . In our model , microtubules originating from centrosomes rapidly switch from shrinking to growth ( i . e . re-nucleate ) , while non-centrosomal microtubules disappears completely ( i . e . no re-nucleation occurs ) . We further assume that mother and daughter microtubules disappear without affecting each other . Indeed , if the collapse of the mother microtubule triggered the collapse of the daughter microtubule ( or vice versa ) , then no net increase in the number of microtubules would be possible in the bounded regime . One consequence of this assumption is that the minus end of a daughter microtubule becomes detached from any other microtubules in the aster following the collapse of the mother microtubule . As a result , minus ends need to be stabilized after nucleation possibly by some additional factors ( Akhmanova and Hoogenraad , 2015 ) and mechanical integrity of the aster should rely on microtubule bundling ( Ishihara et al . , 2014a ) . To check if our model can describe aster growth , we solved Equation ( 3 ) numerically using finite difference methods in an 1D planar geometry . With relatively low nucleation rates and J<0 , microtubule populations reached a steady-state profile confined near the centrosome reminiscent of an aster in the standard model with bounded microtubule dynamics ( Figure 2A left ) . When the nucleation rate was increased , the microtubule populations expanded as a traveling wave with an approximately invariant shape and constant microtubule density at the periphery ( Figure 2A right ) consistent with the growth of interphase asters in our reconstitution experiments ( Figure 2B and Ishihara et al . , 2014a ) . Thus , our model predicted two qualitatively different states: stationary and growing asters . 10 . 7554/eLife . 19145 . 005Figure 2 . Our model captures key features of large aster growth . ( A ) Time evolution of growing plus end density predicted by our model , which we solved via numerical simulations in 1D geometry . In the stationary regime , the microtubule population remained near the centrosome vg=30 , vs=40 , fcat=3 , fres=1 , and r=1 . 0 ( left ) . In contrast , outward expansion of the microtubule population was observed when the nucleation rate was increased to r=2 . 5 , above the critical nucleation rate rc ( right ) . For both simulations , microtubules are in the bounded regime J<0 . ( B ) Experimental measurements confirm that asters expand at a constant rate with time-invariant profiles of the plus end density , as predicted by our model . The plus end densities were estimated as EB1 comet density during aster growth as previously described ( Ishihara et al . , 2014a ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19145 . 005© 2014 Proceedings of the National Academy of Sciences of the United States of America . All Rights Reserved2014Proceedings of the National Academy of Sciences of the United States of AmericaPanel B reprinted with permission from Figure 4C from ( Ishihara et al . , 2014a ) , Proceedings of the National Academy of Sciences of the United States of America . Not covered by the terms of the Creative Commons Attribution 4 . 0 International license . Next , we solved Equation ( 3 ) exactly and obtained the following analytical expression for the aster velocity in terms of model parameters: ( 4 ) V=vg ( vgfres−vsfcat ) 2 ( vg ( vgfres−vsfcat ) ( fres+fcat ) + ( vg+vs ) ( vgfres+vsfcat ) r−2 ( vg+vs ) vgfcatfresr ( vgfres−vsfcat+vsr ) ) , which holds for the parameter range rc<r<fcat . The details of the calculation , including the definition of rc are summarized in SI . Using this expression , we summarize how aster velocity V is affected by the mean polymerization rate J ( Figure 3A ) and nucleation rate r ( Figure 3B ) . In the absence of autocatalytic nucleation ( r=0 ) , our model reduces to the standard model and predicts that asters only grow when J>0 with V=J ( Figure 3A blue line ) . When nucleation is allowed ( r>0 ) , the aster velocity increases with r and asters can grow even when individual microtubules are unstable J<0 ( Figure 3A and B ) . During this collective growth , the aster expands radially because more microtubules are nucleated than lost at the front . In the aster bulk , nucleation is reduced from the carrying capacity , and the aster exists in the dynamic balance between microtubule gain due to nucleation and loss due to depolymerization . Since microtubules are in the bounded regime , their lifetime is short , and they disappear before reaching an appreciable length . In sharp contrast to the standard model , we predict that asters are a dynamic network of short microtubules with properties independent from the distance to the centrosome . Thus , nucleation not only increases the number of microtubules , but also controls the growth rate and spatial organization of asters enabling them to span length scales far exceeding the length of an individual microtubule . 10 . 7554/eLife . 19145 . 006Figure 3 . Explosive transition from stationary to growing asters and other theoretical predictions . Analytical solution ( lines ) and numerical simulations ( dots ) predict that asters either remain stationary or expand at a constant velocity , which increases with the net polymerization rate J ( A ) and nucleation rate r ( B ) . The transition to a growing state is accompanied by a finite jump in the expansion velocity labeled as Vgap . ( A ) The behavior in the standard model ( r=0 ) is shown in blue and our model ( r=1 . 5 ) in red . Note that aster growth commences at J<0 in the presence of nucleation and occurs at a minimal velocity Vgap . Although spatial growth can occur for both J>0 and J<0 the properties of the resulting asters could be very different ( see SI ) . Here , vg=30 , vs=30 , fcat=3 . ( B ) If J<0 , critical nucleation rc is required to commence aster growth . Blue line corresponds to J>0 ( vg=30 , vs=15 , fcat=3 , fres=3 ) and red line to J<0 ( vg=30 , vs=15 , fcat=3 , fres=1 ) . See Materials and methods and SI for the details of the analytical solution and numerical simulations . DOI: http://dx . doi . org/10 . 7554/eLife . 19145 . 00610 . 7554/eLife . 19145 . 007Figure 3—figure supplement 1 . Feedback regulation of catastrophe rate leads to the same explosive transition . ( A ) Time evolution of growing plus end density similar to the simulations in 1D planar geometry as in Figure 2 . This example represents a scenario where the nucleation rate is above the critical nucleation rate ( vg=30 , vs=40 , fcat=3 , fres=1 , r=2 . 5 ) resulting in aster growth . ( B ) Analytical solution ( lines ) and numerical simulations ( dots ) predict aster velocity as a function of nucleation rate similar to Figure 3A . Blue line corresponds to J>0 ( vg=30 , vs=15 , fcat=3 , fres=3 ) and red line to J<0 ( vg=30 , vs=15 , fcat=3 , fres=1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19145 . 00710 . 7554/eLife . 19145 . 008Figure 3—figure supplement 2 . Aster growth by polymer-stimulated nucleation leads to the same explosive transition . ( A ) Time evolution of growing plus end density similar to Figure 2A . Below the critical nucleation rate , asters are stationary ( left , vg=30 , vs=40 , fcat=3 , fres=1 , p=0 . 07 ) . Above the critical nucleation rate , asters grow in radius ( right , p=0 . 18 ) even when microtubules are unstable ( J<0 ) . Here , the critical polymer nucleation rate pc=0 . 0964 . . . as predicted by Equation ( A59 ) . ( B ) Numerical simulations predict aster velocity as a function of J ( fres was varied while keeping vg=30 , vs=15 , fcat=3 , p=0 . 04 ) . ( C ) Numerical simulations predict aster velocity as a function of polymer-stimulated nucleation rate p ( p was varied while keeping vg=30 , vs=15 , fcat=3 , fres=0 . 3 ) . Dashed vertical lines indicate the predicted critical transitions from Equation ( A59 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19145 . 008 When J<0 , a critical nucleation rate is required for aster growth ( Figure 3B ) . Indeed , microtubules constantly disappear as their length shrinks to zero , and the nucleation of new microtubules needs to occur frequently enough to overcome the microtubule loss . Consistent with this argument , our analytical solution predicts no aster growth below a certain value of nucleation ( SI ) , termed critical nucleation rate rc: ( 5 ) rc=fcat−vgvsfres . The right hand side of this equation is the inverse of the average time that a microtubule spends in the growing state before shrinking to zero-length and disappearing ( SI ) . Thus , aster growth requires , on average , a microtubule to nucleate at least one new microtubule during its lifetime . The dependence of the critical nucleation rate on model parameters is very intuitive . Increasing the parameters in favor of polymerization ( vg and fres ) , lowers the threshold level of nucleation required for aster growth , while increasing the parameters in favor of depolymerization ( vs and fcat ) has the opposite effect . We also find that rc=0 when J = 0 , suggesting that there is no critical nucleation rate for J≥0 . This limit is consistent with the standard model with J>0 and r=0 where the aster radius increases albeit with radial dilution of microtubule density ( Figure 1B ) . The critical nucleation rate conveys the main implication of our theory: the balance between polymerization dynamics and autocatalytic nucleation defines the quantitative condition for continuous aster growth . At the critical nucleation rate r=rc , the aster velocity V takes a positive , nonzero value ( Figure 3 ) , which we refer to as the ‘gap velocity’ ( SI ) : ( 6 ) Vgap≡limr→rcV=−vgvs ( vgfres−vsfcat ) vg2fres+vs2fcat . This finite jump in the aster velocity is a consequence of microtubules with finite length undergoing dynamic instability and is in sharp contrast to the behavior of reaction-diffusion systems , where traveling fronts typically become infinitesimally slow before ceasing to propagate ( Chang and Ferrell , 2013; Hallatschek and Korolev , 2009; Méndez et al . , 2007; van Saarloos , 2003 ) . One can understand the origin of Vgap>0 when microtubules are eliminated after a catastrophe event ( fres=0 , J=−vs ) . In this limit , plus ends always expand with the velocity vg until they eventually collapse . Below rc , this forward expansion of plus ends fails to produce aster growth because the number of plus ends declines on average . Right above rc , the number of plus ends is stable , and the aster grows at the same velocity as every individual microtubule . Indeed , Equation ( 6 ) predicts that Vgap=vg when fres=0 . The dynamics are similar for fres>0 . At the transition , nucleation stabilizes a subpopulation of microtubules expanding forward , and their average velocity sets the value of Vgap . We also find that the magnitude of Vgap is inversely proportional to the mean length of microtubules in the system ( SI ) . Thus , the shorter the microtubules , the more explosive this transition becomes . In the SI , we also show that microtubule density inside the aster is proportional to r-rc . Thus , the density is close to zero during the transition from stationary to growing asters , but quickly increases as the nucleation rate becomes larger . As a result , cells can achieve rapid aster growth while keeping the density of the resulting microtubule network sufficiently low . The low density might be beneficial because of its mechanical properties or because it simply requires less tubulin to produce and energy to maintain . In addition , the explosive transition to growth with Vgap>0 allows the cell to independently control the aster density and growth speed . Model parameters other than the nucleation rate can also be tuned to transition asters from growth to no growth regimes . Similar to Equation ( 5 ) and ( 6 ) , one can define the critical parameter value and gap velocity to encompass all such transitions ( Appendix 4—table 1 ) . In all cases , we find that the onset of aster growth is accompanied by a discontinuous increase in the aster velocity . The finite jump in aster velocity is similarly predicted in a wide range of alternative scenarios including ( i ) feedback regulation of plus end dynamics ( SI and Figure 3—figure supplement 1 ) and ( ii ) aster growth by microtubule polymer-stimulated nucleation ( SI and Figure 3—figure supplement 2 ) . In summary , the gap velocity is a general prediction of the collective behavior of microtubules that are short-lived . Based on our theory , we reasoned that it would be possible to transform a growing interphase aster to a small , stationary aster by tuning polymerization dynamics and/or nucleation via biochemical perturbations in Xenopus egg extract . To this end , we performed reconstitution experiments in undiluted interphase cytoplasm supplied with anti-Aurora kinase A antibody coated beads , which nucleate microtubules and initiate aster growth under the coverslip ( Field et al . , 2014; Ishihara et al . , 2014a ) . We explored perturbation of various regulators for plus end dynamics and nucleation . We settled on perturbation of MCAK/KIF2C , classically characterized as the main catastrophe-promoting factor in the extract system ( Kinoshita et al . , 2001; Walczak et al . , 1996 ) , and imaged aster growth . In control reactions , aster radius , visualized by the plus end marker EB1-mApple , increased at velocities of 20 . 3±3 . 1 μm/min ( n = 21 asters ) . We saw no detectable changes to aster growth with the addition of the wild type MCAK protein . In contrast , addition of MCAK-Q710 ( Moore and Wordeman , 2004 ) decreased aster velocity ( Figure 4A and B ) . At concentrations of MCAK-Q710 above 320 nM , most asters had small radii with very few microtubules growing from the Aurora A beads . In our model , this behavior is consistent with any change of parameter ( s ) that reduces the aster velocity ( Equation 4 ) and arrests growth . 10 . 7554/eLife . 19145 . 009Figure 4 . Titration of MCAK-Q710 slows then arrests aster growth through a discontinuous transition . ( A ) Addition of MCAK-Q710 results in smaller interphase asters assembled by Aurora A beads in Xenopus egg extract . Images were obtained 20 min post initiation with the plus end marker EB1-mApple . Dotted lines indicate the approximate outline of asters . ( B ) Aster velocity decreases with MCAK-Q710 concentration and then abruptly vanishes as predicted by the model . Note a clear gap in the values of the observed velocities and bimodality near the transition , which support the existence of Vgap . Quantification methods are described in methods and Figure 4—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 19145 . 00910 . 7554/eLife . 19145 . 010Figure 4—figure supplement 1 . Aurora A kinase bead asters at different MCAK-Q710 concentrations . ( A ) Measuring aster growth velocities from time-lapse images of asters visualized with the plus end marker EB1-mApple . A linear region is chosen in the radial outward direction ( left ) . The raw fluorescent intensity profiles ( center ) are subjected to a low pass filter ( right ) , and the half-max position was manually selected to define the radius of the aster at different time points . Blue to red lines indicate profiles at two minute intervals . ( B ) At the critical concentration of 320 nM MCAK-Q710 , some asters assembled from Aurora A beads showed slow growth ( top ) while others contained few microtubules which gradually decreased over time ( bottom ) . The latter was scored as zero growth velocity . The reaction was started at time zero by the addition of calcium and beads to the extract . Scale bars 100 μm . ( C ) Aster growth velocities measured at increasing MCAK-Q710 concentrations . Biological replicate of the same experiment as in Figure 4B . DOI: http://dx . doi . org/10 . 7554/eLife . 19145 . 01010 . 7554/eLife . 19145 . 011Figure 4—figure supplement 2 . Pellicle asters at different MCAK-Q710 concentrations . ( A ) Asters assembled by Tetrahymena pellicles as the nucleating center showed aster growth which was slowed down by MCAK-Q710 ( top ) . At higher MCAK-Q710 concentrations , stationary asters that did not change its radius for over 60 min ( bottom ) . ( B ) EB1-mApple fluorescence intensity profile of the stationary aster in panel A for time intervals 70–84 min post calcium addition . Such asters were scored as zero velocity . ( C ) Pellicle aster growth velocities at different MCAK-Q710 concentrations . DOI: http://dx . doi . org/10 . 7554/eLife . 19145 . 01110 . 7554/eLife . 19145 . 012Figure 4—figure supplement 3 . Plus end polymerization rate and catastrophe rate do not significantly change with MCAK-Q710 titration . Measurements were made by imaging and tracking EB1 comets in growing interphase asters assembled by Aurora A beads ( see Materials and methods ) . ( A ) Distribution of plus end polymerization rates at different MCAK-Q710 concentrations . ( B ) Distribution of EB1 comet lifetimes at different MCAK-Q710 concentrations . Inset shows the same data plotted on a semilog scale . ( C ) Summary of measurements from EB1 tracking analysis . The table shows the number of movies ( or asters ) and total number of tracks analyzed for each condition . Errors indicate standard error . The catastrophe rate fcat was derived from a linear fit to the semilog plots of the lifetime distributions in the intervals 5–60 s . Its mean and standard error were calculated by bootstrapping . DOI: http://dx . doi . org/10 . 7554/eLife . 19145 . 012 At 320 nM MCAK-Q710 concentration , we observed bimodal behavior . Some asters increased in radius at moderate rates , while other asters maintained a stable size before disappearing , presumably due to the decrease of centrosomal nucleation over time ( Figure 4—figure supplement 1 and Ishihara et al . , 2014a ) . In particular , we observed no asters growing at velocities between 0 and 9 μm/min ( Figure 4B and Figure 4—figure supplement 1 ) . This gap in the range of possible velocities is consistent with the theoretical prediction that growing asters expand above a minimal rate Vgap . To confirm that the failure of aster growth at high concentrations of MCAK-Q710 is caused by the changes in aster growth rather than nucleation from the beads , we repeated the experiments with Tetrahymena pellicles as the initiating centers instead of Aurora A beads . Pellicles are pre-loaded with a high density of microtubule nucleating sites , and are capable of assembling large interphase asters ( Ishihara et al . , 2014a ) . We found pellicle initiated asters to exhibit a similar critical concentration of MCAK-Q710 compared to Aurora A bead asters . While the majority of Aurora A beads subjected to the highest concentration of MCAK-Q710 lost growing microtubules over time , a significant number of microtubules persisted on pellicles even after 60 min ( Figure 4—figure supplement 2 ) . The radii of these asters did not change , consistent with our prediction of stationary asters . Thus , the pellicle experiments confirmed our main experimental result of small , stationary asters and that the nature of transition is consistent with the existence of a gap velocity . Finally , we asked which parameters in our model were altered in the MCAK-Q710 perturbation . To this end , we measured the polymerization and catastrophe rates in interphase asters assembled by Aurora A beads at various MCAK-Q710 concentrations . We imaged EB1 comets at high spatiotemporal resolution , and analyzed their trajectories by tracking-based image analysis ( Applegate et al . , 2011; Matov et al . , 2010 , Materials and methods ) . Neither the polymerization nor the catastrophe rate changed at the MCAK-Q710 concentrations corresponding to the transition between growing and stationary asters ( Figure 4—figure supplement 3 ) . MCAK-Q710 has been reported to reduce microtubule polymer levels in cells ( Moore and Wordeman , 2004 ) , but its precise effect on polymerization dynamics and/or nucleation remains unknown . Our data are consistent with the following three scenarios for how MCAK-Q710 antagonizes microtubule assembly: ( i ) increased depolymerization rate , ( ii ) decreased rescue rate , and/or ( iii ) decreased nucleation rate .
It has not been clear whether the standard model of aster growth can explain the morphology of asters observed in all animal cells , including those of extreme size ( Mitchison et al . , 2015 ) . To resolve this question , we constructed a biophysical framework that incorporates microtubule polymerization dynamics and autocatalytic nucleation . Numerical simulations and analytical solutions ( Figures 2 and 3 , and Figure 3—figure supplements 1 and 2 ) recapitulated both stationary and continuously growing asters in a parameter-dependent manner . Interestingly , the explosive transition from ‘growth’ to ‘no growth’ was predicted to involve a finite aster velocity , which we confirmed in biochemical experiments ( Figure 4 ) . Our biophysical model offers a biologically appealing explanation to aster growth and allows us to estimate parameters that are not directly accessible: the rescue and autocatalytic nucleation rates . For example , if we assume that MCAK-Q710 decreases the nucleation rate , we may use the Vgap equation for r→rc ( Equation ( 6 ) ) , the equation for aster velocity V ( Equation ( 4 ) ) , and our measurements of vg , vs , fcat , V , and Vgap ( Table 1 ) to simultaneously estimate fres and r . These results are summarized in Table 1 . Our inferred value of autocatalytic nucleation r = 2 . 1 min−1 is comparable to previous estimates: 1 . 5 min−1 ( Clausen and Ribbeck , 2007 ) and 1 min−1 ( Petry et al . , 2013 ) in meiotic egg extract supplemented with RanGTP . In the alternative scenarios , where MCAK-Q710 decreases the catastrophe rate or increases the depolymerization rate , our estimates of r and fres are essentially the same ( Appendix 8—table 1 ) . Thus , our model recapitulates aster growth with reasonable parameter values and offers a new understanding for how asters grow to span large cytoplasms even when individual microtubules are unstable . To date , few studies have rigorously compared the mechanistic consequences of plus-end-stimulated vs . polymer-stimulated nucleation . Above , we presented the theoretical predictions for aster growth by plus-end stimulated nucleation . In the SI , we also provide the results for polymer-stimulated nucleation including the critical nucleation rate Equation A59 . Both models of nucleation have qualitatively similar behavior including the gap velocity and recapitulate experimental observations of asters growing as traveling waves . Thus , in our case , the qualitative conclusions do not depend on the precise molecular mechanism of autocatalytic nucleation . In particular , the explosive transition characterized by the gap velocity is a general prediction of modeling microtubules as self-amplifying elements whose lifetime depends on their length . By carefully defining and quantifying autocatalytic nucleation , future studies may be able to distinguish its precise mechanism . With plus-end-stimulated nucleation , the nucleation rate r has units of min−1 and describes the number of new microtubules generated per existing plus end per minute . With polymer-stimulated nucleation , the nucleation rate has units of μm−1 min−1 , and describes the number of new microtubules generated per micron of existing microtubule per minute . This difference has important implications for the structural mechanism of microtubule nucleation and for the prediction of cell-scale phenomena . For the issue of large aster growth , we propose specific experiments that might be able distinguish these scenarios ( SI ) . How do large cells control aster size during rapid divisions ? We summarize our theoretical findings with a phase diagram for aster growth in Figure 5 . Small mitotic asters are represented by stationary asters found in the regime of bounded polymerization dynamics J<0 and low nucleation rates . These model parameters must change as cells transition from mitosis to interphase to produce large growing asters . Polymerization dynamics becomes more favorable to elongation during interphase ( Belmont et al . , 1990; Verde et al . , 1992 ) . This may be accompanied by an increased autocatalytic nucleation of microtubules . 10 . 7554/eLife . 19145 . 013Figure 5 . Phase diagram for aster growth . Aster morphology is determined by the balance of polymerization dynamics and autocatalytic nucleation . Small , stationary asters ( V=0 ) , as observed during mitosis , occur at low nucleation r and net depolymerization of individual microtubules ( J<0 ) . Net polymerization ( J>0 ) without nucleation ( r=0 ) produces asters that expand at rate V=J with dilution of microtubule density at the periphery and are thus inconsistent with experimental observations . The addition of nucleation to the individual growth regime changes these dynamics only marginally ( yellow region ) ; see SI . Alternatively , the transition from stationary to growing asters can be achieved by increasing the nucleation rate , r , while keeping J negative . Above the critical nucleation rate rc starts the regime of collective growth ( V as in Equation ( 4 ) , which is valid for r<fcat ) that produces asters composed of relatively short microtubules ( red region ) . The transition from stationary aster to collective growth may be achieved by crossing the curve at any location , but always involves an explosive jump in aster velocity , Vgap . The reverse transition recapitulates the results of our experimental perturbation of MCAK activity ( Figure 4 ) and mitotic entry ( solid arrows ) . We propose this unified biophysical picture as an explanation for the cell cycle dependent changes of aster morphology in vivo . DOI: http://dx . doi . org/10 . 7554/eLife . 19145 . 013 According to the standard model , increasing J to a positive value with no nucleation leads to asters in the 'individual growth' regime . A previous study suggested the interphase cytoplasm is in the unbounded polymerization dynamics J>0 ( Verde et al . , 1992 ) , but our measurements of parameters used to calculate J differ greatly ( Table 1 ) . The individual growth regime is also inconsistent with the steady-state density of microtubules at the periphery of large asters in both fish and frog embryos ( Ishihara et al . , 2014a; Wühr et al . , 2008 , 2010 ) . Experiments in egg extracts further confirm the addition of new microtubules during aster growth ( Ishihara et al . , 2014a ) contrary to the predictions of the standard model . Furthermore , the presence of a high density of growing plus ends in the interior of growing asters in egg extract suggests that microtubules must be short compared to aster radius , and J must be negative , at least in the aster interior ( Ishihara et al . , 2014a ) . By constructing a model that incorporates autocatalytic nucleation r>0 , we discovered a new regime , in which continuous aster growth is supported even when microtubules are unstable ( J<0 ) . We call this the ‘collective growth’ regime because individual microtubules are much shorter ( estimated mean length of 16 μm ± 2 μm , Table 1 ) than the aster radius ( hundreds of microns ) . Predictions of this model are fully confirmed by the biochemical perturbation via MCAK-Q710 . The finite jump in the aster velocity ( Figure 4 ) is in sharp contrast to the prediction of the standard models of spatial growth ( Fisher , 1937; Kolmogorov and Petrovskii , 1937; Skellam , 1951; van Saarloos , 2003 ) . Spatial growth is typically modeled by reaction-diffusion processes that account for birth events and random motion , which , in the context of microtubules , correspond to the nucleation and dynamic instability of plus ends . Reaction-diffusion models , however , neglect internal dynamics of the agents such as the length of a microtubule . As a result , such models inevitably predict a continuous , gradual increase in the aster velocity as the model parameters are varied ( Chang and Ferrell , 2013; Hallatschek and Korolev , 2009; Méndez et al . , 2007; van Saarloos , 2003 ) . The observation of finite velocity jump provides a strong support for our model and rules out a very wide class of models that reproduce the overall phenomenology of aster growth including the constant velocity and profile shape ( Figure 2 ) . In particular , the observation of Vgap excludes the model that we previously proposed based on the analogy of aster growth and the Fisher-Kolmogorov equation ( Ishihara et al . , 2014b ) . The implications of Vgap for model selection are further discussed in SI . Our theory allows for independent regulation of aster growth rate and microtubule density through the control of the nucleation rate and microtubule polymerization . Thus , cells have a lot of flexibility in optimizing aster properties and behavior . The existence of a gap velocity results in switch-like transition from quiescence to rapid growth and allows cells to drastically alter aster morphology with a small change of parameters . Importantly , the rapid growth does not require high microtubule density inside asters , which can be tuned independently . Collective growth produces a meshwork of short microtubules with potentially desirable properties . First , the network is robust to microtubule severing or the spontaneous detachment from the centrosome . Second , the network can span arbitrarily large distances yet disassemble rapidly upon mitotic entry . Third , the structure , and therefore the mechanical properties , of the network do not depend on the distance from the centrosome . As a speculation , the physical interconnection of the microtubules may facilitate the transduction of mechanical forces across the cell in a way unattainable in the radial array predicted by the standard model ( Tanimoto et al . , 2016; Wühr et al . , 2010 ) . The regime of collective growth parallels the assembly of other large cellular structures from short , interacting filaments ( Pollard and Borisy , 2003 ) and is particularly reminiscent of how meiosis-II spindles self-assemble ( Burbank et al . , 2007; Brugués et al . , 2012; Brugués and Needleman , 2014 ) . Due to such dynamic architecture , spindles are known to have unique physical properties such as self-repair , fusion ( Gatlin et al . , 2009 ) and scaling ( Good et al . , 2013; Hazel et al . , 2013; Wühr et al . , 2008 ) , which could allow for greater robustness and evolvability ( Kirschner and Gerhart , 1998 ) . Perhaps , collective growth is one of the most reliable ways for a cell to assemble cytoskeletal structures that exceed the typical length scales of individual filaments .
We implemented a finite difference method with fixed time steps to numerically solve the continuum model ( Equation 3 ) . The forward Euler’s discretization scheme was used except exact solutions of advection equations was used to account for the gradient terms . Specifically , the plus end positions were simply shifted by +vgδt for growing microtubules and by -vsδt for shrinking microtubules . Nucleation added new growing microtubules of zero length at a position-dependent rate given by Q ( x ) . The algorithm was implemented using MATLAB ( Mathworks ) . We linearized Equation 3 for small Cg and solved it using Laplace transforms in both space and time . The inverse Laplace transform was evaluated using the saddle point method ( Bender and Orszag , 1999 ) . We found the aster velocity as in Equation 4 . The details of this calculation are summarized in the Supporting Text ( SI ) . Interphase microtubule asters were reconstituted in Xenopus egg extract as described previously with use of p150-CC1 to inhibit dynein mediated microtubule sliding ( Field et al . , 2014; Ishihara et al . , 2014a ) . Fluorescence microscopy was performed on a Nikon 90i upright microscope equipped with a Prior Proscan II motorized stage . EB1-mApple was imaged every 2 min with a 10x Plan Apo 0 . 45 N . A . or a 20x Plan Apo 0 . 75 N . A . objective . For the analysis of the aster growth front , a linear region originating from the center of asters was chosen ( Figure 4—figure supplement 1 ) . A low pass filter was applied to the fluorescence intensity profile and the half-max position , corresponding to the aster edge , was determined manually . The analysis was assisted by scripts written in ImageJ and MATLAB ( Mathworks ) . Univariate scatter plots were generated with a template from ( Weissgerber et al . , 2015 ) . EB1-mApple were purified as in ( Petry et al . , 2011 ) , used at a final concentration of 100 nM . In some experiments , MCAK or MCAK-Q710-GFP ( Moore and Wordeman , 2004 ) proteins were added to the reactions . Protein A Dynabeads coated with anti-Aurora kinase A antibody ( Tsai and Zheng , 2005 ) or Tetrahymena pellicles were used as microtubule nucleating sites . Interphase asters were assembled as described above . Catastrophe rates and plus end polymerization rates were estimated from time lapse images of EB1 comets that localize to growing plus ends ( Matov et al . , 2010 ) . The distributions of EB1 track durations were fitted to an exponential function to estimate the catastrophe rate . Spinning disc confocal microscopy was performed on a Nikon Ti motorized inverted microscope equipped with Perfect Focus , a Prior Proscan II motorized stage , Yokagawa CSU-X1 spinning disk confocal with Spectral Applied Research Aurora Borealis modification , Spectral Applied Research LMM-5 laser merge module with AOTF controlled solid state lasers: 488 nm ( 100 mW ) , 561 nm ( 100 mW ) , and Hamamatsu ORCA-AG cooled CCD camera . EB1-mApple was imaged every 2 s with a 60x Plan Apo 1 . 40 N . A . objective with 2×2 binning . EB1 tracks were analyzed with PlusTipTracker ( Applegate et al . , 2011 ) . A 2 min video abstract of this paper is available at https://youtu . be/jfjA2S-fE9U . | Cells must carefully organize their contents in order to work effectively . Protein filaments called microtubules often play important roles in this organization , as well as giving structure to the cell . Many cells contain structures called asters that are formed of microtubules that radiate out from a central point ( much like a star shape ) . Textbooks generally state that all microtubules in the aster grow outward from its center . If this was the case , the microtubules at the edge of large asters – such as those found in frog egg cells and other extremely large cells – would be spread relatively far apart from each other . However , even at the edges of large asters , the microtubules are quite densely packed . In 2014 , a group of researchers proposed that new microtubules could form throughout the aster instead of all originating from the center . This model had not been tested; it was also unclear under what conditions an aster would be able to grow to fill a large cell . Ishihara et al . – including some of the researchers involved in the 2014 work – have now developed a mathematical theory of aster growth that is based on the assumption that microtubules stimulate the generation of new microtubules . The theory reproduces the key features seen during the growth of asters in large cells , and predicts that the asters may stay at a constant size or grow continuously . The condition required for the aster to grow is simple: each microtubule in it has to trigger the generation of at least one new microtubule during its lifetime . Ishihara et al . have named this process “collective growth” . Experiments performed using microtubules taken from crushed frog eggs and assembled under a cover slip provided further evidence that asters grow via a collective growth process . Future studies could now investigate whether collective growth also underlies the formation of other cellular structures . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"physics",
"of",
"living",
"systems"
] | 2016 | Physical basis of large microtubule aster growth |
The haptoglobin-haemoglobin receptor of the African trypanosome species , Trypanosoma brucei , is expressed when the parasite is in the bloodstream of the mammalian host , allowing it to acquire haem through the uptake of haptoglobin-haemoglobin complexes . Here we show that in Trypanosoma congolense this receptor is instead expressed in the epimastigote developmental stage that occurs in the tsetse fly , where it acts as a haemoglobin receptor . We also present the structure of the T . congolense receptor in complex with haemoglobin . This allows us to propose an evolutionary history for this receptor , charting the structural and cellular changes that took place as it adapted from a role in the insect to a new role in the mammalian host .
Infection of livestock by African trypanosomes has a significant effect on food production in sub-Saharan Africa ( Shaw et al . , 2004 ) . In contrast to human disease , which is caused by a restricted set of subspecies of Trypanosoma brucei ( Laveran , 1902; Pays and Vanhollebeke , 2009 ) , livestock disease is caused by at least six distinct species of African trypanosome , the most prevalent being T . congolense and T . vivax ( Rotureau and Van Den Abbeele , 2013 ) . While all share some common features , including antigenic variation and transmission by the tsetse fly , one of the most obvious differences between species is variation in the developmental cycle in the fly ( Hoare , 1972 ) and in particular the location of the epimastigote developmental stage . T . brucei epimastigotes attach to the epithelium in the salivary glands away from the digestive tract , whereas T . congolense and T . vivax attach in the proboscis within the digestive tract ( Hoare , 1972; Peacock et al . , 2012; Jefferies et al . , 1987 ) . The basis for these different tissue tropisms is not known . The cell surface of an African trypanosome acts as the molecular interface with its host , and the developmental transitions of the life cycle involve radical changes in cell surface composition , presumably as adaptations to different host niches . The best understood stage is the mammalian bloodstream form where the cell surface is covered with a dense layer of variant surface glycoprotein ( VSG ) , which acts to protect the plasma membrane , enhancing survival of individual cells and allowing antigenic variation to ensure population survival ( Schwede and Carrington , 2010; Horn , 2014 ) . The density of packing of the VSG molecules on the surface of the bloodstream form trypanosome is thought to approach the maximum possible ( Grünfelder et al . , 2002 ) . In contrast , the developmental stages found inside insects , including the procyclic and epimastigote forms , have less densely packed surface coats , and contain different sets of surface proteins , including GARP in T . congolense and procyclins and BARP in T . brucei ( Bayne et al . , 1993; Roditi et al . , 1989; Urwyler et al . , 2007; Beecroft et al . , 1993 ) . Other cell surface proteins , including receptors and transporters , must operate in the context of these different cell surface architectures ( Lane-Serff et al . , 2014; Stødkilde et al . , 2014 ) . The haptoglobin-haemoglobin receptor of Trypanosoma brucei ( TbHpHbR ) is the best characterised trypanosome receptor . It is expressed in the mammalian bloodstream form and is used for the uptake of haptoglobin-haemoglobin complexes ( HpHb ) for haem acquisition ( Vanhollebeke et al . , 2008 ) . In humans and some other primates , TbHpHbR also plays a role in innate immunity . Human serum contains two complexes , trypanolytic factors-1 and -2 ( TLF1 and TLF2 ) , which cause trypanosome lysis ( Rifkin , 1978; Hajduk et al . , 1989; Tomlinson et al . , 1995; Raper et al . , 1996 ) . TLF1 and TLF2 both contain the apolipoprotein L1 toxin ( Vanhamme et al . , 2003 ) and a complex of haemoglobin bound to haptoglobin-related protein ( HprHb ) ( Raper et al . , 1996 ) . It is the binding of HprHb to TbHpHbR that provides the uptake route for TLF1 into the trypanosome ( Vanhollebeke et al . , 2008 ) . High-resolution structures of TbHpHbR , both alone and in complex with HpHb , have shown how it can function within the densely packed VSG layer ( Lane-Serff et al . , 2014; Stødkilde et al . , 2014 ) . The N-terminal domain of TbHpHbR is formed from an extended three α-helical bundle with a small , membrane-distal head , and is attached to the plasma membrane by a glycophosphatidylinosotol-anchor at the C-terminus of a small C-terminal domain . HpHb binds along the membrane-distal half of the helical bundle . A striking feature of this helical bundle is a ~50° kink , which lies between the HpHb binding site and the membrane attachment point . This kink is likely to result in separation of the VSG molecules on either side of the receptor , holding them apart and presenting the HpHb binding site to the extracellular environment . It also allows two receptors to contact a single HpHb dimer , allowing greater avidity for the ligand and more efficient uptake into trypanosomes ( Lane-Serff et al . , 2014 ) . Specificity for HpHb results from direct , simultaneous contact between the receptor and both haptoglobin and the β-subunit of haemoglobin . Neither isolated haemoglobin nor haptoglobin binds significantly ( Vanhollebeke et al . , 2008 ) . In addition , the propionate chains of haem directly contact the receptor and contribute to binding , with a significant reduction in binding affinity for HpHb that lacks haem ( Stødkilde et al . , 2014 ) . Therefore TbHpHbR has evolved specific adaptations to function in the context of the VSG layer and to selectively bind to haem-loaded HpHb complexes . The T . congolense receptor ( TcHpHbR ) was identified by sequence homology to the receptor from T . brucei and also binds to HpHb with low micromolar affinity ( a KD of 8 μM for TcHpHbR compared with 1 μM for TbHpHb ) ( Higgins et al . , 2013 ) . Mutagenesis studies showed that TcHpHbR and TbHpHbR use overlapping binding sites to interact with HpHb ( Higgins et al . , 2013 ) . The structure of TcHpHbR has a similar architecture to TbHpHbR , with a long three α-helical bundle and small membrane-distal head . However , one striking difference is that the helical bundle of TcHpHbR lacks the kink found in TbHpHbR . With the kink suggested to play a critical role in the operation of TbHpHbR in the context of the VSG layer , it was surprising that TcHpHbR lacks this evolutionary adaption if it too operates in a similar , densely packed environment . For this reason , we set out to investigate the structure and function of the T . congolense HpHbR , to understand its ligand binding specificity , its site of action and the evolutionary adaptations undergone by haptoglobin-haemoglobin receptors from different species .
To gain an understanding of ligand binding by TcHpHbR , we aimed to determine its structure in complex with HpHb . Human haemoglobin in complex with the SP domain of human haptoglobin ( HpSP ) was mixed with a small molar excess of TcHpHbR prior to gel filtration chromatography , in order to purify receptor-ligand complex . However , analysis of the resulting fractions from the chromatography column revealed a higher molecular weight peak containing TcHpHbR and Hb , but no HpSP . A lower molecular weight peak , at the elution volume expected for unliganded TcHpHbR , contained excess free TcHpHbR and HpSP ( Figure 1A ) . Therefore the presence of TcHpHbR led to disassembly of the HpSPHb complex and the formation of a complex containing TcHpHbR and Hb . 10 . 7554/eLife . 13044 . 003Figure 1 . T . congolense and T . vivax HpHbRs are haemoglobin receptors . ( A ) Analytical gel filtration chromatography analysis showing the consequence of mixing HpSPHb and TcHpHbR . A mixture of TcHpHbR and HpSPHb was loaded onto the column , resulting in two peaks . The gel shows that peak 1 contains a complex of TcHpHbR bound to Hb , while peak 2 contains free TcHpHbR and free HpSP . ( B ) Surface plasmon resonance analysis of the binding of HpHbRs from T . b . brucei , T . congolense and T . vivax to either Hb or HpHb . ( C ) Isothermal titration calorimetry shows the binding of two TcHpHbR to one Hb tetramer . ( D ) A phylogenetic tree indicating the evolutionary history of the trypanosome strains under study ( Stevens et al . , 2001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 00310 . 7554/eLife . 13044 . 004Figure 1—figure supplement 1 . TcHpHbR does not disassemble HpHb . ( A ) Analytical gel filtration chromatography analysis showing the consequence of mixing HpHb and TcHpHbR . A mixture of TcHpHbR and HpHb was loaded onto the column , resulting in two peaks . The gel shows that peak 1 contains HpHb while peak 2 contains free TcHpHbR . This shows that TcHpHbR did not disassemble HpHb and that the TcHpHbR:HpHb affinity is not sufficient for the complex to remain stable during the course of a gel filtration experiment . ( B ) Analytical gel filtration chromatography analysis showing the consequence of mixing Hb and TcHpHbR . A mixture of TcHpHbR and Hb was loaded onto the column , resulting in two peaks . The gel shows that peak 1 contains the TcHpHbR:Hb complex while peak 2 contains free TcHpHbR , showing the formation of a stable complex . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 00410 . 7554/eLife . 13044 . 005Figure 1—figure supplement 2 . TcHpHbR binding to bovine Hb . Surface plasmon resonance analysis of the binding of T . congolense HpHbR to bovine Hb . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 005 While native haptoglobin is cleaved to form α- and β-subunits , the recombinant HpSP used in this study is not cleaved . We therefore repeated the gel filtration experiment using native Hp in an HpHb complex . In this case , we found that the HpHb complex is not disassembled by incubation with the receptor , suggesting that the receptor does not disassemble native HpHb complexes during its physiological function ( Figure 1—figure supplement 1A ) . However , the complex that it forms with the receptor is not sufficiently strong to remain intact during gel filtration chromatography and HpHb and the receptor elute in separate peaks , confirming that the receptor has a low affinity for HpHb . The discovery that TcHpHbR binds to Hb was unexpected as TbHpHbR had previously been shown to bind HpHb but not to free Hb ( Vanhollebeke et al . , 2008 ) . Surface plasmon resonance ( SPR ) was therefore used to investigate the ligand specificity of HpHbRs from T . brucei , T . congolense and T . vivax . The receptors were immobilised on the chip surface through a C-terminal biotin moiety , causing them to be presented in the same orientation as on the cell surface . Binding was then measured for both Hb and HpHb . As expected , HpHbR from T . brucei interacted with HpHb , but not with Hb alone ( Figure 1B ) . However , HpHbRs from both T . congolense and T . vivax interacted with either HpHb or free Hb . Indeed , both receptors formed a far more stable complex with Hb than with HpHb , as shown by the lower off-rate ( Figure 1B ) . In addition when Hb was mixed with TcHpHbR and subjected to gel filtration chromatography , the primary peak was a complex of TcHpHbR bound to Hb , indicating that TcHpHbR forms a more stable complex with Hb than with HpHb ( Figure 1—figure supplement 1B ) . T . congolense is not a human infective pathogen , but is found in numerous livestock species . We therefore also tested the binding of TcHpHbR to bovine haemoglobin and observed a strong interaction with a slow off rate ( Figure 1—figure supplement 2 ) . Therefore TcHpHbR has a high affinity for Hb , while TbHpHbR does not bind to Hb alone . A similar change in specificity is seen in the mammalian scavenger receptor CD163 , as mouse CD163 binds to Hb while human CD163 binds to HpHb alone ( Etzerodt et al . , 2013 ) . As haemoglobin is a symmetrical tetramer of two α and two β subunits , each tetramer could potentially bind to two receptors and the SPR measurements described above would result from a mixture of monovalent and bivalent binding . Isothermal titration calorimetry ( ITC ) was therefore used to measure the monovalent KD and the stoichiometry of the interaction between TcHpHbR and Hb ( Figure 1C ) . This revealed that two receptors interact with one Hb tetramer . This follows the same pattern as the TbHpHbR:HpHb complex , where two receptors bind to each dimeric HpHb complex ( Lane-Serff et al . , 2014; Stødkilde et al . , 2014 ) . The KD for the TcHpHbR:Hb interaction was estimated to be 3 nM by ITC . This is around 1000-fold tighter than the affinity of the same receptor for HpHb , previously measured by ITC to be 3 μM ( Higgins et al . , 2013 ) . T . congolense and T . brucei diverged from each other after their last common ancestor had diverged from T . vivax ( Figure 1D ) ( Stevens et al . , 1999; Hamilton et al . , 2004; Kelly et al . , 2014 ) . As HpHbRs from both T . congolense and T . vivax bind to Hb , it is most likely that this property was lost from T . brucei after it diverged from T . congolense rather than separately gained in both T . congolense and T . vivax . The ~1000-fold higher affinity of TcHpHbR for Hb than for HpHb also suggests that Hb binding is the major evolved function of this receptor in T . congolense and T . vivax . Together these observations suggest that the ancestor of the HpHbRs was primarily a haemoglobin receptor and that evolutionary changes that have taken place during the evolution of T . brucei have led to an alteration in its binding specificity . The finding that TcHpHbR is a haemoglobin receptor was unexpected in the light of our current knowledge of TbHpHbR . In T . brucei , the receptor is expressed in the mammalian bloodstream stage of the parasite , where it functions in the acquisition of haem as a nutrient ( Vanhollebeke et al . , 2008 ) . In mammalian blood , free haemoglobin levels are extremely low due to the presence of haptoglobin as a scavenger . Haptoglobin binds free Hb , allowing it to be removed from the serum by endocytosis of HpHb complexes into macrophages , reducing the potential for oxidative damage caused by haem ( Kristiansen et al . , 2001 ) . Therefore , except under conditions of exceptional haemolysis , there is little haemoglobin present in the blood , bringing into question the requirement for a haemoglobin receptor . We therefore assessed the life cycle stage of T . congolense in which the receptor is expressed . While T . congolense will not be exposed to free haemoglobin in the mammalian bloodstream , developmental forms in the tsetse fly will be exposed to haemoglobin derived from the bloodmeal . A previous proteomic analysis comparing different developmental stages of T . congolense did not detect TcHpHbR protein in bloodstream forms but did detect it as an abundant protein in epimastigotes ( Eyford et al . , 2011 ) . In addition , a transcriptome analysis of various developmental forms of T . vivax indicated that TvHpHbR mRNA was most abundant in epimastigotes ( Jackson et al . , 2015 ) . To investigate further , T . congolense epimastigotes were generated from procyclic form cultures ( Coustou et al . , 2010 ) . Western blot analysis of three independently generated epimastigote-containing cultures showed expression of the TcHpHbR in these populations , while expression was below detectable levels in the original procyclic cells or in bloodstream form T . congolense ( Figure 2; Figure 2—figure supplement 1 ) . Therefore differentiation to the epimastigote form was associated with expression of TcHpHbR . 10 . 7554/eLife . 13044 . 006Figure 2 . Western analysis reveals high levels of HpHbR expression in epimastigote-enriched cultures . ( A ) Three populations of T . congolense epimastigotes were generated in vitro by maintaining procyclic form cultures of T . congolense IL3000 cells at stationary phase ( Coustou et al . , 2010 ) . The majority of epimastigote forms are adherent and stick to the culture flask while the culture supernatant is predominantly trypomastigotes with a low percentage of detached epimastigote forms . Cell lysates where generated from culture supernatants and subject to western blot analysis . No TcHpHbR protein expression was detected in procyclic form cultures ( PCF ) whereas TcHpHbR expression was observed in all epimastigote-containing cultures . The protein is observed above the expected 32 kDa , probably due to the GPI-anchor and N-glycosylation affecting mobility as has been observed for the TbHpHbR ( Vanhollebeke et al . , 2008 ) . Loading control is anti-TbPRF2 . ( B ) No TcHpHbR protein expression was detected in T . congolense bloodstream forms ( BSF ) or procyclic form cultures ( PCF ) by western blot , whereas expression was detected in epimastigote-containing cultures ( Epi ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 00610 . 7554/eLife . 13044 . 007Figure 2—figure supplement 1 . Validation of TcHpHbR antisera and quantification of TcHpHbR protein expression in T . congolense epimastigotes . ( A ) Antisera was raised and purified against recombinant TcHpHbR . The antisera recognized two bands in T . congolense epimastigotes that run at a higher molecular weight on an acrylamide gel , as expected of a GPI-anchored and N-glycosylated protein and as observed for the TbHpHbR in T . brucei bloodstream forms , BSFs ( Vanhollebeke et al . , 2008 ) , with the largest band representing the fully processed form of the protein . To confirm this was indeed the TcHpHbR , T . brucei TbHpHbR KO cells were transfected with an inducible copy of the TcHpHbR in an over-expression vector . Upon induction with doxycycline ( Dox ) , the antisera detected two bands that again ran at a higher molecular weight than the recombinant TcHpHbR . While the upper band represents the minority of protein in this case , this is probably due to differential modification of the GPI-anchor in bloodstream and insect cell forms ( Ferguson et al . , 1993 ) . ( B ) Known quantities of recombinant TcHpHbR protein were loaded next to known numbers of TcHpHbR expressing cell equivalents ( calculated by IFA analysis on mixed epimastigote and procyclic populations ) . By two independent western blots 8 . 35 x 104 cell equivalents was estimated to be equivalent to between 2 . 25 and 4 . 5 ng of recombinant protein ( an average of between 5 and 9 x 105 molcules per cell ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 007 Immunofluorescence analysis ( IFA ) was carried out next to determine the sub-cellular localisation of TcHpHbR using epimastigote forms generated in vitro from procyclic form cultures . The key distinguishing feature resulting from differentiation to epimastigotes is a change in the relative positions of the nucleus and kinetoplast along the anterior-posterior axis of the cell . In trypomastigote forms , such as procyclics , the kinetoplast is positioned posterior to the nucleus while in epimastigote forms it is juxtanuclear or anterior to the nucleus ( Peacock et al . , 2012; Vickerman , 1984 ) . In all cells in which kinetoplast repositioning to that found in the epimastigote form had occurred , TcHpHbR protein was detected and localised across the whole cell surface , including the flagellum , in both permeabilised and non-permeabilised cells ( Figure 3A and B ) . 10 . 7554/eLife . 13044 . 008Figure 3 . Immunofluorescent analysis of T . congolense epimastigotes reveals TcHpHbR protein is expressed at high levels across the entire cell surface . ( A ) Immunofluorescence analysis of paraformaldehyde-fixed ( non-permeabilised ) in vitro generated T . congolense WG81 epimastigotes with rabbit anti-TcHpHbR antisera and an Alexa488 conjugated anti-rabbit secondary antibody . TcHpHbR was readily detected on the surfaces of all cells where kinetoplast repositioning had occurred . ( Arrow highlights a trypomastigote , arrowhead highlights an epimastigote ) ( B ) This was also observed with immunofluorescence analysis of methanol-fixed ( permeabilised ) in vitro generated T . congolense Il3000 epimastigotes . Occasionally TcHpHbR expression was detected in cells that did not display kinetoplast repositioning . ( Arrow highlights a TcHpHbR positive cell without associated kinetoplast repositioning . ) ( C ) T . congolense Gam2 cells were harvested from the midgut ( top panel ) , proventriculus ( second panel ) and proboscis ( lower three panels ) of tsetse flies . Cells were fixed with methanol and immunofluorescence analysis was carried out as above . Trypanosomes harvested from the midgut ( top panel ) were always negative for TcHpHbR and those harvested from the proventiculus ( second panel ) were mostly negative for TcHpHbR , although occasional cells were identified with a faint positive signal . Epimastigotes harvested from the proboscis ( lower three panels , arrowheads ) were always strongly positive for TcHpHbR . Trypomastigotes from the proboscis ( lower three panels , arrows ) showed a faint or negative signal for TcHpHbR ( arrows ) . All scale bars represents 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 00810 . 7554/eLife . 13044 . 009Figure 3—figure supplement 1 . TcHpHbR is expressed on a cell undergoing asymmetric division . A TcHpHbR-expressing cell was identified that was undergoing asymmetric division , where one daughter cell will be an epimastigote and the other will be a trypomastigote , as occurs during metacyclogenesis in T . brucei . Scale bar represents 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 009 A minority of cells that expressed TcHpHbR did not have the characteristic kinetoplast and nuclear positioning characteristic of epimastigotes ( Figure 3B , arrow ) . This observation may indicate that: ( i ) TcHpHbR expression in epimastigotes occurs immediately prior to kinetoplast repositioning , and/or ( ii ) epimastigote formation in vitro is incomplete and/or ( iii ) the epimastigotes were differentiating further into metacyclic forms and that these , at least initially , retain TcHpHbR expression . The last possibility is supported by the identification of a TcHpHbR-expressing cell undergoing asymmetric division , where one daughter cell will be an epimastigote and the other will not , as occurs during metacyclogenesis in T . brucei ( Rotureau and Van Den Abbeele , 2013 ) ( Figure 3—figure supplement 1 ) and by the proteomic detection of low levels of TcHpHbR in metacyclic populations ( Eyford et al . , 2011 ) . To confirm that the TcHpHbR is indeed an epimastigote-stage protein in vivo , Glossina pallidipes tsetse flies were infected with T . congolense Gam2 . Trypanosomes were harvested from the midgut , proventriculus and proboscis of infected flies 40 days post infection and expression of TcHpHbR was investigated by immunofluorescence . Trypomastigote forms harvested from the midgut ( procyclics ) were all negative for TcHpHbR staining ( Figure 3C , top panel ) . Trypomastigotes from the proventriculus were also typically negative for TcHpHbR staining ( Figure 3C , second panel ) , although some cells were identified with faint positive signal ( data not shown ) . Trypanosomes collected from the tsetse proboscis included both trypomastigotes ( proventricular trypomastigotes , pre-metacyclics or metacyclic forms ) and epimastigotes ( Figure 3C , lower three panels , arrows highlight trypomastigotes and arrowheads highlight epimastigotes ) . All epimastigotes identified had high levels of TcHpHbR expression , whereas trypomastigotes were either negative or weakly positive ( Figure 3C , lower three panels ) . Therefore , TcHpHbR is highly expressed in the T . congolense epimastigote life-stage in vivo , with some upregulation of expression occurring prior to kinetoplast repositioning . TcHpHbR expression was readily detected over the entire cell surface of epimastigotes , suggesting expression levels were higher than those of receptors previously characterized in bloodstream forms of T . brucei . The average copy number of TcHpHbR was therefore estimated using western blots to compare cell lysates from in vitro generated epimastigotes with known quantities of recombinant protein , with an adjustment for the percentage of cells in the culture expressing the protein as determined by immunofluorescence . This suggested an average of ~5–9 x 105 TcHpHbR molecules to be present per TcHpHbR-expressing cell ( Figure 2—figure supplement 1 ) . For comparison , T . brucei bloodstream forms express approximately 200 to 400 TbHpHbR molecules per cell ( Vanhollebeke et al . , 2008; Drain et al . , 2001 ) . Therefore the T . congolense HpHbR is an abundant protein expressed in epimastigotes , with around a 1000-fold more receptors per cell than are found in the T . brucei bloodstream form . To determine if TcHpHbR functioned in receptor-mediated endocytosis of Hb , ligand uptake was monitored in a live cell assay using culture-derived epimastigotes of T . congolense WG81 . These cultures contained both trypomastigotes and epimastigotes . The trypanosomes were incubated with either 10 nM Alexa488-labelled Hb or 10 nM Alexa488-labelled BSA . Internalisation of Hb , but not BSA , was observed specifically in epimastigote forms but not in trypomastigote forms ( Figure 4 ) . Therefore , the T . congolense developmental form that highly expresses the TcHpHbR on its surface is indeed able to internalise Hb at low nanomolar concentrations . 10 . 7554/eLife . 13044 . 010Figure 4 . T . congolense epimastigotes internalise 488-labelled Hb . Assay for uptake of Alexa488-labelled Hb ( Hb-488 ) and Alexa488-labelled BSA ( BSA-488 ) into T . congolense WG81 epimastigotes was monitored by microscopy . Uptake of Hb-488 was detected at 10 nM in epimastigotes ( lower panel , arrowheads ) but not in trypomastigotes ( lower panel , arrow ) . No fluid phase uptake of BSA-488 at 10 nM in any cells ( centre panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 010 To investigate the molecular basis for haemoglobin binding by TcHpHbR , we mixed the receptor with HpSPHb . As described above ( Figure 1A ) , this generated a complex of TcHpHbR bound to Hb . This complex crystallised with a well solution of 8% PEG 8000 , 0 . 1 M sodium citrate pH 5 . 0 in 18 hr , and a complete data set was collected to 3 . 2 Å resolution . A molecular replacement solution was determined using T . congolense HpHbR ( pdb: 4E40 ) and human Hb ( pdb: 1HHO ) as search models . This allowed placement of two receptor molecules and a single haemoglobin tetramer in the asymmetric unit of the crystal ( Figure 5 , Table 1 ) . Each receptor makes the same interactions with haemoglobin , and the receptor conformation is unaltered from that of unliganded receptor ( Higgins et al . , 2013 ) , with a root-mean-square deviation of 0 . 6 Å . This lack of conformational change on ligand binding matches that seen in the structures of T . brucei HpHbR alone and in the presence of HpHb ( Lane-Serff et al . , 2014 ) . The haemoglobin is in the oxygenated conformation with a root-mean-square deviation of 0 . 7 Å from the search model . 10 . 7554/eLife . 13044 . 011Figure 5 . The structural basis of haemoglobin binding by TcHpHbR . ( A ) The structure of a complex of TcHpHbR bound to Hb . ( B ) The structure of the TbHpHbR:HpSPHb complex ( Lane-Serff et al . , 2014 ) . ( C ) The contents of the asymmetric unit of the TcHpHbR:Hb crystals , showing two receptors binding to a single haemoglobin tetramer . ( D ) A close up view of the interaction of TcHpHbR with the β-chain of haemoglobin showing the direct contacts made with the haem group . ( E ) A close up view of the interaction of TcHpHbR with the α-chain of haemoglobin showing the direct contacts made with the haem group . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 01110 . 7554/eLife . 13044 . 012Figure 5—figure supplement 1 . TcHpHbR-binding residues are conserved across mammalian haemoglobins . Residues in green interact with TcHpHbR in the TcHpHbR:Hb structure . Residues in blue interact with TbHpHbR in the TbHpHbR:HpSPHb structure . Residues in orange interact with both TcHpHbR and TbHpHbR . These residues are conserved in human haemoglobin and also in haemoglobins from livestock species that can be infected by T . congolense . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 01210 . 7554/eLife . 13044 . 013Table 1 . Crystallographic statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 013BeamlineDiamond I03Space GroupP22121Cell parameters ( Å ) a=72 . 75 , b=127 . 3 , c=172 . 42Resolution ( Å ) 101 . 89-3 . 2Wavelength ( Å ) 0 . 976RPIM ( % ) 7 . 3 ( 41 . 0 ) I/ σ ( I ) 7 . 3 ( 2 . 4 ) Completeness ( % ) 99 . 0 ( 99 . 1 ) Multiplicity2 . 7 ( 3 . 0 ) Resolution ( Å ) 3 . 2No . reflections25191Rwork / Rfree ( % ) 20 . 44 / 23 . 51No . of protein residues in model1063rmsd bond lengths ( Å ) 0 . 010rmsd bond angles ( ° ) 1 . 21Ramachandran plotPreferred region93 . 1%Allowed region6 . 9%Outliers0% The interaction surface can be divided into two subsites , with interactions made with the α-subunit of one haemoglobin dimer and the β-subunit of the second haemoglobin dimer ( Figure 5A ) . This suggests that the receptor either binds selectively to haemoglobin tetramers or that two haemoglobin dimers will assemble together with the receptor . All receptor-ligand interactions are mediated by features that are absolutely conserved between haemoglobin from human and livestock species ( Table 2 , Figure 5—figure supplement 1 ) . It is therefore highly likely that the haemoglobin molecules of most of the mammals bitten by tsetse flies will bind to the receptor . The membrane-distal binding site interacts with the α-subunit of haemoglobin ( Figure 5E ) . This contains more than half of the total interaction surface area ( ~735 Å2 of 1385 Å2 ) . It is predominantly mediated by hydrogen bonds and electrostatic contacts , with a significant component from a direct interaction between the receptor and the propionate chains of haem . This binding site is similar in location , size and chemical nature to the membrane-distal binding site of TbHpHbR for HpHb , in which the same region of TbHpHbR makes direct contacts with the β-subunit of haemoglobin ( Figure 5B ) . 10 . 7554/eLife . 13044 . 014Table 2 . Description of interactions between TcHpHbR and Hb . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 014TcHpHbRHbResidueGroupChainResidueGroupInteractionHbαS29backbone COC/EH46side chainHydrogen bondS29side chainC/EHaemO1Hydrogen bondI30side chainC/EPatchHydrophobicR37side chain NH1/NH2C/EL92backbone COHydrogen bondK127side chainC/EP45backbone COHydrogen bondK130side chainC/EPatchHydrophobicT166side chainC/EHaemO3Hydrogen bondY168side chainC/EPatchHydrophobicY168backbone COC/EK91side chainHydrogen bondD169side chainC/EK91side chainSalt bridgeHbβI41side chainD/FPatchHydrophobicR42side chain NH2D/FHaemO1Hydrogen bondA44side chainD/FPatchHydrophobicT45side chainD/FPatchHydrophobicE47side chain OE2D/FK96side chainSalt bridgeF48side chainD/FPatchHydrophobicK52side chainD/FHaemO3Hydrogen bond The smaller , membrane-proximal part of the binding site contacts the β-subunit of haemoglobin with a total contact surface area of 650 Å2 ( Figure 5D ) . This interface is more hydrophobic in nature than the membrane-distal site . However , once again , the propionate chains of haem directly interact with the receptor . This membrane distal binding site is distinct in position and chemical nature from the region of TbHpHbR that contacts haptoglobin , which is closer to the membrane surface and is smaller ( ~505 Å2 ) . The more extensive contact site between TcHpHbR and the β-subunit of Hb , than that between TbHpHbR and Hp , is likely to be a major contributor to the increased affinity of TcHpHbR for Hb . TcHpHbR also interacts with HpHb , albeit with a significantly lower affinity . To understand the molecular basis for this interaction , we generated a model of the TcHpHbR:HpHb complex ( Figure 6A ) . One surprise from a comparison of the TcHpHbR:Hb structure with that of TbHpHbR:HpHb is that the membrane distal binding site of TcHpHbR interacts with the Hb α-subunit while the equivalent site of TbHpHbR interacts with the Hb β-subunit of HpHb . We modeled the TcHpHbR:HpHb structure by docking either Hb subunit from the structure of HpHb onto the interacting Hb α-subunit of the TcHpHbR:Hb complex . When the Hb α-subunit is docked onto the membrane-distal binding site , Hp makes no contacts with the receptor . In contrast , when the HpHb is ‘flipped’ so that the Hb β-subunit contacts the membrane-distal binding site , this positions Hp adjacent to the receptor ( Figure 6A ) . In addition , the residues from the Hb α-subunit that contact the membrane-distal binding site are replaced by chemically similar residues in the Hb β-subunit , allowing equivalent interactions to occur . The interaction between TbHpHbR and Hp is predominantly hydrophobic in nature , and a similar , but smaller , hydrophobic region in TcHpHbR , involving phenylalanine 48 is similarly positioned to interact with Hp . We therefore predict that HpHb interacts with TcHpHbR with a similar binding mode to that observed in the TbHpHbR:HpHb complex with the Hb β-subunit contacting the membrane-distal binding site , allowing haptoglobin to bind to a membrane-proximal site . 10 . 7554/eLife . 13044 . 015Figure 6 . Understanding HpHbR ligand specificity . ( A ) A model of TcHpHbR bound to HpSPHb , based on the TcHpHbR:Hb structure . ( B , C ) The TcHpHbR:Hb and TbHpHbR:HpSPHb complexes have been aligned , with the haemoglobin subunit that interacts with the membrane distal binding site used for the alignment . This shows that a change in the path of the helical bundle of TbHpHbR ( blue ) prevents the interaction that occurs between TcHpHbR ( green ) and the membrane proximal haemoglobin subunit . This disruption of the membrane proximal binding site has caused TbHpHbR to lose affinity for HbDOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 015 Finally , a comparison of the TcHpHbR:Hb structure with that of TbHpHbR:HpHb allows us to rationalise the changes have taken place in TbHpHbR to cause the loss of Hb binding . While an alignment of these structures shows little alteration in the membrane-distal binding site , the path of helix I of the receptor is altered in TbHpHbR so that it no longer forms the membrane-proximal binding site for the Hb β-subunit , thereby reducing the total interaction surface area by nearly half ( Figure 6B and C ) . This change in the helical pathway could be a consequence of the adoption of a kink in TbHpHbR , which allows the receptor to function in the context of the VSG layer . Alternatively , it could be a result of changes that lead to the increase in the affinity of the receptor for HpHb , from 8 μM in T . congolense to 1 µM in T . brucei , as determined by SPR ( Higgins et al . , 2013 ) .
The haptoglobin-haemoglobin receptor of T . brucei has been extensively characterised because of its role in the uptake of haptoglobin-haemoglobin into the bloodstream form parasite ( Vanhollebeke et al . , 2008; Lane-Serff et al . , 2014 ) . This receptor also mediates uptake of trypanolytic factor-1 ( Vanhollebeke et al . , 2008 ) and to a lesser extent the uptake of trypanolytic factor-2 ( Bullard et al . , 2012 ) , contributing to the inability of most isolates of T . brucei to survive in human serum . It was a reasonable expectation that orthologues in other African trypanosome species , such as T . congolense and T . vivax , would have a similar function . However , here it is shown that there are significant differences between the receptors from T . brucei and T . congolense . Firstly , T . congolense HpHbR has an approximately 1000-fold greater affinity for haemoglobin than for haptoglobin-haemoglobin and developmental forms expressing TcHpHbR are able to internalise Hb at low nanomolar concentrations . Secondly , the T . congolense receptor is expressed in the epimastigotes with a copy number approximately 1000-fold greater than that of T . brucei HpHbR in the bloodstream form . Finally , T . congolense HpHbR is distributed over the whole cell surface , whereas in T . brucei it is concentrated in the flagellar pocket ( Vanhollebeke et al . , 2008 ) . Similar findings are seen for T . vivax , as TvHpHbR also binds haemoglobin preferentially over haptoglobin-haemoglobin and is , at the mRNA level , preferentially expressed in epimastigotes ( Jackson et al . , 2015 ) . This , together with the evolutionary history of the trypanosomes , suggests that the receptors from T . vivax and T . congolense represent the ancestral form , while the T . brucei receptor has adopted a modified function and cellular distribution . The location of T . congolense epimastigotes in the tsetse fly mouthparts provides the parasite with the opportunity to acquire nutrients from bloodmeals obtained by the fly . The presence of TcHpHbR will allow epimastigotes to scavenge haem that is present in haemoglobin molecules released from lysed erythrocytes , or to bind more weakly to haptoglobin-haemoglobin . It is noteworthy that the predominant form of bovine haptoglobin adopts higher order multimers more complex than the dimeric HpHb used in this study ( Lai et al . , 2007 ) and it is possible that this may increase the avidity for TcHpHbR , as observed for multimeric human HpHb binding to scavenger receptor CD163 ( Kristiansen et al . , 2001 ) , making uptake of HpHb more efficient . Endocytosis of TcHpHbR bound to either Hb or HpHb would then provide the parasite with haem . The expression level of TcHpHbR in the epimastigote form is surprising , with over 5 x 105 copies per cell . This is approximately 1000-fold greater than the level of TbHpHbR in the bloodstream form ( Vanhollebeke et al . , 2008; Drain et al . , 2001 ) . One possible explanation for this abundant expression is that the receptor must capture haemoglobin as it periodically and transiently flows over the cell surface when the blood meal passes through the tsetse fly mouthparts . This is in contrast with TbHpHbR expressed in the T . brucei bloodstream stage , which will be constantly exposed to its ligand . It remains unknown how T . congolense bloodstream forms acquire haem . It may be that sufficient haptoglobin-haemoglobin enters the cell through fluid phase endocytosis or perhaps through an alternative receptor , such as the ortholog of the LHR1 haem transpoter utilized by Leishmania amazonensis ( Huynh et al . , 2012 ) . Our knowledge of the structure and function of HpHbR in both T . congolense and T . brucei allows us to propose an evolutionary history for the changes that took place in the development of the T . brucei receptor . First , perhaps to evade toxic components of the blood meal or to avoid niche competition with other trypanosome species in the proboscis , the developmental cycle of T . brucei altered , with the epimastigotes adopting a new location in the salivary glands , rather than developing in the mouthparts . As no haemoglobin is available from bloodmeals in the salivary glands , the receptor became redundant . A new pattern of expression then evolved in which the receptor was expressed in bloodstream forms instead of in epimastigotes . This switch conferred the ability to more efficiently acquire haem more efficiently in the bloodstream form . That this provides an advantage is evidenced by the attenuation of growth of a TbHpHbR null mutant in a mouse model ( Vanhollebeke et al . , 2008 ) . However , free haemoglobin is not normally present in blood , where it rapidly assembles into HpHb complexes ( Wada et al . , 1970; Deiss and Lee , 1999 ) . Evolutionary changes therefore took place in TbHpHbR that resulted in an increase in its affinity for HpHb , from 8 μM in T . congolense to 1 µM in T . brucei ( Higgins et al . , 2013 ) . A decrease in expression levels and a change in primary location from the cell surface to being concentrated in the flagellar pocket ( Vanhollebeke et al . , 2008 ) was most likely a final adaptation , perhaps driven by the need to avoid detection by the mammalian acquired immune system . The change in the developmental stage of expression also had significant effects on the structure of the receptor . On the epimastigote surface , TcHpHbR is free to interact with its ligand relatively unimpeded by other surface proteins . In addition , with a 3 nM affinity for Hb , monovalent binding will allow efficient uptake , and simple tilting of the receptor around its GPI-anchor will position the binding site to allow simultaneous binding of two receptors to one Hb if required ( Figure 7 ) . A switch to expression in the bloodstream form forced the receptor to function within the densely packed VSG layer . In addition , switching to a ligand with approximately 1000-fold weaker binding made bivalent binding important for efficient uptake . To operate in this new context , the T . brucei receptor evolved by gaining both a novel C-terminal domain that probably increases the distance of the ligand binding site from the plasma membrane and by evolving a significant kink between the ligand-binding site and the membrane surface ( Lane-Serff et al . , 2014; Stødkilde et al . , 2014 ) . This kink pushes the VSG molecules apart and presents the ligand-binding site at the surface . It also allows two receptors , both coupled to the membrane surface , to simultaneously bind to a single dimer of haptoglobin-haemoglobin , increasing avidity and uptake efficiency . A consequence of these changes was the loss of haemoglobin binding , which was no longer under positive selection . 10 . 7554/eLife . 13044 . 016Figure 7 . A comparison of ligand binding by HpHbRs from different species . Space filling models of ( A ) Two TcHpHbR molecules binding to a single haemoglobin tetramer . ( B ) Two TbHpHbRs bound to a single haptoglobin-haemoglobin dimer . DOI: http://dx . doi . org/10 . 7554/eLife . 13044 . 016 The evolution of the receptor has continued as some primates have acquired innate immune factors that kill trypanosomes . These trypanolytic factors , TLF1 and TLF2 , exploit TbHpHbR to increase the efficiency of TLF uptake into T . brucei ( Bullard et al . , 2012 ) . For the majority of African trypanosomes this has had a minor effect on parasite survival as non-primates , which lack TLFs , are their predominant hosts ( Hoare , 1972 ) . However , in T . brucei gambiense , the one subspecies that has evolved to infect humans as the main host , the receptor has responded to this new selection pressure through a point mutation that reduces affinity for TLF1 and HpHb ( Higgins et al . , 2013; DeJesus et al . , 2013; Symula et al . , 2012 ) . Therefore , the haptoglobin-haemoglobin receptor of African trypanosomes has undergone a remarkable set of adaptations in its co-evolution with its hosts . It has changed from an epimastigote-expressed haemoglobin receptor into a haptoglobin-haemoglobin receptor , expressed in the bloodstage of T . brucei and has adapted to function efficiently in its new surface environment . With an important role at the host-parasite interface , and as a target of innate immunity , it continues to evolve and adapt , allowing it to provide the parasite with a source of haem , while evading destruction by innate immunity factors .
The gene TvY486 0024220 ( tritrypdb . org ) was identified as the closest homologue to T . congolense HpHbR using Blastp . The putative N-terminal signal sequence cleavage site was identified using SignalP ( Petersen et al . , 2011 ) and the putative C-terminal GPI-anchor addition site was identified by comparison with other trypanosome GPI-anchored proteins . A coding sequence for the mature polypeptide open reading frame plus a Tobacco Etch Virus ( TEV ) protease site at the N-terminus was synthesised with codons optimised for expression in E . coli using the manufacturer’s software ( Geneart , Thermo Fisher Scientific , Waltham MA ) and was cloned into the NdeI and BamHI sites of pET15b . The T . b . brucei HpHbR N-terminal domain and T . congolense HpHbR had been previously cloned for expression into a modified pET-15b to include N-terminal hexahistidine tags and cleavage sites for TEV protease ( Lane-Serff et al . , 2014; Higgins et al . , 2013 ) . To prepare receptors suitable for coupling to a surface plasmon resonance chip , sequences encoding biotin acceptor peptides ( BAP ) were cloned onto the C-termini of TbHpHbR , TcHpHbR and TvHpHbR . All three receptors were expressed in E . coli Origami B cells . These were induced with 1 mM IPTG ( Melford , UK ) and incubated for 3 hr at 30°C for TcHpHbR and TvHpHbR , and overnight at 18°C for TbbHpHbR . The protein was purified by Ni2+-NTA affinity chromatography , followed by gel filtration using a Superdex 75 16/60 column ( GE Healthcare , UK ) in 20 mM HEPES pH 7 . 5 , 150 mM NaCl . Protein used in crystallography experiments was cleaved overnight with His-tagged TEV protease at 4°C in 20 mM sodium phosphate pH 7 . 4 , 150 mM NaCl , 3 mM oxidised glutathione , 0 . 3 mM reduced glutathione to remove the N-terminal His-tag . This was followed by reverse Ni2+-NTA affinity chromatography prior to gel filtration . The SP domain of human haptoglobin had been previously cloned into a modified pAcGP67A vector to generate a polypeptide with an N-terminal hexahistidine tag and a cleavage site for TEV protease . This was expressed in Sf9 insect cells and purified by Ni2+-NTA affinity chromatography and gel filtration as described previously ( Lane-Serff et al . , 2014 ) . Full length , dimeric haptoglobin 1–1 was purchased ( Sigma Aldrich , St Louis , MO ) . To purify haemoglobin , human blood was sonicated , followed by anion exchange chromatography using a Mono Q column ( GE Healthcare ) . HpHb and HpSPHb were assembled and purified as described previously ( Lane-Serff et al . , 2014 ) . The assembly of complexes containing TcHpHbR and HpSPHb was assessed using analytical gel filtration chromatography . 0 . 2 mg of TcHpHbR and 0 . 3 mg of HpSPHb were mixed ( ~4:3 molar ratio ) before loading onto a Superdex 200 10/300 GL column ( GE Healthcare ) . This was run using an ÄKTApurifier ( GE Healthcare ) in 20 mM HEPES pH 7 . 5 , 150 mM NaCl . T . congolense IL3000 bloodstream form cells were grown in TcBSF-1 media at 37°C with 5% CO2 . T . congolense IL3000 procyclic form cells were grown in TcPCF-3 media at 27°C with 5% CO2 ( Coustou et al . , 2010 ) . T . congolense procyclic and epimastigote form cells derived from isolates Gam 2 and WG81 were grown in Cunningham’s medium ( CM ) at 27°C . In order to generate epimastigotes , procyclic form cultures were maintained at stationary phase by replacing half of the culture medium every three to four days ( Coustou et al . , 2010 ) . Differentiation to epimastigotes occurred in these cultures after 1–3 months . Epimastigotes were identified by adherence to the culture flask and repositioning of the kinetoplast from posterior and distal to the nucleus , to a position either proximal or anterior to the nucleus . Attempts to harvest the adherent epimastigotes using a cell scraper resulted in damaged/destroyed cells . Some epimastigotes could be dislodged into the supernatant by washing the flask several times with the culture supernatant . Western blots were therefore carried out on the supernatant of these cultures containing both trypomastigote and epimastigote forms . T . brucei TbHpHbR KO bloodstream form cells ( Lane-Serff et al . , 2014 ) were grown in HMI-9 with 10% FCS at 37°C with 5% CO2 ( Hirumi and Hirumi , 1989 ) . The TcHpHbR was inducibly overexpressed in T . brucei TbHpHbR KO BSFs transfected with pSMOX ( Poon et al . , 2012 ) and a modified version of pDEX777 ( also Poon et al . , 2012 ) where the GFP ORF was replaced with the TcHpHbR ORF . Cells were induced with 10 µg/ml doxycycline for 24 hr before protein was harvested and analysed by western blot . Experimental tsetse flies ( Glossina pallidipes ) were caged in groups of 15 , kept at 25°C and 70% relative humidity , and fed on sterile defibrinated horse blood supplemented with 1 mM dATP ( Galun and Margalit , 1969 ) via a silicone membrane . Male and female flies were used for experiments , being given the infective bloodmeal for their first feed 24–48 hr post-eclosion . The infective feed contained approximately 1 x 106 T . congolense Gam 2 trypanosomes ml-1 from the supernatant of epimastigote cultures in washed red blood cells supplemented with 10 mM L-glutathione ( MacLeod et al . , 2007 ) to increase infection rates . Flies were dissected 40–42 days post infection . Heads were removed and proboscides dissected directly into a drop of vPBS on assay slides , carefully separating apart the labrum , hypopharynx and labium . Whole tsetse alimentary tracts were dissected and the proventriculus and midgut placed into separate drops of vPBS . Western blot analysis was carried out on cell lysates using standard methods . Bloodstream and procyclic form cell lysates were harvested from T . congolense Il3000 cells from log-phase cultures . Epimastigote cell lysates were collected from three independently generated epimastigote-containing cultures . Antibodies were raised by injecting recombinant TcHpHbR into rabbits ( Covalab , France ) and purified using affinity chromatographywith TcHpHbR agarose . Quantification of the copy number of TcHpHbR was carried out by western blot and comparison between cell lysates and recombinant protein . To determine the number of cells expressing TcHpHbR in the samples , six IFAs were carried out ( as described below ) and 500 cells per IFA were scored as positive or negative for TcHpHbR expression . A total of 50/3000 cells , or 1 . 67% of the population , were positive for TcHpHbR expression . By comparison with known quantities of recombinant TcHpHbR protein on two independent western blots it was observed that 8 . 35 x 104 TcHpHbR-expressing cells was equivalent to 2 . 25–4 . 5 ng protein , or 4 . 6 and 9 . 3 x 105 molecules per cell , using a molecular weight of 35 kDa for calculations . Immunofluorescent analysis of culture-generated epimastigotes was carried out on culture supernatants as described above or on cells grown on glass coverslips to enrich for the adherent epimastigotes . Cells were either fixed with 4% paraformaldehyde at room temperature for 30 min and then blocked with 10 mM methylamine-HCl pH 8 . 0 for 30 min or fixed by air-drying and then incubating in ice-cold methanol for 30–60 min . Samples were blocked with 5% donkey serum in PBS for 1 hr . Cells were then incubated for 1 hr with rabbit anti-TcHpHbR polyclonal antisera raised against recombinant TcHpHbR protein followed by an Alexa488 donkey anti-rabbit secondary antibody diluted in 5% serum in PBS , also for 1 hr . Cells were stained with 1 µg/ml DAPI for 5 min , washed and mounted with Calbiochem FluorSave Reagent ( Merck Millipore , Billerica , MA ) . For immunofluorescent analysis of tsetse-derived T . congolense , dissected samples were air-dried and fixed in ice-cold methanol for 30 min , then processed as above . Microscopy was carried out on a Zeiss Imager M1 microscope and analysed with AxioVision Rel 4 . 8 software . Hb and BSA were labelled with Alexa Fluor 488 using a protein labelling kit ( Thermo Fisher Scientific ) . T . congolense WG81 epimastigote-containing cultures ( generated as above ) were grown on coverslips overnight in serum-free Cunningham’s media supplemented with 5 mg/ml BSA , 1 mM hypoxanthine and 0 . 16 mM thymidine . Incubation in serum-free media was required to remove competing Hb ligand from the media . Coverslips were moved onto poly-l-lysine slides and incubated with no ligand , 10 nM Hb-488 or 10 nM-BSA at 27°C for 4 hr . At 2 hr post-addition of ligand , 2 µM protease inhibitor FMK-024 was added . Cells were fixed in 4% paraformaldehyde for 30 min at room temperature , washed 3x in PBS , stained with 1 µg/ml DAPI for 5 min and mounted . Microscopy was carried out as above . HpSPHb and TcHpHbR were mixed in equimolar ratios to a final total concentration of 12 . 5 mg/ml in 20 mM HEPES pH 7 . 5 , 150 mM NaCl and were subjected to crystallisation trials . Crystals were obtained after 18 hr in a sitting drop format with a well solution containing 0 . 1 M sodium citrate pH 5 , 8% w/v PEG 8000 . After cryoprotection by transfer into well solution with the addition of 30% glycerol , the crystals were cryo-cooled . Data were collected on beamline I03 at the Diamond light source and were indexed and scaled using iMosflm ( Battye et al . , 2011 ) and Scala ( Evans et al . , 1993 ) respectively . Phaser ( McCoy et al . , 2007 ) was used to determine a molecular replacement model , using the known structures of TcHpHbR ( pdb: 4E40 , Higgins et al . , 2013 ) and human oxygenated Hb ( pdb: 1HHO , Shaanan , 1983 ) as search models . Refinement and rebuilding was completed using Buster ( Bricogne et al . , 2011 ) and Coot ( Emsley et al . , 2010 ) respectively . Receptors were coupled to an SPR chip using a biotin attached to the C-terminal BAP tag . This strategy was designed to allow them to be immobilised with an orientation matching that found on the parasite surface , and to generate a surface that could be readily regenerated . Purified receptors were biotinylated by mixing 1 mg of protein at 30 μM with 20 μg of BirA ( Sigma Aldrich ) , 5 mM ATP ( Sigma Aldrich ) and 300 μM biotin ( Sigma Aldrich ) . They were incubated at room temperature overnight , before desalting using a PD10 column ( GE Healthcare ) to remove excess biotin . SPR experiments were carried out on a Biacore T200 instrument ( GE Healthcare ) . All experiments were performed in 20 mM HEPES pH 7 . 5 , 150 mM NaCl , 0 . 005% Tween-20 at 25°C . Two-fold dilution series of human Hb , human HpHb or bovine Hb ( Sigma ) were prepared for injection over a receptor-coated chip with upper concentrations of 1 μM . For each cycle , biotinylated recombinant receptor was immobilised on a CAP chip using the Biotin CAPture Kit ( GE Healthcare ) to a total loading of ~250 RU . Binding partners were injected for 240 s with a dissociation time of 300 s . The chip was regenerated between cycles using regeneration solution from the Biotin CAPture Kit . The specific binding response of the ligands to receptors was determined by subtracting the response given by Hb or HpHb from a surface to which no receptor had been coupled . As both Hb and HpHb have the capacity to simultaneously interact with two receptors , this data was not fitted to obtain affinity measurements . ITC measurements were carried out on a MicroCal iTC200 System ( Malvern , UK ) . Samples were dialysed for 15 hr into 20 mM HEPES pH 7 . 5 , 150 mM NaCl at 4°C . Experiments were performed at 25°C with 50 μl of T . congolense HpHbR at 250 μM titrated into a cell containing 300 μl of Hb at 18 μM . The titrant was injected in 20 injections of 2 μl . Data were integrated and fit by nonlinear least-squares fitting using Origin ITC Software ( Malvern ) . | Trypanosomes are single-celled parasites that infect a range of animal hosts . These parasites need a molecule called haem to grow properly and are mostly spread by insects that feed on the blood of mammals . Most haem in mammals is found in red blood cells and is bound to a protein called haemoglobin . When it is released from these cells , haemoglobin forms a complex with another protein called haptoglobin as well . The best-studied trypanosomes from Africa have a receptor protein on their surface that recognizes the haptoglobin-haemoglobin complex and allows the parasites to obtain haem from their hosts . An African trypanosome called T . brucei causes sleeping sickness in humans , and has a receptor that can only recognize haemoglobin when it is in complex with haptoglobin . However , few trypanosome receptors have been studied to date , and so it was not clear if they all work in the same way . Trypanosoma congolense is a trypanosome that has a big impact on livestock farmers in sub-Saharan Africa and infects cattle , pigs and goats . Lane-Serff , MacGregor et al . now report that the receptor protein from T . congolense can bind to haemoglobin on its own . A technique called X-ray crystallography was used to reveal the three-dimensional structure of the T . congolense receptor and haemoglobin in fine detail . Further experiments then confirmed that the receptor actually binds more strongly to haemoglobin than it does to the haptoglobin-haemoglobin complex . Experiments with living parasites showed that T . congolense produces its receptor when it is in the mouthparts of its insect host , the tsetse fly . This is unlike what occurs in T . brucei , which only produces its receptor while it is in the bloodstream of its mammalian host . Lane-Serff , MacGregor et al . suggest that T . congolense’s receptor is more like the receptor found in ancestor of the trypanosomes . This means that , at least once during the evolution of these parasites , this receptor evolved from being a haemoglobin receptor produced in the tsetse fly to a haptoglobin-haemoglobin receptor produced in an infected mammal . The next step is to investigate the details of the role played by the T . congolense receptor when the parasite is in the tsetse fly . It will also be important to understand how this parasite is still able to grow in the mammalian host’s bloodstream even though it does not produce much of the receptor during this stage . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2016 | Evolutionary diversification of the trypanosome haptoglobin-haemoglobin receptor from an ancestral haemoglobin receptor |
Despite the central role that antibodies play in the adaptive immune system and in biotechnology , much remains unknown about the quantitative relationship between an antibody’s amino acid sequence and its antigen binding affinity . Here we describe a new experimental approach , called Tite-Seq , that is capable of measuring binding titration curves and corresponding affinities for thousands of variant antibodies in parallel . The measurement of titration curves eliminates the confounding effects of antibody expression and stability that arise in standard deep mutational scanning assays . We demonstrate Tite-Seq on the CDR1H and CDR3H regions of a well-studied scFv antibody . Our data shed light on the structural basis for antigen binding affinity and suggests a role for secondary CDR loops in establishing antibody stability . Tite-Seq fills a large gap in the ability to measure critical aspects of the adaptive immune system , and can be readily used for studying sequence-affinity landscapes in other protein systems .
During an infection , the immune system must recognize and neutralize invading pathogens . B-cells contribute to immune defense by producing antibodies , proteins that bind specifically to foreign antigens . The astonishing capability of antibodies to recognize virtually any foreign molecule has been repurposed by scientists in a wide variety of experimental techniques ( immunofluorescence , western blots , ELISA , ChIP-Seq , etc . ) . Antibody-based therapeutic drugs have also been developed for treating many different diseases , including cancer ( Chan and Carter , 2010 ) . Much is known about the qualitative mechanisms of antibody generation and function ( Murphy et al . , 2008 ) . The antigenic specificity of antibodies in humans , mice , and most jawed vertebrates is primarily governed by six complementarity determining regions ( CDRs ) , each roughly 10 amino acids ( aa ) long . Three CDRs ( denoted CDR1H , CDR2H , and CDR3H ) are located on the antibody heavy chain , and three are on the light chain . During B-cell differentiation , these six sequences are randomized through V ( D ) J recombination , then selected for functionality as well as against the ability to recognize host antigens . Upon participation in an immune response , CDR regions can further undergo somatic hypermutation and selection , yielding higher-affinity antibodies for specific antigens . Among the CDRs , CDR3H is the most highly variable and typically contributes the most to antigen specificity; less clear are the functional roles of the other CDRs , which often do not interact with the target antigen directly . Many high-throughput techniques , including phage display ( Smith , 1985; Vaughan et al . , 1996; Schirrmann et al . , 2011 ) , ribosome display ( Fujino et al . , 2012 ) , yeast display ( Boder and Wittrup , 1997; Gai and Wittrup , 2007 ) , and mammalian cell display ( Forsyth et al . , 2013 ) , have been developed for optimizing antibodies ex vivo . Advances in DNA sequencing technology have also made it possible to effectively monitor both antibody and T-cell receptor diversity within immune repertoires , e . g . in healthy individuals ( Boyd et al . , 2009; Weinstein et al . , 2009; Robins et al . , 2009 , 2010; Mora et al . , 2010; Venturi et al . , 2011; Murugan et al . , 2012; Zvyagin et al . , 2014; Elhanati et al . , 2014; Qi et al . , 2014; Thomas et al . , 2014; Elhanati et al . , 2015 ) , in specific tissues ( Madi et al . , 2014 ) , in individuals with diseases ( Parameswaran et al . , 2013 ) or following vaccination ( Jiang et al . , 2013; Vollmers et al . , 2013; Laserson et al . , 2014; Galson et al . , 2014; Wang et al . , 2015 ) . Yet many questions remain about basic aspects of the quantitative relationship between antibody sequence and antigen binding affinity . How many different antibodies will bind a given antigen with specified affinity ? How large of a role do epistatic interactions between amino acid positions within the CDRs have on antigen binding affinity ? How is this sequence-affinity landscape navigated by the V ( D ) J recombination process , or by somatic hypermutation ? Answering these and related questions is likely to prove critical for developing a systems-level understanding of the adaptive immune system , as well as for using antibody repertoire sequencing to diagnose and monitor disease . Recently developed ‘deep mutational scanning’ ( DMS ) assays ( Fowler and Fields , 2014 ) provide one potential method for measuring binding affinities with high enough throughput to effectively explore antibody sequence-affinity landscapes . In DMS experiments , one begins with a library of variants of a specific protein . Proteins that have high levels of a particular activity of interest are then enriched via one or more rounds of selection , which can be carried out in a variety of ways . The set of enriched sequences is then compared to the initial library , and protein sequences ( or mutations within these sequences ) are scored according to how much this enrichment procedure increases their prevalence . Multiple DMS assays have been described for investigating protein-ligand binding affinity . But no DMS assay has yet been shown to provide absolute quantitative binding affinity measurements , i . e . , dissociation constants in molar units . For example , one of the first DMS experiments ( Fowler et al . , 2010 ) used phage display technology to measure how mutations in a WW domain affect the affinity of this domain for its peptide ligand . These data were sufficient to compute enrichment ratios and corresponding sequence logos , but they did not yield quantitative affinities . Analogous experiments have since been performed on antibodies using yeast display ( Reich et al . , 2015; Kowalsky et al . , 2015 ) and mammalian cell display ( Forsyth et al . , 2013 ) . Yeast-display-based DMS assays have also proven particularly useful for mapping protein epitopes that are targeted by specific antibodies of interest ( Kowalsky et al . , 2015; Doolan and Colby , 2015; Van Blarcom et al . , 2015 ) . Still , none of these approaches provides quantitative affinity values . SORTCERY ( Reich et al . , 2015 , ) , a DMS assay that combines yeast display and quantitative modeling , has been shown to provide approximate rank-order values for the affinity of a specific protein for short unstructured peptides of varying sequence . Determining quantitative affinities from SORTCERY data , however , requires separate low-throughput calibration measurements ( Reich et al . , 2014 ) . Moreover , it is unclear how well SORTCERY , if applied to a library of folded proteins rather than unstructured peptides , can distinguish sequence-dependence effects on affinity from sequence-dependent effects on protein expression and stability . Other recent work has described a DMS assay , again based on yeast display , for measuring fold-changes in affinity relative to a reference protein ( Kowalsky and Whitehead , 2016 ) . This method , however , does not provide absolute values for dissociation constants , is vulnerable to the confounding effects of sequence-dependent expression and protein stability , and was observed to have only a 10-fold dynamic range . To enable massively parallel measurements of absolute binding affinities for antibodies and other structured proteins , we have developed an assay called ‘Tite-Seq . ’ Tite-Seq , like SORTCERY , builds on the capabilities of Sort-Seq , an experimental strategy that was first developed for studying transcriptional regulatory sequences in bacteria ( Kinney et al . , 2010 ) . Sort-Seq combines fluorescence-activated cell sorting ( FACS ) with high-throughput sequencing to provide massively parallel measurements of cellular fluorescence . In the Tite-Seq assay , Sort-Seq is applied to antibodies displayed on the surface of yeast cells and incubated with antigen at a wide range of concentrations . From the resulting sequence data , thousands of antibody-antigen binding titration curves and their corresponding absolute dissociation constants ( here denoted KD ) can be inferred . By assaying full binding curves , Tite-Seq is able to measure affinities over many orders of magnitude ( We note that Kowalsky et al . ( 2015 ) have described yeast display DMS experiments performed at multiple concentrations . These data , however , were not used to reconstruct titration curves or infer quantitative KD values ) . Moreover , the resulting affinity values provided by Tite-Seq are not confounded by the ( rather substantial ) effect that sequence variation can have on either ( a ) the amount of protein expressed on the surface of cells or ( b ) the specific activity of displayed proteins ( i . e . , the fraction of protein molecules that are functional ) . We demonstrated Tite-Seq on a protein library derived from a well-studied single-chain variable fragment ( scFv ) antibody specific to the small molecule fluorescein ( Boder and Wittrup , 1997; Boder et al . , 2000 ) . Mutations were restricted to CDR1H and CDR3H regions , which are known to play an important role in the antigen recognition of this scFv ( Boder et al . , 2000; Midelfort et al . , 2004 ) . The resulting affinity measurements were validated with binding curves for a handful of clones measured using standard low-throughput flow cytometry . Our Tite-Seq measurements reveal both expected and unexpected differences between the effects of mutations in CDR1H and CDR3H . These data also shed light on structural aspects of antigen recognition that are independent of effects on antibody stability .
Our general strategy is illustrated in Figure 1 . First , a library of variant antibodies is displayed on the surface of yeast cells ( Figure 1A ) . The composition of this library is such that each cell displays a single antibody variant , and each variant is expressed on the surface of multiple cells . Cells are then incubated with the antigen of interest , bound antigen is fluorescently labeled , and fluorescence-activated cell sorting ( FACS ) is used to sort cells one-by-one into multiple ‘bins’ based on this fluorescent readout ( Figure 1B ) . Deep sequencing is then used to survey the antibody variants present in each bin . Because each variant antibody is sorted multiple times , it will be associated with a histogram of counts spread across one or more bins ( Figure 1C ) . The spread in each histogram is due to cell-to-cell variability in antibody expression , and to the inherent noisiness of flow cytometry measurements . Finally , the histogram corresponding to each antibody variant is used to compute an ‘average bin number’ ( Figure 1C , dots ) , which serves as a proxy measurement for the average amount of bound antigen per cell . 10 . 7554/eLife . 23156 . 003Figure 1 . Schematic illustration of Tite-Seq . ( A ) A library of variant antibodies ( various colors ) are displayed on the surface of yeast cells ( tan ) . ( B ) The library is exposed to antigen ( green triangles ) at a defined concentration , cell-bound antigen is fluorescently labeled , and FACS is used to sort cells into bins according to measured fluorescence . ( C ) The antibody variants in each bin are sequenced and the distribution of each variant across bins is computed ( histograms; colors correspond to specific variants ) . The mean bin number ( dot ) is then used to quantify the typical amount of bound antigen per cell . ( D ) Binding titration curves ( solid lines ) and corresponding KD values ( vertical lines ) can be inferred for individual antibody sequences by using the mean fluorescence values ( dots ) obtained from flow cytometry experiments performed on clonal populations of antibody-displaying yeast . ( E ) Tite-Seq consists of performing the Sort-Seq experiment in panels A–C at multiple antigen concentrations , then inferring binding curves using mean bin number as a proxy for mean cellular fluorescence . This enables KD measurements for thousands of variant antibodies in parallel . We note that the Tite-Seq results illustrated in panel E were simulated using three bins under idealized experimental conditions , as described in Appendix 1 . The inference of binding curves from real Tite-Seq data is more involved than this panel might suggest , due to the multiple sources of experimental noise that must be accounted for . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 003 It has previously been shown that KD values can be accurately measured using yeast-displayed antibodies by taking binding titration curves , i . e . , by measuring the average amount of bound antigen as a function of antigen concentration ( VanAntwerp and Wittrup , 2000; Gai and Wittrup , 2007 ) . The median fluorescence f of labeled cells is expected to be related to antigen concentration via ( 1 ) f=Acc+KD+B where A is proportional to the number of functional antibodies displayed on the cell surface , B accounts for background fluorescence , and c is the concentration of free antigen in solution . Figure 1D illustrates the shape of curves having this form . By using flow cytometry to measure f on clonal populations of yeast at different antigen concentrations c , one can infer curves having the sigmoidal form shown in Equation 1 and thereby learn KD . Such measurements , however , can only be performed in a low-throughput manner . Tite-Seq allows thousands of binding titration curves to be measured in parallel . The Sort-Seq procedure illustrated in Figure 1A–C is performed at multiple antigen concentrations , and the resulting average bin number for each variant antibody is plotted against concentration . Sigmoidal curves are then fit to these proxy measurements , enabling KD values to be inferred for each variant . We emphasize that KD values cannot , in general , be accurately inferred from Sort-Seq experiments performed at a single antigen concentration . Because the relationship between binding and KD is sigmoidal , the amount of bound antigen provides a quantitative readout of KD only when the concentration of antigen used in the labeling procedure is comparable in magnitude to KD . However , single mutations within a protein binding domain often change KD by multiple orders of magnitude . Sort-Seq experiments used to measure sequence-affinity landscapes must therefore be carried out over a range of concentrations large enough to encompass this variation . Furthermore , as illustrated in Figure 1C and D , different antibody variants often lead to different levels of functional antibody expression on the yeast cell surface . If one performs Sort-Seq at a single antigen concentration , high affinity ( low KD ) variants with low expression ( blue variant ) may bind less antigen than low affinity ( high KD ) variants with high expression ( orange variant ) . Only by measuring full titration curves can the effect that sequence has on affinity be deconvolved from sequence-dependent effects on functional protein expression . To test the feasibility of Tite-Seq , we used a well-characterized antibody-antigen system: the 4-4-20 single chain variable fragment ( scFv ) antibody ( Boder and Wittrup , 1997 ) , which binds the small molecule fluorescein with KD=1 . 2 nM ( Gai and Wittrup , 2007 ) . This system was used in early work to establish the capabilities of yeast display ( Boder and Wittrup , 1997 ) , and a high resolution co-crystal structure of the 4-4-20 antibody bound to fluorescein , shown in Figure 2A , has been determined ( Whitlow et al . , 1995 ) . An ultra-high-affinity ( KD=270 fM ) variant of this scFv , called 4m5 . 3 , has also been found ( Boder et al . , 2000 ) . In what follows , we refer to the 4-4-20 scFv from Boder and Wittrup ( 1997 ) as WT , and the 4m5 . 3 variant from Boder et al . ( 2000 ) as OPT . 10 . 7554/eLife . 23156 . 004Figure 2 . Yeast display construct and antibody libraries ( A ) Co-crystal structure of the 4-4-20 ( WT ) antibody from Whitlow et al . ( 1995 ) ( PDB code 1FLR ) . The CDR1H and CDR3H regions are colored blue and red , respectively . ( B ) The yeast display scFv construct from Boder and Wittrup ( 1997 ) that was used in this study . Antibody-bound antigen ( fluorescein ) was visualized using PE dye . The amount of surface-expressed protein was separately visualized using BV dye . Approximate location of the CDR1H ( blue ) and CDR3H ( red ) regions within the scFv are illustrated . ( C ) The gene coding for this scFv construct , with the six CDR regions indicated . The WT sequence of the two 10 aa variable regions are also shown . ( D ) The number of 1- , 2- , and 3-codon variants present in the 1H and 3H scFv libraries . Figure 2—figure supplement 1 shows the cloning vector used to construct the CDR1H and CDR3H libraries , as well as the form of the resulting expression plasmids . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 00410 . 7554/eLife . 23156 . 005Figure 2—figure supplement 1 . Cloning strategy . ( A ) The iRA11 amplicon library , which was prepared from microarray-synthesized oligos containing variant CDR1H or variant CDR3H regions . This amplicon is flanked by inward-facing BsaI restriction sites . ( B ) The pRA10 cloning vector , which contains the ccdB selection gene within a cassette flanked by outward-facing BsmBI restriction sites . ( C ) The pRA11 plasmid library , which was cloned by ligating BsaI-digested iRA11 amplicons and BsmBI-digest pRA10 vector . ( D ) The sequencing amplicon that was amplified from sorted cells after Tite-Seq and Sort-Seq experiments and submitted for ultra-high-throughput DNA sequencing . Appendix 3 provides more details about iRA11 amplicons , the pRA10 vector , and the pRA11 plasmid library . Appendix 4 provides more information about the creation of sequencing amplicons . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 005 The scFv was expressed on the surface of yeast as part of the multi-domain construct illustrated in Figure 2B and previously described in Boder and Wittrup ( 1997 ) . Following ( Boder et al . , 2000 ) , we used fluorescein-biotin as the antigen and labeled scFv-bound antigen with streptavidin-RPE ( PE ) . The amount of surface-expressed protein was separately quantified by labeling the C-terminal c-Myc tag using anti-c-Myc primary antibodies , followed by secondary antibodies conjugated to Brilliant Violet 421 ( BV ) . See Appendix 2 for details on this labeling procedure . Two different scFv libraries were assayed simultaneously . In the ‘1H’ library , a 10 aa region encompasing the CDR1H region of the WT scFv ( see Figure 2C ) was mutagenized using microarray-synthesized oligos ( see Appendix 3 for details ) . The resulting 1H library consisted of all 600 single-codon variants of this 10 aa region , 1100 randomly chosen 2-codon variants , and 150 random 3-codon variants ( Figure 2D ) . An analogous ‘3H’ library was generated for a 10 aa region containing the CDR3H region of this scFv . In all of the Tite-Seq experiments described below , these two libraries were pooled together and supplemented with WT and OPT scFvs , as well with a nonfunctional scFv referred to as Δ . Tite-Seq was carried out as follows . Yeast cells expressing scFv from the mixed library were incubated with fluorescein-biotin at one of eleven concentrations: 0 M , 10-9 . 5 M , 10-9 M , 10-8 . 5 M , 10-8 M , 10-7 . 5 M , 10-7 M , 10-6 . 5 M , 10-6 M , 10-5 . 5 M , and 10-5 M . After subsequent PE labeling of bound antigen , cells were sorted into four bins using FACS ( Figure 3A ) . Separately , BV-labeled cells were sorted according to measured scFv expression levels ( Figure 3B ) . The number of cells sorted into each bin is shown in Figure 3C . Each bin of cells was regrown and bulk DNA was extracted . The 1H and 3H variable regions were then PCR amplified and sequenced using paired-end Illumina sequencing , as described in Appendix 4 . The final data set consisted of an average of 2 . 6×106 sequences per bin across all 48 bins ( Figure 3D ) . Three independent replicates of this experiment were performed on three different days . 10 . 7554/eLife . 23156 . 006Figure 3 . Details of our Tite-Seq experiments . ( A ) Gates used to sort cells based on PE fluorescence , which provides a readout of bound antigen . Cells were labeled at the eleven different antigen concentrations . Shades of red indicate the four fluorescence gates used to sort cells; these correspond to bins 0 , 1 , 2 , and 3 ( from left to right ) . ( B ) Gates , indicated in shades of purple , used to sort cells based on BV fluorescence , which provides a readout of antibody expression . ( C ) The number of cells sorted into each bin . ( D ) The number of Illumina reads obtained from each bin of sorted cells after quality control measures were applied . The data shown in this figure corresponds to a single Tite-Seq experiment . Figure 3—figure supplement 1 and Figure 3—figure supplement 2 show data for two independent replicates of this experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 00610 . 7554/eLife . 23156 . 007Figure 3—figure supplement 1 . Tite-Seq experiment , replicate 2 . Analog of Figure 3 in the main text , but for the replicate 2 Tite-Seq experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 00710 . 7554/eLife . 23156 . 008Figure 3—figure supplement 2 . Tite-Seq experiment , replicate 3 . Analog of Figure 3 in the main text , but for the replicate 3 Tite-Seq experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 008 For each variant scFv gene , a KD value was inferred by fitting a binding curve to the resulting Tite-Seq data , with separate curves independently fit to data from each Tite-Seq experiment ( Figure 4A ) . As illustrated in Figure 1E , this fitting procedure uses the sigmoidal function in Equation 1 to model mean bin number as a function of antigen concentration . However , the need to account for multiple sources of noise in the Tite-Seq experiment necessitates a more complex procedure than Figure 1E might suggest; the details of this inference procedure are described in Appendix 5 . 10 . 7554/eLife . 23156 . 009Figure 4 . Accuracy and precision of Tite-Seq . ( A ) Binding curves and KD measurements inferred from Tite-Seq data . ( B ) Mean fluorescence values ( dots ) and corresponding inferred binding curves ( lines ) obtained by flow cytometry measurements for five selected scFvs ( WT , OPT , C5 , C45 , and C107 ) . In ( A , B ) , values corresponding to 0 M fluorescein are plotted on the left-most edge of the plot , dotted lines show the upper ( 10-5 M ) and lower ( 10-9 . 5 M ) limits on KD sensitivity , vertical lines show inferred KD values , and different shades correspond to different replicate experiments . ( C ) Comparison of the Tite-Seq-measured and flow-cytometry-measured KD values for all clones tested . Colors indicate different scFv protein sequences as follows: WT ( purple ) , OPT ( green ) , Δ ( black ) , 1H clones ( blue ) , and 3H clones ( red ) . Each KD value indicates the mean log10KD value obtained across all replicates , with error bars indicating standard error . Clones with KD outside of the affinity range are drawn on the boundaries of this range , which are indicated with dotted lines . The coefficient of determination ( R2 ) between log Tite-Seq values and log flow KDvalues includes clones outside of the affinity range; in such cases , the corresponding boundary value ( 10-9 . 5 M or 10-5 . 0 M ) has been used . The amino acid sequences and measured KD values for all clones tested are provided in Table 1 . Figure 4—figure supplement 1 provides plots , analogous to panels A and B , for all of the assayed clones . Figure 4—figure supplement 2 compares KD and E values obtained across all three Tite-Seq replicates . Figure 4—figure supplement 3 quantifies measurement error using synonymous mutants . Figure 4—figure supplement 4 provides information about library composition . Figure 4—figure supplement 5 illustrates the poor correlation between scFv enrichment and Tite-seq measured KD values . Figure 4—figure supplement 6 shows a 2-fold difference in the specific activities of OPT and WT scFvs . Figure 4—figure supplement 7 illustrates the simulations we used in Figure 4—figure supplement 8 to validate the ability of our analysis to infer correct KD values . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 00910 . 7554/eLife . 23156 . 010Figure 4—figure supplement 1 . Binding curves for all clones . Binding curves , measured using ( A ) Tite-Seq or ( B ) flow cytometry , for all clones analyzed in this paper and described in Table 1 . Plots are drawn as in Figure 4 , panels A and B . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 01010 . 7554/eLife . 23156 . 011Figure 4—figure supplement 2 . Concordance between replicate experiments . Density plots of ( A ) Tite-Seq-measured KD values and ( B ) Sort-Seq-measured E values between all pairs of replicate experiments . Measurements for these quantities that were judged to be of low precision due to low sequence counts are not plotted . f indicates the percentage of total assayed sequences plotted; r is the Pearson correlation and includes clonal measurements outside the boundaries of our measurable ranges ( 10-9 . 5-10-5 M for KD , 0–2 for expression ) . Clones outside of these ranges were given values at the closest boundary . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 01110 . 7554/eLife . 23156 . 012Figure 4—figure supplement 3 . Error estimates from synonymous mutants . Density plots for ( A ) Tite-Seq-measured log10KD standard deviation and average log10KD and ( B ) Sort-Seq-measured E standard deviation and average E are shown for each scFv sequence with more than one synonymous mutant for each of the replicate experiments . The KD error peaked between 10-7-10-6 M . The expression error peaked at or above WT expression ( i . e . 1 ) levels . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 01210 . 7554/eLife . 23156 . 013Figure 4—figure supplement 4 . Composition of scFv libraries . ( A ) Comparison of library composition between all pairs of replicate experiments . ( B ) Zipf plots showing the library composition in each replicate experiment . In both panels , the prevalence of each scFv sequence in each replicate experiment was determined as part of the Tite-Seq curve fitting procedure , as described in Appendix 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 01310 . 7554/eLife . 23156 . 014Figure 4—figure supplement 5 . Sort-Seq enrichment correlates poorly with Tite-Seq-measured affinity . To assess how well simple enrichment calculations might reproduce the KD values measured by Tite-Seq , we did the following calculation . For each of the two libraries ( 1 H and 3 H ) , we partitioned scFvs into seven groups based on their measured KDs ( columns ) . For each group at each antigen concentration ( rows ) , we then computed the enrichment of each scFv in the high PE bins ( bins 2 , 3 ) relative to the low PE bins ( bins 0 , 1 ) . In these enrichment calculations , the number of counts in each bin was re-weighted to accurately reflect the fraction of library cells falling within the fluorescence range of that bin . This figure shows the resulting Spearman rank correlation ( ρ ) between enrichment and log KD values computed for each scFv group at each antigen concentration . In both libraries , we see that correlation values above background ( which can be assessed from the values in the 0 M fluorescein row ) only occur close to the diagonal , i . e . , when KD is close to the fluorescein concentration used . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 01410 . 7554/eLife . 23156 . 015Figure 4—figure supplement 6 . Differing specific activities of OPT and WT . 2D flow cytometry histograms showing both OPT- and WT-expressing cells labeled with PE and BV after incubation at 2 μM fluorescein . At this fluorescein concentration , nearly all functional WT and OPT scFvs are bound . Regression lines ( fixed to have slope 1 ) were fit to data points with BV signal between 104 . 5 and 105 . The vertical shift of the OPT data relative to the WT data indicates a factor of 2 . 03±0 . 07 difference ( computed from four replicate experiments ) in the amount labeled antigen . This difference is not due to a difference in the number of surface-displayed scFvs , as this would cause the OPT and WT clouds to lie along the same diagonal . Rather , this difference between WT and OPT is due to variation in specific activity . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 01510 . 7554/eLife . 23156 . 016Figure 4—figure supplement 7 . Realistic Tite-Seq simulations . Realistic Tite-Seq data were simulated separately for each distinct pair of affinity ( KD ) and amplitude ( A ) values , as described in Appendix 7 . This figure shows simulated data , akin to the data displayed in Figure 4—figure supplement 6 , for WT values of KD and A . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 01610 . 7554/eLife . 23156 . 017Figure 4—figure supplement 8 . Validation of analysis pipeline . KD values were inferred for Tite-Seq data simulated using ( green ) the same number of cells , ( light green ) 10-3 times as many cells , or ( black ) 104 times as many sorted cells as in our experiments . Areas indicate approximately plus or minus one standard deviation in the fitted KD values obtained for each true KD value . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 01710 . 7554/eLife . 23156 . 018Table 1 . Clones measured using flow cytometry and Tite-Seq . List of scFv clones , ordered by their flow-cytometry-measured KD values . With the exception of OPT and Δ , these clones differed from WT only in their 1H and 3H variable regions . WT amino acids within these regions are capitalized; variant amino acids are shown in lower case . No sequence is shown for Δ because this clone contained a large deletion , making identification of the 1H and 3H variable regions meaningless . KD values saturating our lower detection limit of 10-9 . 5 M or upper detection limit of 10-5 . 0M are written with a ≲ or ≳ sign to emphasize the uncertainty in these measurements . Tite-Seq KD values indicate mean and standard errors computed across the three replicate Tite-Seq experiments; they are not averaged across synonymous variants . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 018Name1H variable region3H variable regionNo . replicates ( flow ) KD [M] ( flow ) KD [M] ( Tite-Seq ) OPTTFghYWMNWVGasYGMeYlG3≲10−9 . 5≲10−9 . 5C107TFSDYWMNWVGaYYGMDYWG310−9 . 28±0 . 0410−9 . 18±0 . 11C112TFSDYWMNWVGSYYGMDYcG310−8 . 95±0 . 0710−9 . 19±0 . 14WTTFSDYWMNWVGSYYGMDYWG1010−8 . 61±0 . 0710−8 . 92±0 . 10C144vFSDYWMNWVGSYYGMDYWG310−8 . 57±0 . 0310−8 . 86±0 . 04C133aFSDYWMNWVGSYYGMDYWG310−8 . 55±0 . 0610−8 . 62±0 . 09C132TFmDYWlNWVGSYYGMDYWG310−8 . 48±0 . 0810−8 . 38±0 . 29C94TFSDYWMNWVGSYYGMDsWG310−8 . 46±0 . 0610−8 . 50±0 . 04C5TFSDYWiNWVGSYYGMDYWG310−8 . 34±0 . 1010−8 . 55±0 . 09C93TFSDYWMNWVGSYrGMDYWG310−7 . 35±0 . 0810−7 . 60±0 . 70C39TFSDYWMNWVGSYYGMDYWa310−7 . 08±0 . 2010−7 . 28±0 . 17C102TFSDYWMNWVsSkYGMDYWG310−5 . 76±0 . 1610−7 . 25±0 . 60C22ssSDYWMNWVGSYYGMDYWG310−5 . 69±0 . 3110−7 . 53±0 . 07C7hFSDYWMNWlGSYYGMDYWG310−5 . 53±0 . 1810−5 . 39±0 . 18C45TFSDYWMNWVGSYdGnDYWG310−5 . 40±0 . 24≳10−5 . 0C103TFSDYWMNWVGSYYGMDlWG310−5 . 15±0 . 4710−5 . 44±0 . 55C3TFSDYWMsWVGSYYGMDYWG3≳10−5 . 0≳10−5 . 0C18TFSDYsMNWVGSYYGMDYWG3≳10−5 . 0≳10−5 . 0Δ––12≳10−5 . 0≳10−5 . 010 . 7554/eLife . 23156 . 019Table 2 . Primers . Oligonucleotide sequences are written 5′ to 3′ . Bold sequences indicate variable regions . The ‘1H library’ and ‘3H library’ primers respectively contained the 1H and 3H variable regions ( bold ) analyzed in this paper . These primer libraries were synthesized by LC Biosciences using microarray-based DNA synthesis . All other primers were ordered from Integrated DNA Technologies . The ‘[XX]’ portion of L1AF_XX and L1AR_XX indicates the location of each of 64 different barcodes ( i . e . , XX = 01 , 02 , … , 64 ) , which ranged in length from 7 bp to 10 bp and which differed from each other by at least two substitution mutations . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 019NameSequence1H libraryGTGTTGCCTCTGGATTCACTTTTAGTGACTACTGGATGAACTGGGTCCGCCAGTCTCCAGA3H libraryGTGACTGAGGTTCCTTGACCCCAGTAGTCCATACCATAGTAAGAACCCGTACAGTAATAGATACCCAToRAL10TTCTGAGGAGACGGTGACTGAGGTTCCTTGoRAR10TGAAGACATGGGTATCTATTACTGTACGoRAL11CAGTCCTTTCTCTGGAGACTGGCGoRAR11ATGAAACTCTCCTGTGTTGCCTCTGGATTC3H1FTTCTGAGGAGACGGTGACT3H2RTGAAGACATGGGTATCTATTACTGTAC1H2FCAGTCCTTTCTCTGGAGACTG1H1RATGAAACTCTCCTGTGTTGCCToRA10GCATATCTAAGGTCTCGTTCTGAGGAGACGGTGACoRA11GCCGATTGTTGGTCTCCATGAAACTCTCCTGTGTTGCPE1v3extAATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGPE2v3AAGCAGAAGACGGCATACGAGATCGGTCTCGGCATTCCTGCTL1AF_XXACACTCTTTCCCTACACGACGCTCTTCCGATCT[XX]AGTCTTCTTCAGAAATAAGCL1AR_XXCTCGGCATTCCTGCTGAACCGCTCTTCCGATCT[XX]GCTTGGTGCAACCTG Separately , the Sort-Seq data obtained by sorting the BV-labeled libraries were used to determine the expression level of each scFv . Specifically , we use E to denote ( for each scFv in the library ) the mean bin number that results from this expression-based sorting; this E value provides a measurement of the surface expression level of that scFv . All E values have been scaled so that the mean of such measurements for all synonymous WT scFv gene variants is 1 . 0 . To judge the accuracy of Tite-Seq , we separately measured binding curves for individual scFv clones as described for Figure 1D . In addition to the WT , OPT , and Δ scFvs , we assayed eight clones from the 1H library ( named C3 , C5 , C7 , C18 , C22 , C132 , C133 and C144 ) and eight clones from the 3H library ( C39 , C45 , C93 , C94 , C102 , C103 , C107 , C112 ) . Each clone underwent the same labeling procedure as in the Tite-Seq experiment , after which median fluorescence values were measured using standard flow cytometry . KD values were then inferred by fitting binding curves of the form in Equation 1 using the procedure described in Appendix 6 . These curves , which can be directly compared to Tite-Seq measurements ( Figure 4A ) , are plotted in Figure 4B; at least three replicate binding curves were measured for each clone . See Figure 4—figure supplement 1 for the titration curves of all the tested clones . Figure 4C reveals a strong correspondence between the KD values measured by Tite-Seq and those measured using low-throughput flow cytometry . The robustness of Tite-Seq is further illustrated by the consistency of KD values measured for the WT scFv . Using Tite-Seq , and averaging the results from the 33 synonymous variants and over all three replicates , we determined KD=10-8 . 87±0 . 02 M for the WT scFv . These measurements are largely consistent with the measurement of KD=10-8 . 61±0 . 07 M obtained by averaging low-throughput flow cytometry measurements across 10 replicates , and coincides with the previously measured value of 1 . 2 nM =10-8 . 9 M reported in ( Gai and Wittrup , 2007 ) . The three independent replicate Tite-Seq experiments give reproducible results as measured by direct comparison ( Figure 4—figure supplement 2 ) , from synonymous mutant variation ( Figure 4—figure supplement 3 ) and library composition Figure 4—figure supplement 4 ) with Pearson coefficients ranging from r=0 . 82 to r=0 . 89 for all the measured KD values between replicates; note that KD values outside of the sensitivity range are included in the calculation of these Pearson coefficients as described in the Figure 4 caption . The error bars for KD values in Figure 4C calculated from the variability of the fits to different replicates therefore support the reproducibility of the experiment . The main discrepancy in these error bar calculations occurred for clones c22 and c102 ( see also Figure 4—figure supplement 1 ) . The reason for this discrepancy is currently unclear . We note that Tite-Seq-measured KD values for these two clones are close to 10-7 M , and that the analysis of synonymous variants ( Figure 4—figure supplement 3 ) found that Tite-Seq-measured KDs in this region exhibited the largest variations . The necessity of performing KD measurements over a wide range of antigen concentrations is illustrated in Figure 4—figure supplement 5 . At each antigen concentration used in our Tite-Seq experiments , the enrichment of scFvs in the high-PE bins correlated poorly with the KD values inferred from full titration curves . Moreover , at each antigen concentration used , a detectable correlation between KD and enrichment was found only for scFvs with KD values close to that concentration . Figure 4—figure supplement 6 suggests a possible reason for the weak correlation between KD values and enrichment in high-PE bins . We found that , at saturating concentrations of fluorescein ( 2μM ) , cells expressing the OPT scFv bound twice as much fluorescein as cells expressing the WT scFv . This difference was not due to variation in the total amount of displayed scFv , which one might control for by labeling the c-Myc epitope as in Reich et al . ( 2015 ) . Rather , this difference in binding reflects a difference in the specific activity of displayed scFvs . Yeast display experiments performed at a single antigen concentration cannot distinguish such differences in specific activity from differences in scFv affinity . To further test the capability of Tite-Seq to infer dissociation constants from sequencing data over a wide range of values , as well as to validate our analysis procedures , we simulated Tite-Seq data in silico and analyzed the results using the same analysis pipeline that we used for our experiments . Details about the simulations are given in Appendix 7 . The simulated data is illustrated in Figure 4—figure supplement 7 . KD values inferred from these simulated data agreed to high accuracy with the KD used in the simulation ( Figure 4—figure supplement 8 ) , thus validating our analysis pipeline . Figure 5 shows the effect that every single-amino-acid substitution mutation within the 1H and 3H variable regions has on affinity and on expression; histograms of these effects are provided in Figure 5—figure supplement 1 . In both regions , the large majority of mutations weaken antigen binding ( 1H: 88%; 3H: 93% ) , with many mutations increasing KD above our detection threshold of 10-5 M ( 1H: 36%; 3H: 52% ) . Far fewer mutations reduced KD ( 1H: 12%; 3H: 7% ) , and very few dropped KD below our detection limit of 10-9 . 5 M ( 1H: 0%; 3H: 3% ) . Histograms of the effect of two or three amino acid changes relative to WT , shown in Figure 5—figure supplement 2A , reveal that multiple random mutations tend to further reduce affinity . We also observed that mutations within the 3H variable region have a larger effect on affinity than do mutations in the 1H variable region . Specifically , single amino acid mutations in 3H were seen to increased KD more than mutations in 1H ( 1H median KD=10-6 . 84; 3H median KD≳10-5 . 0P=4 . 7×10-4; P=4 . 7×10−4 , one-sided Mann-Whitney U test ) . This result suggests that binding affinity is more sensitive to variation in CDR3H than to variation in CDR1H , a finding that is consistent with the conventional understanding of these antibody CDR regions ( Xu and Davis , 2000; Liberman et al . , 2013 ) . 10 . 7554/eLife . 23156 . 020Figure 5 . Effects of substitution mutations on affinity and expression . Heatmaps show the measured effects on affinity ( A , B ) and expression ( C , D ) of all single amino acid substitutions within the variables regions of the 1H ( A , C ) and 3H ( B , D ) libraries . Purple dots indicate residues of the WT scFv . Green dots indicate non-WT residues in the OPT scFv . Figure 5—figure supplement 1 provides histograms of the non-WT values displayed in panels A–D . Figure 5—figure supplement 2 compares the effects on KD of both single-point and multi-point mutations . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 02010 . 7554/eLife . 23156 . 021Figure 5—figure supplement 1 . Histograms of substitution effects on affinity and expression . ( A , B ) Histogram showing the KD values measured for all substitution mutations in the 1 H ( A ) and 3 H ( B ) libraries . Note that these are the values plotted in panels A and B of Figure 5 , except that the WT KD value is not included . Dashed lines indicate the KD of the WT scFv; dotted lines indicate thresholds just within our detection boundaries , 10-9 . 49 M and 10-5 . 01 M , while the colored bars outside this interval indicate the number of substitution mutations with KD above ( blue ) and below ( red ) this range . ( C , D ) Histogram of E values for all single-substitution variants in the 1 H ( C ) or 3 H ( D ) libraries . These values , save those of the WT scFv , are plotted in panels C and D of Figure 5 . Dashed lines indicate the WT expression level of E=1 . 0 . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 02110 . 7554/eLife . 23156 . 022Figure 5—figure supplement 2 . Effects of multi-point mutations on affinity and expression . The effect of 1 , 2 , or three mutations on ( A ) Tite-Seq-measured KD values or ( B ) Sort-Seq-measured E values . Plots show the relative probability density ( over 30 bins along the KD or E axes ) observed for variants in each class . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 022 Our observations are thus fully consistent with the hypothesis that the amino acid sequences of the CDR1H and CDR3H regions of the WT scFv have been selected for high affinity binding to fluorescein . We know this to be true , of course; still , this result provides an important validation of our Tite-Seq measurements . To further validate our Tite-Seq affinity measurements , we examined positions in the high affinity OPT scFv ( from [Boder et al . , 2000] ) that differ from WT and that lie within the 1H and 3H variable regions . As illustrated in Figure 5A and B , five of the six OPT-specific mutations reduce KD or are nearly neutral . Previous structural analysis ( Midelfort et al . , 2004 ) has suggested that D106E , the only OPT mutation that we find significantly increases KD , may indeed disrupt antigen binding on its own while still increasing affinity in the presence of the S101A mutation . Next , we used our measurements to build a ‘matrix model’ ( also known as a ‘position-specific affinity matrix , ’ or PSAM [Foat et al . , 2006] ) describing the sequence-affinity landscape of these two regions . Our model assumed that the log10KD value for an arbitrary amino acid sequence could be computed from the log10KD value of the WT scFv , plus the measured change in log10KD produced by each amino acid substitution away from WT . We evaluated our matrix models on the 1H and 3H variable regions of OPT , finding an affinity of 10-9 . 16 M . Our simple model for the sequence affinity landscape of this scFv therefore correctly predicts that OPT has higher affinity than WT . The quantitative affinity predicted by our model does not match the known affinity of the OPT scFv ( KD=10-12 . 6 M ) , but this is unsurprising for three reasons . First the OPT scFv differs from WT in 14 residues , only 6 of which are inside the 1H and 3H variable regions assayed here . Second , one of the OPT mutations ( W108L ) reduces KD below our detection threshold of 10-9 . 5 M; in building our matrix model , we set this value equal to 10-9 . 5 , knowing it would likely underestimate the affinity-increasing effect of the mutation . Third , our additive model ignores potential epistatic interactions . Still , we thought it worth asking how likely it it would be for six random mutations within the 1H and 3H variable regions to reduce affinity as much as our model predicts for OPT . We therefore simulated a large number ( 107 ) of variants having a total of 6 substitution mutations randomly scattered across the 1H and 3H variable regions . The fraction of these random sequences that had an affinity at or below our predicted affinity for OPT was 4 . 7×10-5 . This finding is fully consistent with the fact that the mutations in OPT relative to WT were selected for increased affinity , an additional confirmation of the validity of our Tite-Seq measurements . The sequence-expression landscape measured in our separate Sort-Seq experiment yielded qualitatively different results ( Figure 5C and D ) . We observed no significant difference in the median effect that mutations in the variable regions of 1H ( median E=0 . 826 ) versus 3H ( median E=0 . 822 ) have on expression ( P=0 . 96 , two-sided Mann-Whitney U test ) ; see also Figure 5—figure supplement 1 . The variance in these effects , however , was larger in 3H than in 1H ( P=9 . 9×10-16 , Levene’s test ) . These results suggest two things . First , the 3H variable region appears to have a larger effect on scFv expression than the 1H variable region has . At the same time , since we observe fewer beneficial mutations in 1H ( Figure 5C ) than in 3H ( Figure 5D ) , the WT sequence appears to be more highly optimized for expression in CDR1H than in CDR3H . The effect of double or triple mutations further reduced expression in both CDRs ( Figure 5—figure supplement 2B ) , similar to what was observed for affinity . We asked if the sensitivity of the antibody to mutations could be understood from a structural perspective . To quantify sensitivity of affinity and expression at each position i , we computed two quantities: ( 2 ) SKi=⟨ ( log10KDia−log10KDWT ) 2⟩a|i , ( 3 ) SEi=⟨ ( Eia−EWT ) 2⟩a|i . Here , KDWT and EWT respectively denote the dissociation constant and expression level measured for the WT scFv , KDia and Eia denote analogous quantities for the scFv with a single substitution mutation of amino acid a at position i , and ⟨⋅⟩a|i denotes an average computed over the 19 non-WT amino acids at that position . Figure 6A shows the known structure ( Whitlow et al . , 1995 ) of the 1H and 3H variable regions of the WT scFv in complex with fluorescein . Each residue is colored according to the SK and SE values computed for its position . To get a better understanding of what aspects of the structure might govern affinity , we plotted SK values against two other quantities: the number of amino acid contacts made by the WT residue within the antibody structure ( Figure 6B ) , and the distance between the WT residue and the antigen ( Figure 6C ) . We found a strong correlation between SK and the number of contacts , but no significant correlation between SK and distance to antigen . By contrast , SE did not correlate significantly with either of these structural quantities ( Figure 6D and E ) . 10 . 7554/eLife . 23156 . 023Figure 6 . Structural context of mutational effects . ( A ) Crystal structure ( Whitlow et al . , 1995 ) of the CDR1H and CDR3H variable regions of the WT scFv in complex with fluorescein ( green ) . Each residue ( CDR1H: positions 28–37; CDR3H: positions 100–109 ) is colored according to the SK and SE values computed for that position . These variables , SK and SE , respectively quantify the sensitivity of KD and E to amino acid substitutions at each position , with larger values corresponding to greater sensitivity; see Equations 2 and 3 for definitions of these quantities . ( B , C ) For each position in the CDR1H and CDR3H variable regions , SK is plotted against either ( B ) the number of contacts the WT residue makes within the protein structure , or ( C ) the distance of the WT residue to the fluorescein molecule . ( D , E ) Similarly , SE is plotted against either ( D ) the number of contacts or ( E ) the distance to the antigen . R2 is the coefficient of determination . DOI: http://dx . doi . org/10 . 7554/eLife . 23156 . 023
We have described a massively parallel assay , called Tite-Seq , for measuring the sequence-affinity landscape of antibodies . The range of affinities measured in our Tite-Seq experiments ( 10-9 . 5 M to 10-5 . 0 M ) includes a large fraction of the physiological range relevant to affinity maturation ( 10-10 M to ~10−6 M ) ( Batista and Neuberger , 1998; Foote and Eisen , 1995; Roost et al . , 1995 ) . Expanding the measured range of affinities below 10-9 . 5 M might require larger volume labeling reactions , but would be straight-forward . Tite-Seq therefore provides a potentially powerful method for mapping the sequence-affinity trajectories of antibodies during the affinity maturation process , as well as for studying other aspects of the adaptive immune response . The details of our Tite-Seq experiments ( e . g . , 11 antigen concentrations , four sorting bins per concentration , etc . ) were chosen largely for experimental convenience . The effects of varying these parameters have not been systematically explored , and a future investigation of these effects might be valuable . Figure 4—figure supplement 8 does illustrate , via simulation , the effect of read depth on the precision of measured KD values . These simulations , along with an analysis of synonymous variants ( Figure 4—figure supplement 3 ) , suggest that the primary source of noise in our experiments came not from a lack of sorted cells or Illumina reads , but rather from the inefficient post-sort recovery of antibody sequences . We therefore suggest that improvements to our post-sort DNA recovery protocol might substantially improve the resolution of Tite-Seq . Tite-Seq fundamentally differs from prior DMS experiments in that full binding titration curves , not two-bin enrichment statistics , are used to determine binding affinities . The measurement of binding curves provides three major advantages . First , binding curves provide absolute KD values in molar units , not just rank-order affinities , like those provided by SORTCERY ( Reich et al . , 2015 ) , or relative affinity ratios , like those provided by the method of Kowalsky and Whitehead ( 2016 ) . Second , because ligand binding is a sigmoidal function of affinity , DMS experiments performed at a single ligand concentration ( e . g . , [Kowalsky and Whitehead , 2016] ) are insensitive to receptor KDs that differ substantially from this ligand concentration . Binding curves , by contrast , integrate measurements over a wide range of concentrations and are therefore sensitive to a wide range of KDs . The third advantage of measuring binding curves pertains to the fact that protein sequence determines not just ligand-binding affinity , but also the quantity and specific activity of surface-displayed proteins . Our data ( Figure 4—figure supplement 5 and Figure 4—figure supplement 6 ) suggest that these confounding effects can be large and that they can distort yeast display affinity measurements computed from enrichment statistics gathered at a single antigen concentration . Strong sequence-dependent effects on both the expression and specific activity of yeast-displayed proteins has been reported by other groups as well ( e . g . , [Burns et al . , 2014] ) , although the absence of such effects has also been reported ( e . g . , [Kowalsky and Whitehead , 2016] ) . Ultimately , the magnitude of these effects is likely to vary substantially from protein to protein . It should also be noted that many DMS studies using yeast display ( e . g . , epitope mapping studies [Kowalsky et al . , 2015; Doolan and Colby , 2015; Van Blarcom et al . , 2015] ) might not suffer from these potentially confounding effects , and in such cases it probably makes sense to employ a simpler experimental design than is required for Tite-Seq . Nevertheless , either Tite-Seq or other experimental methods that assay full binding curves are probably essential if one wants to quantitatively and reliably measure KD values in a massively parallel fashion . We wish to emphasize , more generally , that changing a protein’s amino acid sequence can be expected to change multiple biochemical properties of that protein . Our work illustrates the importance of designing massively parallel assays that can disentangle these effects . Tite-Seq provides a general solution to this problem for massively parallel studies of protein-ligand binding . Indeed , the Tite-Seq procedure described here can be readily applied to any protein binding assay that is compatible with yeast display and FACS . Many such assays have been developed ( Liu , 2015 ) . We expect that Tite-Seq can also be readily adapted for use with other expression platforms , such as mammalian cell display ( Forsyth et al . , 2013 ) . Our Tite-Seq measurements reveal interesting distinctions between the effects of mutations in the CDR1H and CDR3H regions of the anti-fluorescein scFv antibody studied here . As expected , we found that variation in and around CDR3H had a larger effect on affinity than did variation in and around CDR1H . We also found that CDR1H is more optimized for protein expression than is CDR3H , an unexpected finding that appears to be novel . Yeast display expression levels are known to correlate with thermostability ( Shusta et al . , 1999 ) . Our data is limited in scope , and we remain cautious about generalizing our observations to arbitrary antibody-antigen interactions . Still , this finding suggests the possibility that secondary CDR regions ( such as CDR1H ) might be evolutionarily optimized to help ensure antibody stability , thereby freeing up CDR3H to encode antigen specificity . If this hypothesis holds , it could provide a biochemical rationale for why CDR3H is more likely than CDR1H to be mutated in functioning receptors ( Liberman et al . , 2013 ) and why variation in CDR3H is often sufficient to establish antigen specificity ( Xu and Davis , 2000 ) . Tite-Seq can also potentially shed light on the structural basis for antibody-antigen recognition . By comparing the effects of mutations with the known antibody-fluorescein co-crystal structure ( Whitlow et al . , 1995 ) , we identified a strong correlation between the effect that a position has on affinity and the number of molecular contacts that the residue at that position makes within the antibody . By contrast , no such correlation of expression with this number of contacts is observed . Again , we are cautious about generalizing from observations made on a single antibody . If our observation were to hold for other antibodies , however , it would suggest that the functional geometry of paratopes might be governed by networks of residues whose positions and orientations are strongly interdependent .
Tite-Seq was performed as follows . Variant 3H and 1H regions were generated using microarray-synthesized oligos ( LC Biosciences , Houston TX . USA ) . These were inserted into the 4-4-20 scFv of ( Boder and Wittrup , 1997 ) using cassette-replacement restriction cloning as in ( Kinney et al . , 2010 ) ; see Appendix 3 . Yeast display experiments were performed as previously described ( Boder et al . , 2000 ) with modifications; see Appendix 2 . Sorted cells were regrown and bulk DNA was extracted using standard techniques , and amplicons containing the 1H and 3H variable regions were amplified using PCR and sequenced using the Illumina NextSeq platform; see Appendix 4 . Three replicate experiments were performed on different days . Raw sequencing data has been posted on the Sequence Read Archive under BioProject ID PRJNA344711 . Low-throughput flow cytometry measurements were performed on clones randomly picked from the Tite-Seq library . Sequence data and flow cytometry data were analyzed using custom Python scripts , as described in Appendices 5 and 6 . Processed data and analysis scripts are available at github . com/jbkinney/16_titeseq . | Antibodies are proteins produced by cells of the immune system to tag or neutralize potential threats to the body , such as foreign substances and disease-causing microbes . Antibodies do this by binding to target molecules called antigens . An antibody’s ability to bind to an antigen depends on the sequence of amino acids – the building blocks of proteins – that make up the antibody . Through a process that randomizes this sequence of amino acids , the immune system generates a vast pool of antibodies that are able to target almost any foreign antigen that exists in nature . Currently , little is understood about how the sequence of amino acids in an antibody determines how strongly that antibody binds to its antigen target – a property referred to as the antibody’s binding affinity . Answering this fundamental question requires techniques that can measure the affinities of many different antibodies at the same time . However , previous high-throughput methods have been unable to provide quantitative measurements of binding affinities . These kinds of measurements are difficult because an antibody’s amino acid sequence governs more than just binding affinity: it also affects how easy it is to produce that antibody , and what fraction of antibody molecules work properly . Adams et al . now describe a new method , named “Tite-Seq” , that overcomes these issues . First , thousands of different antibodies are displayed on the surface of yeast cells , with each cell carrying a single kind of antibody . These cells are then incubated with fluorescently labeled antigen at a wide range of different concentrations . Next , the yeast cells are sorted based on how brightly they glow; brighter cells have more antigen bound to them , and so it is possible to calculate how much of the antigen is bound to each kind of antibody at each concentration . Plotting these data provides a “binding curve” for each antibody , which is then used to read off the antibody’s binding affinity in a way that is not affected by the factors that have plagued other high-throughput methods . Tite-Seq is thus able to measure the binding affinities for thousands of different antibodies at the same time . This will potentially allow researchers to address many fundamental and yet unanswered questions about how the immune system works . Tite-Seq can also be used to measure how amino acid sequence affects the binding affinity of proteins other than antibodies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2016 | Measuring the sequence-affinity landscape of antibodies with massively parallel titration curves |
The Ccr4-Not complex removes mRNA poly ( A ) tails to regulate eukaryotic mRNA stability and translation . RNA-binding proteins contribute to specificity by interacting with both Ccr4-Not and target mRNAs , but this is not fully understood . Here , we reconstitute accelerated and selective deadenylation of RNAs containing AU-rich elements ( AREs ) and Pumilio-response elements ( PREs ) . We find that the fission yeast homologues of Tristetraprolin/TTP and Pumilio/Puf ( Zfs1 and Puf3 ) interact with Ccr4-Not via multiple regions within low-complexity sequences , suggestive of a multipartite interface that extends beyond previously defined interactions . Using a two-color assay to simultaneously monitor poly ( A ) tail removal from different RNAs , we demonstrate that Puf3 can distinguish between RNAs of very similar sequence . Analysis of binding kinetics reveals that this is primarily due to differences in dissociation rate constants . Consequently , motif quality is a major determinant of mRNA stability for Puf3 targets in vivo and can be used for the prediction of mRNA targets .
Shortening or removal of mRNA poly ( A ) tails ( deadenylation ) represses gene expression in eukaryotes . The Ccr4-Not complex is a conserved deadenylase that can initiate cytoplasmic mRNA decay and reduce translation by releasing poly ( A ) -binding protein ( Pab1/PABPC1 ) ( Parker , 2012; Tucker et al . , 2001 ) . Ccr4-Not contains seven core subunits , including two poly ( A ) -specific exonucleases , Ccr4/CNOT6/CNOT6L and Caf1/Pop2/CNOT7/CNOT8 ( Wahle and Winkler , 2013 ) . It can be specifically recruited to mRNAs by RNA-binding proteins and the RNA-induced silencing complex ( RISC ) . Ccr4-Not recruitment can also be modulated by codon optimality and covalent modifications of the RNA sequence ( Du et al . , 2016; Eulalio et al . , 2009; Wahle and Winkler , 2013; Webster et al . , 2018 ) . Diverse biological processes depend on deadenylation , including embryogenesis , germline maintenance , cell proliferation , and attenuation of the immune response ( Carballo et al . , 1998; Leppek et al . , 2013; Subtelny et al . , 2014; Yamaji et al . , 2017 ) . RNA-binding proteins such as Pumilio/Puf and Tristetraprolin/TTP can recognize sequence motifs in mRNAs , and also bind Ccr4-Not to target those RNAs for deadenylation ( Wahle and Winkler , 2013 ) . The set of transcripts targeted for repression and decay by a given RNA-binding protein is therefore defined by the distribution of its binding motifs across the transcriptome , and often includes groups of mRNAs that encode functionally related proteins . This generates RNA regulatory networks , structured such that a single protein can regulate the expression of numerous genes ( Joshi et al . , 2011; Keene , 2007; Lapointe et al . , 2017; Wilinski et al . , 2015 ) . Targets of RNA-binding proteins often include factors that regulate gene expression themselves , thereby forming gene expression cascades ( Barckmann and Simonelig , 2013; Lapointe et al . , 2017; Shyu et al . , 1991; Wells et al . , 2012 ) . Identification of the mRNAs within these networks allows a better understanding of gene expression and the roles of RNA-binding proteins that regulate them . Recently , three sequence elements that strongly promote mRNA destabilization were identified in an unbiased screen in zebrafish embryogenesis: the Pumilio-response element ( PRE ) , the AU-rich element ( ARE ) and the miR-430 seed sequence ( Rabani et al . , 2017 ) . PRE- and ARE-mediated mRNA destabilization is conserved from yeast to human but the set of mRNAs that contain these motifs are different across species ( Brooks and Blackshear , 2013; Wells et al . , 2015; Wickens et al . , 2002; Wilinski et al . , 2017 ) . Consequently , the biological processes regulated by orthologous RNA-binding proteins are diverse across organisms . The PRE sequence motif is recognized by a conserved RNA-binding domain found within the Pumilio protein family . A direct interaction between the Pumilio RNA-binding domain and the CNOT7/Caf1 exonuclease has been reported ( Van Etten et al . , 2012 ) , but it is unknown if this is the only site of interaction on Ccr4-Not . The PRE that interacts with human PUM1 and PUM2 ( UGUANAUA ) is found in mRNAs encoding proteins involved in signaling pathways and neuronal processes ( Bohn et al . , 2018 ) . In Saccharomyces cerevisiae ( Sc ) , there are six Pumilio proteins . Of these , ScPuf3 is most closely related to homologs in higher eukaryotes ( Wickens et al . , 2002 ) . The RNA motif recognized by ScPuf3 resembles that of the human PUM proteins but , additionally , cytosine is preferred in the −1 and −2 positions ( Gerber et al . , 2004; Kershaw et al . , 2015; Lapointe et al . , 2015; Wickens et al . , 2002; Zhu et al . , 2009 ) . The closest homolog to human PUM in Schizosaccharomyces pombe ( Puf3 ) also contains residues that define this additional selectivity pocket , and is therefore predicted to bind sequences containing an upstream cytosine ( Qiu et al . , 2012 ) . S . cerevisiae Puf3 is a key regulator of mitochondrial function ( Lee and Tu , 2015; Saint-Georges et al . , 2008 ) . Consistent with this , its mRNA targets encode proteins localized to the mitochondria and involved in the oxidative phosphorylation pathway ( Gerber et al . , 2004; Kershaw et al . , 2015; Lapointe et al . , 2015; Lapointe et al . , 2018 ) . Many transcripts proposed to be regulated by ScPuf3 do not , however , show a clear link to this role ( Kershaw et al . , 2015 ) . It is therefore unclear whether the function of ScPuf3 is limited to mitochondrial regulation . Puf3 homologs in other organisms ( including S . pombe ) do not necessarily regulate mitochondrial function ( Hogan et al . , 2015 ) . The nine-nucleotide ARE motif ( UUAUUUAUU ) was identified in unstable mRNAs encoding cytokines and lymphokines in human cells ( Caput et al . , 1986; Chen and Shyu , 1995 ) . This sequence is recognized by tandem zinc finger ( TZF ) proteins including tristetraprolin ( TTP ) . TTP attenuates immunological responses by binding directly to Ccr4-Not via the CNOT1 and CNOT9 subunits , targeting ARE-containing transcripts for degradation ( Brooks and Blackshear , 2013; Bulbrook et al . , 2018; Fabian et al . , 2013 ) . Phosphorylation of TTP occurs as part of the p38 MAPK pathway , and this disrupts Ccr4-Not recruitment ( Marchese et al . , 2010 ) . The S . pombe Zfs1 protein is homologous to TTP , and recognizes the same RNA motif ( Cuthbertson et al . , 2008; Wells et al . , 2015 ) . An interaction between Zfs1 and Ccr4-Not has not been characterized in fission yeast and the Ccr4-Not-interacting amino acid sequences of TTP are not clearly conserved in Zfs1 . Understanding the molecular basis of accelerated deadenylation has been limited by the lack of a biochemical system containing purified components that reconstitutes this process . Previous studies have shown that a purified domain of the budding yeast Pumilio protein Mpt5 stimulates the activity of immunoprecipitated Ccr4-Not ( Goldstrohm et al . , 2006 ) , and that isolated Caf1 is stimulated by addition of purified BTG2 and PABPC1 ( Stupfler et al . , 2016 ) . We recently purified the complete seven-subunit S . pombe Ccr4-Not complex after overexpression of the subunits in insect cells ( Stowell et al . , 2016 ) . Biochemical assays revealed that this recombinant complex was substantially more active than the isolated nuclease enzymes ( Stowell et al . , 2016; Webster et al . , 2018 ) . Co-expression of the conserved subunits of Ccr4-Not with Mmi1 , an RNA-binding protein found in fission yeast , generated a complex that deadenylated Mmi1-target RNAs more rapidly than non-target RNAs ( Stowell et al . , 2016 ) . Here , we reconstitute accelerated and selective deadenylation of PRE- and ARE-containing RNAs using recombinant S . pombe proteins . We find that Puf3 and Zfs1 act as molecular tethers capable of inducing accelerated and RNA-selective deadenylation by Ccr4-Not in vitro . Biochemical and biophysical analyses of Puf3 binding to RNA reveal a high degree of sequence selectivity . Correspondingly , in S . cerevisiae RNA motif quality is a critical determinant of the RNAs stably bound by Puf3 in vivo . Collectively , our findings show that a substantially improved understanding of RNA-binding protein regulatory networks can be obtained through detailed analysis of motif quality .
To characterize substrate-selective deadenylation in the presence of RNA-binding proteins , we reconstituted this process using purified proteins . Full-length Zfs1 and Puf3 were expressed recombinantly as N-terminal maltose-binding protein ( MBP ) fusion proteins , and intact recombinant seven-subunit S . pombe Ccr4-Not was purified as described previously ( Stowell et al . , 2016 ) . Pull-down assays revealed that Puf3 and Zfs1 interact with isolated Ccr4-Not ( Figure 1—figure supplement 1A ) . Therefore , like metazoan homologs , S . pombe Puf3 and Zfs1 bind directly to Ccr4-Not . By using a short model RNA substrate , deadenylation can be visualized at single-nucleotide resolution , and can be accurately quantified ( Webster et al . , 2017 ) . To investigate whether Puf3 and Zfs1 stimulate the deadenylation activity of Ccr4-Not in vitro , we designed model RNA substrates containing the sequence motif recognized by either Puf3 or Zfs1 ( PRE UGUAAAUA , or ARE UUAUUUAUU respectively , Figure 1—figure supplement 1B ) . These motifs were embedded within a synthetic 23-nucleotide sequence , upstream of a 30-adenosine poly ( A ) tail . Electrophoretic mobility shift assays confirmed that Puf3 and Zfs1 bind stably to their respective target RNAs ( Figure 1—figure supplement 1C ) . The 1:1 stoichiometric protein-RNA complexes formed under these conditions were used as substrates in assays measuring the deadenylation activity of Ccr4-Not . Ccr4-Not removed the poly ( A ) tails of both Puf3 and Zfs1 substrate RNAs in vitro . Puf3 and Zfs1 each stimulated deadenylation of their target RNAs: In the absence of Puf3 , complete removal of the poly ( A ) tail from PRE-containing RNA occurred in approximately 32 min , whereas in the presence of Puf3 the reaction was complete in less than 2 min ( Figure 1A and Figure 1—figure supplement 1D–E ) . Therefore , addition of Puf3 increased the rate of deadenylation more than 20-fold . Zfs1 similarly increased the rate of deadenylation of its target RNA to less than 2 min ( Figure 1B and Figure 1—figure supplement 1D–E ) . Thus , both Puf3 and Zfs1 have striking effects on the rate of deadenylation . Non-adenosine residues upstream of the poly ( A ) tail were also rapidly removed by Ccr4-Not when Puf3 or Zfs1 were present . The exonucleases stop when they reach nucleotides that correspond to the binding site of the protein , suggesting that either the RNA-binding proteins cannot be released by Ccr4-Not , or that deadenylation becomes very slow after release of the RNA binding protein . This is consistent with a model in which the specificity of Ccr4-Not for adenosine is substantially less evident when it is tethered to the RNA substrate for a prolonged period of time ( Finoux and Séraphin , 2006 ) . In addition , the deadenylation reaction appeared to be more processive in the presence of Puf3 or Zfs1 because RNA lacking a poly ( A ) tail was visible when RNA with an intact tail was still present in the reaction ( Figure 1—figure supplement 1F ) . This indicates that the RNA-binding protein increases the stability of the interaction of Ccr4-Not with the target RNA substrate during the course of the reaction . In the absence of additional protein , Ccr4-Not is approximately 3-fold more active on the ARE-containing substrate than on the PRE-containing substrate ( Figure 1 and Figure 1—figure supplement 1E ) . It is generally believed that additional protein factors confer mRNA-selectivity to Ccr4-Not , but this finding indicates Ccr4-Not may possess a moderate degree of intrinsic mRNA selectivity . We previously observed that Ccr4-Not was less active on RNA substrates with secondary structure in the upstream 3ʹ-UTR ( Stowell et al . , 2016 ) . The free energies of the predicted most stable structures calculated with RNAfold software are −1 . 2 kcal/mol ( Puf3-target RNA ) and −0 . 2 kcal/mol ( Zfs1-target RNA ) . Thus , neither of the substrates tested here is predicted to form stable RNA secondary structure . Our data suggest that the rate at which different mRNAs are deadenylated in vivo may be modulated by sequence-selectivity intrinsic to the Ccr4-Not complex . As mentioned above , specific interactions between Ccr4-Not and Pumilio/TTP proteins had been previously described ( Fabian et al . , 2013; Goldstrohm et al . , 2006; Van Etten et al . , 2012 ) . To dissect the mechanism of accelerated deadenylation and to determine the contributions of different parts of Puf3 and Zfs1 , we made a series of mutations and truncations in these proteins , and tested their abilities to stimulate deadenylation . First , we tested whether the RNA-binding domains alone ( Puf3PUM and Zfs1TZF , Figure 2—figure supplement 1A–B ) affected deadenylation . Addition of Puf3PUM caused a modest acceleration ( complete deadenylation in ~20 min compared to ~32 min; Figure 2A and Figure 2—figure supplement 1C , D ) . This is consistent with a reported interaction between the Pumilio domain and the Caf1/CNOT7 subunit of Ccr4-Not , that tethers RNA directly to the nuclease subunit to accelerate deadenylation ( Goldstrohm et al . , 2006; Van Etten et al . , 2012 ) . Since stimulation by Puf3PUM was much less than with full-length Puf3 , regions outside the Pumilio domain likely play an important role in the recruitment of Ccr4-Not . Interestingly , Zfs1TZF was inhibitory to deadenylation ( Figure 2A and Figure 2—figure supplement 1C ) . This may be because this domain conceals sequence elements that confer the higher level of intrinsic activity on the ARE-containing RNA . To confirm that RNA binding is essential to the function of Puf3 and Zfs1 , we generated full-length proteins with mutations in the RNA-binding interface ( Puf3Mut: Y449A/Y671A and Zfs1Mut: F349A/F387A ) . Compared to the wild-type proteins , these mutants had a substantially reduced ability to bind their substrate RNAs , and to accelerate deadenylation ( Figure 2A and Figure 2—figure supplement 1B–E ) . Together , these data show that the strong stimulatory effect of Puf3 and Zfs1 depends on RNA binding , but the RNA-binding domains alone are not sufficient . In addition to a C-terminal RNA-binding domain , both Puf3 and Zfs1 contain an extended N-terminal region of low amino acid sequence complexity that is predicted to be predominantly disordered ( Figure 2B ) . We hypothesized that short sequence motifs within the low-complexity regions of Puf3 and Zfs1 interact with Ccr4-Not , as has been demonstrated for other interaction partners ( Bhandari et al . , 2014; Fabian et al . , 2013; Raisch et al . , 2016 ) . Thus , we generated a series of N-terminal truncations and measured their effects on the deadenylation activity of Ccr4-Not ( Figure 2B and Figure 2—figure supplement 1F–G ) . For both proteins , no single truncation accounted for the stimulatory effect on deadenylation . Instead , as the length of low-complexity sequence was reduced , the stimulatory effect on the deadenylation rate was also reduced . It is therefore likely that multiple sequence elements , distributed throughout the low-complexity regions , interact with Ccr4-Not . Two of the seven core subunits of Ccr4-Not are poly ( A ) -selective exonucleases , Caf1 and Ccr4 . To determine whether Puf3 and Zfs1 can stimulate deadenylation in the absence of the non-enzymatic subunits , we performed reactions with a purified dimeric subcomplex comprised of Caf1 and Ccr4 ( Stowell et al . , 2016 ) . Puf3 accelerated the rate of deadenylation by Caf1-Ccr4 substantially less than it accelerated intact Ccr4-Not ( 2-fold and >20 fold respectively , Figure 2C left compared to 1A ) . Fully-deadenylated and non-deadenylated RNA was observed simultaneously during both reactions , suggestive of a processive reaction . Furthermore , non-adenosine nucleotides were also readily removed by the Caf1-Ccr4 complex . This suggests that Puf3 interacts with Caf1-Ccr4 , but association occurs more slowly when these subunits are not incorporated into the intact Ccr4-Not complex . Zfs1 did not substantially affect the deadenylation activity of the Caf1-Ccr4 dimeric complex ( Figure 2C , right ) . Thus , our data are consistent with Puf3 and Zfs1 recruiting the Ccr4-Not complex primarily through an interaction between their low-complexity regions and the non-enzymatic subunits of Ccr4-Not . Poly ( A ) -binding protein ( Pab1 ) contributes to translation initiation , and its release from mRNAs can repress translation ( Parker , 2012 ) . Ccr4-Not efficiently releases Pab1 from the poly ( A ) tail through removal of the nucleotides to which it is bound ( Webster et al . , 2018 ) . It is not known , however , how Pab1 influences accelerated deadenylation . We therefore examined deadenylation of Pab1-bound RNA in the presence of Puf3 . Electrophoretic mobility shift assays showed that stoichiometric amounts of Puf3 and Pab1 could be simultaneously loaded onto the RNA substrate ( Figure 2—figure supplement 2A ) . In deadenylation assays , the addition of Puf3 increased the rate at which the Pab1-bound poly ( A ) tail was removed by ~8 fold ( Figure 2D and Figure 2—figure supplement 2B ) . Therefore , Puf3 can promote release of Pab1 by accelerating deadenylation by Ccr4-Not . Still , Pab1 slows the overall rate of Puf3-accelerated deadenylation , suggesting that the rate of Pab1 release limits the very fast rate of deadenylation that occurs when Ccr4-Not is recruited by Puf3 . We previously reported that Caf1 and Ccr4 have similar activities within the purified Ccr4-Not complex ( Stowell et al . , 2016 ) . More recently , we showed that the presence of Pab1 differentiates the activity of Ccr4 and Caf1: only Ccr4 removes adenosines bound by Pab1 ( Webster et al . , 2018 ) . It is not known , however , whether Caf1 and Ccr4 are differentially active when RNA-binding proteins recruit Ccr4-Not to specific mRNAs . We tested the contribution of Ccr4 and Caf1 to Puf3-accelerated deadenylation using purified Ccr4-Not variants with point mutants that abolish the catalytic activity of either nuclease . Using our in vitro assays , Puf3 accelerated deadenylation by both Ccr4-inactive and Caf1-inactive complexes ( Figure 2—figure supplement 2C ) . This indicates that deadenylation in the presence of Puf3 is mediated by either nuclease . In the presence of Puf3 and Pab1 , Ccr4-inactive Ccr4-Not did not proceed beyond ~20 adenosines ( Figure 2—figure supplement 2C ) . This is similar to when Puf3 was not present ( Webster et al . , 2018 ) . Thus , Pab1 , and not Puf3 , differentiates the two nucleases of Ccr4-Not . Having established that Puf3 and Zfs1 are capable of accelerating Ccr4-Not activity , we sought to test whether they are also sufficient for substrate-selectivity . We began by examining whether the presence of the target motif recognized by each protein was essential for accelerated deadenylation . The effect of Puf3 and Zfs1 on its own target RNA was compared with its effect on a non-target RNA . For the latter , we used the RNA target of the other protein ( Puf3 with Zfs1-target RNA , and Zfs1 with Puf3-target RNA ) . We were surprised to find that each protein stimulated deadenylation of an RNA that did not contain its target motif ( Figure 3A ) . Quantification of deadenylation rates showed that on-target deadenylation was approximately 3–4 fold faster than off-target deadenylation , which was in turn 2–5 fold faster than when no additional protein was present ( Figure 3—figure supplement 1A ) . Hence , the correct target RNA motif contributes to the ability of an RNA-binding protein to accelerate deadenylation in vitro , but is not absolutely required . Since Puf3 with a mutated RNA-binding domain still stimulates deadenylation , allosteric activation of Ccr4-Not by the Puf3 low complexity region may contribute to this accelerated deadenylation of off-target RNAs ( Figure 2—figure supplement 1D ) . However , mutant Zfs1 that does not bind RNA doesn’t accelerate deadenylation . Thus , allosteric activation is not likely to be a general mechanism . Instead , we hypothesized that off-target stimulation of deadenylation in our in vitro experiments is due to off-target RNA binding . To determine the binding affinities of full-length Puf3 and Zfs1 for the RNA substrates , we performed fluorescence polarization RNA binding assays with RNAs containing the target motif of Puf3 or Zfs1 , as well as 30-mer polyadenosine ( Figure 3B ) . Each RNA-binding protein bound to its respective target RNA sequence with high affinity ( KD = 1–2 nM ) . In contrast , Puf3 bound to the off-target ARE sequence with KD of 120 nM , while Zfs1 bound the off-target PRE sequence with KD of 850 nM . In both cases , poly ( A ) was bound with ~100 fold lower affinity than the target RNA . Hence , both Puf3 and Zfs1 are selective in RNA binding , with >100 fold higher affinity for their target RNA than off-target RNAs . Still , Puf3 and Zfs1 bind off-target sequences with nanomolar affinities and this could account for their abilities to stimulate deadenylation of off-target RNAs . We hypothesized that Puf3 and Zfs1 would select their targets from a mixture of target and non-target RNAs to promote targeted deadenylation . To test this , we developed a system to directly examine selective deadenylation by Ccr4-Not . Two polyadenylated RNA substrates labelled with different fluorophores were combined in stoichiometric quantities before the addition of Ccr4-Not and analysis of deadenylation activity . The same gel could therefore be used to monitor deadenylation of both the fluorescein-labelled Puf3-target substrate , and the Alexa647-labelled Zfs1-target substrate , displayed in blue and red respectively ( Figure 3C ) . In this two-color assay , the Zfs1-target RNA was deadenylated by Ccr4-Not 3-fold faster than the Puf3-target RNA in the absence of additional protein . This is consistent with intrinsic selectivity of Ccr4-Not that was observed when each RNA was examined in isolation ( Figure 1 ) . We then performed a series of reactions using three different concentrations of Puf3 or Zfs1 , and each RNA at 100 nM . With 50 nM Puf3 or Zfs1 , deadenylation was accelerated only on the target RNA ( Figure 3D , top; blue for Puf3-target and red for Zfs1-target ) . A fraction of the target RNA was unaffected , consistent with incomplete occupancy of the RNA-binding protein on its target . With 250 nM Puf3 or Zfs1 , deadenylation of all target RNA was substantially increased while deadenylation of the non-target RNA was unchanged ( Figure 3D , middle ) . Thus , Puf3 and Zfs1 promote selective , accelerated deadenylation . Selectivity was lost when an excess of the RNA-binding protein was added ( 1000 nM ) and all RNA was deadenylated rapidly ( Figure 3D , bottom ) . To examine RNA binding by Puf3 and Zfs1 under the same conditions , we performed electrophoretic mobility shift assays . Consistent with deadenylation assays , target RNA binding is complete and selective at 250 nM , but is non-selective at 1000 nM ( Figure 3—figure supplement 1B ) . In conclusion , Puf3 and Zfs1 are sufficient for substrate-selective deadenylation by Ccr4-Not , within a given concentration window . At higher concentrations , they also bound to and accelerated deadenylation on RNAs that do not contain their target motif . In the cell , RNAs with very similar sequences are present but RNA-binding proteins distinguish between these to act preferentially on specific targets . Yeasts possess multiple Pumilio proteins , and these recognize RNA motifs similar in sequence ( Wickens et al . , 2002 ) . For example , an RNA target of S . cerevisiae ( Sc ) Puf3 differs from a target of ScPuf5/Mpt5 in two nucleotide positions within a core 8-nt motif ( Goldstrohm et al . , 2006 ) ( Figure 4A , top ) . Given that Puf3 binds off-target RNAs at higher concentrations ( Figure 3—figure supplement 1B ) , we investigated whether the fission yeast Puf3 distinguishes between the very similar motifs of ScPuf3 and ScPuf5 in vitro . We used fluorescence polarization experiments to examine RNA binding and found that S . pombe Puf3PUM binds to both the ScPuf3 target and the ScPuf5 target with high affinity ( Figure 4A , bottom ) . The affinity for the ScPuf3 motif is ~2 fold higher than for the ScPuf5 motif , but the binding affinity in both cases is in the picomolar range . Surprisingly , in deadenylation activity assays with 100 nM Puf3 , we found that the ScPuf3 target was selectively deadenylated in preference to the ScPuf5 target ( Figure 4B ) . Thus , Puf3 can distinguish between very similar RNAs and selectively target them for deadenylation by Ccr4-Not . Selective deadenylation only occurred within a narrow range of Puf3 concentrations where the quantity of Puf3 protein is less than or equal to the quantity of ScPuf3-target RNA ( Figure 4B , C ) . With less Puf3 ( 50 nM Puf3 ) , not all ScPuf3-target RNA was rapidly deadenylated . Conversely , when a molar excess of Puf3 was present ( 200 nM Puf3 ) , both the ScPuf3- and ScPuf5-target RNAs were deadenylated rapidly . We performed RNA EMSAs and found that the ability of Puf3 to stimulate deadenylation correlated with RNA binding: binding to ScPuf5-target RNA occurred only when Puf3 protein was in molar excess to the ScPuf3-target RNA ( Figure 4D ) . Complete and selective binding by Puf3 , and selective deadenylation by Ccr4-Not , therefore only occurs within a narrow window of Puf3 protein concentration . We next examined whether Puf3 can distinguish between sequences that are even more similar . Within the 8-nucleotide motif recognized by ScPuf3 , the fifth position is the most variable , and A or U are thought to be tolerated ( Kershaw et al . , 2015 ) . This is a property shared by the human homologs PUM1 and PUM2 ( Galgano et al . , 2008; Morris et al . , 2008 ) . In the previous experiments , the RNA motif contained A at position 5 ( now called ScPuf3-target ( 5A ) ) . We tested whether there is a preference for A or U at position five using the two-color deadenylation assay and found that the ScPuf3-target ( 5A ) RNA was deadenylated in preference to ScPuf3-target ( 5U ) RNA ( Figure 4E ) . Thus , Puf3 has a remarkable capacity to distinguish between sequences within the current definition of its target motif . The identity of the nucleotide in position five likely affects ‘motif quality’ , and Puf3 interacts with sequences of different quality in a priority-based manner . To gain insight into the selective RNA binding by Puf3 that allows it to distinguish between RNAs of very similar sequences , we analyzed the binding kinetics using SwitchSENSE ( Knezevic et al . , 2012 ) . Switchable nanolevers were functionalized with RNAs containing each of the three RNA motifs tested in the deadenylation reactions: ScPuf3-target ( 5A ) , ScPuf3-target ( 5U ) , and ScPuf5-target . We measured the rates of association ( kon ) and dissociation ( koff ) of fission yeast Puf3PUM binding to each of these , as well as to polyadenosine as a negative control . Rates of association followed the same pattern of priority binding that we had identified by deadenylation assays and EMSA ( Figure 5A ) : Association with the ScPuf3-target ( 5A ) was ~25% faster than with ScPuf3-target ( 5U ) and ~three fold faster than with the ScPuf5-target . Association with polyadenosine was even slower ( ~seven fold ) . Thus , RNA sequence affects the rate of Puf3 association . Importantly , however , all sequences were bound rapidly: at a concentration of 20 nM , binding to all RNAs took place on the timescale of seconds ( Figure 5A , right ) . When we examined rates of dissociation , we found that Puf3 remained bound to the ScPuf3-target with A at position five for twice as long as it remained bound to the RNA with U at position 5 , and ~three times as long as it was bound to the ScPuf5-target RNA ( Figure 5B ) . Dissociation from polyadenosine occurred more than 100-fold faster . Thus , rates of Puf3 dissociation also reflected the preference order that we had observed in deadenylation experiments . Combining our association and dissociation measurements , sequence-selective binding by Puf3 is expected to occur by the combination of rapid sampling of all RNAs with slower dissociation from sequences with higher motif quality . The ability of Puf3 to discriminate between similar sequences is likely to play a role in defining the mRNAs within its regulatory network . To examine the contribution of RNA motif quality to Puf3 binding in vivo , we analyzed several published studies of S . cerevisiae Puf3 . The mRNAs that contain Puf3 motifs are different between species , but we expected motif quality to be similarly important in all organisms ( Hogan et al . , 2015 ) . Thus , we applied our in vitro findings using S . pombe proteins to a computational analysis of S . cerevisiae Puf3 . We compiled a list of 1331 putative ScPuf3 mRNA targets from three published sources . The first study of transcriptome-wide ScPuf3 binding identified 220 targets by RNA immunoprecipitation followed by microarray analysis ( RIP-chip ) ( Gerber et al . , 2004 ) . A more recent study using RNA immunoprecipitation followed by sequencing ( RIP-seq ) identified 1132 significantly enriched transcripts ( Kershaw et al . , 2015 ) : the larger number of targets identified was attributed to the improved sensitivity of deep sequencing . A third study detected in vivo binding of ScPuf3 to 476 mRNAs using an RNA-tagging method ( Lapointe et al . , 2015 ) . While there is substantial overlap between the binding targets from each list , 74% of mRNAs were identified in only one of the three studies ( Figure 6A ) . The set of mRNAs that are direct targets of ScPuf3 regulation therefore remains unclear . We began by computationally identifying the most enriched sequence motif in the complete set of 1331 mRNAs using MEME ( Bailey et al . , 2015 ) . This had seven conserved nucleotides within an 8-nt motif ( UGUANAUA ) , and was present in 627 mRNAs ( 41% of total ) ( Figure 6B , top ) . The nucleotide at position five is variable , consistent with previous analyses of transcriptome-wide data ( Gerber et al . , 2004; Kershaw et al . , 2015 ) . The best-characterized ScPuf3 binding site is within the COX17 mRNA ( Olivas and Parker , 2000 ) . This mRNA was also the most highly enriched following RNA-immunoprecipitation ( Kershaw et al . , 2015 ) . Interestingly , it has sequence features outside the seven core nucleotides that contribute to the particularly high affinity binding of ScPuf3: cytosine at position –2 and adenosine at position +5 ( Zhu et al . , 2009 ) . Having identified that fission yeast Puf3 has the capacity to distinguish between similar sequences , we hypothesized that these additional sequence features might provide a better definition of ScPuf3 targets transcriptome-wide . Thus , we categorized target mRNAs by the presence of a ScPuf3-target motif as well as the nucleotide identity at positions –two and +5 . We ranked these four groups by motif quality based on previously measured affinity values ( Figure 6B and Supplementary file 1 ) ( Zhu et al . , 2009 ) . Many of the predicted ScPuf3-target transcripts that do not fit into one of these four groups do not contain any of these sequences but have a motif in which one of the core seven nucleotides are altered ( denoted here as ‘Δcore’ ) . mRNAs that increase in steady-state abundance upon deletion of PUF3 represent high-confidence targets of ScPuf3-mediated mRNA decay . 82 such mRNAs were identified previously ( Kershaw et al . , 2015 ) , and we found that motif 1 , 2 or 3 is present in 73 of these ( 89% ) ( Figure 6C and Supplementary file 1 ) . Furthermore , in 71 of the mRNAs that are increased in abundance upon PUF3 deletion , the motif is located within the 3ʹ UTR . Hence , the quality and the location of ScPuf3 motifs both contribute to mRNA-destabilization . We examined whether motif quality is correlated with the amount of ScPuf3 bound to mRNAs in vivo and found that transcripts that were most enriched in ScPuf3 immunoprecipitation experiments ( Kershaw et al . , 2015 ) had motif 1 , 2 or 3 located in their 3ʹ-UTR ( Figure 6D and Supplementary file 1 ) . In contrast , transcripts containing a group 4 motif , or a mismatch within the core seven nucleotides , generally had less ScPuf3 bound . Interestingly , mRNAs with high quality motifs located within the coding sequence were detected but were consistently bound by less ScPuf3 . Whether this reflects a lower occupancy of ScPuf3 , or a reduced stability of the interaction , is unknown . In summary , RNA sequence is an excellent predictor of ScPuf3 binding in vivo . In S . cerevisiae , Puf3 is a regulator of mitochondrial function ( Lee and Tu , 2015; Saint-Georges et al . , 2008 ) . Consistent with this , targets of ScPuf3 include mRNAs that encode proteins localized to the mitochondria and factors involved in the oxidative phosphorylation pathway ( Gerber et al . , 2004; Kershaw et al . , 2015; Lapointe et al . , 2015; Lapointe et al . , 2018 ) . Many transcripts identified as targets , however , show no clear link to this role ( Kershaw et al . , 2015 ) . An important question therefore remains whether the role of ScPuf3 is limited to mitochondrial regulation . We examined the link between the quality of ScPuf3-target motifs in mRNAs and the functions of the proteins they encode . A more detailed examination of gene function revealed that of the functionally annotated mRNAs containing motifs 1 , 2 or 3 in their 3ʹ UTR , 245 of 278 ( 88% ) were related to mitochondrial biogenesis ( Figure 6E and Supplementary file 2 ) . This includes 74 of the 80 subunits of the mitochondrial ribosome ( Bieri et al . , 2018 ) , but none of the cytoplasmic ribosome . Furthermore , the 14 mitochondrion-specific tRNA aminoacylation enzymes , but none of the five that are both mitochondrial and cytoplasmic , have a high quality ScPuf3 motif . Interestingly , transcripts encoding proteins that assemble the respiratory chain complexes , but not nuclear-encoded subunits of the complexes subunits themselves , are ScPuf3 targets . Most components of the mitochondrial protein import complexes TIM22 , TIM23 and TOM; mitochondrial chaperones; and membrane insertion factors are also targets . In contrast , mRNAs containing group four motif or Δcore were not generally related to mitochondrial function . Compared to previous lists of ScPuf3 target candidates , analysis of motif quality yields a list of mRNA targets that is shorter and more functionally coherent ( Figure 6—figure supplement 1A ) . To assess the limitations of predicting RNA-binding protein targets directly from mRNA sequence motif we examined ScPuf3 binding and the sequence features of transcripts that had motif 1 , 2 or 3 in the 3ʹ-UTR but encoded proteins not evidently related to mitochondrial biogenesis ( 33 in total ) , as well as mRNAs related to mitochondrial biogenesis but with motif 4 in the 3ʹ-UTR ( 33 transcripts ) . These represent possible false-positives and false-negative predicted targets respectively . mRNAs with motif 1 , 2 or 3 that were not related to mitochondrial biogenesis had significantly less ScPuf3 association than those that were related to mitochondria ( Figure 6—figure supplement 1B ) . This suggests that despite the presence of a high-quality motif , ScPuf3 binding is less enriched . Similarly , mRNAs with motif 4 that were related to mitochondrial biogenesis had higher levels of ScPuf3 , suggesting that despite sequences with lower motif quality , ScPuf3 is selectively bound . Analysis of sequences surrounding the ScPuf3-target motif revealed no consistent differences between functional groups ( Figure 6—figure supplement 1C ) . Therefore , factors other than motif quality also influence ScPuf3 binding and this is correlated with encoded function .
RNA-binding proteins that interact with the mRNA decay machinery typically contain a canonical RNA-binding domain , and regions of low sequence complexity that are predicted to be structurally disordered ( Jonas and Izaurralde , 2013 ) . Short linear motifs within these regions can adopt secondary structure , and this may facilitate interaction with other proteins , including Ccr4-Not ( Bhandari et al . , 2014; Fabian et al . , 2013; Raisch et al . , 2016; Sgromo et al . , 2017; Stowell et al . , 2016 ) . Recent work showed that Drosophila Nanos utilizes two short linear motifs to contact the NOT1 and NOT3 subunits of Ccr4-Not ( Raisch et al . , 2016 ) . Likewise , the N- and C-terminal regions of human TTP interact with the CNOT1 and the CNOT9 subunits respectively ( Bulbrook et al . , 2018; Fabian et al . , 2013; Lykke-Andersen and Wagner , 2005 ) . We show that multiple regions within fission yeast Puf3 and Zfs1 contribute to Ccr4-Not recruitment , and these are distributed throughout the 300–400 amino acid region that is predicted to be predominantly disordered . Thus , the multi-partite mode of interaction for Puf3 and Zfs1 may be a general feature of RNA-binding proteins that interact with Ccr4-Not . The in vitro deadenylation assays described here represent a powerful tool for testing the importance of particular sequence regions in interaction partners and evaluating their redundancy . The Pumilio family of RNA-binding proteins appeared to represent an exception to the pattern that Ccr4-Not recruitment occurs exclusively via short linear motifs within intrinsically disordered regions . The RNA-binding Pumilio domain of human PUM1 and S . cerevisiae Puf4 and Puf5 interacts directly with Caf1/CNOT8 ( Goldstrohm et al . , 2006 ) . Consistent with this , in our study the S . pombe Puf3 PUM domain stimulated the rate of deadenylation by Ccr4-Not ( Figure 2—figure supplement 1C ) , and full-length Puf3 stimulated deadenylation by a Caf1-Ccr4 subcomplex ( Figure 2C ) . Importantly , however , both of these effects were weak relative to the stimulation of intact Ccr4-Not by full-length Puf3 ( Figure 1A ) . Thus , the low-complexity region of S . pombe Puf3 and the non-enzymatic subunits of Ccr4-Not play important roles in the interaction . The Pumilio proteins are therefore likely similar to other Ccr4-Not interaction partners . Intrinsically disordered regions are common sites of post-translational modification such as phosphorylation . TTP is a phosphorylation target in the p38 MAPK pathway , and this disrupts the recruitment of Ccr4-Not ( Marchese et al . , 2010 ) . Similarly , phosphorylation occurs throughout the low-complexity region of S . cerevisiae Puf3 , and this stops the rapid decay of transcripts to which it is bound ( Lee and Tu , 2015 ) . Given that the low-complexity region is involved in the recruitment of Ccr4-Not , it is likely that phosphorylation regulates this interaction . Rates of mRNA decay may be tuned by the extent of phosphorylation within the extended interface that contacts Ccr4-Not . The correct recognition of mRNA targets by RNA-binding proteins is critical for recruitment of effector complexes , such as Ccr4-Not . We found that Puf3 and Zfs1 stimulate deadenylation of off-target RNAs , but they can discriminate between mixtures of RNA substrates to select their target . Titration of Puf3 under conditions where two similar RNA sequences were present showed that binding of RNA occurs by a model of priority: Puf3 binds to an optimal target when it is available , but after saturation of all available sites , it will readily associate with a sub-optimal target ( Figure 7B ) . This model of RNA target selection , where an RNA-binding protein has a distribution of affinities for different RNA sequences , challenges the model where there is a sharp distinction between affinities for specific and non-specific RNAs ( Jankowsky and Harris , 2015 ) . The relative abundance of RNA-binding proteins and RNA therefore determines occupancy , and is integral to the definition of the target motif . We propose that tight control of RNA-binding protein abundance in the cell is essential for maintenance of a stable set of target mRNAs . Interestingly , the mRNA encoding Puf3 in S . cerevisiae is itself a target of Puf3 ( Supplementary file 1 ) . This predicts a potential feedback mechanism by which ScPuf3 protein abundance is auto-regulated . Furthermore , the ScPuf3-recognition motif within the ScPuf3 mRNA is within group 4 ( Figure 6B ) and this is at the boundary that separates targets from non-targets . The ScPuf3 mRNA may only become destabilized when ScPuf3 protein abundance becomes high enough that it begins to destabilize mRNAs containing motifs of lower quality that are not related to mitochondrial biogenesis . Increasing the level of human PUM1 by overexpression causes a number of mRNAs to be repressed , indicative of off-target RNA binding , and PUM1 availability may be regulated by a long noncoding RNA ( Lee et al . , 2016; Tichon et al . , 2018 ) . In mice , PUM1 is expressed constitutively , but heterozygotes that contain only a single gene copy display reduced PUM1 protein levels ( Gennarino et al . , 2015 ) . This haploinsufficiency produces severely impaired neuronal function that is due , at least in part , to impaired destabilization of the ATAXIN1 mRNA . Thus , consistent with our model of RNA binding , the dosage of RNA-binding proteins is also important to maintain on-target mRNA expression . Even when the consensus motif recognized by an RNA-binding protein is known , it is often difficult to predict which transcripts are targeted by that protein in vivo: A high affinity interaction is not sufficient for the sequence to be a preferred target , as is exemplified by the tight binding of fission yeast Puf3 to the RNA motif recognized by ScPuf5 ( Figure 4A ) . Our kinetic analysis of Puf3 binding to RNA revealed a fast rate of association with RNA that does not contain the optimal sequence motif . We conclude that Puf3-RNA dissociation rates are likely major contributors to the rates of deadenylation . In vivo , Puf3 likely samples numerous mRNAs before remaining stably associated with an mRNA containing an optimal sequence motif . An important consequence of this is that RNA-immunoprecipitation ( and especially use of covalent crosslinking ) will detect mRNAs that are not genuine targets . An indication of in vivo dissociation rates of ScPuf3 were previously obtained using RNA tagging ( Lapointe et al . , 2015 ) . In this method , an enzyme that adds uridines to the 3′ end of RNA is attached to the RNA-binding protein of interest . The length of the U-tail on an mRNA correlated with whether the mRNAs are regulated by ScPuf3 in the cell and is also expected to correlate with the duration the RNA-binding protein is associated with it ( Lapointe et al . , 2015 ) . This study differentiated between Puf3 ‘sampling’ of RNAs where the U-tails were short and the off-rate was too fast to allow regulation ( deadenylation ) , and Puf3 ‘regulation’ where the U-tails were longer , the off-rate was slower , and the interaction leads to a productive output ( deadenylation ) . In addition to cellular concentration of RNA binding proteins and their targets , competition among RNA binding proteins with similar sequence preferences will also influence this kinetic model for productive interaction ( Lapointe et al . , 2017 ) . Taken together , there is a striking convergence between the results of in vitro ( this study ) and in vivo ( Lapointe et al . , 2015 ) studies . Our analysis also correlates well with a more recent study that combined multiple methods to identify 91 Puf3 targets ( Lapointe et al . , 2018 ) . A major limitation in all experimental approaches is that the threshold separating genuine from non-genuine targets is difficult to define . Our analysis of the mRNA targets of S . cerevisiae Puf3 demonstrates , however , that accurate prediction may be possible from examination of the 3ʹ-UTR sequence . If , as our data suggests , the function of S . cerevisiae Puf3 is limited to mitochondrial biogenesis , ~88% of mRNAs with a high-quality motif are genuine targets . The workflow we have used should therefore be informative in defining the regulatory network of other RNA-binding proteins ( Figure 7C ) . How specific mRNAs can be rapidly and specifically targeted for degradation in response to cellular signals remains a major question in understanding eukaryotic gene expression . Our priority model for RNA binding to an adapter protein that acts as a tether to the deadenylation machinery provides new mechanistic insight and brings us closer to being able to predict RNA stability from sequence .
Intact Ccr4-Not complex was purified as described previously after overexpression of the seven core subunits of the Schizosaccharomyces pombe complex ( Ccr4 , Caf1 , Not1 , Not2 , Not3 , Not4/Mot2 and Rcd1/Caf40 ) in Sf9 cells ( Stowell et al . , 2016 ) . The Caf1-Ccr4 heterodimeric complex was prepared as described previously from the Sf9 lysate used for Ccr4-Not expression ( Webster et al . , 2018 ) . S . pombe Pab1 ( amino acids 80–653 ) was prepared as described previously ( Webster et al . , 2018 ) . DNA sequences encoding full-length S . pombe Puf3 and Zfs1 were synthesized with codon optimization for E . coli ( GenScript ) . Primers Puf3_Fwd and RBP_Rev were used to amplify the Puf3 sequence , and primers Zfs1_Fwd and RBP_Rev were used to amplify the Zfs1 sequence ( Supplementary file 3 ) . DNA was inserted into pMAL-c2X expression vector using an In-Fusion HD Cloning Kit ( Clontech ) . DNA encoding variants with N-terminal truncations were amplified using primers Puf3_res{63 , 125 , 181 , 240 , 300 , 378}_Fwd and RBP_Rev for Puf3 and Zfs1_res{34 , 71 , 182 , 248 , 322}_Fwd and RBP_Rev for Zfs1 . These were also inserted into pMAL-c2X . PCR-based site-directed mutagenesis was performed with a Quikchange Lightning Mutagenesis Kit ( Agilent ) to generate mutations in the Puf3 RNA-binding domain ( Puf3Mut: Y449A/Y671A ) and the Zfs1 RNA-binding domain ( Zfs1Mut: F349A/F387A ) . Primers used to introduce mutations ( Puf3_Quikchange and Zfs1_Quikchange ) are listed in Supplementary file 3 . Sequences of all expression vectors were confirmed by Sanger sequencing with primers pMAL_Seq_Fwd and pMAL_Seq_Rev . Puf3 and Zfs1 were expressed as N-terminal MBP fusions in BL21 Star ( DE3 ) E . coli ( Thermo Fisher Scientific ) . Transformed cells were grown at 37°C to an A600 nm of 0 . 6 before the temperature was reduced to 18°C and protein expression induced by the addition of IPTG to 0 . 5 mM . Growth was continued for 18 hr before cells were harvested by centrifugation and flash frozen for storage at −80°C . Cells were defrosted and lysed by sonication in lysis buffer ( 50 mM Tris-HCl pH 8 , 250 mM KCl , 1 mM TCEP ) supplemented with protease inhibitor cocktail ( Roche ) and DNase I ( 5 μg/ml ) ( Sigma ) . The lysate was cleared by centrifugation at 40 , 000 × g for 30 min at 4°C . Supernatant was applied to amylose affinity resin ( 1 ml bed volume per 2 l culture; New England Biolabs ) , and incubated with rotation for 1 hr at 4°C . Resin was separated from lysate with an Econo-Column ( Bio-Rad ) and washed three times with 50 ml of RBP buffer ( 20 mM Tris-HCl pH 8 , 50 mM NaCl , 0 . 1 mM TCEP ) . Protein was eluted from the resin with RBP buffer supplemented with 50 mM maltose . Sample was loaded onto a HiTrap Q HP 5 ml column ( GE Healthcare ) , and the protein was eluted with a gradient of 20 column volumes into RBP buffer containing 1 M NaCl . Protein in the elution peak was concentrated to 0 . 5 ml with an Amicon Ultra concentrator ( 4 ml , 50 kDa MWCO , Millipore ) by centrifugation at 3000 × g at 4°C . Sample was then applied to a Superdex 200 10/300 GL size-exclusion column ( GE Healthcare ) equilibrated in RBP buffer . The column was run at 0 . 5 ml/min at 4°C . Full-length protein eluted at 10–13 ml , typically together with smaller protein species identified by tandem mass spectrometry to be truncated protein lacking C-terminal residues . The sample was applied to a HiTrap Heparin HP 5 ml column ( GE Healthcare ) , and the protein was eluted with a gradient of 20 column volumes into RBP buffer containing 1 M NaCl . Full-length protein eluted at ~250 mM NaCl , while shorter fragments eluted at 100–200 mM NaCl . Peak fractions were concentrated with an Amicon Ultra 50 kDa MWCO centrifugal concentrator ( Millipore ) and stored at −80°C . The RNA-binding PUM domain of S . pombe Puf3 was defined as residues 378–714 , while the RNA-binding TZF domain of S . pombe Zfs1 was defined as residues 699–769 . DNA sequences encoding the RNA-binding domains of Puf3 ( Puf3PUM ) and Zfs1 ( Zfs1TZF ) were amplified from vectors encoding full-length proteins using primers Puf3_PUM_Fwd and Puf3_PUM_Rev or Zfs1_TZF_Fwd and Zfs1_TZF_Rev ( Supplementary file 3 ) . These were inserted into pGEX-6P-2 expression vector using an In-Fusion HD Cloning Kit ( Clontech ) . Puf3PUM and Zfs1TZF were expressed as N-terminal GST fusions in BL21 Star ( DE3 ) E . coli ( Thermo Fisher Scientific ) . Expression and cell lysis were performed as described for full-length MBP-Puf3 and MBP-Zfs1 . Clarified lysate was applied to glutathione sepharose 4B affinity resin ( 1 ml bed volume per 2 l culture; New England Biolabs ) , and incubated with rotation for 1 hr at 4°C . Resin was separated from lysate with an Econo-Column ( Bio-Rad ) and washed three times with 50 ml of RBP buffer . Protein was eluted from the resin in 20 ml of glutathione elution buffer ( 20 mM HEPES pH 7 . 4 , 150 mM NaCl , 0 . 1 mM TCEP , 50 mM glutathione ) . The GST tag was cleaved by incubation with 1 mg of 3C protease for 16 hr at 4°C . Cleaved protein was buffer-exchanged into 20 mM Bis-Tris pH 6 . 5 , 50 mM NaCl , 0 . 1 mM TCEP with a HiPrep 26/10 desalting column ( GE Healthcare ) . The sample was applied to a Mono S 5/50 GL cation-exchange column ( GE Healthcare ) and eluted over 20 column volumes into buffer supplemented to 1 M NaCl . Pure protein eluted at ~200 mM NaCl , and samples were pooled and stored at −80°C . Deadenylation activity was measured as previously described ( Webster et al . , 2017 ) . Puf3 and Zfs1 were prepared in 20 mM HEPES pH 7 . 4 , 50 mM NaCl , 0 . 1 mM TCEP and added to a final concentration of 250 nM in the reaction , unless otherwise indicated . RBPs were incubated with RNA for 10 min at 22°C prior to the addition of enzyme to allow protein-RNA binding to reach equilibrium . Puf3-target-A30 and Zfs1-target-A30 RNA were synthesized with a 5ʹ 6-FAM fluorophore label ( Integrated DNA Technologies ) . The RNAs were designed with a consensus Puf3 or Zfs1 binding site ending six nucleotides upstream of a poly ( A ) tail , within a 23 nt non-poly ( A ) sequence . 200 nM RNA was used in each reaction . Ccr4-Not and Caf1-Ccr4 were prepared at 1 μM ( 10× ) in 20 mM HEPES pH 7 . 4 , 400 mM NaCl , 2 mM Mg ( OAc ) 2 , 0 . 1 mM TCEP and added to a final concentration of 100 nM in the reaction . Reactions were performed in 20 mM PIPES pH 6 . 8 , 10 mM KCl , 45 mM NaCl , 2 mM Mg ( OAc ) 2 , 0 . 1 mM TCEP ( includes components added with protein factors ) at 22°C . Reactions were stopped at the indicated time points by addition of 2 × denaturing loading dye ( 95% formamide , 10 mM EDTA , 0 . 01% w/v bromophenol blue ) . Samples were applied to TBE ( Tris-borate-EDTA ) , 20% polyacrylamide gels containing 7 M urea and run at 400 V in 1 × TBE running buffer . Gels were scanned with a Typhoon FLA-7000 with excitation at 478 nm . Densitometric analysis was performed with ImageJ as described previously ( Schneider et al . , 2012; Webster et al . , 2017 ) . Poly ( A ) tail lengths were calibrated using RNA markers with no tail and tails of known length . Intermediate tail lengths were calculated by counting bands on gels with single-nucleotide resolution as described previously ( Webster et al . , 2017 ) . Deadenylation rates were calculated from a plot of the size of the most abundant RNA species versus time . Statistical significance was calculated by one-way ANOVA in GraphPad Prism . For RNA competition experiments , Zfs1-target-A30 and ScPuf3 ( 5U ) -target-A30 RNAs were synthesized with a 5ʹ-Alexa647 fluorophore label , and Puf5-target-A30 with a 5ʹ-Cy3 fluorophore label ( Integrated DNA Technologies ) . Reactions contained 100 nM of each RNA , mixed prior to addition of protein factors . Gels were scanned twice with a Typhoon FLA-7000 with excitation at 478 nm ( fluorescein detection ) and either 633 nm ( Alexa647 detection ) or 573 nm ( Cy3 detection ) . Images were superimposed with color overlay . Binding reactions ( 10 μl ) were prepared by adding protein at the indicated concentration to RNA ( 200 nM ) in 20 mM PIPES pH 6 . 8 , 10 mM KCl , 90 mM NaCl , 2 mM Mg ( OAc ) 2 , 0 . 1 mM TCEP . For RNA competition experiments , 100 nM of each RNA was mixed prior to addition of protein . The sample was incubated for 15 min at 22°C before the addition of 6 × loading dye ( 30% glycerol and 0 . 2% w/v orange G ) . Samples were loaded onto 6% TBE-polyacrylamide non-denaturing gels and electrophoresis was performed at 100 V in 1 × TBE running buffer . Gels were scanned with a Typhoon FLA-7000 with excitation at 478 nm ( fluorescein detection ) and 633 nm ( Alexa647 detection ) . A two-fold protein dilution series was prepared in 20 mM HEPES pH 7 . 4 , 150 mM NaCl . Proteins were incubated for 2 hr at 22°C with 0 . 2 nM 5ʹ 6-FAM RNA ( synthesized by IDT ) . Fluorescence polarization was measured with a PHERAstar Plus microplate reader ( BMG Labtech ) . Dissociation constants were estimated by non-linear regression in GraphPad Prism 6 . Error bars indicate the standard deviation in three replicate measurements . Kinetic measurements were performed using a DRX series instrument with a MPC-48–2-Y1 chip ( Dynamic Biosensors ) . Hybrid oligonucleotides were synthesized ( IDT ) with the RNA of interest ( Table 1 ) at the 5ʹ end followed by single-stranded DNA complementary in sequence to the fluorescently labelled oligonucleotide on the chip . Annealing was performed by flowing 500 nM oligonucleotide over the chip for 4 min in a buffer of 20 mM HEPES pH 7 . 4 , 40 mM NaCl and 0 . 001% Tween-20 . Purified Puf3PUM ( residues 378–714 ) was applied to the chip in binding buffer ( 20 mM HEPES pH 7 . 4 , 150 mM NaCl and 0 . 001% Tween-20 ) at 20°C with a flow rate of 30 μl/min . For association measurements , a series of protein concentrations was tested: 5 , 10 , 15 , 20 , 25 , 30 , 50 , 70 nM for ScPuf3- and ScPuf5-target RNA; 50 , 100 , 150 , 200 nM for poly ( A ) . The dynamic response represents the change in nanolever switch speed on the timescale 0–4 μsec as described previously ( Langer et al . , 2013 ) . The observed association rate ( Kobs ) was estimated by fit of an exponential curve to data points collected with a sampling rate of 1 s with GraphPad Prism 6 . Error bars indicate the standard error in three replicate measurements . Association rate ( kon ) was calculated by linear regression analysis of the protein concentration series . For analysis of dissociation rate , 20 nM Puf3PUM was loaded onto RNA as described above . Binding buffer was then flowed across the chip at 20°C with a flow rate of 30 μl/min . Data points from dissociation experiments with ScPuf3-target and ScPuf5-target RNAs were averaged in 10 s intervals to improve the signal-to-noise . Dissociation rates ( koff ) were estimated by fitting of an exponential function with GraphPad Prism 6 . Saccharomyces cerevisiae mRNAs that are candidates for Puf3 regulation were collated from published sources ( Gerber et al . , 2004; Kershaw et al . , 2015; Lapointe et al . , 2015 ) . Proportional Venn diagram of gene overlap was generated with Biovenn ( Hulsen et al . , 2008 ) and redrawn for publication . 3ʹ-UTR sequences were defined using poly ( A ) sites previously identified ( Ozsolak et al . , 2010 ) . The most statistically enriched RNA motif from the combined set of 1331 mRNAs was identified with MEME ( Bailey et al . , 2015 ) . Analysis of sequence context around the target motifs ( 10 nucleotides in each direction ) was performed with WebLogo 3 ( Crooks et al . , 2004 ) . Motifs were identified in mRNA sequences with Geneious v 7 . 0 . 6 ( Kearse et al . , 2012 ) with the following general expressions: motif 1 ( CNUGUAAAUA ) , motif 2 ( CNUGUANAUA ) , motif 3 ( UGUAAAUA ) and motif 4 ( UGUAAAUA ) . Sequences within motif category 1 are by definition within 2 , 3 and 4 . Once it was evident that mRNAs with motif 1 were significantly enriched in ScPuf3 relative to the remainder of mRNAs in categories 2 , 3 and 4 , they were assigned to a distinct category . No significant differences in ScPuf3 enrichment were detected between mRNAs with other permutations not presented ( −2: A , U or G; −1 any nucleotide;+5 C , G or U ) . For this reason , these sequences were treated as equivalent and the mRNAs containing them analyzed as a group . We do not exclude the possibility that differences in ScPuf3 binding exist because statistically evaluation is limited to the number of endogenous genes containing each permutation . Motifs were assigned to the category 3ʹ-UTR if at least part of the motif contained nucleotides 3ʹ of the coding sequence ( including the stop codon ) . Statistical significance was evaluated with a Mann-Whitney rank comparison test in GraphPad Prism . Gene ontology ( GO ) analysis was performed with PANTHER over-representation test using the dataset released on 2018-02-02 ( Mi et al . , 2017 ) . Comprehensive GO term analysis of significant over-representation is available in Supplementary file 2 . Detailed analysis of individual gene function was performed with the Saccharomyces Genome Database ( http://yeastmine . yeastgenome . org ) and the references contained therein . | When a cell needs to make a particular protein , it first copies the instructions from the matching gene into a molecule known as a messenger RNA ( or an mRNA for short ) . The more mRNA copies it makes , the more protein it can produce . A simple way to control protein production is to raise or lower the number of these mRNA messages , and living cells have lots of ways to make this happen . One method involves codes built into the mRNAs themselves . The mRNAs can carry short sequences of genetic letters that can trigger their own destruction . Known as “destabilising motifs” , these sequences attract the attention of a group of proteins called Ccr4-Not . Together these proteins shorten the end of the mRNAs , preparing the molecules for degradation . But how does Ccr4-Not choose which mRNAs to target ? Different mRNAs carry different destabilising motifs . This means that when groups of mRNAs all carry the same motif , the cell can destroy them all together . This allows the cell to switch networks of related genes off together without affecting the mRNAs it still needs . What is puzzling is that the destabilising motifs that control different groups of mRNAs can be very similar , and scientists do not yet know how Ccr4-Not can tell the difference , or what triggers it to start breaking down groups of mRNAs . To find out , Webster et al . recreated the system in the laboratory using purified molecules . The test-tube system confirmed previous suggestions that a protein called Puf3 forms a bridge between Ccr4-Not and mRNAs . It acts as a tether , recognising a destabilising motif and linking it to Ccr4-Not . Labelling different mRNAs with two colours of fluorescent dye showed how Puf3 helps the cell to choose which to destroy . Puf3 allows Ccr4-Not to select specific mRNAs from a mixture of molecules . Puf3 could distinguish between mRNAs that differed in a single letter of genetic code . When it matched with the wrong mRNA , it disconnected much faster than when it matched with the right one , preventing Ccr4-Not from linking up . The ability to destroy specific mRNA messages is critical for cell survival . It happens when cells divide , during immune responses such as inflammation , and in early development . Understanding the targets of tethers like Puf3 could help scientists to predict which genes will switch off and when . This could reveal genes that work together , helping to unravel their roles inside cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2019 | RNA-binding proteins distinguish between similar sequence motifs to promote targeted deadenylation by Ccr4-Not |
Carotenoids are essential in oxygenic photosynthesis: they stabilize the pigment–protein complexes , are active in harvesting sunlight and in photoprotection . In plants , they are present as carotenes and their oxygenated derivatives , xanthophylls . While mutant plants lacking xanthophylls are capable of photoautotrophic growth , no plants without carotenes in their photosystems have been reported so far , which has led to the common opinion that carotenes are essential for photosynthesis . Here , we report the first plant that grows photoautotrophically in the absence of carotenes: a tobacco plant containing only the xanthophyll astaxanthin . Surprisingly , both photosystems are fully functional despite their carotenoid-binding sites being occupied by astaxanthin instead of β-carotene or remaining empty ( i . e . are not occupied by carotenoids ) . These plants display non-photochemical quenching , despite the absence of both zeaxanthin and lutein and show that tobacco can regulate the ratio between the two photosystems in a very large dynamic range to optimize electron transport .
Carotenoids form a large class of natural pigments responsible for the yellow , orange , and red colors of fruits and leaves ( Stange , 2016 ) . In the photosynthetic membranes , they are mainly associated with proteins , forming pigment–protein complexes . Their large absorption cross-section in the blue region of the solar spectrum makes them ideal light-harvesting pigments , especially for aquatic organisms ( Croce and van Amerongen , 2014 ) . However , the primary role of carotenoids in photosynthesis is photoprotection . Their capacity to quench chlorophyll ( Chl ) triplets ( thus avoiding their reaction with molecular oxygen and the production of singlet oxygen ) , and to scavenge singlet oxygen make them essential for the survival of the organism ( Frank and Cogdell , 1996; Borth , 1975; Havaux , 1998 ) . In addition , carotenoids are involved in the quenching of singlet excited state Chls in a process known as non-photochemical quenching ( NPQ ) , which controls the level of excited states in the membrane , thus protecting the photosynthetic apparatus from high light damage ( Ruban et al . , 2012 ) . Two species of carotenoids are present in the photosynthetic membranes: carotenes and their oxygenated derivatives , xanthophylls . The main carotene , β-carotene ( β-car ) , is associated with the core of photosystems I and II ( Umena et al . , 2011; Qin et al . , 2015 ) , and is present in all organisms performing oxygenic photosynthesis . The xanthophylls ( in plants mainly lutein ( Lut ) , neoxanthin ( Neo ) , violaxanthin ( Vio ) and zeaxanthin ( Zea ) ) , instead , are bound to the light-harvesting complexes ( LHCs ) that act as peripheral antennae increasing the absorption cross-section of both photosystems ( Qin et al . , 2015; Su et al . , 2017 ) . LHCs are able to accommodate different xanthophylls , but they cannot fold in the presence of β-carotene only ( Croce et al . , 1999; Phillip et al . , 2002 ) . Also , PSII assembly has been suggested to require the presence of carotenes ( Masamoto et al . , 2004 ) , while PSI is stable also in the absence of carotenoids ( Masamoto et al . , 2004; Santabarbara et al . , 2013 ) . While mutants lacking individual or all xanthophylls but still containing carotenes have been identified for several organisms ( e . g . Dall'Osto et al . , 2013; Ware et al . , 2016; Niyogi et al . , 1997; Pogson et al . , 1998; Domonkos et al . , 2013; Schäfer et al . , 2005 ) , mutants lacking carotenes have only been isolated in cyanobacteria and the green alga Chlamydomonas reinhardtii when these organisms were grown heterotrophically ( Santabarbara et al . , 2013; Sozer et al . , 2010; Tóth et al . , 2015 ) . In these mutants , no PSII was assembled . This finding , together with the fact that no PSII complexes without carotenes have ever been observed , have suggested that carotenes have a vital role not only in photosynthesis but also for the survival of the plant cell ( Dall’Osto et al . , 2014 ) . However , this assumption could never be verified , because available mutants without carotenes completely lack carotenoids . In this work , we have analyzed tobacco ( Nicotiana tabacum ) plants in which the carotenoid biosynthetic pathway was engineered ( by stable transformation of the chloroplast genome ) to only produce the ketocarotenoid astaxanthin ( Lu et al . , 2017; Figure 1 ) . The physiological characteristics and the autotrophic growth of these plants demonstrate that photosynthesis without carotenes is possible , at least when plants are grown in laboratory conditions .
The leaves of the tobacco mutant are orange at an early stage and become greener with age ( see Figure 1 , Table 1 ) . This is likely due to the high expression of the chloroplast genome in young leaves ( Edwards et al . , 2010 ) , which in the mutant results in the massive production of astaxanthin . Most of the astaxanthin is present in the form of crystals or aggregates in the chloroplast ( Lu et al . , 2017 ) . The high-level synthesis of astaxanthin uses a substantial part of the plant’s energy budget and fixed carbon and may contribute to the slow growth of the plants . This negative effect on growth is likely exacerbated by the fact that astaxanthin absorbs most of the incident light , decreasing the number of photons available for photosynthesis . Indeed , the greening of the leaves corresponds to an increase in the growth rate of the plants . Both young and mature leaves of the mutant plants contain only 20% of the Chls per fresh weight as compared to the wild-type ( WT ) , but have a similar ( mature leaves ) or far higher ( young leaves ) carotenoid content ( Table 1 ) . However , the mutant , at all stages of growth , only contains astaxanthin and traces of by-products of astaxanthin synthesis ( Hasunuma et al . , 2008 ) and does not accumulate ( <0 . 005 times the WT ) the carotenoids that are usually present in the WT ( Figure 1 ) . This result is different from the analysis of previously generated astaxanthin-producing plants that still contained WT carotenoids , although in reduced amounts ( Hasunuma et al . , 2008; Fujii et al . , 2016; Röding et al . , 2015 ) . Thus , our engineered tobacco ( hereafter referred to as Asta ) represents the first organism showing autotrophic growth in the virtual absence of carotenes . In the following , we report the experiments performed on mature leaves , which have a Chl/car similar to the WT . Since violaxanthin and lutein are considered to be necessary for the folding of the antenna complexes ( Dall'Osto et al . , 2006 ) , and β-carotene was thought to be required for PSII assembly and photosynthetic activity ( Santabarbara et al . , 2013; Sozer et al . , 2010; Tóth et al . , 2015 ) , we analyzed the effect of their absence on the composition and organization of the photosynthetic apparatus . 2D gel electrophoresis ( Figure 2 ) and immunoblot analyses ( Figure 2—figure supplement 1 ) of thylakoid membranes show that all of the main photosynthetic proteins are present in Asta plants , but the PSII/PSI ratio is far higher than in the WT ( Figure 2—figure supplement 1 ) . The LHC/PSII ratio is , however , similar , except for the antenna protein Lhcb5 that is strongly reduced , and for PsbS , the main protein involved in NPQ ( Li et al . , 2000 ) , which is increased in the mutant ( Figure 2—figure supplement 1 ) . PSI-LHCI , ATP synthase and cytochrome b6f have the same mobility in native gels as the WT complexes , indicating that they are stable and have the same supramolecular organization . By contrast , the stability of PSII seems to be affected as the bands corresponding to PSII supercomplexes and LHCII trimers , which are well defined in the WT , are substituted by a smear in the mutant , suggesting that the PSII complexes are more heterogeneous , incompletely assembled or less stable than in the WT ( Figure 2 ) . Next , we investigated the effects of the change in carotenoid composition on the properties of the individual complexes that were isolated from thylakoid membranes and separated by sucrose density gradient ultracentrifugation ( Figure 3—figure supplement 1 ) . Pigment analysis ( Table 2 and Figure 3—figure supplement 2 ) confirmed that astaxanthin is the only carotenoid associated with all pigment-binding complexes in Asta plants , while β-carotene is present in PSI in a highly substoichiometric amount ( 0 . 15 β-carotene molecules per complex ) and it is virtually absent in PSII ( 0 . 03 molecules per complex ) . This means that most of the PSI and PSII complexes in the mutant plants do not contain β-carotene at all . Normalized to Chl , Asta-LHCs contain the same number of carotenoids as the WT monomers , but instead of binding lutein , neoxanthin and violaxanthin , they only bind astaxanthin , indicating that all carotenoid-binding sites are promiscuous and can accommodate different xanthophylls . The pigment analysis also showed that β-carotene can be substituted by astaxanthin in both PSII and PSI cores . However , the higher Chl/car ratio in the isolated Asta complexes compared to the WT complexes indicates that not all sites that are occupied by β-carotene in the WT are occupied by astaxanthin in the mutant complexes , but some are left ‘empty’ in that they are not occupied by carotenoids . Although we cannot exclude the possibility that some of the astaxanthin molecules are more loosely bound and thus are lost during purification , the fact that both PSI and PSII complexes can be purified with a large number of ‘empty’ sites indicates that their occupancy by carotenoids is not crucial for the stability of the complexes . Interestingly , absorption ( Figure 3—figure supplement 3A ) and circular dichroism ( Figure 3—figure supplement 3B ) spectra of LHCs and PSII core complexes from the WT and the mutant are very similar ( see Figure 3—figure supplement 3 for a more detailed explanation ) in the Chl absorption regions . This indicates that there are no significant changes in the pigment organization of the complexes and thus in their three-dimensional structure . The only exception is Asta-PSI-LHCI , the fluorescence emission of which showed a 6 nm shift to shorter wavelengths as compared to the WT complex ( Figure 3—figure supplement 3C and Figure 3—figure supplement 4 ) . Since the PSI emission at 77 K originates mostly from two specific Chls ( called far-red Chls ) of Lhca3 and Lhca4 ( Morosinotto et al . , 2003 ) , we can conclude that the interaction between these Chls is slightly changed in the mutant . Carotenoids are known to be required for the stability of the pigment-binding holoproteins ( Paulsen et al . , 1993 ) . Our data measured on the isolated complexes show that the difference in composition between WT and mutant complexes influences the denaturation temperature by only 5–10°C ( Figure 3A ) . This is surprising considering that several of the carotenoid-binding sites in the isolated PSI and PSII are not occupied by carotenoids and indicates that only some of them play a crucial role in protein stability . Photoprotection via Chl triplet quenching and singlet oxygen scavenging is the primary role of carotenoids in photosynthesis ( Siefermann-Harms , 1987 ) . Photobleaching experiments ( Figure 3B ) show that , while the photostability of LHCs and PSII core is only partially affected by the change in carotenoid composition , Asta-PSI-LHCI is far more sensitive to light than the WT complex . It is likely that this effect on PSI is due to the reduced number of carotenoid molecules associated with the complex , which results in less efficient Chl triplet quenching . However , it is worth noting that , even with a large part of the carotenoid-binding sites not occupied by carotenoids , PSI is more photostable than PSII-WT , in agreement with the fact that , in PSII , carotenoids cannot provide protection by quenching singlet oxygen formed via P680 triplet because of the very high oxidizing potential of PSII ( Telfer , 2014 ) . Next , we investigated the effect of the substitution of carotenes with astaxanthin on the light-harvesting and trapping properties of the photosynthetic complexes in vivo by performing time-resolved fluorescence measurements at 20°C on intact leaves ( Figure 4 and Figure 4—figure supplement 1 ) . The PSI kinetics is very similar in the WT ( 70 ps ) and in the mutant ( 65 ps ) , and the small difference can be ascribed to the reduced far-red Chl content of Asta-PSI-LHCI , which is known to influence the PSI trapping time ( Croce and van Amerongen , 2013 ) . The PSII kinetics in the mutant leaves changes in the presence versus absence of photochemistry ( measurements performed with the reaction center ( RC ) open and closed , respectively ) as it does in the WT , indicating that excitation energy transfer occurs in the mutant and the harvested energy is used for photochemistry . However , all the kinetics are faster and the difference between closed and open RC is smaller in mutant than in WT leaves , suggesting that the antenna complexes of the mutant plants are statically quenched in vivo . Measurements on isolated Asta-Lhcb show that this is indeed the case: these complexes are strongly quenched ( lifetime of 0 . 87 ns vs . 3 . 5 ns in the WT; Figure 4—figure supplement 2 ) due to the presence of astaxanthin ( Liguori et al . , 2017 ) . It has also been shown that part of the astaxanthin population can transfer excitation energy to the chlorophylls , thus also acting as light-harvesting pigment ( Liguori et al . , 2017 ) . The presence of excitation energy transfer from the antenna to the RC in mutant leaves indicates that , although the interactions between the building blocks of the PSII supercomplex are not strong enough to survive purification ( see Figure 2 ) , the supercomplexes are functional in vivo meaning that , in the membrane , the subunits are close enough to each other to ensure the delivery of the harvested energy to the reaction center . Indeed , the short excited state lifetime of the antenna ( indicative of a constitutively quenched antenna ) in the mutant can fully account for the lower maximum quantum efficiency of PSII ( FV/FM; Table 3 ) in mutant plants , which is mainly the result of low fluorescence emission in the absence of photochemistry ( FM ) . Finally , we analyzed the photosynthetic performance of the Asta plants . Electrochromic shift ( ECS ) of the carotenoid absorption is commonly used to study the function of all major photosynthetic complexes ( Bailleul et al . , 2015 ) . We verified that the mutant plants exhibit an ECS signal and we determined its light-induced difference spectrum , which agreed with the prediction that astaxanthin is solely responsible for this in vivo Stark effect ( Figure 5—figure supplement 1 ) . Using ECS , we observed that the functional PSII/PSI RC ratio was far larger in the mutant than in the WT , in qualitative agreement with our biochemical data ( Table 3; Figure 2—figure supplement 1 ) . The difference in the values obtained with the two methods is partially due to the limited quantitative power of immunoblots , but also suggests that some of the PSII cores are not functional . The high PSII/PSI ratio in the mutant seems to be a compensation mechanism for the decrease in the relative functional PSII/PSI antenna size ( measured with two independent methods; Table 3 and Figure 5—figure supplement 2 ) observed in the Asta plants , which is due to the presence of static quenching . Indeed , comparison of the steady-state photochemical yields of PSII and PSI revealed that , at all light intensities , in both WT and mutant plants , the balance between PSII and PSI photochemistry is maintained ( Figure 5A and Figure 5—figure supplements 3 and 4 ) , meaning that the plants are able to compensate for the strong decrease in the PSII functional antenna size by decreasing the PSI/PSII ratio . This means that these plants have the capacity to modulate the PSI/PSII ratio in a large dynamic range . Finally , transient QA reduction and reoxidation kinetics suggest that no significant differences in the PSII electron transfer occur in the mutant plants ( Figure 5—figure supplements 2 and 5 ) . The full operational capacity of the electron transport chain permitted us to verify whether the photoprotective regulation is maintained in the mutant plants . As expected , the NPQ amplitude was largely reduced in the mutant ( Figure 5B and Figure 5—figure supplement 6 ) , because the ΔpH-induced , PsbS-dependent quenching has to compete with the strong , constitutive astaxanthin quenching in these plants . Note that the difference in NPQ level ( 1 . 8 in the WT vs . 0 . 3 in the mutant ) can be fully ascribed to the presence of the static quencher in the mutant , which strongly reduces the maximal fluorescence in both dark ( FM ) and light ( FM’ ) states . This is supported by the NPQ ( t ) calculation , which permits to correct the apparent NPQ for the presence of a pre-existing quenching , assuming that a decrease of the FV/FM value is solely due to this static quenching . The data show that NPQ ( t ) is even larger in the mutant than in the WT ( Figure 5—figure supplement 7 ) . Importantly , despite the difference in apparent NPQ amplitude , the kinetics of onset and recovery are identical to those of the WT ( inset in Figure 5B ) and consistent with qE characteristics . This outcome is particularly striking if one considers that Asta plants lack both lutein and zeaxanthin , which are believed to be essential for NPQ ( Niyogi et al . , 1998 ) . It is likely that the high amount of PsbS in the mutant ( Figure 2—figure supplement 1 ) can compensate for the lack of the xanthophyll cycle , or that astaxanthin can also be responsible for the dynamic quenching . Whatever the reason for the presence of NPQ in the mutant , our Asta plants clearly show that lutein and zeaxanthin are not absolutely necessary for it . In summary , we have shown that the carotenoid-binding sites of the core complexes of PSI and PSII are promiscuous . Although they bind carotenes in all known photosynthetic organisms , our data demonstrate that they can also accommodate xanthophylls . This is at variance with the LHCs that can bind various xanthophylls but cannot fold with carotenes . More importantly , we show that both PSI and PSII are stable while most of their carotenoid-binding sites are not occupied by carotenoids and the rest is occupied by an alien xanthophyll . These results indicate that the core complexes are even more robust than the outer antennae and can endure radical changes even in one of their main components . In this respect , it is important to realize that the difference in growth rate between WT and mutant plants is not due to the absence of carotenes , but rather to the presence of astaxanthin that stabilizes the LHCs in a quenched conformation . In conclusion , the substitution of carotenes with the xanthophyll astaxanthin does not impair the functional assembly of the photosynthetic apparatus , nor does it impede efficient electron transfer and NPQ , demonstrating that carotenes are not essential neither for the biosynthesis of the photosynthetic apparatus nor for its function . This finding has important implications not only for our understanding of the structure and function of the photosynthetic apparatus but also for future efforts to design synthetic photosystems with novel and improved properties .
Seeds from mutant and WT plants were sown on moist filter paper and synchronized at 4°C for 2–3 days before being moved to room temperature until germination . The seedlings were transferred to soil and grown at 22°C under 150–200 μmol photons m−2 s−1 for the WT and 80–120 μmol photons m−2 s−1 for the mutant with 14 hr of light per day . Plants were fed with commercial fertilizer each week . Leaves from WT ( 5–6 weeks old ) and Asta plants ( around 20 weeks for younger leaves , 24–30 weeks for older leaves ) were used for physiological measurements and thylakoid isolation . WT thylakoid isolation was performed as described in Xu et al . , 2015 . The centrifuge speed was increased to 4000 g for the first step of the isolation from mutant tobacco . Pigments from isolated protein–pigment complexes or leaves were extracted with 80% acetone . HPLC was performed as in Xu et al . , 2015 with the modification that buffer B was linearly increased from 0 to 100% in 9 . 2 min . Chlorophyll a/b ratios and chlorophyll/carotenoid ratios were calculated by fitting their individual absorption spectra to measured spectra ( Xu et al . , 2015 ) . Examples of the fitting of total thylakoids and isolated Lhcbs are shown in Figure 3—figure supplement 2 panels C and D . Blue-native gels were performed as described in Järvi et al . , 2011 with the modifications described in Bielczynski et al . , 2016 . The second dimension and the SDS-PAGE were performed as described in Schägger , 2006 . For immunoblot analysis , total protein extracts were separated by SDS-PAGE and transferred to a Protran 0 . 45 mm nitrocellulose membrane . Specific primary antibodies ( Agrisera ) were used to detect the target proteins . Chemiluminescence was detected using an ImageQuant LAS 4000 imaging system . For sucrose density gradient fractionation , thylakoids equivalent to 0 . 2 mg total chlorophyll were washed with 5 mM EDTA and resuspended in 200 μL 10 mM Hepes ( pH 7 . 5 ) . An equal volume of 1 . 2% α-DDM was added , mixed gently , and the solubilized thylakoids were centrifuged at 14 , 000 rpm for 10 min at 4°C . The supernatant was loaded on a 0–1 M sucrose gradient ( 10 mM Hepes , pH 7 . 5 , 0 . 03% α-DDM ) and centrifuged at 288 , 000 g for 17 hr . The separated bands were collected with a syringe . Time-resolved fluorescence measurements on leaves were done using a time-correlated single photon-counting ( TCSPC ) setup as described previously ( Chukhutsina et al . , 2019 ) . Excitation at 650 nm was used to excite Chl b preferentially . Detached plant leaves were placed between two glass plates and mounted in the rotation cuvette ( diameter: 10 cm; thickness: 1 mm ) . The cuvette was rotated at 1400 rpm while oscillating sideways . Fluorescence was measured in a front-face arrangement from the upper side of the leaves . Time-resolved fluorescence decays were measured at multiple detection wavelengths ( between 675 and 690 nm with a wavelength step of 5 nm , and between 700 and 760 nm with a maximal wavelength step of 10 nm ) . The measurements were done in the presence and in the absence of PSII photochemistry ( open ( F0 ) and closed ( FM ) states , respectively ) . The measurement time at a single wavelength was limited to 10 min , to avoid changes in the leaves due to prolonged measurement in the rotating cuvette . To perform an experiment in one state took 2–3 hr . All in vivo measurements were performed at 20°C . The obtained fluorescence decay traces were analyzed globally with the ‘TRFA Data Processing Package’ of the Scientific Software Technologies Center ( Belarusian State University , Minsk , Belarus ) ( Digris et al . , 1999 ) . The global analysis methodology is described in van Stokkum et al . , 2004 . In short , a number of parallel , non-interacting kinetic components was used as a kinetic model , so the total dataset was fitted with function f ( t , λ ) as follows:∑1 , 2 . . . NDASi ( λ ) exp ( −tτi ) ⊕irf ( t , λ ) , where the decay-associated spectrum ( DASi ) is the amplitude factor associated with a decay component i having a decay lifetime τi , and irf ( t , λ ) was measured using scattering light . Typical full-width at half-maximum ( FWHM ) values were 28 ± 2 ps . Time-resolved fluorescence measurements on isolated LHCII were performed on a FluoTime200 setup ( Picoquant ) . The samples were diluted to an OD of 0 . 05 cm−1 at the maximum in the Qy region and measured in a 3 . 5 mL cuvette with a path length of 1 cm at 283 K . Excitation was provided by a 468 nm laser diode ( preferential Chl b excitation ) operating at 10 MHz repetition rate . The instrument response function ( IRF ) was obtained by measuring the decay of a pinacyanol iodide dye dissolved in methanol , which has a six ps fluorescence lifetime ( van Oort et al . , 2008 ) . The resulting IRF FWHM was ∼88 ps . The fluorescence decay kinetics was detected at 680 nm with a channel time spacing of 8 ps . Data analysis was performed by the TRFA DATA software as described above . The ECS light-induced difference spectrum was determined according to Bailleul et al . , 2015 using a JTS-10 spectrophotometer ( BioLogic , Grenoble , France ) . In brief , the leaf was subjected to a saturating pulse of red light ( 3000 µmol photons m−2 s−1; 80 ms ) , and the absorption changes at each wavelength after the pulse were recorded without additional actinic light . The baseline obtained without the saturating pulse was subtracted , and the values between 100 and 200 ms after the pulse ( to avoid the contribution of signals due to rapid redox-changes of cytochromes ) were averaged . The obtained spectrum closely matches the theoretical ECS spectrum of pure astaxanthin , which is [1- ( dfdx ) ] of astaxanthin-detergent solution spectrum ( Bailleul et al . , 2010 ) . The PSII:PSI RC ratio was determined using the JTS-10 spectrophotometer using saturating single-turnover laser flashes ( five ns duration ) provided by a dye laser pumped with a Nd:YAG laser ( Minilite , Continuum ) using the protocol described in Nawrocki et al . , 2016 but adapted for leaves . For the PSII+PSI signal , the leaf was infiltrated with water , and to obtain a pure PSI signal the leaf was infiltrated with hydroxylamine ( HA , 1 mM ) and 3- ( 3 , 4-dichlorophenyl ) −1 , 1-dimethylurea ( DCMU , 10 µM; both from Sigma ) , after a systematic verification that no variable fluorescence , and thus PSII activity , remained in the leaf . ECS was detected at 546 nm ( Asta ) and 520–546 nm ( WT ) using weak white light LED pulses filtered with a 10 nm FWHM interference filter . The peak amplitude at 546 nm allows the detection at the isosbestic point of cyt . b6f haems ( Alric et al . , 2005 ) . The functional antenna size was measured as described in Nawrocki et al . , 2016 but with 300 µmol photos m−2 s−1 red actinic light ( 630 nm peak ) and detecting light as described above . The quantities of active PSII were corrected for the ~20% slowly-opening RCs accumulating after actinic light . | Most life on Earth depends on photosynthesis , the process used by plants and many other organisms to store energy from sunlight and produce oxygen . The first steps of photosynthesis , the capture and conversion of sunlight into chemical energy , happen in large assemblies of proteins containing many pigment molecules called photosystems . In plants , the pigments involved in photosynthesis are green chlorophylls and carotenoids . In addition to harvesting light , carotenoids have an important role in preventing damage caused by overexposure to sunlight There are over one thousand different carotenoids in living beings , but only one , β-carotene , is present in every organism that performs the type of photosynthesis in which oxygen is released , and is thought to be essential for the process . However , this could never be proved because it is impossible to remove β-carotene from cells using typical genetic approaches without affecting all other carotenoids . Xu et al . used genetic engineering to create tobacco plants that produced a pigment called astaxanthin in place of β-carotene . Astaxanthin is a carotenoid from salmon and shrimp , not normally found in plants . These plants are the first living things known to perform photosynthesis without β-carotene and demonstrate that this pigment is not essential for photosynthesis as long as other carotenoids are present . Xu et al . also show that the photosystems can adapt to using different carotenoids , and can even operate with a reduced number of them . Xu et al’s findings show the high flexibility of photosynthesis in plants , which are able to incorporate non-native elements to the process . These results are also important in the context of increasing the photosynthetic efficiency , and thus the productivity of crops , since they show that a radical redesign of the photosynthetic machinery is feasible . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology"
] | 2020 | Photosynthesis without β-carotene |
Individual recognition ( IR ) is essential for maintaining various social interactions in a group , and face recognition is one of the most specialised cognitive abilities in IR . We used both a mating preference system and an electric shock conditioning experiment to test IR ability in medaka , and found that signals near the face are important . Medaka required more time to discriminate vertically inverted faces , but not horizontally shifted faces or inverted non-face objects . The ability may be comparable to the classic ‘face inversion effect’ in humans and some other mammals . Extra patterns added to the face also did not influence the IR . These findings suggest the possibility that the process of face recognition may differ from that used for other objects . The complex form of recognition may promote specific processing adaptations , although the mechanisms and neurological bases might differ in mammals and medaka . The ability to recognise other individuals is important for shaping animal societies .
In a social group , the ability to recognise other individuals correctly is essential for maintaining various social interactions in animals , such as pair-bonding , hierarchy , inbreeding avoidance , and recognition of offspring , nest mates , or neighbours ( Tibbetts and Dale , 2007; Wiley , 2013 ) . For example , some territorial birds can remember specific neighbours for a long period of time ( Godard , 1991 ) , and king penguins can identify their chick from thousands of conspecifics ( Aubin and Jouventin , 1998 ) . Receivers associate different types of identity signals , such as odour , sound , tactile , motion , electric or morphological cues , with certain individuals ( Sherman et al . , 2009 ) and identify them afterwards when necessary . In addition to looking at how animals recognise conspecifics , their mental representations of specific individuals can also give hints that allow us to judge their cognitive abilities . For example , hamsters have various odours for different body parts , and an unfamiliar hamster will categorise them as multiple individuals , while a previously interacted hamster can associate the odours to the specific individual ( Johnston and Bullock , 2001 ) . Animals may have complicated mechanisms to link multiple identity signals to different types of fitness-related tasks , or may use simpler rules to remember an individual . Among all of the individual recognition ( IR ) systems , face recognition is one of the most specific abilities , and is reported in animals from a number of distinct evolutionary lineages ( Kendrick and Baldwin , 1987; McKone et al . , 2007; Van der Velden et al . , 2008 , Coulon et al . , 2009; Racca et al . , 2010; Sheehan and Tibbetts , 2011 ) . How faces are recognised , and whether the processes involved differ from those used to perceive other objects , is a main topic of interest in the field of cognitive psychology and biology . In humans and some other mammals , faces are specially processed in cognitive , developmental and functional ways ( Calder , 2011 ) . Human infants are hypothesised to be attracted to faces innately ( Morton and Johnson , 1991 ) , but also develop face recognition skills and specific brain regions for processing faces during childhood . A familiar face can be individuated in 250 ms ( Jacques and Rossion , 2006 ) , and we can possibly remember more faces than other visual stimuli with similar variations in details and features . Studies of a neuropsychological disorder known as prosopagnosia or face blindness , in which individuals are unable to recognise faces but have no difficulty in recognising individuals by other modalities ( such as voice ) or in discriminating non-face objects ( Meadows , 1974; Behrmann et al . , 2005 ) , have shown that facial recognition proceeds through specific cognitive and neural pathways ( Valentine , 1988; McKone et al . , 2007 ) . In addition , the increase in recognition difficulty associated with inversion of faces is greater than that for the inversion of other types of visual stimuli ( Yin , 1969 ) . The so-called face-inversion effect indirectly indicates that faces are perceived configurally rather than only by specific features ( such as the eyes , nose , or mouth ) , and that once inverted , such a global configuration is difficult to match and passes through routes which are used for recognising other objects ( Bartlett and Searcy , 1993; Haxby et al . , 1999; Boutsen et al . , 2006 ) . Likewise , the Thatcher illusion found in both humans ( Thompson , 1980 ) and monkeys ( Adachi et al . , 2009; Dahl et al . , 2010 ) , in which the eyes and mouth are inverted relative to the face , becomes difficult to detect when upside down , further demonstrating that configural perception is interrupted when orientation is inverted . Some other animals , ranging from mammals , birds and fish to invertebrates , have also been reported to use faces for IR ( Brown and Dooling , 1992; Kendrick et al . , 1995; Bovet and Vauclair , 2000; Van der Velden et al . , 2008 , Kohda et al . , 2015; Parr and Hecht , 2011; Tibbetts , 2002 ) . Scientists have long argued that the face-specific processes are unique to humans or shared only by quite closely related species ( Tate et al . , 2006 ) . However , such specialised ability may also have evolved in distinct animal taxa when selection force associated with complicated , repeated social interactions strongly favours IR . The face inversion effect is the method most widely used in animals to test whether faces may be processed specifically , and researchers have identified this ability in some non-human primates ( Overman and Doty , 1982; Tomonaga , 1994; Parr et al . , 1998; Vermeire and Hamilton , 1998; Neiworth et al . , 2007 ) and in sheep ( Kendrick et al . , 1996 ) . Some monkeys failed to show such oriented-specific face-processing ( Rosenfeld and Van Hoesen , 1979; Bruce , 1982; Dittrich , 1990; Parr et al . , 1999; Weiss et al . , 2001; Gothard et al . , 2004 ) , but many studies lacked the use of non-face signals as controls , making it difficult to interpret the results ( Parr et al . , 1999 ) . Specialised neural systems for face recognition have been found in some non-human primates and in sheep ( Kendrick and Baldwin , 1987; Kanwisher and Yovel , 2006; Tsao et al . , 2006 ) , providing great opportunities to interpret how these animals perceive faces perceptually and mechanically for comparative research . Other than the inversion effect , sheep , chimpanzees , and wasps exhibit better discrimination of conspecific faces than of non-face objects ( Kendrick et al . , 1996; Parr et al . , 1998; Sheehan and Tibbetts , 2011 ) . The difference between decision speed and accuracy in discriminating faces and non-facial stimuli is hypothesized to be due to face-specific perception ( Sheehan and Tibbetts , 2011 ) . In the present study , we used a popular freshwater animal model , the medaka fish ( Figure 1A ) , to test IR ability and to examine whether these animals perceive faces differently from non-face stimuli . Researchers have only recently found that fish can use facial pattern to individuate others . Manipulation using digital models demonstrated that two species of cichlid fish use facial patterns , but not body colouration , to recognise familiar individuals ( Kohda et al . , 2015; Satoh et al . , 2016 ) . A species of reef fish uses UV patterns on the face for species recognition , but there is no evidence of IR ( Siebeck et al . , 2010 ) . Medaka are shoaling fish with diverse social behaviours that has become a popular model in genetic and neural research . Medaka females prefer males with larger body sizes ( Howard et al . , 1998 ) and longer fins ( Fujimoto et al . , 2014 ) , or familiar males . Visual contact for 5 hr can shorten the time to mate for a pair of medaka , and a certain extrahypothalamic neuromodulatory system alters the preference in response to familiarity ( Okuyama et al . , 2014 ) . Nonetheless , the cues used for medaka IR and the cognitive basis that underlies IR remain unknown . Here , we investigated the identity signals used for medaka IR , and whether the process of recognising other individuals differs from that for other objects . We propose that medaka can become a powerful model for understanding IR systems for many reasons . First , abundant closely related species with different social behaviours are available , allowing us to test the evolutionary background that promotes strict IR . Second , the social behaviours within the species are also variable . Medaka from different geographic regions or different inbred strains behave uniquely ( Tsuboko et al . , 2014 ) , allowing us to investigate how ecological factors influence the use of identity signals , as well as the mechanisms behind these signals . Moreover , rich genetic techniques such as genome editing and epigenetic methods are available for medaka ( Kirchmaier et al . , 2015 ) , providing powerful tools with which to solve complex questions . 10 . 7554/eLife . 24728 . 003Figure 1 . Morphological differences between individual medaka fish . ( A ) Medaka individuals may differ in pattern , colour or body shape . The colour and pattern may change based on lighting conditions , physiological conditions and stress level . ( B ) Mean ± SEM relative reflectance of fish body trunks from five individuals from Figure 1A . Each colour represents one individual fish . Even though the fish look similar under human vision , their reflectance spectra can be very different . DOI: http://dx . doi . org/10 . 7554/eLife . 24728 . 003 The first aim of this study was to identify the cues used for medaka IR . We tested whether visual and odour cues are part of the identity signals , and whether the cues work collaboratively . We also investigated which visual components ( such as appearance , motion and different body parts ) are necessary for IR , as well as the extent to which the signals can be manipulated ( extra pattern added or image inverted ) without affecting IR . The second aim was to test whether the mechanism of face recognition differs from that for non-face objects using the classic face-inversion paradigm and the accuracy of discriminating faces and non-face objects . We used both ecologically realistic settings ( mating test ) and a conditioned test ( electric shock two-alternative forced-choice [TAFC] design ) to assess strict IR in medaka . Understanding the cues that animals use to recognise others , as well as their cognitive basis , can help us to elucidate how animals connect to each other in their social world .
First , we tested cues from different modalities to determine which cue is important for medaka IR , using a mating paradigm in which females more quickly accept familiar males . We exposed females to cues from males through different modalities ( visual , odour , both visual and odour , and no cue ) for at least 5 hr , and placed the pair of fish together for a mating test . Female medaka took significantly less time to accept a male when familiarised with his visual cues or with both his visual and odour cues before mating , compared with an unfamiliar male ( ANOVA , F3 , 76=5 . 35 , p=0 . 002; Tukey’s HSD , p<0 . 05; Figure 2A ) . When different males were substituted in after visual familiarisation , the females were able to recognise the difference and required more time to accept the substituted male than a familiar male ( F3 , 57=6 . 49 , p=0 . 003; Tukey’s HSD , p<0 . 05; Figure 2B ) . Females were also able to discriminate between individual males by conditioning with electric shock . We used a TAFC design in which two unfamiliar males were placed at two ends of the setup , and the female was given an electric shock when she entered the side containing the ‘incorrect male’ ( Figure 2C ) . In the last six trials of the experiment , females made significantly more correct choices than in the first six trials ( paired t-test , t38=4 . 68 , p<0 . 0001; Figure 2D ) . After 24 hr , females were able to discriminate the males and made significantly more correct choices than on the first day during the first six trials ( paired t-test , t38=5 . 35 , p<0 . 0001 Figure 2D ) . 10 . 7554/eLife . 24728 . 004Figure 2 . Mating test and electric shock two-alternative forced-choice ( TAFC ) test were used to examine medaka individual recognition ( IR ) . ( A ) Females were familiarised with different types of male cues for more than 5 hr and then the males and females were placed together for mating tests . Grey lines indicate log transformed mean ± SEM time for females to mate . Different letters indicate statistically significant differences after a Tukey’s post hoc test ( p<0 . 05 ) . Each dot represents an individual female . With visual cues alone , the females were able to accept males as familiar mates and required less time to mate . ( B ) Log transformed time to mate for familiar males ( females familiarised with visual cues ) , unfamiliar males ( females given no cue ) , and exchanged males ( females familarised with visual cues from a different male ) . After substituting the males , the females were able to detect the change and required more time to accept the males . ( C ) Setup of the electric shock TAFC experiment . The side views of the males were covered . Females were allowed to choose between two unfamiliar males , and when the female entered the area containing the ‘incorrect’ male , she was given an electric shock . When the female remained in the ‘correct’ side for more than 3 min , it was considered that she had made a correct choice , and no shock was given . ( D ) We tested whether medaka females could discriminate different males with the electric shock-conditioned test . The figure shows the mean ± SEM percentage of correct choices in the electric shock task for two consecutive days . Females were able to distinguish individual males associated with electric shock and performance was improved in the last six trials on the first day . Even after 24 hr , the females could still remember the males and made significantly more correct choices than in the first six trials on the first day . DOI: http://dx . doi . org/10 . 7554/eLife . 24728 . 004 We tested how much time female medaka required to accept a male as a familiar mate , and how long the effect lasted . We visually familiarised male medaka with females for 1 , 2 , 3 , and >5 hr , and then put them together for a mating test . We also visually familiarised pairs of medaka for more than 5 hr , and separated them for 1 , 2 , 3 , and 24 hr before the mating test . After visual familiarisation for 3 hr , the females accepted the familiarised males significantly faster than the unfamiliar males ( F4 , 95=5 . 39 , p=0 . 0006; Tukey’s HSD , p<0 . 05; Figure 3A ) , and the effect lasted for at least 3 hr after separation , but not for 24 hr ( F5 , 113=6 . 84 , p<0 . 0001; Tukey’s HSD , p<0 . 05; Figure 3B ) . Thus , the outcome differed from that of the electric shock experiment , in which female fish remembered an individual male even after 24 hr . 10 . 7554/eLife . 24728 . 005Figure 3 . Illustration of the experimental protocol and the time required for female medaka to mate . Grey lines indicate log-transformed mean ± SEM time required for females to mate with different groups of visually familiarised males . Letters represent significant differences after analysis of variance tests ( ANOVAs ) and Tukey’s post hoc tests . Dots indicate individual fish . ( A ) Female medaka were visually familiarised with a male for different durations . The effect of visual familiarisation was significant after 3 hr of habituation . ( B ) Pairs of medaka were separated for different durations after being visually familiarised for >5 hr . Even after separation for 3 hr , the females still treated the males as familiar mates; this was no longer the case after 24 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 24728 . 005 We examined whether motion was involved in medaka IR , and we also looked at the importance of cues from different body parts . We familiarised female medaka to the movements of male medaka using semi-transparent films to obscure their appearance but not their movements . The response of females to the motion-familiarised males did not differ significantly from the response to unfamiliar males ( t-test , t28=−0 . 03 , p=0 . 97 ) . Familiarisation with only the appearance and not the motion of the males ( males fixed in a transparent container ) was sufficient for the females to require significantly less time to accept the males ( F4 , 70=3 . 85 , p=0 . 007; Tukey’s HSD , p<0 . 05; Figure 4A ) ; however , females required significantly more time to accept head-covered males than tail-covered males ( t-test , t32=−2 . 33 , p=0 . 03; Figure 4B ) . Even when black spots were added to the faces of the males after visual familiarisation , females still accepted the males as familiar mates ( F2 , 42=0 . 22 , p=0 . 80; Figure 4C ) . 10 . 7554/eLife . 24728 . 006Figure 4 . Which morphological traits are important for medaka individual recognition , and how they can be modified . Grey lines indicate log transformed mean ± SEM time to mate with different groups of visually familiarised males . Dots indicate individual fish and asterisks indicate p<0 . 05 . Letters represent significant differences after ANOVA and Tukey’s post hoc tests . ( A ) Females were visually familiarised with different types of male visual cues , including appearance and motion . Females were able to recognise the males as familiar mates on the basis of appearance alone . ( B ) When the head of the male medaka was covered , females were not able to recognise the familiar male and the time to mate was increased . Photos show head-covered and tail-covered medaka . ( C ) Signals proximate to the head are important for medaka individual recognition . Females were still able to recognise the males after extra spots were painted on the faces of males after visual familiarisation . In the control group , the males were painted by brush with no ink on the face . Photos show one medaka before and after black ink was painted on the face . ( D ) Images of the males were manipulated with a prism during visual familiarisation , which was followed by mating tests . When familiarised with vertically shifted images , females did not treat the males as familiar mates . DOI: http://dx . doi . org/10 . 7554/eLife . 24728 . 006 We tested whether medaka can recognise inverted faces using both mating tests ( Figure 5A ) and electric shock TAFC tasks . In the mating tests , the time to mate was significantly longer for females familiarised with the vertically flipped images of the males , compared with those familiarised with horizontally flipped and upright images ( F2 , 42=5 . 00 , p=0 . 01; Tukey’s HSD , p<0 . 05; Figure 4D ) . We tested the face inversion effect with the electric shock TAFC tasks as well , also using non-face objects as a control . Fish were trained to discriminate between two individuals or between two sets of non-face objects that differed in familiarity ( Figure 5B ) . The fish were exposed to the familiar non-face objects from hatching . Fish were able to discriminate between two fish , two non-face objects , and two familiar non-face objects . They made significantly more correct choices ( pair t test , fish: t38=2 . 87 , p=0 . 007; non-face objects: t38=3 . 09 , p=0 . 004; non-face familiar objects: t18=2 . 72 , p=0 . 014 ) for fish/object presented in the upright position for the last six trials ( mean ± standard deviation percentage correct choices , fish: 57 . 50 ± 16 . 64; non-face objects: 54 . 17 ± 16 . 99; familiar objects: 61 . 67 ± 17 . 67 ) than for those in the first six trials ( fish: 44 . 17 ± 12 . 49; non-face objects: 39 . 17 ± 13 . 55; familiar objects: 40 . 00 ± 17 . 92 ) . We examined the effects of visual stimuli type and stimuli orientation on the percentage of correct choices using two-way ANOVA ( Figure 5C ) . There was a significant interaction between stimulus type and orientation ( F2 , 94=3 . 68 , p=0 . 03 ) . Therefore , simple main-effect analysis for stimulus type was performed with a Bonferroni adjustment . All pairwise comparisons were run for each simple main effect . In the fish discrimination group , the correct choices decreased significantly after the image was inverted ( F1 , 94=12 . 26 , p=0 . 001 ) , but this was not the cases when sets of two objects were used as stimuli ( non-face objects: F1 , 94=0 . 25 , p=0 . 62; familiar non-face objects: F1 , 94=0 . 22 , p=0 . 64 ) . There was a significant difference in correct choices between the three types of upright stimuli ( F2 , 94=4 . 68 , p=0 . 01 ) . The correct choices were significantly more frequent for the upright fish stimuli compared to the upright non-face objects ( p=0 . 01 ) , but not for familiar objects ( p=0 . 24 ) . There was no significant difference between two sets of non-face objects ( p=1 . 0 ) . 10 . 7554/eLife . 24728 . 007Figure 5 . We tested how medaka fish recognise inverted fish and objects . ( A ) Setup for the prism glass test . L , left; R , right; D , dorsal; V , ventral . ( B ) Two sets of non-face-object stimuli were used in the electric shock two-alternative forced-choice ( TAFC ) tasks . The fish had been exposed to the familiar objects since hatching . ( C ) Box plots of percentage correct choices from 6 trials before and after the signals were inverted in the TAFC tasks . Fish were trained to discriminate between two fish or two sets of non-face objects for discrete 36 trials , in addition to 6 inverted trials . The ends of the whiskers represent the minimum and maximum of all of the data . DOI: http://dx . doi . org/10 . 7554/eLife . 24728 . 007
We demonstrate here that medaka fish are able to perform strict IR in both an ecologically relevant paradigm ( mating test ) and a conditioning setting ( electric shock test ) . IR is a complex form of recognition and may require strong evolutionary force . For example , nesting penguins use simpler parameters for parent–chick recognition than do non-nesting species ( Jouventin and Aubin , 2002 ) . Without nest-site information , non-nesting penguins may face higher selection pressure for specific IR ability . Wild medaka are frequently observed in high-density groups ( more than hundreds in one pond , Wang and Takeuchi , personal observation ) and without obvious , constant nest site or territory . Also , medaka spawn every day and appear to have complex social interactions such as courtship ( Walter and Hamilton , 1970 ) , mate-guarding behaviour ( Weir et al . , 2011; Yokoi et al . , 2015 ) , dynamic group-forming and social learning ( Ochiai et al . , 2013 ) . Such frequent and repeated social interactions may induce strict IR ( Tibbetts , 2004 ) and sophisticated cognitive/neural adaptation . Medaka and related species provide an excellent model for investigating how the identity signals and recognition ability has evolved , and how animals link multiple identity signals to different social interactions . They are also widely used as genetic and developmental models for social interaction; for example , TN-GnRH3 neurons function as a gateway for activating mate preferences ( Okuyama et al . , 2014 ) , but we do not yet know whether these neurons regulate sensory perception or the decision-making process after signals are perceived . Here , we tested how medaka fish link identity signals to mating partner or conditioned punishment , both of which are rarely described in animal IR literature . Even though medaka are not monogamous , they still have an astonishing ability to recognise mates , suggesting that IR in mating system may be more common than previously thought . Medaka can successfully differentiate individuals using visual cues alone . More specifically , they use signals around the face for IR . Few animals other than mammals use faces to discriminate individuals , and these species are considered to use relatively simple mechanisms to encode facial features . Two species of cichlid fish use the face to recognise shoal mates or mating partners , and when the facial patterns are exchanged with those of other individuals using digital models , the recognition was found to be based on facial features alone ( Kohda et al . , 2015; Satoh et al . , 2016 ) . A species of wasp uses a number of facial spots to rank dominance , and this ranking can be artificially altered by adding extra patterns ( Tibbetts and Dale , 2004 ) . In our study , even after spots were painted onto the faces of the male medaka , the females still treated the male as a familiar mate . This suggests that medaka are able to tolerant some level of local change during IR . More interestingly , medaka showed the classical face-inversion effect , with fish taking a longer time or failing to recognise the inverted faces , but not the inverted non-face objects . To the best of our knowledge , this is the first study to explore the face-inversion effect in animals other than mammals . The inversion effect is indirect evidence for configural/holistic face processing and has been found in humans ( Maurer et al . , 2002 ) , chimpanzees ( Parr et al . , 1998 ) and sheep ( Kendrick et al . , 1996 ) . These animals not only perceive faces by using internal features , but also make use of configural cues which combine the sum of a number of parts ( Diamond and Carey , 1986; Peirce et al . , 2000; Tomonaga , 2007; McKone and Yovel , 2009 ) . When the faces are upside-down , this configural recognition is impaired , and discrimination times and accuracy deteriorate . The configural recognition process is generally unique to faces and does not appear in other stimuli , although a few special cases have been reported ( Diamond and Carey , 1986 ) . In addition , specialised neural systems are found to encode faces in humans and some other mammals . In humans , inverted faces delay the neural correlates of faces and increase the activity in object-processing areas ( Aguirre et al . , 1999; Haxby et al . , 1999 ) . We do not know whether medaka have specific face processing pathways , but regardless of whether medaka sharecommon mechanisms with mammals , these fish can be an important comparative model . Dorsal parts of the telencephalon ( pallium ) in teleost fish are hypothesised to be related to the mammalian cerebral cortex , including the hippocampus and the pallial amygdala ( O'Connell and Hofmann , 2012 ) , and thus could be a possible candidate brain region for face-recognition processing . It is worth noting that the face inversion effect is not direct evidence for holistic processing ( Valentine , 1988 ) . Further experiments such as the composite task ( Young et al . , 1987 ) , the part-whole task ( Tanaka and Farah , 1993 ) and the part-in-spacing-changed-whole task ( Tanaka and Sengco , 1997 ) are necessary to test whether the animal uses holistic cues to process faces . But at least , we show the possibility that the mechanism for detecting faces may be different from that for other stimuli . Other hypotheses should also take into consideration , such as the within-class discrimination ( Damasio et al . , 1982 ) or the expertise hypotheses ( Diamond and Carey , 1986 ) . The within-class discrimination hypothesis proposes that the special property of faces is due to individual-level discrimination within one type of stimuli , which is a relatively difficult task . For example , discriminating between individual dogs is more difficult than discriminating a dog from a set of mammals such as cats , sheep and monkeys . Although the hypothesis is mostly rejected in humans , it is still possible that it applies to other animals . Another ongoing debate is that humans are face experts and generally use facial stimuli more often than other objects , so the face-specific mechanism is actually expertise-specific . Here , we tested a pair of familiar non-face objects to which the fish had been exposed constantly since they had hatched , in order to control the familiarity level with fish faces . The medaka had a similar level of accuracy when discriminating between familiar objects or medaka faces in the upright orientation , which shows that the familiar objects are sufficient to control for the familiarity and task difficulty in the inversion experiment . The accuracy of discrimination was significantly lower for non-familiar objects compared to that for faces , but we do not know whether the difference was due to task difficulty or familiarity level . In humans , our ability to match unfamiliar faces is surprisingly low . More studies are necessary to understand how familiarity level influences medaka IR , and under which circumstances they can perform IR . One study demonstrated that medaka failed to discriminate fish from their own strain under monochromatic light , whereas they showed strong preference for same-strain mates under normal lighting conditions ( Utagawa et al . , 2016 ) . The males from both strains were familiar to the females , so colours may be important for identifying between strains . We do not know whether medaka show the face inversion effect for IR under monochromatic light as do humans and monkeys ( Dittrich , 1990; Yovel and Duchaine , 2006; McKone and Yovel , 2009 ) . When being successfully recognised by others is favoured by selection , distinctive traits among individuals may evolve to increase the possibility of being identified ( Tibbetts and Dale , 2007 ) . Even though medaka may distinguish each other with visual cues , we cannot find obvious visual traits that vary substantially between individuals . One possible explanation for this is that medaka have eight types of cone opsins , and maximum wavelength absorbance ranging from 356 nm to 562 nm ( Matsumoto et al . , 2006 ) , whereas human vision has just three types of opsins with absorbance ranging from 430 nm to 560 nm . Although difficult to detect for human eyes , there is some level of individual difference in reflectance spectra from medaka bodies ( Figure 1B ) and craniofacial morphology ( Kimura et al . , 2007 ) . On the other hand , even individuals in a group are not especially easy to discriminate; when the signal receivers can benefit from successful individuation , the ability for IR can be favoured by selection ( Johnstone , 1997; Dale et al . , 2001 ) . Specific identification abilities , such as face recognition , may also evolve under strong selection pressure . Another possible explanation is that medaka do not have to link many individuals with fitness-related tasks , or do not have to remember the individual for long time , which may decrease the resources necessary for IR . Distinguishing faces from other species can also be difficult . For example , sheep can discriminate sheep faces with only 5–10% differences ( Tate et al . , 2006 ) , which is difficult for non-experienced humans . Thus , arguing that medaka lack individual-level variation from human’s point of view may be inappropriate . Future research should look at whether medaka can link one or more individuals to multiple ecological tasks , and at whether they can connect identity signals to other fitness-related information such as mate quality or health condition . Rich inbreed lines that are available in medaka could also be useful tools for investigating the heritability of identity signal recognition . For example , in human twin studies , face discrimination ability is heritable for upright faces , but not for inverted faces or other objects ( Mckone and Coltheart , 2010 ) . In the mating test , females accepted familiar males after 3 hr of separation , but not after 24 hr . Even after 24 hr , however , females were able to discriminate between males in the electric shock experiment . One possibility is that the females were able to remember the males , but chose not to accept them . This demonstrates that the mate preferences of medaka females are influenced by familiarity level , but with some flexibility . Another explanation is that repeated conditioning strengthens the memory formation , but as yet there is no evidence related to learning ability in either setting . Some other fish species also prefer familiar mates ( Korner et al . , 1999; Sogabe , 2011; Boyle and Tricas , 2014 ) , and the preference for familiar individuals has been reported to be formed after 4 min ( Dugatkin and Alfieri , 1991 ) or 12 days ( Griffiths and Magurran , 1997 ) and can last for 2 months ( Brown and Smith , 1994 ) , depending on the ecological necessity . In medaka , mate-guarding behaviour by males before and after mating has been reported ( Weir et al . , 2011; Yokoi et al . , 2015 ) and dominant males are able to remain close to the females , which is a possible explanation for such an ephemeral female preference . Medaka females can spawn daily; thus , flexibility in mate preference can ensure mating with the strongest male at the time . In addition , many types of parasitic mating are reported in medaka ( Koya et al . , 2013 ) , which could facilitate strict IR for recognising the correct mate . Although humans and chimpanzees can differentiate faces immediately ( Parr et al . , 2000 ) , other animal models generally require a longer period for conditioning identity signals to reward or punishment . For example , sheep require at least 30 to 40 trials to condition an unfamiliar face to food reward ( Kendrick et al . , 1996 ) . With similar numbers of trials , medaka are able to condition one individual with electric shock punishment . It should be taken into consideration that almost all animal IR experiments are linked with an ecologically relevant task ( e . g . mating in this study ) or a conditioning test ( either positive or negative conditioning ) . Other unavoidable factors such as the physiological condition or stress level of the animals can influence recognition . For example , mating is a sensitive and bipolar procedure which is influenced by the condition and interaction of both individuals . Multiple paradigms with suitable controls may be useful to assess the evidence of convergent IR in animals . Overall , our data suggest that medaka can perform strict IR and that , as in humans , specific visual features such as the face may be more important for IR than others . Medaka also show the classic face-inversion effect , which could indicate that specific processes are involved in recognising faces . It is likely that the mechanism underlying medaka face recognition differs from that in mammals . The application of the rich genetic toolset available for this species , which includes genome editing tools ( such as CRISPR/Cas9 ) and epigenetic methods , will allow more detailed investigation of the specialised cognitive abilities ( Ansai and Kinoshita , 2014; Nakamura et al . , 2014 ) . Even a brain as small as that of the medaka is able to manage such a complex cognitive task . A understanding of how faces are perceptually encoded in simpler models will provide concepts that may be exploited both for the development of new machine face-recognition systems and to explain how the brain processes highly homogenous social information in general . The advantages and limitations of this model compared to mammalian models in face recognition will allow interesting future investigations of convergent systems from phylogenetically distant groups . Other than looking to provide a comparative view on the neurobiology of faces , our future direction will focus on IR in the real world . For example , we would like to study how medaka link individuals to multiple ecological-related topics and how IR shapes their societies and group forming ( Wang et al . , 2015 ) . The evidence gathered in such studies will indicate the evolutionary background in which such sophisticated cognitive process were formed , which is important for all social animals .
A total of 569 adult medaka fish ( Oryzias latipes , drR strain ) , aged 6–18 months , were tested in this study . Fish were maintained as described in Okuyama et al . ( 2014 ) . The animal experiments were performed as approved by the Animal Care and Use Committee of the University of Tokyo ( permit number: 12–07 ) . All efforts were made to minimise suffering according to the NIH Guide for the Care and Use of Laboratory Animals . We sacrificed the fish using a −20° freezer and placed them in a Petri dish for measurement . The reflectance spectra of the body trunks from five medaka were measured by a spectrometer ( FLAME-S-UV-VIS-ES , Ocean Optics , Inc . FL , US ) . A light source ( DH-MINI ) providing UV to visible light output illuminated the probe ( R400-7-SR ) under an angle of 45° to the fish trunk . The reflectance spectra of the fish were recorded with a resolution of 1 nm relative to a white standard ( WS-1 ) with OCEANVIEW software ( Ocean Optics , Inc . FL , US ) . | Being able to recognize each other is crucial for social interactions in humans , as well as many other animals . To humans , faces are the most important body part to differentiate between one another . Humans read the face as a whole , rather than look at parts of the face , which is why it is harder to recognise a face when we see it upside-down , but not when we see an upside-down object . Some other mammals also identify each other by the face and take longer to recognise an upside-down face , but this ability has never been observed in animals other than mammals . Previous research has shown that some fish species can distinguish between individuals . For example , female medaka fish prefer males they have seen before to ‘strangers’ . However , until now , it was not known if they can recognize individual faces , nor how they distinguish a specific male from many others . To see if medaka fish use vision , smell or both cues to recognise mates , Wang and Takeuchi familiarised the fish before the mating test in different settings . In the first group , the male and the female could see each other but were kept in different tanks; in the second group to test odour cues , the male and the female were in the same tank but could not see each other; in the third group , the fish were in the same tank and could see each other; the fish in the fourth group were kept in different tanks and could not see each other . To make sure the fish can recognise and distinguish between fish or objects , Wang and Takeuchi also performed negative conditioning experiments , in which the females had to learn to form an association between a negative stimulus and a specific situation . Wang and Takeuchi found that medaka fish use both vision and smell to distinguish between other fish , but could recognise each other based on vision alone . More specifically , the fish looked at the faces to tell others apart , and even when spots were added to their faces , the fish could still recognise the other . The mekada fish were also able to discriminate between two fish and two objects , but failed the task when the fish images were presented upside-down . However , when two objects were inverted , they were still able to tell the difference . This suggests that just like humans , faces may be special for fish too . This is the first study that shows the face inversion effect in animals other than mammals . A next step will be to compare the different mechanisms between species , and identify the underlying genes and nerve cells responsible for face recognition . This will enable us to better understand social interactions in fish , and enhance our knowledge of how our own ability to recognize faces has changed from an evolutionary point of view . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"neuroscience"
] | 2017 | Individual recognition and the ‘face inversion effect’ in medaka fish (Oryzias latipes) |
Peptide transport plays an important role in cellular homeostasis as a key route for nitrogen acquisition in mammalian cells . PepT1 and PepT2 , the mammalian proton coupled peptide transporters ( POTs ) , function to assimilate and retain diet-derived peptides and play important roles in drug pharmacokinetics . A key characteristic of the POT family is the mechanism of peptide selectivity , with members able to recognise and transport >8000 different peptides . In this study , we present thermodynamic evidence that in the bacterial POT family transporter PepTSt , from Streptococcus thermophilus , at least two alternative transport mechanisms operate to move peptides into the cell . Whilst tri-peptides are transported with a proton:peptide stoichiometry of 3:1 , di-peptides are co-transported with either 4 or 5 protons . This is the first thermodynamic study of proton:peptide stoichiometry in the POT family and reveals that secondary active transporters can evolve different coupling mechanisms to accommodate and transport chemically and physically diverse ligands across the membrane .
Secondary active transporters are integral membrane proteins that couple the energy stored in an ion gradient to drive the uptake of a solute against its concentration gradient ( Nicholls and Ferguson , 2013 ) . This can be accomplished through either a symport mechanism , with the solute being moved in the direction of the driving ion , or antiport , where the solute movement is counter to that of the driving ion ( Shi , 2013 ) . The ion gradients utilised by secondary active transporters include proton ( ΔμH+ ) , sodium ( ΔμNa+ ) , or chloride ( ΔμCl− ) gradients , which are in turn established through the action of the primary ATP driven P- , F- , V- , and A-type ion pumps ( Voet and Voet , 2011 ) . A fundamental characteristic of these systems is that , in general , transport is strictly coupled; the movement of solutes and ions is obligatory and one cannot be transported without the other . If these systems were to operate in a decoupled manner , they would act as leaks and dissipate the ion gradients across the membrane , quickly leading to cell death . Given the strict requirement for coupling solute binding and transport to ion movement , the stoichiometry of these mechanisms is normally a fixed ratio . Examples include the Escherichia coli lactose transporter , LacY , which transports lactose in a symport mechanism with one proton ( Kaback et al . , 2011 ) and EmrE , the small multidrug extrusion transporter , which moves both monovalent and divalent substrates in a 1:2 drug:proton stoichiometry ( Rotem and Schuldiner , 2004 ) . A number of membrane proteins have been identified that recognise multiple structurally and chemically diverse solutes ( Koepsell , 2013; Pelis and Wright , 2014 ) . Prominent among these are the proton coupled oligopeptide transporters or POTs ( Hillgren et al . , 2013; Smith et al . , 2013 ) . POT family transporters are widely distributed within bacterial , fungal , and plant genomes where they are responsible for the uptake of di- and tri-peptides from the external environment ( Daniel et al . , 2006 ) . Mammals contain four POT family transporters , PepT1 ( SLC15A1 ) , PepT2 ( SLC15A2 ) , PHT1 ( SLC15A4 ) , and PHT2 ( SLC15A3 ) . PepT1 and PepT2 are expressed at the plasma membrane , whereas PHT1 and PHT2 are found in lysosomal membranes ( Daniel and Kottra , 2004 ) . Throughout the POT family the transport mechanism and peptide binding site are highly conserved , with bacterial counterparts sharing ∼80% identity to human PepT1 and PepT2 within their peptide binding sites ( Terada and Inui , 2012; Newstead , 2014 ) . All POT family members studied to date transport their substrates into the cell in a coupled symport mechanism , driven by the proton electrochemical gradient . While a number of mutational studies on the mammalian PepT1 and PepT2 transporters address peptide recognition ( Terada et al . , 1996; Fei et al . , 1997 , 1998; Yeung et al . , 1998; Uchiyama et al . , 2003; Luckner and Brandsch , 2005; Kulkarni et al . , 2007; Pieri et al . , 2009 ) , the question of how many protons are coupled to peptide transport remains unresolved; early studies using Caco-2 cell lines derives a ratio of greater than two protons per peptide ( Thwaites et al . , 1993 ) . However due to experimental design , narrowing this figure to a more precise stoichiometry was not possible ( Kottra et al . , 2002 ) . Electrophysiological studies using two electrode voltage clamping ( TEVC ) in Xenopus oocytes in tandem with radio ligand transport assays on non hydrolysable peptide ( D-Phe-L-Gln/Glu/Lys or Gly-Sar ) have reported stoichiometry ratios of 1:1 and 2:1 proton:peptide for neutral/basic and acidic di-peptides respectively for PepT1 ( Fei et al . , 1994; Steel et al . , 1997; Chen et al . , 1999 ) . Similar experiments on PepT2 have given different ratios either D-Phe-L-ala of 2:1 and for D-Phe-L-Glu 3:1 ( Chen JBC 1999 ) or 1:1 for D-Phe-L-Gln/Glu or Lys ( Fei et al . , 1999 ) . Recently , we reported two crystal structures of a bacterial POT family transporter , PepTSt , from Streptoccocus thermophilus , which revealed di- and tri-peptides interacting differently within the binding site ( Lyons et al . , 2014 ) . Whereas the di-peptide L-Ala-L-Phe binds in a horizontal position with respect to the plane of the membrane , whilst the tri-peptide L-Ala-L-Ala-L-Ala resides in a vertical orientation and makes subtly different interactions within the binding site . This raised the interesting and to our knowledge unique proposition , that two different transport mechanisms may have evolved within the same binding site as a way to accommodate a diverse library of peptide ligands , >8000 ( Ito et al . , 2013 ) . To address whether PepTSt could indeed operate using distinct mechanisms to drive di- and tri-peptides , we explored the coupling mechanism between protons and peptide in a reconstituted system , determining the coupling stoichiometries of protons and peptides . We show that whilst tri-peptide import is coupled to three protons , the mechanism for di-peptide import requires at least four and possibly five protons . These results provide further biochemical evidence that POT family transporters do operate via multiple mechanisms for coupling peptide transport to the proton gradient . The ability to couple different numbers of protons to structurally and chemically diverse ligands may explain how the POT family is able to accommodate and transport such a large library of di- and tri-peptides .
To address the question of stoichiometry , we developed a sensitive and robust assay to follow the proton movement during the transport cycle . Previously published peptide transport assays tend to follow peptide uptake using a radiolabeled peptide substrate . To instead follow proton movement , we monitored the internal pH with the ratiometric pH sensitive fluorophore , pyranine ( Figure 1A and Figure 1—figure supplement 1 ) . We first performed a number of control experiments to see whether our system could indeed follow proton coupled peptide transport into a liposome . Acidification of the lumen was only observed in the presence of peptide and a large hyperpolarized ( negative inside ) membrane potential imposed by adding the potassium ionophore valinomycin in the presence of a K+ gradient ( Figure 1B , C and Figure 1—figure supplement 2 ) . We did not see such acidification either in the absence of valinomycin or in the presence of amino acid ( alanine ) or tetra peptide ( Ala-Ala-Ala-Ala ) , confirming that this transporter is indeed specific for di- and tri-peptide substrates ( Figure 1C ) . PepTSt has been shown previously to transport Ala–Ala with a 10-fold higher activity than Ala-Ala-Ala as judged by IC50 values using tritiated Ala–Ala as a reporter ( Solcan et al . , 2012 ) . We confirmed this 10-fold difference in uptake by using our proton-based assay , which can now report the direct uptake of any transported substrate rather than inferring substrate specificity through inhibition of Ala–Ala ( Figure 1—figure supplement 3 ) . This assay is also useful to study poor competitors of di-alanine for example , di-lysine where no competition could be observed previously ( Solcan et al . , 2012 ) . Using this assay , we can now observe uptake of this substrate indirectly by monitoring the coupled proton movement ( Figure 1C ) . 10 . 7554/eLife . 04273 . 003Figure 1 . Monitoring peptide-coupled proton transport using the pH sensitive dye , pyranine . ( A ) Experimental setup to monitor proton flux . PepTSt is reconstituted into liposomes loaded with pyranine and a high concentration of potassium ions ( 120 mM ) , the external solution contains peptide and a low potassium concentration ( 1 . 2 mM ) . On addition of valinomycin , the membrane becomes highly potassium permeable , generating a hyperpolarised membrane potential ( negative inside ) this drives the uptake of peptide with protons , protonating the pyranine dye and altering its fluorescent properties . ( B ) Representative pyranine fluorescence traces produced from the set up described in ( A ) indicating that acidification of the liposomal lumen only occurs in the presence of a valinomycin ( black arrow ) -induced membrane potential with peptide . The Y axis indicates the fluorescence ratio as stated in the methods . ( C ) PepTSt can only transport di- and tri-peptides . To initiate transport , valinomycin was added to all experiments at the time indicated by the black arrow and the external substrate concentration was 0 . 2 mM . Data were normalised to the first time point for ease of comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 00310 . 7554/eLife . 04273 . 004Figure 1—figure supplement 1 . Representative raw data of the pyranine fluorescence traces . Raw data produced from the set up described in ( Figure 1A ) indicating that acidification of the liposomal lumen only occurs in the presence of a valinomycin ( black arrow ) -induced membrane potential with peptide , this data were used to make Figure 1B . ( B ) Shows the data from ( A ) but shown as a ratio of the spectra obtained at 460/415 . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 00410 . 7554/eLife . 04273 . 005Figure 1—figure supplement 2 . PepTSt POPE:POPG proteoliposomes can hold a pH gradient of 1 unit . ( A ) Potential proton leakage was monitored using pyranine in a system with internal pH at 6 . 8 and external pH at 5 . 8 . The proton gradient was collapsed by the addition of CCCP . ( B ) as ( A ) but internal pH was 7 . 5 and external pH was 6 . 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 00510 . 7554/eLife . 04273 . 006Figure 1—figure supplement 3 . Transport strength of Ala–Ala vs Ala-Ala-Ala . Increasing concentrations of peptide both Ala-Ala-Ala ( Left ) and Ala-Ala ( right ) lead to an increase in the fluorescence change until a maximum level is reached . The mid point of the fluorescence change between maximum decrease and that with no peptide for Ala-Ala-Ala is ∼125 µM and 10 µM for Ala–Ala , an approximate 10-fold difference . ( Concentrations used for tri-ala were 0 , 0 . 1 , 1 , 5 , 10 , 25 , 50 , 125 , 250 , 500 , 1000 , 2000 µM and di-ala 0 , 0 . 01 , 0 . 1 , 1 , 5 , 10 , 25 , 50 , 100 , 250 , 1000 , 2000 µM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 006 Since PepTSt is an electrogenic transporter , we predict that imposition of a membrane potential in the presence of a pH gradient should drive uphill substrate transport . By loading the liposomes with high concentration of peptide and imposing a large hyperpolarising membrane potential ( negative inside ) , we observe acidification of the lumen , indicating that the voltage can drive PepTSt-mediated transport against a 100-fold peptide gradient ( Figure 2A ) . Importantly , we can also manipulate this system to see protons leaving the liposomal lumen as would be expected under an oppositely orientated membrane potential ( positive inside ) . With an assay system set up where we could drive transport in predicted directions , we were now in a position to assess proton:peptide stoichiometry by measuring the equilibrium potential for proton flux using pyranine at a series of membrane voltages , set at the start of the assay with the appropriate potassium ion concentration gradient and the addition of valinomycin . This type of assay was used previously to address the stoichiometry of a lysosomal Cl−/H+ antiporter , CLC-7 ( Graves et al . , 2008 ) . 10 . 7554/eLife . 04273 . 007Figure 2 . Peptide transport depends on the imposed voltage . Proteoliposomes in the presence of a 100-fold peptide gradient ( 0 . 1 mM outside and 10 mM inside ) and no pH gradient ( pH 6 . 8 both inside and outside ) . No proton flux occurs until the transmembrane potential is shunted by addition of valinomycin ( green trace , val added at black arrow ) . PepTSt can drive transport of peptides ( and protons ) against this gradient into the interior of the liposome using the proton electrochemical gradient when a negative voltage is imposed by valinomycin addition . ( negative inside—yellow line ) . Further addition of peptide ( 0 . 5 mM ) results in additional uptake ( grey arrow ) . When a large ( inside ) positive voltage is applied ( same peptide and pH gradients ) , peptides ( and protons ) exit the liposome ( pink line ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 00710 . 7554/eLife . 04273 . 008Figure 2—figure supplement 1 . Derivation of the transport equation for a proton peptide transporter . We assume that transport occurs with a fixed stoichiometry as reflected in ( A ) . Pep indicates the neutral peptide and the subscripts , ‘out’ and ‘in’ refer to peptide or protons outside or inside the liposome membrane , ‘n’ is the number of protons transported per cycle and ‘m’ is the number of peptides . We seek to determine n/m the stoichiometric ratio of protons:peptide . For the reaction shown in ( B ) , μ is the chemical potential of the species , R is the universal gas constant , T is the temperature in ( K ) , F is the Faraday constant , ZH is the proton charge , ΔΨ is the voltage difference across the membrane , where ΔΨ = Ψin − Ψout . This is equivalent to the sign convention that the outside of the liposome is defined as ground ( Ψout = 0 ) . At equilibrium ( C ) , the equation can be rearranged ( D ) to yield the equilibrium potential , ΔΨ , in terms of the values of pH and [peptide] and the stoichiometric ratio . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 008 We assume that PepTSt operates via the coupled mechanism , nHout + mPepout ⇔nHin + mPepin , where the relative stoichiometry of protons:peptide is n/m . For this coupled system , the equilibrium potential is defined as the voltage at which there is no net substrate flux . This voltage , also known as the reversal potential ( ΔΨ ) , is independent of the reaction mechanism . Rather , it depends only on the concentrations of protons and substrate and on the coupling stoichiometry with the form: ΔΨ = 60{[pHin − pHout] − m/n log ( [Pep]in/[Pep]out ) } , where m and n are the stoichiometric coefficients in the chemical reaction above and ΔΨ is in mV ( see derivation in Figure 2—figure supplement 1 ) . For a given combination of pH and peptide gradients this equation predicts a voltage at which no net pH change will occur; voltages above and below that value should produce inward or outward proton flux , depending on the voltage . Conversely , if at a series of voltages ( set using K+/valinomycin ) , we observe acidification/no flux/alkalinzation , we can derive the relative stoichiometry of PepTSt for protons and peptide . We performed such experiments for the neutral peptide , Ala-Ala-Ala with no net pH difference between the inside and outside of the liposome and a 100-fold peptide gradient ( higher concentration inside ) and observed an absence of proton flux at a membrane potential of−40 mV ( inside negative ) which corresponds to a 3:1 proton:peptide stoichiometry ( Figure 3A and Figure 3—source data 1 ) . In contrast , voltages corresponding to reversal potentials for stoichiometries of 2:1 and 4:1 produced clearly distinguishable inward and outward fluxes respectively , strongly pointing to a 3:1 stoichiometry for the transporter . A very different combination of proton and peptide gradients , where we now included a proton gradient , also ( pH more acidic outside ) gave the same stoichiometry of 3:1 for the same tri-alanine peptide ( Figure 3B ) . We also obtained this three proton:peptide stoichiometry for a different tri-peptide substrate , Ala-Leu-Ala ( Figure 3C ) . 10 . 7554/eLife . 04273 . 009Figure 3 . Tri-peptides are co-transported with three protons . The potassium gradient across the liposomes was varied in order to set the desired voltages ( on valinomycin addition , black arrow ) to achieve no net proton movement ( indicated by the black line at FR of 1 . 0 ) . The number next to the voltages are the proton:peptide stoichiometry that would reverse at that voltage . The internal peptide concentration ( 10 mM ) ( Tri-ala for A and B , and Ala-Leu-Ala for C ) was 100-fold above that of the external concentration and the pH was inside 6 . 8 , outside 6 . 8 ( for A and C ) and 6 . 0 ( for B ) . Representative traces are shown for each experiment , which were repeated at least three independent times . The line graph for each experiment represents the mean change in fluorescence at time point 60 s and S . E . M . is indicated . Grey arrows indicate the addition of more peptide , to show that the system can be further manipulated . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 00910 . 7554/eLife . 04273 . 010Figure 3—source data 1 . Table showing the conditions used to calculate the stoichiometry of transport . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 010 We went on to determine the proton:peptide stoichiometry of PepTSt when transporting di-peptide substrates . However , when we performed experiments with the same gradients as our initial tri-peptide measurements only now with the neutral peptide Ala–Ala , we still observed transport at a voltage of −40 mV , where before we saw no net proton flux for tri-alanine . Transport is still also occurring at −30 mV which under these conditions would correspond to a proton:peptide stoichiometry of 4:1 and where we saw proton influx for tri-alanine ( Figure 4A ) . Therefore , surprisingly , the stoichiometry of protons to peptides in PepTSt appears to be different for tri-Ala as compared with di-Ala . Further experiments to try to pin down the number of protons being co-transported with di-alanine lead to slightly ambiguous results , as increasing proton:peptide stoichiometries predict decreasing increments in reversal potential . These are , in turn , harder to generate reproducibly with our valinomycin/K+ system . We performed experiments with voltages set at reversal potentials predicted for symport ratios of 5 and 6 protons:peptide and proton flux was minimal at −20 mV ( 6 protons ) but it is hard to be able to fully distinguish . Importantly , the system can still be driven in the opposite direction with higher voltages ( 0 mV , Figure 4B ) . Regardless of whether the actual stoichiometry is 5 or 6 protons:peptide , these experiments strongly suggest that the stoichiometry is higher for di-peptides than for tri-peptides . We confirmed this higher ratio for di-peptide transport using the substrate Ala–Phe ( Figure 4C ) . Again transport is clearly observed at a membrane potential at −30 mV , so the stoichiometry for di-peptide transport by PepTSt is greater than four protons , clearly different from that of tri-peptides . 10 . 7554/eLife . 04273 . 011Figure 4 . Di-peptide transport requires more protons than tri-peptide . The potassium gradient across the liposomes was varied in order to set the desired voltages ( on addition of valinomycin , black arrow ) to achieve no net proton movement ( indicated by the black line at FR of 1 . 0 ) . Only voltages that indicate a stoichiometry of proton:peptide of greater than 5:1 showed either no net movement of protons or reversal for both Ala–Ala ( A , B ) and Ala–Phe ( C ) di-peptides . Representative traces are shown for each experiment , which was repeated at least three independent times . The line graph for each experiment represents the mean change in fluorescence at time point 60 s and S . E . M . is indicated . Grey arrows indicate the addition of more peptide , to show that the system can be further manipulated . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 011 The reversal potential equation above predicts that if ΔΨ = 0 , then [Pepin]/[Pepout] = ( [H+out]/[H+in] ) n , where n is the number of protons transported per peptide . Therefore , if our results are truly indicative of higher coupling ratios for di-peptides than for tri-peptides , the same proton electrochemical gradient should accumulate di-peptides to a higher steady-state concentration than tri-peptides . We tested this prediction by measuring the uptake of radiolabeled di- and tri-peptides in the presence of a fixed , 1-unit pH gradient ( acid outside ) . As shown in Figure 5 , we find dramatically higher uptake of the di-peptide in this gradient , conclusively supporting the idea that different length peptides couple to the proton gradient with differing stoichiometries . 10 . 7554/eLife . 04273 . 012Figure 5 . Steady-state accumulation of di- vs tri-alanine . Peptide transport was driven by an inwardly directed proton gradient in saturating amounts of peptide . Uptake was measured via scintillation counting using radiolabeled peptides ( 3H for di-alanine and 14C for tri-alanine ) and converted to pmols peptide transported per unit time . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 012 A different coupling stoichiometry for di- vs tri-peptides raises an interesting question of whether different amino-acid side chains within the transporter are required for di-peptide vs tri-peptide transport . PepTSt contains six-protonatable side chains within its binding site ( Glu 22 , 25 , 299 , 300 , 400 , and K126 , Figure 6A ) . All of these with the exception of Glu299 are conserved across the PTR family from bacteria through to mammalian PepT1 and PepT2 . Previous biochemical studies have shown that in PepTSt this non-conserved residue is likely to be involved in structural and/or stability features specific to this protein as mutation of this residue results in no expression of the protein . Glu400 and Lys126 are likely to form a salt bridge that stabilises the outward open confirmation of the transporter , a feature that would be conserved regardless of the substrate . Glu300 has been shown to interact with both a di-peptide and a tri-peptide substrate and therefore likely to be involved in the transport mechanism for both di- and tri-peptides and has been shown to be important for di-alanine transport previously ( Solcan et al . , 2012; Doki et al . , 2013 ) . This leaves Glu22 and Glu25 as candidates for differential effects on di- and tri-peptide transport . Previously these residues have been shown to be important for proton coupling for di-alanine , however , here we also found that mutating either of these residues to alanine yielded proteins unable to catalyse proton coupled tri-alanine transport ( Figure 6B ) . Therefore , despite different proton:peptide stoichiometries are apparently required for di- and tri-alanine transport , all five protonatable side chains within the binding site of PepTSt are likely to be important for the transport mechanism . 10 . 7554/eLife . 04273 . 013Figure 6 . Model for proton:peptide symport . ( A ) Crystal structure of PepTSt with bound Ala–Phe peptide ( orange ) ( PBD 4D2C ) showing the protonatable side chains within the binding site ( green ) . Helices TM5 and TM8 have been removed for clarity . ( B ) E22A and E25A variants of PepTSt are unable to couple proton movement to the transport either di-alanine or tri-alanine . ( C ) Ala–Phe and Ala-Ala-Ala adopt different orientations within the binding site of PepTSt . Two views of PepTSt shown in the plane of the membrane and rotated 90° , with Ala-Ala-Ala ( blue ) and Ala–Phe ( magenta ) shown as sticks . Helices TM5 and TM8 have been removed for clarity . ( D ) Model for proton:peptide symport in PepTSt . Di-peptides transport requires at least four protons , whereas tri-peptides require only three , suggesting this is the lowest number of protons required to drive the conformational changes required for alternating access transport . Essential residues are indicated; residues involved in the potential stabilising salt bridges are labelled in blue and red , whereas protonatable side chains are labelled in purple . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 01310 . 7554/eLife . 04273 . 014Figure 6—figure supplement 1 . Model of the outward facing state of PepTSt with bound Ala-Ala-Ala . The position of the tri-alanine peptide ( mesh ) would obstruct the formation of the predicted salt bridge between residue Lys126 and Glu400 . DOI: http://dx . doi . org/10 . 7554/eLife . 04273 . 014
A fundamental aspect of any transport mechanism is its coupling stoichiometry , how many ions are moved for each molecule of solute , as this information is necessary to generate reasonable mechanistic models for the transporter under study . Previous electrophysiological recordings on PepT1 and PepT2 have focused their attention on di-peptide substrates and suggest a 1:1 proton:peptide stoichiometry in PepT1 for neutral peptides and either 1:1 or > in PepT2 ( Smith et al . , 2013 ) . Recent crystal structures and functional data on a bacterial POT family transporter , PepTSt , a homologue of PepT1 and PepT2 , revealed that peptides could adopt different orientations within the binding site ( Figure 6C ) . Whereas tri-alanine was observed adopting a vertical position and coordinated by 4 hydrogen bonds , the di-peptide L-Ala-L-Phe was held in a more horizontal position and coordinated through a more extensive network of interactions involving electrostatic interactions with conserved side chains from the N- and C-terminal bundles ( Lyons et al . , 2014 ) . This raised the possibility that PepTSt could transport peptides using two different mechanisms operating within the same binding site and has important implications for understanding proton coupled transport more generally within the POT family . Here , we used a reconstituted proteoliposome system to accurately measure reversal potentials for peptide-coupled proton fluxes and found that di- and tri-peptides are transported using different proton stoichiometries . Assuming that our measurements on two sets of distinct di- and tri-peptides reflect the stoichiometries in general , our new data add to our previous multiple binding mode model by showing that tri-peptides are transported using three protons , whereas di-peptides are transported using four or possibly even five protons per cycle . It is important at this stage to highlight that our experiments cannot discriminate between protons that come through the transporter and those that may come through bound to the peptide . However , even if some protons are being moved on the peptide as opposed to being required to rearrange interaction networks during transport , our results still demonstrate that different numbers of protons are moved during the coupled transport of neutral di- vs tri-peptides , which we ascertain establishes a fundamental difference in the way this protein handles these two ligands . Interestingly , all five of the protonatable residues within the binding site are required for transport as none could be mutated and still allow for the transport of either peptide . This could be due to movement of the three protons within the binding site to different side chains , perhaps to help re-orientate the tri-peptide as the transporter transitions through its conformational cycle . Indeed if you overlay the crystal structure of the tri-alanine structure with a model of the outward open structure ( Figure 6—figure supplement 1 ) , tri-alanine in this position would obstruct the closing of the intracellular gate through disruption of the intracellular stabilising salt bridge ( formed between Lys126 ( TM4 ) and Glu400 ( TM10 ) ) . Therefore , we suggest that for tri-alanine to be transported across the membrane , it is likely to undergo a change in its vertical binding position , to allow for closure of the transporter . The question then arises as to the functional role of the protons in the transporter . We have previously identified six-protonatable side chains present in the binding site of PepTSt ( Figure 6A ) , with all but Glu299 being conserved across the POT family ( Solcan et al . , 2012 ) . It would be tempting therefore , on the basis of simplicity , to predict that our data suggest only three side chains are protonated during tri-peptide transport , and four/five in di-peptide transport . However , we do not believe this to be the case . Our attempts to systematically remove the protonable side chains Glu22 and Glu25 on TM1 , which form part of the conserved ExxERF motif but do not interact with either of the di-peptide or tri-peptide in the crystal structures ( Lyons et al . , 2014 ) , resulted in inactive transporters ( Figure 6B ) . Previous studies have shown that mutating any of the remaining side chains; Lys126 , Glu300 , and Glu400 also result in inactive proton coupled transport ( Solcan et al . , 2012 ) . We conclude from these results that all of the protonatable side chains are required for transport regardless of substrate . However , the observation that tri-peptides can be moved using only three protons delineates the minimal number of de-protonation events that are required to drive the conformational changes that re-orientate the binding site . In this model , the remaining protonatable groups remain proton-bound throughout the transport cycle when tri-peptides are transported . On the basis of the new data presented here , we can add further mechanistic insight into our earlier model for peptide transport that we have summarised in Figure 6D . Alternating access within the POT family is physiologically driven by the proton electrochemical gradient , with defined conformational states stabilised through conserved pairs of salt bridges that act to coordinate the opening and closing of the intracellular and extracellular gates ( Newstead , 2014 ) . Starting from the outward open conformation the binding site is accessible to the extracellular side of the membrane , and the intracellular gate is closed and stabilised by a possible salt bridge between Lys126 ( TM4 ) and Glu400 ( TM10 ) . Functional studies have revealed that in another bacterial POT family transporter , from Geobacillus kaustophilus , GkPOT , that the equivalent glutamate to Glu300 is protonated and may be required to allow the binding of peptide ( Doki et al . , 2013 ) . Considering the minimal three-proton model for tri-peptide transport we propose , it seems reasonable to suggest that proton transfer from Glu300 to Glu400 during transport may occur to couple closing of the extracellular gate with opening of the intracellular one . This would account for one proton . The other two we suggest may come from Lys126 and either of Glu22 or Glu25 . Our evidence is that in previous functional studies we showed that either of these side chains could be mutated to alanine with only a slight reduction in counterflow transport but complete loss of proton coupled peptide uptake ( Solcan et al . , 2012 ) , behaviour classically used to identify side chains required for proton coupled uptake . In the case of di-peptides , additional deprotonation is clearly required . We conjecture that this is the result of the tighter coordination observed in the di-peptide complex structure compared to that for the tri-peptide ( Figure 6C ) and maybe one reason why this binding site is so sensitive to mutation in our assays . An adaptable coupling mechanism , such as we propose , might have been an important component that enabled the POT family to adapt its binding site to accommodate structurally and chemically diverse molecules for nutritional assimilation . Whilst in bacterial , fungi , and mammals POT family homologues are responsible for peptide uptake , in plants this family has evolved to recognise widely diverse molecules , including nitrate , glucosinylates , hormones , and peptides ( Léran et al . , 2014; Sun et al . , 2014; Parker and Newstead , 2014 ) . This may explain why mammals use the POT family homologues PepT1 and PepT2 , as coupling peptide transport to the proton gradient appears to facilitate a promiscuous binding site that can adapt to chemically diverse side chain groups more easily than sodium coupled transporters , which require well-defined binding sites for the cation .
For reconstitution , PepTSt purified in the detergent DM ( Solcan et al . , 2012 ) was mixed in a 60:1 ratio ( lipid:protein ) with lipid vesicles composed of a mixture of POPE and POPG ( in a 3:1 ratio ) . These lipids were chosen as they had been previously reported to form proton tight liposomes ( Tsai and Miller , 2013 ) . We confirmed this in our liposomes ( Figure 1—figure supplement 1 ) . The protein:lipid mix was diluted into a large volume of reconstitution buffer ( 50 mM potassium phosphate 6 . 8 ) , and proteoliposomes were harvested by ultracentrifugation ( >200 , 000×g ) for 3 hr . Pelleted liposomes were resuspended at 0 . 5 µg/µl ( protein ) and dialysed extensively against reconstitution buffer ( 24 hr with two changes of buffer ) . Proteoliposomes were recovered and subjected to three rounds of freeze thawing before storage at −80°C . Proteoliposomes were harvested and resuspended in inside transport buffer ( 5 mM Hepes pH 6 . 8 , 2 mM MgSO4 , 1 mM Pyranine ( trisodium 8-hydroxypyrene-1 , 3 , 6-trisulfonate ) also containing the desired potassium concentration ( KCl ) and peptide concentration ) and subjected to three rounds of freeze thaw in liquid nitrogen and then extruded through a 0 . 4-µm membrane . Pyranine is a fluorescent pH indicator dye which is water soluble and can be trapped within liposomes . Acidification of the lumen of the liposome is indicated by a decrease in the ratio of fluorescence measured at 510 when excited at either 460 or 415 . After extrusion the liposomes were harvested and excess pyranine removed through gel filtration using a superdex-25 column pre-equilibrated in inside transport buffer without pyranine . For the assays the liposomes were diluted into external transport buffer in a 0 . 85-ml micro cuvette with a small magnetic flea ( 5 mM HEPES pH 6 . 8 or 5 mM MES pH 6 . 0 , 2 mM MgSO4 and the desired amount of KCl to obtain the desired potassium gradient , ionic strength was kept equal across the liposome using NaCl ) . Transport was initiated using 1 µM valinomycin , and fluorescence was read at excitation 460 and 415 emission 510 in a Cary eclipse fluorimeter with continual stirring . To examine the data , the data were exported into Graphpad and the fluorescence was measured at 510 excitation 460 divided by that measured at 415 excitation , indicated at FR in the results . To compare multiple conditions , the data were normalised to 1 ( from the first reading ) for each experiment . Representative raw data are shown in Figure 1—figure supplement 2 . For each individual experiment , the mean value was calculated from 55 to 65 s and this was repeated for each replicate ( minimum of three ) to generate an overall mean and S . E . M , which is plotted as a line graph on each figure . Proteoliposomes were harvested and resuspended in inside buffer ( 5 mM HEPES pH 7 . 5 , 2 mM MgSO4 , 75 mM KCl ) and subjected to three rounds of freeze thaw in liquid nitrogen and then extruded through a 0 . 4-µm membrane . For the assays , the liposomes were diluted into external transport buffer ( 5 mM MES pH 6 . 5 , 2 mM MgSO4 75 mM KCl ) . Peptide , to a final concentration of 0 . 5 mM containing a tracer amount of either 3H-di-alanine ( specific activity 30 Ci/mmol ) or 14C-tri-alanine ( specific activity 55 mCi/mmol ) , was added with 1 µM valinomycin and time points taken . The assays were performed at 22°C . Time points were taken and stopped by addition into 2 ml 0 . 1 M LiCl and filtering immediately through a 0 . 4-µm membrane in a vacuum manifold . The filters were washed twice with 2 ml of LiCl prior to scintillation counting in Ultima Gold ( Perkin elmer ) . The amount of peptide transported into the liposomes was calculated based on specific activity for each peptide as detailed by the manufacturer and counting efficiency for the radioisotope in Ultima Gold counted in a Wallac scintillation counter ( 3H 45% counting efficiency , 14C 98% ) . Experiments were performed four times to generate an overall mean and S . E . M . | The cell membrane encases cells and functions as a protective barrier . Although this has the benefit of preventing harmful substances from entering a cell , it also keeps beneficial molecules out . The cell membrane therefore contains a system of different ‘gates’ , called transporters , through which selected supplies can pass . One large family of transporters , found in bacteria , mammals , and plants , is the ‘proton coupled oligopeptide transporter’ family , called POTs for short . These transport over 8000 types of small peptide molecule , each of which is made up of two or three smaller molecules called amino acids . The energy for this transport process is gained by simultaneously transporting charged ions called protons with the peptides . Because these transporters also recognize and transport various drugs , they are currently being investigated to discover whether they could be manipulated to increase how much of a drug is taken up into cells . It remains unknown how the POT family of transporters imports so many different small peptides across the cell membrane , or how many protons are needed to transport a peptide . A study published earlier in 2014 has nevertheless provided some hints: it appears that small peptides adopt different shapes when bound to a bacterial POT transporter depending on whether they consist of two or three amino acids . This suggests that two different transport mechanisms operate from the same binding site , which may account for the wide variety of molecules that can be transported . In a follow up to this work , Parker et al . , including some of the researchers involved in the earlier 2014 work , now look in detail at how many protons this bacterial transporter uses to import these small peptides . This reveals that while the transport of peptides made of three amino acids requires three protons to also be moved through the transporter , the transport of peptides containing two amino acids requires four , or possibly five , protons . This challenges previous findings that these transporters transport one peptide for every proton , and further supports the idea that a single transporter can use more than one method to bind to and transport molecules . Whether other membrane transporters , particularly the human versions of the POT family , share this ability remains an open question . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2014 | Thermodynamic evidence for a dual transport mechanism in a POT peptide transporter |
The 96-nm axonemal repeat includes dynein motors and accessory structures as the foundation for motility of eukaryotic flagella and cilia . However , high-resolution 3D axoneme structures are unavailable for organisms among the Excavates , which include pathogens of medical and economic importance . Here we report cryo electron tomography structures of the 96-nm repeat from Trypanosoma brucei , a protozoan parasite in the Excavate lineage that causes African trypanosomiasis . We examined bloodstream and procyclic life cycle stages , and a knockdown lacking DRC11/CMF22 of the nexin dynein regulatory complex ( NDRC ) . Sub-tomogram averaging yields a resolution of 21 . 8 Å for the 96-nm repeat . We discovered several lineage-specific structures , including novel inter-doublet linkages and microtubule inner proteins ( MIPs ) . We establish that DRC11/CMF22 is required for the NDRC proximal lobe that binds the adjacent doublet microtubule . We propose that lineage-specific elaboration of axoneme structure in T . brucei reflects adaptations to support unique motility needs in diverse host environments .
Flagella ( also called cilia ) are hair-like structures that protrude from the surface of eukaryotic cells and perform motility and signaling functions ( Smith and Rohatgi , 2010 ) . These activities are essential for health , development and reproduction in humans and other multicellular organisms and to power movement of protists , including microbial pathogens that afflict nearly one billion people worldwide and present an economic burden as agricultural pests ( Langousis and Hill , 2014; Gerdes et al . , 2009; Ibanez-Tallon , 2003; Anvarian et al . , 2019 ) . The structural basis for the flagellum is the axoneme , and in motile flagella the axoneme typically has a ‘9+2’ arrangement , consisting of 9 doublet microtubules ( DMTs ) arrayed symmetrically around a pair of singlet microtubules , with radial spokes ( RS ) extending inward from each DMT contacting the central pair ( Khan and Scholey , 2018 ) . Axoneme beating is driven by dynein motors and associated structures arranged in a repeating unit of 96-nm periodicity along each DMT . This 96-nm axonemal repeat is thus the foundational unit of motility for eukaryotic flagella . Canonical features of the repeat are four outer arm dyneins ( OAD ) ( each having two or three motor domains , depending on species ) , seven inner arm dyneins ( IAD ) ( one , IAD-f , having two motor domains and the others having a single motor domain ) , the nexin dynein regulatory complex ( NDRC ) inter-doublet linkage , and two or three RS ( Porter and Sale , 2000 ) . The most proximal IAD in the 96nm repeat , IAD-f , is distinguished from other IADs by having two motor domains , a large Intermediate Chain/Light Chain ( IC/LC ) complex that connects to the OAD and the NDRC , and extra connections to the A-tubule ( Nicastro et al . , 2006; Heuser et al . , 2012a ) . Within each 96-nm repeat , dynein motors are permanently affixed to the A-tubule of one DMT and use ATP-dependent binding , translocation and release of the B-tubule on the adjacent DMT to drive microtubule sliding ( Gibbons and Rowe , 1965 ) . DMT attachment to the basal body at one end , together with ATP-independent connections , called nexin links , between adjacent DMTs , limits sliding and therefore causes DMTs to bend in response to dynein activity ( Satir , 1968; Satir et al . , 2014; Holwill and Satir , 1990 ) . Precise , spatiotemporal coordination of dynein activity on different DMTs enables the bend to be propagated along the length of the axoneme , giving rise to axonemal beating ( Satir , 1968; Lin and Nicastro , 2018 ) . RS , together with the NDRC and the IAD-f-IC/LC complex , are thought to provide a means for transmitting mechanochemical signals across the axoneme as part of a complex and as yet incompletely understood system for regulating dynein activity ( Porter and Sale , 2000; Satir et al . , 2014; King , 2018; Viswanadha et al . , 2017 ) . Recent advances in cryo electron tomography ( cryoET ) have made high-resolution , 3D structural analyses of the 96-nm repeat possible , providing insights into mechanisms of axoneme assembly and motility ( Nicastro et al . , 2006; Lin and Nicastro , 2018; Bui et al . , 2009; Oda et al . , 2014a; Jordan et al . , 2018 ) . However , such analyses have been limited to a restricted number of cell types and phylogenetic lineages . In particular , there has been no such analysis of the 96-nm repeat in any member of the Excavates ( Figure 1 ) , which includes several human and agricultural pathogens of importance to global public health . Consequently , we lack understanding of the full range of structural foundations for axoneme assembly and motility , and what structural variations underlie lineage-specific beating patterns observed in different organisms . For pathogens , such variations present potential therapeutic targets . African trypanosomes , Trypanosoma brucei ( T . brucei ) and related species , are parasitic protists in the Euglenozoa branch of the Excavates ( Figure 1 ) ( Koonin , 2010 ) . They are medically and economically important pathogens of humans and other mammals ( Langousis and Hill , 2014 ) . Critical to T . brucei infection of a mammalian host ( Shimogawa et al . , 2018 ) and to their transmission via a tsetse fly vector ( Rotureau et al . , 2014 ) , is motility of these parasites within and through host tissues . Motility of trypanosomes is driven by a single flagellum that is laterally connected to the cell body along most of its length ( Figure 2A ) ( Langousis and Hill , 2014; Heddergott et al . , 2012 ) . The T . brucei flagellum consists of a 9+2 axoneme and a lineage-specific extra-axonemal structure , termed the paraflagellar rod ( PFR ) , which runs alongside the axoneme for most of its length ( Langousis and Hill , 2014; Hughes et al . , 2012; Koyfman et al . , 2011; Cachon et al . , 1988 ) . While the PFR exerts influence on the axoneme ( Koyfman et al . , 2011; Santrich et al . , 1997 ) , motility itself is driven by axoneme beating , which is transmitted directly to the cell , deforming the cell membrane and underlying cytoskeleton as the waveform propagates along the axoneme ( Sun et al . , 2018 ) . Unlike most organisms , trypanosome axoneme beating propagates from the distal tip to proximal end in a helical wave , creating torsional strain and causing the cell to rotate on its long axis as it translocates with the flagellum tip leading ( Heddergott et al . , 2012; Walker , 1961; Walker and Walker , 1963; Rodríguez et al . , 2009 ) ( Videos 1 and 2 ) . In essence , the entire cell rotates like an auger as it moves forward . This distinctive form of locomotion provides advantages for moving in viscous environments ( Jahn and Bovee , 1968; Bargul et al . , 2016 ) such as within human and fly tissues , and gives the genus its name , as Trypanosoma combines the Greek words for auger ( trypanon ) and body ( soma ) ( Gruby , 1843 ) . The combination of unusual locomotion mechanism , unique connections to other structures , and adaptation to diverse environmental conditions , suggests that the 96-nm repeating unit of the trypanosome axoneme might harbor lineage-specific elaborations . To investigate this possibility , we employed cryoET and sub-tomogram averaging to determine the 3D structure of the T . brucei 96-nm axonemal repeat . We report the 96-nm axonemal repeat structure for wild type parasites in bloodstream ( BSF ) and procyclic ( PCF ) stages , and for an RNAi knockdown targeting the CMF22/DRC11 subunit of the NDRC . Our results reveal lineage-specific adaptations , including novel inter-doublet linkages and microtubule inner proteins ( MIPs ) . We also identify an NDRC subunit involved in inter-doublet connections between adjacent DMTs . We propose that lineage-specific adaptations to the 96-nm repeat may support the unique motility needs of these pathogens .
A critical element of defining any structure is to ensure the sample is pristine . Our analyses demonstrated that flagellar skeletons purified from bloodstream form ( BSF ) trypanosomes are intact , including intact PFR , basal body and distal tip with uniform length distribution and a mean length of 25 . 2 , + /- 3 . 5 µm ( Figure 2B–F ) . Next it is critical that freezing does not distort the sample . A single zero-degree tilt image of a flagellum embedded in ice demonstrated that the axoneme , PFR and axoneme-PFR connectors remain intact following plunge freezing ( Figure 2G ) . Having established high quality of vitrified samples , tilt series were collected from the center part of full-length flagella , spanning the middle third between the basal body and tip ( Figure 2B ) . Major axonemal and PFR structures were resolved in slices through a single tomogram ( Figure 2H , I , Video 3 ) , indicating the 3D structure is well-preserved and relatively uncompressed ( Figure 3—figure supplement 1 ) . Sub-volumes , that is particles , encompassing the 96-nm repeat of DMTs were extracted from 10 tomograms and averaged as described in Materials and methods . In total , 763 particles were averaged to determine the 3D structure of the axonemal repeat ( Figure 3A–D , Video 4 ) . The average resolution of the entire structure is 21 . 8 Å based on the 0 . 143 Fourier shell correlation criterion ( Figure 3—figure supplement 2A ) . The resolutions at different regions vary based on visual inspection , and assessments by both local Fourier shell correlation ( FSC ) and ResMap ( Kucukelbir et al . , 2014 ) calculations ( Figure 3—figure supplement 2A , C–F ) ; the resolution of DMT region with MIPs reached 19 . 0 Å based on local FSC calculation ( Figure 3—figure supplement 2A ) . The 3D structure of the 96-nm repeat clearly resolved the expected major substructures , including OAD , IAD , RS , the IC/LC complex of IAD-f and the NDRC ( Figure 3B–E ) . Individual protofilaments are well-resolved and even alpha and beta tubulin monomers within protofilaments are clearly resolved ( Figure 3F ) . Several MIPs are also observed ( Figure 3B ) . At this resolution , we observed a filamentous structure on the outside of the DMT that spans the entire 96-nm repeat ( Figure 3E–G , red and white arrows ) . The location and extended conformation of this structure lead us to propose it to be the FAP59/172 molecular ruler described in Chlamydomonas that defines the 96-nm repeat ( Oda et al . , 2014a ) . Supporting this idea , the structure makes direct contact with RS , whose position depends on the FAP59/172 ruler ( Oda et al . , 2014a ) . The position of this ruler was previously determined in Chlamydomonas through mass-tagging , but the structure itself was not resolved ( Oda et al . , 2014a ) . We also observed a novel globular structure outside the B-tubule , between protofilaments B7 and B8 , having a periodicity of 8 nm ( Figure 3—figure supplement 3A , B blue arrow ) . The function of this structure is unknown , but it might influence dynein binding , because the microtubule binding domain of OADα contacts the B-tubule at this position ( see Figure 4E red arrow ) , and its 8 nm periodicity is in the range of estimated step size for dynein and kinesin motors ( Kikkawa , 2013; Reck-Peterson et al . , 2006; Coy et al . , 1999 ) . Two holes were observed in the inner junction between the A- and B-tubules ( red arrows in Figure 3C ) . We termed these ‘proximal’ and ‘distal’ holes , based on their position relative to the proximal end of the axoneme . The distal hole is near the site of NDRC attachment to the DMT and corresponds to the hole reported in other organisms ( Nicastro et al . , 2011; Pigino et al . , 2012 ) . The distal hole in Chlamydomonas is dependent on the presence of the NDRC on the external face of the DMT ( Heuser et al . , 2012b ) . The proximal hole is specific to T . brucei . Unlike the distal hole , there are no structures on the external face of the DMT at the site where the proximal hole is located . This indicates the proximal hole reflects structural properties imparted by proteins of the inner junction or inside the microtubules and is not dependent on the presence of external structures . Interconnections were observed between substructures on the A-tubule , including between individual OADs ( Figures 3D and 4 ) , between OAD and the IAD-f complex ( Figure 3B , D , Figure 3—figure supplement 3A , D ) . Particularly noteworthy are extensive contacts between RS3 , IAD-d , and the A and B-tubules ( Figure 3C , Figure 3—figure supplement 3C , D ) . At the base of RS3 we observed a structure that extends over four A-tubule protofilaments and attaches to the inner junction . Unlike the case for Chlamydomonas ( Nicastro et al . , 2006 ) , the NDRC did not make direct contact with the OAD in T . brucei ( Figure 3—figure supplement 3D ) , suggesting differences in mechanisms for coordinating inner and outer dynein motor activities . An earlier cryoET study of the T . brucei axoneme revealed the expected 4 OADs/repeat but did not resolve individual dynein motors ( Hughes et al . , 2012 ) . With sub-tomogram averaging , the beta and alpha OAD motors are now clearly resolved ( Figures 3B , D and 4A ) . This result provides the first direct demonstration that OADs contain two motor domains in T . brucei , making it the first protist shown to have two motors per OAD and correcting a misconception that all protists contain three motors ( Lin and Nicastro , 2018 ) . Together with three radial spokes per repeat , the entire arrangement of the T . brucei axoneme determined here therefore resembles that of humans more so than does Chlamydomonas or Tetrahymena , which are used as models for human cilium structure and function ( Figure 5 ) ( Pigino et al . , 2012; Owa et al . , 2019; Lin et al . , 2014 ) . Axoneme motility is driven by rotation of the dynein AAA+ ring relative to the linker and tail domains , causing translocation of adjacent DMTs as the dynein transitions from pre-powerstroke to post-powerstroke position ( Lin and Nicastro , 2018; Kikkawa , 2013; Burgess et al . , 2003 ) . The AAA+ ring , linker and tail domains were resolved in the OAD-beta dynein and are in the post-power stroke position ( Figure 4B , C ) , consistent with the fact that samples were prepared without exogenous ATP . This result thus supports structural assignments in the averaged structure . The dynein stalk domain , which contacts the adjacent DMT is visible ( Figure 4E ) . Six IADs were well-resolved ( Figure 3C , D ) and annotated f , a , b , e , g , and d , according to standard nomenclature ( Bui et al . , 2012 ) . Notably , IAD-c , which is important for movement of Chlamydomonas in high viscosity ( Yagi et al . , 2005 ) , is absent from the trypanosome structure . This finding is notable , given the very viscous environments experienced by trypanosomes during movement through tissues of the mammalian host ( Heddergott et al . , 2012; Bargul et al . , 2016; Capewell et al . , 2016; Trindade et al . , 2016 ) and tsetse fly vector ( Schuster et al . , 2017 ) . Nexin links are connections between adjacent DMTs , that are visible in axoneme TEM thin sections . They stabilize the axoneme and are a fundamental component of the sliding filament model for axoneme motility ( Satir , 1968; Satir et al . , 2014; Viswanadha et al . , 2017 ) . Prior studies indicate the NDRC is the only nexin link in Chlamydomonas ( Figure 5A ) ( Heuser et al . , 2009 ) . In T . brucei , however , we identified two prominent inter-doublet connections , the NDRC and the IC/LC complex of IAD-f ( Figure 3B–D ) . We term this second connection the ‘f-connector’ . The NDRC and f-connector each extend from the A-tubule of one DMT to contact near protofilament B9 of the adjacent DMT . NDRC contact is through the proximal and distal lobes defined by Heuser et al . ( 2009 ) and extends approximately 31 nm . The f-connector contact region extends approximately 11 nm . A structure analogous to the f-connector is observed between neighboring DMTs of three specific DMT pairs in Chlamydomonas ( Bui et al . , 2009 ) . However , the prominence of the f-connector observed here in T . brucei suggests it is present between neighboring DMTs of most and perhaps all DMTs , a conclusion supported by analysis of individual DMTs ( see below ) , indicating that nexin links in T . brucei include both the NDRC and the f-connector , as well as the OAD inter-doublet connector described below . This distinguishes the T . brucei axoneme from 3D axoneme structures from other organisms so far reported ( Figure 5 ) ( Pigino et al . , 2012; Owa et al . , 2019; Lin et al . , 2014 ) . A conspicuous structure not previously reported in any organism is a large protrusion at the junction between the tail and stalk domains of OAD-alpha ( Figure 4D , F ) . This protrusion , which we termed the ‘OAD inter-doublet connector’ , extends to the space between protofilament B6 and B7 of the adjacent DMT . The OAD inter-doublet connector is thus distinguished from the OAD-alpha stalk , which extends from the AAA+ ring to the space between protofilament B7 and B8 of the adjacent DMT ( Figure 4E ) . The OAD inter-doublet connector is present on all four OAD-alpha motors in the 96-nm repeat but is not observed in OAD-beta . The 96-nm repeat structure described above represents an average of all nine DMTs and does not reflect heterogeneity that may distinguish individual DMTs , as reported for Chlamydomonas ( Bui et al . , 2012 ) . To address this , we did sub-tomogram averaging on each DMT separately . The PFR restricts axoneme orientations on the EM grid and consequently , individual DMT structures suffer from the missing wedge . This was most severe for DMT 3 and 7 and we therefore cannot comment on these DMTs ( Figure 6—figure supplement 1F–G ) . For the remaining seven DMTs , distortion due to the missing-wedge problem obscured some details , particularly MIPs and OADs . However , main features of the 96-nm repeat were resolved ( Figure 6—figure supplement 1B–E ) . Each DMT was distinct , but careful examination revealed some similarities , particularly in the region of IAD-b , between DMTs 1+5 , 2+6 and 8+9 ( Figure 6—figure supplement 1C–E ) . Therefore , to reduce the impact of the missing wedge , we averaged DMTs within these pairs together . We recognize that this approach may still mask some features of a single doublet , but it nonetheless reveals heterogeneity between doublets . As shown in Figure 6 and Figure 6—figure supplement 1 , we identified doublet-specific structures that were not evident in the entire averaged structure . DMT 8 and 9 are distinguished from all other DMTs in that they do not have an IAD-b . In the place of IAD-b is a previously undescribed arch-like structure that extends upward from between RS1 and RS2 , which we termed ‘arch’ ( Figure 6D ) . DMT 1 and 5 are distinguished by the presence of a novel inter-doublet connecter , which we termed ‘b-connector’ , that connects IAD-b to the adjacent DMT and includes a ‘tail’ domain that connects with the ‘Modifier of Inner Arms’ MIA complex ( Yamamoto et al . , 2013 ) ( Figure 6B ) . DMT 2 and 6 contain a b-connector that lacks the tail domain ( Figure 6C ) . DMT 4 , 8 and 9 lack the b-connector . Structural variation of the b-connector on different DMTs explains why it was not evident in the entire averaged structure . DMTs 1 , 4 , 5 , 6 , 8 and nine each have an f-connector structure . DMT two does not have a clear f-connector , but this may reflect a missing wedge artifact since the density of the NDRC connection is also reduced ( Figure 6—figure supplement 1D ) . The analysis of individual DMTs supports the interpretation that the f-connector is present on most DMTs . Additionally , this analysis identified a new lineage specific inter-doublet connection not present in other organisms , the b-connector . The PFR is attached to DMT 4 , 5 , 6 and 7 and we therefore considered whether this attachment alters the 96-nm repeat . As detailed above , two PFR-attached DMTs , DMT 5 and 6 , each show similarities to non-attached DMTs , DMT 1 and 2 , that are not shared by each other ( Figure 6—figure supplement 1A , C–D ) . Therefore , PFR attachment does not seem to correlate with specific structural changes in the 96-nm repeat , at least at the current resolution . PFR-attachment complexes themselves , have a 56 nm periodicity ( Hughes et al . , 2012; Koyfman et al . , 2011 ) and therefore would not be resolved in our 96-nm repeat structure . The NDRC functions in axoneme stability and motility and these functions are thought to be mediated in part through inter-doublet connections ( Viswanadha et al . , 2017; Olbrich et al . , 2015; Wirschell et al . , 2013; Ralston and Hill , 2006 ) . The NDRC is composed of at least 11 subunits and some of these have been positioned within the complex ( Heuser et al . , 2009; Yamamoto et al . , 2013; Ralston et al . , 2006; Nguyen et al . , 2013; Kabututu et al . , 2010; Bower et al . , 2013; Lin et al . , 2011; Huang et al . , 1982; Song et al . , 2015; Oda et al . , 2014b ) . However , subunits that contact the B-tubule of the adjacent DMT are unknown . We identified CMF22 as a subunit of the T . brucei NDRC ( Nguyen et al . , 2013 ) , and the Chlamydomonas CMF22 orthologue is DRC11 ( Bower et al . , 2013 ) . RNAi knockdown of CMF22/DRC11 abolishes forward motility in T . brucei , demonstrating the importance of DRC11 in axoneme motility ( Video 5 and Video 6 ) ( Nguyen et al . , 2013 ) . The position of CMF22/DRC11 in the NDRC is unknown , but biochemical data indicate it may be within the proximal or distal lobe structures that contact the adjacent DMT ( Nguyen et al . , 2013; Bower et al . , 2013; Awata et al . , 2015 ) . We therefore used cryoET and sub-tomogram averaging to determine the structural basis of the CMF22/DRC11 RNAi knockdown . We used procyclic culture form ( PCF ) T . brucei , because loss of axonemal components is lethal in bloodstream forms ( Ralston and Hill , 2006; Broadhead et al . , 2006; Ralston and Hill , 2008 ) . The 96-nm repeat of WT PCF ( Figure 7A ) axonemes was very similar to that of BSF ( Figures 3 and 4 ) , including the presence of the novel OAD inter-doublet connector and the f-connector , as well as the missing IAD-c . In the CMF22 knockdown , the only structure clearly affected is the NDRC ( Figure 7C–E ) . The entire structure of the complex is mostly preserved , but the proximal lobe of the linker region is severely reduced ( Figure 7E ) . The affected structures encompass a large portion of the inter-doublet contact area for the T . brucei NDRC and include both regions reported to contact the adjacent DMT in the Chlamydomonas NDRC ( Heuser et al . , 2009 ) . The remaining NDRC domains , including dynein contacts were not grossly affected , although connection from NDRC to the MIA complex ( Yamamoto et al . , 2013 ) might be altered . Therefore , inter-doublet connection mediated by the NDRC is critical for axoneme motility . One major advance resulting from cryoET studies is the discovery that protein structures inside the microtubule , first observed in trypanosomes based on transmission EM studies more than fifty years ago ( Vickerman , 1969; Anderson and Ellis , 1965 ) , are ubiquitous in axonemal microtubules ( Nicastro et al . , 2011; Ichikawa et al . , 2017 ) . A striking feature of T . brucei axonemal microtubules is the presence of extensive MIP complexes not only in the A-tubule , but also in the B-tubule ( Figures 3B and 8A and Supplementary file 1 ) . Figure 8A shows a cross-section view of the averaged 96-nm repeat looking from the proximal end of the axoneme , with MIPs colored and external structures removed for clarity . The B-tubule is on top and the A-tubule is below , with 13 protofilaments of the A-tubule and 10 protofilaments of the B-tubule labeled according to convention ( Figure 8A ) . The shape , position and periodicity of the structure inside the B-tubule , next to the inner junction between the A- and B-tubules ( Figure 8A , B ) , indicate that this structure corresponds to MIP3 described in other organisms ( Nicastro et al . , 2011; Ichikawa et al . , 2017 ) . Notably however , the relationship of other MIPs in T . brucei to previously described MIPs is unclear and most TbMIPs in both the A- and B-tubules appear to be trypanosome-specific ( Figure 5 ) . When viewed in longitudinal section from within the B-tubule , TbMIP3 consists of two lobes , 3a and 3b ( Figure 8B ) , as reported for Chlamydomonas and Tetrahymena ( Nicastro et al . , 2011; Ichikawa et al . , 2017 ) . There are six such TbMIP3 structures in each 96-nm repeat . Subtle structural variations in the sizes of lobe 3b and connections to lobe 3a yield a 48 nm repeating pattern of three adjacent TbMIP3 structures , colored red , gold and orange ( Figure 8B ) . These TbMIP3 variations coincide with other structural variations within the microtubule , such as presence of inner junction holes ( arrows in Figure 8B ) , unique contacts to Snake MIP ( see Snake MIP description below ) , and attachment to a structure identified as MIP3c in Chlamydomonas ( Owa et al . , 2019 ) ( asterisks in Figure 8B ) . Variation in lobe 3b between the two gold TbMIP3 structures could suggest a 96-nm repeat unit , but this variation probably results from interference from the DRC base plate on the outside of the DMT at the site of the distal hole . Facing TbMIP3 , on the opposite side of the B-tubule lumen , are several trypanosome-specific MIPs , MIP B5 , B4 , B2 and a MIP that extends across the entire lumen , thus corresponding to the ponticulus structure previously observed in classical thin section TEM ( Figure 8C ) ( Vickerman , 1969; Anderson and Ellis , 1965; Vaughan et al . , 2006 ) . To our knowledge , the ponticulus was the first structure observed within the microtubule lumen in any organism and is the only structure so far described to extend across the entire microtubule . Our 3D structure shows that the ponticulus is not a single structure , but rather is comprised of 3 discrete MIPs , which we termed Pa , Pb and Pc ( Figure 8C–F ) . Each ponticulus MIP extends across the entire B-tubule lumen , connecting the A-tubule lattice to a different B-tubule protofilament . Pa , Pb and Pc connect protofilament A12 to protofilaments B3 , 5 and 4 , respectively and exhibit 48 nm periodicity ( Figure 8C–F ) . The ponticulus is assembled after construction of the axoneme ( Vaughan et al . , 2006 ) . Therefore , proteins comprising these structures must be delivered into a fully formed DMT . The A tubule also contains a diverse cohort of MIPs each with a repeating unit of 48 nm ( Supplementary file 1 , Figure 8A , Figure 9—figure supplement 1 ) . Rather than constituting several isolated structures however , TbMIPs form a network of interconnected complexes , similar to , but more extensive than , that reported for Tetrahymena ( Ichikawa et al . , 2017 ) . Two A-tubule MIPs are particularly notable . One , which we termed ‘ring MIP’ , is unique among MIPs so far described because it forms a ring structure protruding into the microtubule lumen ( Figure 9B ) . The ring MIP is attached to the protofilaments A8 and 9 and contacts another MIP complex on the protofilaments A8-12 termed ‘Ring Associated MIP’ ( RAM ) ( Figure 9B , C ) . Another MIP , which we termed ‘snake MIP’ , presents as a serpentine structure that appears to weave in and out of the A and B-tubules ( Figure 10 and Video 7 ) . The continuity of this density suggests it might be a contiguous structure , extending 48 nm and spanning multiple tubulin subunits , although we cannot rule out the possibility that protofilament subunits contribute to this structure .
The ciliary axoneme is one of the most iconic features of eukaryotic cells and is considered to have been present in the last eukaryotic common ancestor ( LECA ) ( Khan and Scholey , 2018 ) . To date , however , high-resolution structures of the 96-nm axoneme repeat have only been reported for two of the three eukaryotic supergroups . Here we report the 3D ultrastructure of the T . brucei 96-nm axonemal repeat . This is the first such structure reported for any pathogenic organism and first representative from the eukaryotic lineage of Excavates , a basal group that includes many pathogens of global importance to human health and agriculture ( Hampl et al . , 2009; Dawson and Paredez , 2013 ) . Our studies indicate the diversity of structures comprising the 96-nm repeat is under appreciated , give insight into principles of axoneme structure and function , and identify pathogen-specific features that may support unique motility needs of trypanosomes . The genus Trypanosoma was discovered more than 175 years ago and named for its unique cell motility ( Gruby , 1843 ) , which is driven by a single flagellum . The functional unit of the eukaryotic flagellum is the 96-nm axonemal repeat , which encompasses dynein motors and regulatory proteins that direct flagellum beating ( Porter and Sale , 2000 ) . In trypanosomes , the PFR exerts influence on the axoneme ( Koyfman et al . , 2011; Santrich et al . , 1997; Bastin et al . , 1998 ) , but motility is powered by the axoneme , which is the focus of the current work . Despite intense study for several decades , axoneme structures that underpin the parasite’s unique mechanism of cell propulsion remained hitherto unclear . A main finding from our studies is the discovery of lineage-specific features of the T . brucei 96-nm axonemal repeat , including extensive and novel MIP structures and novel inter-doublet connections between adjacent DMTs ( Figures 3–6 and 8–11 ) . Figure 11 shows a schematic overview of the overall 96-nm structure , previously undescribed features are labeled in panel B . We hypothesize these parasite-specific structures support unique motility needs of trypanosomes and thereby contribute to the transmission and pathogenic capacity of these organisms . The T . brucei axoneme is distinguished by mechanical strain experienced due to lateral attachment to the PFR and cell body , vigorous helical beating , encounter with host tissues and frequent reversals of beat direction ( Shimogawa et al . , 2018; Koyfman et al . , 2011; Santrich et al . , 1997; Bargul et al . , 2016 ) . MIPs have been shown to stabilize the axoneme in other organisms ( Owa et al . , 2019; Ichikawa and Bui , 2018; Stoddard et al . , 2018 ) and the expanded and MIP network of T . brucei may therefore help maintain stability of individual DMTs . Likewise , novel inter-doublet connections are expected to help maintain axoneme integrity under these conditions , analogous to the role of NDRC inter-doublet links in maintaining alignment of DMTs in Chlamydomonas ( Bower et al . , 2013 ) . The diversity and placement of T . brucei MIPs are suggestive of functions beyond stability . It is difficult to imagine for example , how a ring structure like the RingMIP , protruding into the microtubule lumen , would solely provide stability . MIPs in other organisms have been demonstrated to modulate axoneme beating ( Owa et al . , 2019; Stoddard et al . , 2018 ) . Given the presence of numerous trypanosome-specific MIPs , together with MIP differences reported between other species ( Figure 5 ) , we suggest that lineage-specific MIPs may provide a mechanism for fine-tuning the beating of axonemes between species that otherwise share a basic architecture . Extra connections between DMTs can also influence axoneme beating . It has been suggested that vortical beating of nodal cilia in vertebrates axoneme may involve transmission of regulatory signals from DMT to DMT , circumferentially around the axoneme ( King , 2018 ) . Extensive inter-doublet connections identified in our studies provide a means for direct interaction between DMTs and could thus contribute to helical beating that is a hallmark of T . brucei motility . Finally , given the recent demonstration that motility is critical for T . brucei virulence ( Shimogawa et al . , 2018 ) , parasite-specific features of the 96-nm repeat , which is the foundational unit of motility , may present novel therapeutic targets . Future work to identify novel T . brucei MIP and connector proteins will allow these ideas to be tested directly . By defining the structural basis of the motility defect in the CMF22/DRC11 knockdown , we demonstrate a specific requirement for inter-doublet connections in axoneme motility because the defect disrupts inter-doublet connections without affecting dyneins . This contrasts to NDRC mutants analyzed previously in Chlamydomonas , which typically exhibit structural defects in connections to dyneins or in dyneins themselves ( Heuser et al . , 2009; Awata et al . , 2015; Bower et al . , 2018 ) . An exception is sup-pf4 ( Heuser et al . , 2009 ) , but this mutant has only subtle effects on motility and beat frequency ( Awata et al . , 2015 ) , which contrasts to the CMF22/DRC11 knockdown in which propulsive motility is ablated ( Nguyen et al . , 2013 ) . Our CMF22/DRC11 knockdown studies therefore provide several important insights . Firstly , they demonstrate that penetrance of RNAi makes knockdown lines suitable for differential cryoET structural analysis in T . brucei . Secondly , they demonstrate CMF22/DRC11 is required for NDRC proximal lobe assembly and B-tubule attachment and , together with biochemical data ( Nguyen et al . , 2013; Bower et al . , 2013; Awata et al . , 2015 ) , indicate that CMF22/DRC11 is part of the proximal lobe . Thirdly , because inter-doublet contacts are specifically affected , without affecting dyneins , the results demonstrate that the NDRC itself and B-tubule contacts specifically are required for control of axoneme motility . This last point is particularly significant , as dynein-independent connection between adjacent DMTs is considered to be a founding principle of the sliding filament model for axoneme motility ( Satir , 1968; Holwill and Satir , 1990; Viswanadha et al . , 2017 ) , yet direct tests of this idea have been limited . The 96-nm spacing of the axoneme is controlled by a molecular ruler ( Oda et al . , 2014a ) , which is visible in the averaged BSF 96-nm repeat structure . The T . brucei MIP repeating unit is 48 nm , suggesting existence of a separate ruler inside the DMT to guide MIP placement . Such a ruler would need to extend 48 nm , exhibit structural heterogeneity along its length , and form contacts with other MIPs . The snake MIP satisfies these criteria . Notice , for example , that structural heterogeneities along the snake MIP coincide with unique contacts to each TbMIP3a , b structure within the 48 nm repeat ( Figure 8 ) . The snake MIP appears to extend into both the A- and B-tubules , which would make it possible to establish patterns in both tubules . Extensive interconnections between MIPs ( Video 7 ) might allow a single ruler to guide placement of all MIPs , or there might be more than one ruler , as is suggested for the outside of DMTs in Chlamydomonas ( Song et al . , 2018 ) , where the 24 nm repeat of OADs is dictated by something other than the FAP59/172 ruler ( Oda et al . , 2014a ) . Besides the snake MIP , another structure inside the B-tubule ( spine MIP ) appears to exhibit properties required of a 48 nm molecular ruler - forming a contiguous structure , spanning 48 nm and having heterogeneities that make unique contacts to adjacent MIPs ( Figure 8—figure supplement 1 ) .
BSF single marker ( BSSM ) and PCF ( Wirtz et al . , 1999 ) T . brucei cells were used . The CMF22/DRC11 knockdown line is described ( Nguyen et al . , 2013 ) . BSF single marker ( BSSM ) and PCF ( Wirtz et al . , 1999 ) T . brucei cells were cultured as described ( Shimogawa et al . , 2015; Saada et al . , 2014 ) and authenticated based on selective and morphogenetic markers . Cells , 2 × 108 for BSF or 4 × 108 for PCF , were washed three times in sterile 1xPBS . Supernatant was aspirated to ensure all of the PBS is removed . To remove the cell membrane and other soluble proteins and release the DNA , 160 µl Extraction buffer ( 20 mM HEPES pH: 7 . 4 , 1 mM MgCl2 , 150 mM NaCl , 0 . 5% NP40 IGEPAL CA-630 detergent , 2x Protease Inhibitors Cocktail-Sigma EDTA-free ) + 1/10 vol 10x DNase buffer + 1/10 vol DNase ( TURBO , Life Technologies 2 U/μl ) was added and incubated at room temperature for 15 min . In order to solubilize the subpellicular microtubules , 1 mM CaCl2 ( 2 µl of 100 mM CaCl2 ) was added and incubated on ice for 30 min . Then flagellum skeletons ( axoneme with PFR , basal body and FAZ filament ) were centrifuged ( 1500 g at 4°C for 10 min ) and the supernatant was removed . Then flagellum skeletons were purified away from cell body remnants and debris by one further centrifugation step over a 30% sucrose cushion at , 800g at 4°C for 5 min ( Extraction buffer w/o NP-40; 30% w/v sucrose ) . Flagellum skeletons from 200 μl of the upper fraction of the buffer-sucrose interface were collected and washed twice in 200 μl Extraction buffer , centrifugation at 1500 g at 4°C for 10 min , then resuspended in 40 µl buffer . Samples were either mixed with gold beads and plunge frozen immediately , as described below , or assessed directly for sample quality . To assess sample quality , BSF samples were negative-stained and analyzed using an FEI T12 transmission electron microscope equipped with a Gatan 2k × 2 k CCD camera . Samples were intact with uniform length distribution and a mean length of 25 . 2 , + /- 3 . 5 µm ( Figure 2B–F ) . PCF samples were examined by light microscopy to ensure uniform length distribution . BSF or PCF samples in the amount of 40 µl was mixed with either 5 nm ( for BSF ) or 10 nm ( for PCF ) diameter fiducial gold beads in 12:1 ratio . An aliquot of 3 µl of the axoneme-gold beads solution was applied onto Quantifoil ( 3:1 ) holey carbon grids ( for BSF ) or continuous carbon-coated EM grids ( for PCF ) which were freshly glow-discharged for 30 s at −40 mA . Excess of the sample on the grid was blotted away with a filter paper , at a blot force of −4 and blot time of 5 s , and vitrified by immediately plunging into liquid nitrogen-cooled liquid ethane with an FEI Mark IV Vitrobot cryo-sample plunger . Axoneme architectural integrity and gold bead concentration were assessed and plunge-freezing conditions optimized by obtaining low-resolution cryoET tilt series in an FEI TF20 transmission electron microscope equipped with an Eagle 2K HS CCD camera . From these tilt series , cryoET tomograms were evaluated to ensure structural integrity of the axoneme and PFR . Vitrified cryoET grids were stored in liquid nitrogen until use . For high-resolution cryoET tilt series acquisition , vitrified specimens were transferred with a cryo-holder into an FEI Titan Krios 300kV transmission electron microscope equipped with a Gatan imaging filter ( GIF ) and a Gatan K2 Summit direct electron detector . Samples were imaged under low-dose condition using an energy filter slit of 20 eV . CryoET tilt series were recorded with SerialEM ( Mastronarde , 2005 ) by tilting the specimen stage from −60° to +60° with 2° increments . The cumulative electron dosage was limited to 100 ~ 110 e-/Å2 per tilt series . All 4k × 4 k frames were recorded on a Gatan K2 Summit direct electron detector in counting mode with the dose rate of 8–10 e-/pixel/s . For each tilt angle , a movie consisting of 7 to 8 frames was recorded . For the PCF samples , the nominal magnification was x26 , 000 , giving rise to a calibrated pixel size of 6 . 102 Å . The defocus value was targeted at −4 µm . When the BSF samples were ready to be imaged , the same instrument was upgraded with a VPP , allowing us to obtain higher contrast images at closer to focus and higher magnification conditions . To obtain tilt series for the BSF samples with VPP , we follow the procedures previously described ( Fukuda et al . , 2015; Si et al . , 2018 ) and used the same GIF and K2 parameters as indicated above . Before starting each tilt series , we moved to a new VPP slot , waited for 2 min for stabilization , then pre-conditioned the VPP by illumination with a total electron dose of 12 nC for 60 s to achieve a phase shift of ~54° . Tilt series were recorded at a nominal magnification of 53 , 000X ( corresponding to a calibrated pixel size of 2 . 553 Å ) and a targeted defocus value of −0 . 6 µm . For BSF we collected a total of 50 tomograms and selected the 10 best , based on limited axoneme compression for sub-tomogram averaging . Cross sections of these 10 tomograms are shown in Figure 3—figure supplement 1 , and have circularity , measured as ratio of short axis/long axis , ranging from 0 . 92 to 0 . 98 . This yielded 763 particles that were averaged to determine the 3D structure of the BSF axonemal repeat . For WT PCF we collected 27 tomograms , and 17 of them were used for sub-tomogram averaging , resulting in 1177 particles averaged . For DRC11/CMF22 RNAi samples a total of 24 tomograms were collected and 19 of them were used for sub-tomogram averaging , resulting in 1726 particles averaged . For sub-tomogram averaging of individual DMT ( Figure 6 and Figure 6—figure supplement 1 ) , an additional 24 tomograms of BSF axonemes were used , for a total of 34 tomograms , yielding 297 to 339 particles averaged for each DMT ( DMT1 = 339 , DMT2 = 332 , DMT3 = 297 , DMT4 = 327 , DMT5 = 311 , DMT6 = 337 , DMT7 = 316 , DMT8 = 306 , DMT9 = 309 ) . For PCF and BSF samples , frames in each movie of the raw tilt series were drift-corrected , coarsely aligned and averaged with Motioncorr ( Li et al . , 2013 ) , which produced a single image for each tilting angle . The tilt series images were reconstructed into 3D tomograms by weighted back projections using the IMOD software package ( Kremer et al . , 1996 ) in six steps . Micrographs in a tilt series were coarsely aligned by cross-correlation ( step 1 ) and then finely aligned by tracking selected gold fiducial beads ( step 2 ) . The positions of each bead in all micrographs of the tilt series were fitted into a specimen-movements mathematical model , resulting in a series of predicted positions . The mean residual error was recorded to facilitate bead tracking and poorly-modeled-bead fixing ( step 3 ) . With the boundary box reset and the tilt axis readjusted ( step 4 ) , images were realigned ( step 5 ) . Finally , tomograms were generated by weighted back projection ( step 6 ) . Contrast transfer function ( CTF ) was corrected with the ctfphaseflip program ( Xiong et al . , 2009 ) of IMOD in step five above . The defocus value for each micrograph was determined by CTFTILT ( Mindell and Grigorieff , 2003 ) , and the estimated defocus value was used as input for ctfphaseflip . Note , one of the benefits of using a phase plate is that the CTF is insensitive to the sign of the defocus value being negative ( underfocus ) or positive ( overfocus ) ( Fan et al . , 2017 ) . To improve the signal-to-noise ratio and enhance the resolution , sub-tomograms containing the 96-nm axonemal repeated units along each DMT were extracted/boxed out from the raw tomograms . Sub-tomogram averaging and the missing-wedge compensation were performed using PEET program ( Nicastro et al . , 2006; Heumann et al . , 2011 ) as detailed previously ( Si et al . , 2018 ) , except for a new script we wrote to pick sub-volumes as outlined in the subsequent paragraphs . In our sub-tomogram averaging scheme , each particle is defined as the 96-nm repeating unit of the DMT . We developed a MATLAB script , autoPicker , to semi-automatically pick particles and calculate their location and orientation based on axoneme geometry . Briefly , we represent the 9+2 axoneme as a cylinder . For each axoneme in a tomogram , we used IMOD to visually pinpoint 11 points and save their coordinates into a file . The first two points , pa and pb , are the center points of the two bases of the cylinder . The remaining 9 points ( pi , i=1…9 ) identify the centers of the nine DMTs ( particles ) within the first 96-nm length at one end of the selected axoneme . The center is defined as the intersection point of a DMT with the middle of the three radial spokes along each particle’s 96-nm unit length . Our script reads the coordinates of the 11 points , calculates vector papb→ that defines the orientation of the cylinder , determine the center coordinates of all other particles within this axoneme based on the following formula: pij=pi−L⋅j⋅papb→|papb→| . , where i = 1 , 9; j = 1 to |papb→|/L , L is the unit length ( 96nm ) In order to uniquely identify the orientation of each particle , autoPicker also calculates a second point , p*ij for each pij . p*ij corresponds to the middle radial spoke’s end near the central pair . This is accomplished by solving the following linear algebraic equations that both p*ij and pij must satisfy ( see illustrations in Figure 3—figure supplement 4 ) :{papb→ ⋅ pijpij∗→=0 ( papb→ ×papij→ ) ⋅ pijpij∗→=0|pijpij∗→|=Length of the radial spoke ( 60nm ) We ran autoPicker for each axoneme in our tomograms to generate a PEET mod file that contains a list of the above described pij and p*ij pairs for all particles in that axoneme . Program stalkInit in PEET then read this mod file and generate an initial motive list file , a RotAxes file and three model files containing the coordinates for each particle . PEET then read the coordinate and orientation information from these files and automatically extracted the particles from the tomograms to perform iterative sub-tomogram averaging until no further improvement can be obtained . Sub-tomogram averaging of the individual DMTs was performed in two steps . Step1: particles ( 96-nm repeat units ) , picked from all 9 DMTs were classified into nine classes , corresponding to the DMT from which each particle was picked , DMT 1–9 . Step 2: for particles in each of the nine classes , sub-tomogram averaging was performed using PEET . The resolutions of the sub-tomogram averages were evaluated by two different approaches , one based on Fourier shell correlation ( FSC ) calculated by simpleFSC in PEET ( Nicastro et al . , 2006; Heumann et al . , 2011 ) and the other by ResMap ( Kucukelbir et al . , 2014 ) . To calculate FSC curves , we split all particles into two of equal-sized subsets following the PEET tutorial . Specifically , particles are separated into two subsets with the PEET specific motive list file by designating each sub-volume as either ‘1’ or ‘2’ so that it would be placed into one of the two sub-sets . PEET then performed sub-tomogram averages independently for particles in each of the two equal-sized sub-sets , yielding two sub-tomogram averages of the 96-nm axonemal structure . These two independently calculated sub-tomogram averages were then used as the input maps of the simpleFSC program in the PEET package to calculate the FSC curve for the entire 96-nm axonemal repeat ( Figure 3—figure supplement 2A ) . We also calculate FSC curves for local regions encompassing DMT with MIPs , OAD , IAD , NDRC or RS . To do so , a cuboid mask was used in ChimeraX ( Goddard et al . , 2018 ) to extract two corresponding local density regions that primarily containing either DMT with MIPs , or OAD , or IAD , or NDRC or RS from the two sub-tomogram averages . Each set of two corresponding cuboid volumes ( Figure 3—figure supplement 2B ) was then used as the input maps of the simpleFSC program in the PEET package to calculate an FSC curve for the local region , which is plotted as a function of spatial frequency ( Figure 3—figure supplement 2A ) . Local resolution across the entire averaged 96-nm axonemal repeat was also evaluated with ResMap ( Kucukelbir et al . , 2014 ) using the above two independently calculated sub-tomogram averages as input maps and the result is visualized from different views in Figure 3—figure supplement 2C ) . IMOD ( Kremer et al . , 1996 ) was used to visualize the reconstructed tilt-series and the 2D tomographic slices of the sub-tomogram averages . UCSF ChimeraX ( Goddard et al . , 2018 ) was used to visualize the resulting sub-tomogram averages in their three dimensions . Segmentation of densities maps and surface rendering for the different components of the 96-nm repeated unit were performed by the tools volume tracer and color zone in UCSF Chimera ( Pettersen et al . , 2004 ) . GIMP 2 . 8 . 18 ( GNU Image Manipulation Program ) was used to color regions of interest ( Figures 5 , 6B–D , 8B–F and 9B–C; Figure 3—figure supplement 3C–D , Figure 8—figure supplement 1B , Figure 9—figure supplement 1B; Supplementary file 1 ) . For rendering , no filters were applied on MIPS but we applied low pass filters on the other components to improve the clarity of individual structures described in the text . For the structures in Figure 3C–E; Figure 4A , B , D; Figure 7A–E , we filtered the DMT , NDRC , RS , IC/LC , OAD and IAD to 30 Å . For the structures in Figure 5; Figure 6; Figure 6—figure supplement 1 , we filtered the entire map to 50 Å ) . Motility videos of BSF cells were obtained exactly as described in Kisalu et al . ( 2014 ) . Motility videos of PCF cells were obtained exactly as described in Nguyen et al . ( 2013 ) . All videos were recorded and played back at 30 frames per second . The PCF tetracycline-inducible DRC11/CMF22 RNAi knockdown line has been described previously ( Nguyen et al . , 2013 ) . WT and mutant PCF videos correspond to this knockdown line cultured in the absence ( WT ) or presence ( mutant ) of 1 μg/ml tetracycline to induce RNAi . All data generated or analyzed during this study are included in the manuscript and supporting files . Source data files have been provided for Figure 2F and Figure 3—figure supplement 4 . The cryoET sub-tomogram average maps have been deposited in the EM Data Bank under the accession codes EMD-20012 , EMD-20013 and EMD-20014 , for the wild-type bloodstream form , wild-type and DRC11-knock-down procyclic form , respectively . | The parasites that cause African sleeping sickness , known as trypanosomes , propel themselves forward using a structure called a flagellum , a bit like the tail of a human sperm . But rather than connect to the body of the cell just at the base , like in a sperm , the parasite flagellum runs along the side of the cell . This means that , when it beats , the whole cell twists in a screw-like motion . The parasite flagellum beats vigorously , changes direction often , and puts the cell under lots of mechanical stress . This unusual motion likely helps the parasites to move through a thick and sticky fluid like blood . The similarities between the parasite flagellum and the flagellum on a human sperm are down to a shared evolutionary history . Both structures contain the same basic molecular skeleton , known as the axoneme . The axoneme contains a combination of supporting proteins and molecular motors , and the molecular motors essentially pull on the supports to bend the flagellum . The unusual movement of trypanosome parasites suggests that their axonemes may have unique structural features . But the three-dimensional structure of trypanosome axonemes had previously not been studied in great detail . Imhof , Zhang et al . now address this gap in knowledge using a technique called “cryo electron tomography” and showed that axoneme structure in trypanosomes does share many features with those of other organisms but it has extra proteins and connections for support , which could help to protect the flagellum from mechanical stress . The similarities and differences between human and trypanosome flagella could indicate new drug targets that could be used to protect us against these parasites . A better understanding of how flagella work in general could also give insights into human genetic diseases that involve problems with these structures . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2019 | Cryo electron tomography with volta phase plate reveals novel structural foundations of the 96-nm axonemal repeat in the pathogen Trypanosoma brucei |
Canine transmissible venereal tumour ( CTVT ) is a clonally transmissible cancer that originated approximately 11 , 000 years ago and affects dogs worldwide . Despite the clonal origin of the CTVT nuclear genome , CTVT mitochondrial genomes ( mtDNAs ) have been acquired by periodic capture from transient hosts . We sequenced 449 complete mtDNAs from a global population of CTVTs , and show that mtDNA horizontal transfer has occurred at least five times , delineating five tumour clades whose distributions track two millennia of dog global migration . Negative selection has operated to prevent accumulation of deleterious mutations in captured mtDNA , and recombination has caused occasional mtDNA re-assortment . These findings implicate functional mtDNA as a driver of CTVT global metastatic spread , further highlighting the important role of mtDNA in cancer evolution .
The canine transmissible venereal tumour ( CTVT ) is a transmissible cancer that is contagious between dogs via the transfer of living cancer cells during coitus . The disease usually manifests as localised tumours involving the genital mucosa in both male and female domestic dogs . CTVT first arose from the somatic cells of an individual dog that lived approximately 11 , 000 years ago; it subsequently survived beyond the death of this original animal by metastasising to new hosts ( Murgia et al . , 2006; Rebbeck et al . , 2009; Murchison et al . , 2014; Decker et al . , 2015 ) . CTVT is found in dog populations worldwide , and is the oldest and most prolific cancer lineage known in nature ( Murchison et al . , 2014; Strakova and Murchison , 2014; Strakova and Murchison , 2015 ) . The clonal evolution of CTVT renders this lineage a unique genetic tag with which to trace historical global dispersals of dogs together with their human companions . Furthermore , the extreme longevity of this lineage , its serial colonisation of genetically distinct allogeneic hosts and its occasional uptake of host mitochondrial DNA ( mtDNA ) by horizontal transfer ( Rebbeck et al . , 2011 ) , provide opportunities to probe genetic vulnerabilities in cancer and to identify novel host-tumour interactions . We analysed 449 complete mtDNAs in CTVT and used these to investigate the frequency and timing of mtDNA horizontal transfer in this lineage; furthermore , we assessed the contribution of selection to CTVT mtDNA evolution and searched for evidence of mtDNA recombination .
To investigate the global CTVT population structure and estimate the frequency and timing of mtDNA horizontal transfer , we performed low-coverage whole genome sequencing ( ~0 . 3X whole genome coverage ) on 449 CTVT tumours and 338 matched hosts collected from 39 countries across six continents ( Materials and methods ) ( Figure 1—figure supplement 1 , Supplementary file 1 ) . MtDNA was sequenced at ~70X coverage , indicating that each CTVT cell carries approximately 470 mtDNA copies ( Figure 1—figure supplement 2 , Supplementary file 2 , Materials and methods ) . CTVT was confirmed by identification of a characteristic rearrangement involving a long interspersed nuclear element ( LINE ) near the MYC locus ( Katzir et al . , 1985; 1987 ) ( Supplementary file 3 ) . We identified 1005 single point substitution variants and 27 short insertions and deletions ( indels ) in the CTVT mtDNA population ( Supplementary files 4 , 5 , 6 , 7 , 12 , 13 ) . CTVT mtDNA somatic substitution mutations ( see Materials and methods ) had the characteristic profile that is observed in human cancers , dominated by C>T and T>C mutations showing a striking strand bias ( Ju et al . , 2014 ) ( Figure 2—figure supplement 1 ) . This mutational process is probably replication-coupled , and mutations associated with this process appear to accumulate at a roughly constant rate in human cancers ( Ju et al . , 2014 ) . A maximum likelihood phylogenetic tree constructed with mtDNA sequences from CTVT , matched hosts and 252 additional dogs ( see Supplementary file 8 ) revealed that CTVT mtDNAs cluster in five distinct groups within dog mtDNA haplogroup A1 ( Figure 1A , Figure 1—figure supplement 3 , Figure 1—source data 1 ) . These data suggest that CTVT mtDNAs have at least five independent origins , demarcating five groups that we have named CTVT clades 1 to 5 . 10 . 7554/eLife . 14552 . 003Figure 1 . CTVT has acquired mtDNA by horizontal transfer at least five times . ( A ) Maximum likelihood phylogenetic tree constructed with complete mtDNA sequences from 449 CTVT tumours and 590 dogs . Coloured and black dots represent CTVT and dog mtDNA respectively . Scale bar indicates base substitutions per site . ( B ) Number of somatic substitution mutations per CTVT tumour . Coloured bars indicate somatic mutations acquired by each tumour since mtDNA capture . Grey bars indicate substitutions absent from normal dog mtDNA haplotypes but common to all tumours within a clade; thus the early somatic or rare germline status of these variants is unknown . ( C ) Geographical distribution of clades . Coloured dots represent locations from which one or more CTVT tumours were collected . ( D ) Simplified representation of maximum likelihood phylogenetic trees for each clade . Trees illustrate nodes with bootstrap support >60 , and shaded triangles represent coalescence of individual branches within each group . Two tumours were collected in the United States and the Netherlands respectively from dogs imported from Guatemala and Romania . Discontinuous grey lines represent contributions of substitutions absent from normal dog mtDNA haplotypes but common to all tumours within a clade . Assuming a constant accumulation of mutations within and between clades , approximate number of somatic mutations and estimated timing is shown . Maximum likelihood trees upon which these representations are based are found in Figure 1—source data 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 00310 . 7554/eLife . 14552 . 004Figure 1—source data 1 . Maximum likelihood phylogenetic tree of CTVT mtDNA . Maximum likelihood phylogenetic tree constructed using 449 complete CTVT mitochondrial genomes and 590 complete dog mitochondrial genomes . All sequences are labelled with sample identifier , country , breed and haplotype name . The sample identifier for CTVT hosts is the sample name ( Supplementary file 1 ) , the sample identifier for the publicly available dogs is the accession number . Scale bar indicates base substitutions per site . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 00410 . 7554/eLife . 14552 . 005Figure 1—source data 2 . Maximum likelihood phylogenetic trees for CTVT clades 1 to 5 . Maximum likelihood phylogenetic trees for CTVT mtDNA in ( A ) clade 1 ( n = 170 ) ( B ) clade 2 ( n = 252 ) ( C ) clade 3 ( n = 22 ) ( D ) clade 4 ( n = 3 ) and ( E ) clade 5 ( n = 2 ) , rooted with haplotypes CTVT1 to CTVT5 respectively , which contain clade-defining germline and potential somatic substitutions specific to each clade ( Figure 1—figure supplement 4 ) . Bootstrap values were calculated from 100 bootstrap replicates and are shown where bootstrap values ≥60 . Scale bars indicate base substitutions per site . Clade 5 contains only two tumours , which are identical both to each other and to the CTVT5 haplotype; thus the tree for this clade was created separately and does not have a scale bar . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 00510 . 7554/eLife . 14552 . 006Figure 1—figure supplement 1 . Geographical locations and mtDNA clades for CTVT tumours and hosts . Each dot represents the location of ( A ) CTVT tumours , coloured by CTVT mtDNA clade; or ( B ) CTVT hosts , coloured by dog mtDNA clade . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 00610 . 7554/eLife . 14552 . 007Figure 1—figure supplement 2 . mtDNA copy number in CTVT . MtDNA copy number was estimated by normalising mtDNA sequence coverage to whole genome sequence coverage ( Supplementary file 2A ) . Each point represents an individual tumour ( labelled by clade ) or host . MtDNA copy number in tumours was not normalised for host contamination . Host and tumour samples with average MT coverage >300X ( see Supplementary file 2A ) were excluded from the analysis and from calculation of average number of mtDNA copies per cell . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 00710 . 7554/eLife . 14552 . 008Figure 1—figure supplement 3 . CTVT mtDNA clades 1 to 5 all arose from dog mtDNA clade A . Maximum likelihood phylogenetic tree constructed with complete mtDNA sequences from 449 CTVT tumours and 590 dogs . Coloured and black dots represent CTVT and dog mtDNA respectively ( CTVT mtDNA clade colours are represented as in Figure 1A ) . Dog mtDNA clades A to E are labelled ( Savolainen et al . , 2002; Vila et al . , 1997 ) . Scale bar indicates base substitutions per site . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 00810 . 7554/eLife . 14552 . 009Figure 1—figure supplement 4 . Reconstructed donor haplotypes for CTVT mtDNA clades 1 to 5 . Diagrams representing the likely donor haplotype for each of the CTVT mtDNA clades 1 to 5 . The coordinates for each substitution variant position are shown , and substitutions are colour-coded either as 'germline' ( i . e . they are present in all tumours within a clade and are found in the most closely related dog mtDNA haplotype , which is represented below each of the clade diagrams or they are found in the most closely related dog mtDNA haplotype only ) ; or 'potential somatic' ( i . e . they are present in all tumours within a clade but are not found in the most closely related dog mtDNA haplotype ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 00910 . 7554/eLife . 14552 . 010Figure 1—figure supplement 5 . Sequence contribution of nuclear-encoded mtDNA ( NuMTs ) . Sequence read depth across the MT genome for a representative CTVT tumour ( 146T ) and host ( 100H1 ) sequenced in this study to ~0 . 3X whole genome average coverage . This is compared with sequence read depth for simulated reads from CanFam3 . 1 ( excluding the MT chromosome ) ; reads were simulated to ~0 . 3X whole genome average coverage . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 010 Although CTVT originated about 11 , 000 years ago , whole genome sequences of two CTVT tumours derived from clades 1 and 2 indicated that these two clades shared a common ancestor approximately 460 years ago ( Murchison et al . , 2014 ) . We investigated the relative time since each CTVT mtDNA horizontal transfer event by estimating the number of mtDNA somatic mutations acquired by each clade since mtDNA capture ( Figure 1B ) . This analysis revealed that clade 1 mtDNA carry more than double the number of mtDNA somatic mutations ( 22 . 5 mutations average ) compared with clade 2 mtDNA ( 9 . 4 mutations average ) . By inferring that the clade 2 mtDNA horizontal transfer event occurred no more than 460 years ago , this analysis suggests a maximum time since mtDNA uptake of 1097 years for clade 1 , 244 years for clade 3 , 1690 years for clade 4 and 585 years for clade 5 , assuming a constant somatic accumulation of mutations in CTVT mtDNA ( Materials and methods ) . Importantly , two additional mutation rate estimates , derived using human data ( Ju et al . , 2014 ) , suggested similar timing for CTVT clade origins ( Materials and methods , Supplementary file 9 ) . Thus , this analysis suggests that the original mtDNA , that was present in the founder dog that first spawned CTVT , is not detectable in tumours that we have analysed , and indicates that CTVT cells have captured mtDNA from transient hosts at least five times within the last two thousand years . The geographic distribution and phylogenies of the five CTVT clades reveal the dynamic recent history of the CTVT lineage ( Figure 1C , Figure 1D , Figure 1—figure supplement 1 and Figure 1—source data 2 ) . Clades 1 and 2 , which occur most frequently in the CTVT population that we analysed , both have a global distribution . Tumours that diverged early in the clade 1 lineage occur in Russia , Ukraine , China and India , suggesting an Old World origin for this clade ( Figure 1D ) . Clade 1 tumours in Central and South America share a single common ancestor that probably existed no more than 511 years ago , suggesting introduction of CTVT to the Americas with colonial contact; similarly , our data suggest a single introduction of CTVT to Australia after European arrival ( maximum 116 years ago ) ( Figure 1D , Supplementary file 9 , see Materials and methods ) . The distribution pattern and timing of clade 2 suggest that this clade may have been transported between continents via trans-Atlantic and Indian Ocean trade routes ( Figure 1C and 1D ) . The more recent clade 3 lineage was found in Central and South America and India , and the less frequent clades 4 and 5 occurred only in India and Nigeria respectively ( Figure 1C and 1D ) . The extensive and recent global expansion detected in the CTVT lineage is consistent with signals of widespread admixture observed in worldwide populations of domestic dogs ( Shannon et al . , 2015 ) , highlighting the extent to which canine companions accompanied human travellers on their global explorations . Most somatic mutations in cancer are believed to be selectively neutral , and there is little evidence in human cancers for negative selection operating to safeguard essential cellular processes ( Stratton et al . , 2009 ) . We searched for evidence of mtDNA functionality in CTVT cells by examining CTVT mtDNA for signals of negative selection . If present , negative selection would be expected to operate on mtDNA to prevent homoplasmy of deleterious mutations . Consistent with this prediction , the variant allele fraction ( VAF ) of nonsense substitutions and frameshift indels was significantly lower than VAF for other substitutions and indels ( Figure 2A and B , p=0 . 00019 and p=3 . 03x10-05 respectively , two-sample Kolmogorov-Smirnov test ) . Furthermore , dN/dS for somatic mtDNA mutations in CTVT showed significant deviation from neutrality both for nonsense ( 0 . 187 , p=1 . 02x10-07 ) and missense ( 0 . 748 , p=4 . 18x10-03 ) mutations ( Figure 2C ) . Together with evidence of reduced VAF for truncating mtDNA mutations in human cancers ( Ju et al . , 2014; Stewart et al . , 2015 ) , these findings provide evidence for the activity of negative selection operating to preserve mtDNA function in CTVT and indicate that , at least in some cancers , functional mtDNA contributes to driving cancer . 10 . 7554/eLife . 14552 . 011Figure 2 . Negative selection operates to prevent the accumulation of gene-disrupting mutations in CTVT . Cumulative distribution functions for variant allele fraction ( VAF ) for gene-disrupting ( A ) substitutions and ( B ) indels . P-values were calculated using two-sample Kolmogorov-Smirnov tests . ( C ) dN/dS for somatic nonsense and missense substitutions . P-values were calculated using a likelihood ratio test with parameters estimated using a Poisson model . Error bars indicate 95 percent confidence intervals . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 01110 . 7554/eLife . 14552 . 012Figure 2—figure supplement 1 . CTVT mtDNA somatic mutation spectrum . CTVT somatic mutations displayed by mutation type ( in pyrimidine context ) with 5’ and 3’ context and strand . Each of 96 mutation classes is displayed on the horizontal axis , with mutations occurring on the heavy strand displayed in red on the positive axis , and light strand mutations displayed in blue on the negative axis . The normalised substitution rate represents the ( number of observed ) / ( number of expected ) mutations , given mtDNA genome triplet content . Distinctive peaks are individually labelled . Only mutations on the 'conservative somatic list' were used ( see Materials and methods and Supplementary file 4C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 012 MtDNA is usually assumed to be clonally inherited and recombinationally inert . However , mtDNA recombination has been directly observed in various eukaryotes ( Lunt and Hyman , 1997; Hoarau et al . , 2002; Ladoukakis and Zouros , 2001; Gantenbein et al . , 2005; Ujvari et al . , 2007; Bergthorsson et al . , 2003 ) and has been proposed as a mechanism for mtDNA repair ( Thyagarajan et al . , 1996 ) . Recombination of maternal and paternal mtDNA haplotypes has been observed in rare cases of human biparental mtDNA inheritance ( Kraytsberg et al . , 2004; Zsurka et al . , 2005 ) , and mtDNA recombination activity is present in human cell extracts ( Thyagarajan et al . , 1996 ) . However , mtDNA recombination has not , to our knowledge , been previously detected in cancer . Given the possibility for coexistence of two distinct mtDNA haplotypes in CTVT cells , we searched for evidence of mtDNA recombination in CTVT using recombination-detection algorithms 3seq and SiScan ( Boni et al . , 2007; Gibbs et al . , 2000 ) . Remarkably , these algorithms detected significant evidence for mtDNA recombination in CTVT clade 1 , detecting recombination breakpoints at around MT:5430 and MT:16176 . Maximum likelihood phylogenetic trees constructed using segments MT:1–5429 and MT:5430–16176 derived from clade 1 mtDNA produced distinct topologies ( Figure 3A , Figure 3—source data 1 ) . Further inspection of clade 1 mtDNA haplotypes suggested that recombination replaced MT:1–5429 in a clade 1 mtDNA haplotype that diverged from Central American clade 1 CTVTs and that subsequently colonised areas of South and Central America ( Chile , Colombia , Ecuador , Panama , Paraguay ) ( Figure 3B ) . 10 . 7554/eLife . 14552 . 013Figure 3 . Ancient and modern mtDNA recombination in CTVT . ( A ) Maximum likelihood phylogenetic trees constructed using segments MT:1–5429 and MT:5430–16176 from clade 1 CTVT mtDNAs . Three clade 1 mtDNA haplotype groups are represented by coloured dog silhouettes , and their geographical distributions are colour-coded on the map . Bootstrap values were calculated from 100 iterations . Maximum likelihood trees upon which these representations are based are found in Figure 3—source data 1 . ( B ) Simplified haplotype diagrams for clade 1 CTVT mtDNAs derived from groups shown in ( A ) . Germline variants were present in the donor mtDNA that founded clade 1 , represented by the A1/A1c/A1e dog haplotype ( see Figure 1—figure supplement 4 ) . Region putatively replaced by recombination is outlined with orange box . ( C ) Recombination detected in tumour 559T ( Nicaragua ) . The estimated per cent contribution of each recombined haplotype to the mtDNA population within 559T CTVT cells is shown , and grey arrows indicate likely sites of recombination . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 01310 . 7554/eLife . 14552 . 014Figure 3—source data 1 . Ancient mtDNA recombination in CTVT clade 1 . Maximum likelihood cladograms constructed using clade 1 mtDNA positions ( A ) 1-5429bp and ( B ) 5430-16176bp ( see Materials and methods ) . Trees were constructed with 153 clade 1 CTVT mtDNAs rooted with the CTVT1 haplotype , which contains clade 1 clade-defining germline and potential somatic substitutions ( Materials and methods , Figure 1—figure supplement 4 ) . Bootstrap values were calculated from 100 bootstrap replicates and are shown where bootstrap values ≥60 . DOI: http://dx . doi . org/10 . 7554/eLife . 14552 . 014 These data provide evidence of an mtDNA recombination event in an ancestral CTVT lineage . We searched for evidence of more recent mtDNA recombination by examining outliers on CTVT mtDNA phylogenetic trees ( Figure 1—source data 2 , Supplementary file 10 ) . This analysis identified 559T , a CTVT tumour derived from a male dog in Nicaragua ( Figure 3C ) . Further investigation of mtDNA in 559T revealed what appeared to be a CTVT clade 1 mtDNA haplotype ( CTVT_1B2b1_29 ) superimposed upon a dog mtDNA haplotype ( A1d1a_1 ) , neither of which resembled the mtDNA haplotype found in normal tissues from this host dog , 559H ( B1_1 haplotype ) . Phasing of mtDNA variants in 559T using long sequence reads indicated the presence of at least three distinct mtDNA haplotypes in this tumour , each representing a recombination product apparently derived from mtDNA haplotypes CTVT_1B2b1_29 and A1d1a_1 ( Figure 3C ) . These data suggest that a tumour antecedent of 559T captured haplotype A1d1a_1 mtDNA from its host . Recombination was initiated between mtDNA haplotypes CTVT_1B2b1_29 and A1d1a_1 , and cells containing these recombination products were passed to host 559H . Alternatively it is possible that 559H received a mixture of both normal and CTVT cells from its CTVT donor animal , and mtDNA capture and recombination occurred within 559H . It must also be mentioned that the A1d1a_1 haplotype resembles the CTVT clade 3 donor haplotype ( Figure 1—figure supplement 4 ) ; thus we cannot exclude the possibility that the recombination that we observe in 559T involved horizontal transfer between clade 1 and clade 3 CTVT tumours that occurred within the same animal . Our analysis provides evidence for occasional mtDNA recombination activity in CTVT cells . The mechanism whereby distinct mtDNA molecules are able to interact within the cell , and the nature of the signals that trigger onset of mtDNA recombination are not clear . Further analysis will determine if DNA damage signalling is involved , and it is interesting to observe that a truncating nonsense mutation in COX3 was found in some 559T haplotypes ( Figure 3C ) . Although we could not find evidence of mtDNA recombination in CTVT beyond those described , we cannot exclude the possibility that recombination is more widespread in CTVT mtDNA than detected . It is possible , therefore , that our phylogenetic , mutation rate and selection analyses ( Figure 1 , 2 ) have been influenced by an undetected recombination signal . However , the presence in all ( non-recombining ) CTVT mtDNAs of a set of clade-specific markers ( Figure 1—figure supplement 4 ) , the absence ( beyond 559T ) of distinctive phylogenetic outliers ( Figure 1—source data 2 ) , the very low frequency of back-mutation ( Supplementary file 10 ) , the strong somatic signal identified in the CTVT mtDNA mutational spectrum ( Figure 2—figure supplement 1 ) , and the failure of recombination-detection algorithms to detect further recombination , suggest that , if such a signal is present , it is at a low level . CTVT is the world’s oldest known cancer whose metastatic spread through its global host population provides unique insights into evolutionary processes operating in cancer . Our analysis of CTVT mtDNA has illuminated five mtDNA horizontal transfer events which trace two millennia of CTVT global spread . Negative selection has operated on CTVT to maintain mtDNA integrity at the level of nonsense and missense mutations , and occasional mtDNA recombination has occurred , possibly to repair damaged mtDNA . Evidence of negative selection demonstrates that maintenance of functional mtDNA is important for the biology of CTVT; and the observation of multiple mtDNA horizontal transfer events further supports the possibility that mtDNA capture from hosts is a positively selected adaptive mechanism ( Rebbeck et al . , 2011; Tan et al . , 2015; Spees et al . , 2006 ) . This study highlights the important role of functional mtDNA in cancer and reveals unexpected biological mechanisms that have operated in an ancient mammalian somatic cell lineage .
This study was approved by the Department of Veterinary Medicine , University of Cambridge , Ethics and Welfare Committee ( reference number CR174 ) . Tumour and host ( gonad , skin , blood or liver ) tissue samples were collected into RNAlater solution and stored at 4°C until processing . Genomic DNA was extracted using the Qiagen DNeasy Blood and Tissue extraction kit . Sample information is presented in Supplementary file 1 . Quantitative PCR ( qPCR ) assays were performed to confirm CTVT diagnosis ( Supplementary file 3 ) by detection of the CTVT-specific LINE-MYC genomic rearrangement ( Murgia et al . , 2006; Rebbeck et al . , 2009; Murchison et al . , 2014; Katzir et al . , 1985; 1987 ) . Each qPCR was performed in triplicate with SYBR Select Master Mix ( Life Technologies , Carlsbad , CA ) using an Applied Biosystems 7900HT Fast Real-Time PCR system instrument ( Applied Biosystems , Foster City , CA ) with conditions and primers specified below . PrimerSequenceLINE-MYC primers ( obtained from [Rebbeck , 2007] ) ForwardAGG GTT TCC CAT CCT TTA ACA TTReverseAGA TAA GAA GCT TTT GCA CAG CAAACTB primersForwardCTC CAT CAT GAA GTG TGA CGT TGReverseCGA TGA TCT TGA TCT TCA TTG TGC qPCR master mix reagentsVolume per reaction ( μl ) SYBR Green Mix10Primers ( 5 μM/primer ) 2 . 4DNA ( 20 ng/μl ) 0 . 5Water7 . 1Total volume20 Stage of qPCR amplificationTemperature ( °C ) Time ( s ) Initial denaturation9560040 cycles95156060Final dissociation9515 Standard curves were constructed for each primer set using CTVT tumour 29T1 as reference . Relative DNA input was calculated using standard curves as follows: ( Ct = m ( log10 ( iA ) ) + b ) , with each of the parameters defined as follows: Ct = threshold cycle , m = slope of the standard curve , iA = input amount , b = y-intercept of the standard curve . Relative DNA input for LINE-MYC was then normalised to ACTB . LINE-MYC and ACTB are present in three and two copies respectively in 24T and 79T CTVT tumours ( Murchison et al . , 2014 ) ; however , it is possible that copy number at these loci differs between tumours in the current dataset . Whole genome sequencing libraries with insert size 100 to 400 base pairs ( bp ) were constructed using standard methods according to manufacturer’s instructions and sequenced with 75bp paired end reads on an Illumina HiSeq2000 instrument ( Illumina , San Diego , CA ) to an average whole genome depth of 0 . 3X; average mitochondrial DNA ( mtDNA ) coverage was ~70X . Reads were aligned with the CanFam3 . 1 dog reference genome ( Lindblad-Toh et al . , 2005 ) ( http://www . ensembl . org/Canis_familiaris/Info/Index ) using the BWA alignment tool ( Li and Durbin , 2009 ) . To calculate the mitochondrial copy number , we used the following equation: ( mtCOV/nuclCOV ) *P , where mtCOV = average coverage across the mitochondria , nuclCOV = average coverage across the nuclear genome and P = ploidy . The ploidy used in our calculations was 2 for both CTVT tumours and CTVT hosts ( Murchison et al . , 2014 ) . Host and tumour samples with average MT coverage >300X were excluded from the copy number calculations ( see Supplementary file 2A ) . Samples 1380T and 1381T were sequenced separately , based on the methods described in Pang et al ( Pang et al . , 2009 ) . Complete mitochondrial genomes were amplified using the primers listed in Pang et al ( Pang et al . , 2009 ) with a number of additional primers listed below . The PCR conditions are specified below . PCR master mix reagentsVolume per reaction ( μl ) 1 X PCR LATaq buffer2 . 5Primer forward and reverse ( 10 μM ) 1 . 2 ( each ) DNA ( 100-200 ng/μl ) 1 . 2LATaq DNA polymerase0 . 251X dNTP ( 10 mM ) 4Water14 . 65Total volume25 Stage of PCR amplificationTemperature ( °C ) Time ( s ) Initial denaturation9430012 cycles ( touchdown PCR program , reduce 1°C each cycle ) 946061–5060749025 cycles946052607490Final extension74420 Primer nameSequence ( 5’–3’ ) D0132ACC GTA AGG GAA TGA TGA AD0136TGT AAG TGG TCG TAG AGG TTCD0141AGG CGG ACT AAA TCA AAC TCAD0146GGG GTA TCT AAT CCC AGT TTD0149AAG TTT GGT AGC ACG AAG AT The PCR products were purified using a 1 . 0% agarose gel and sequenced on a 3730xl DNA analyser ( Applied Biosystems ) with a Big Dye Terminator v3 . 1 Sequencing Kit ( Applied Biosystems ) . The sequenced fragments were assembled by Seqman ( DNASTAR , Madison , WI ) and the complete mitochondrial genomes were aligned with the CanFam3 . 1 dog mitochondrial reference genome ( Lindblad-Toh et al . , 2005 ) . Nuclear copies of mtDNA ( NuMTs ) are mtDNA fragments that have been incorporated into the nuclear genome . Over 150 NuMTs have been identified in the canine genome ( Verscheure et al . , 2015 ) . Somatically acquired NuMTs have also been described in human cancer ( Ju et al . , 2015 ) . Given that our study design did not involve purification of cytoplasmic mtDNA genomes , we assessed the possibility that our mtDNA variant analysis has been influenced by NuMTs . The two strands of the mtDNA are known as the heavy and light strands , and the light strand is the reference strand in CanFam3 . 1 . Each mutation on the conservative somatic list ( n=835 , 'Classification of tumour substitutions' , Supplementary file 4C ) was classified as one of six possible substitutions in the pyrimidine context ( C>A , C>G , C>T , T>A , T>C , T>G ) and assigned to a strand relative to the reference ( i . e . pyrimidine mutations ( i . e . C> , T> ) with respect to the reference were defined as light strand mutations; purine mutations ( i . e . A> , G> ) with respect to the reference were defined as heavy strand mutations ) . The immediate 5’ and 3’ sequence contexts for each CTVT mutation was extracted from the dog mitochondrial reference genome for mutations on the heavy and light strands , yielding a maximum of 96 mutation types ( 6 possible substitutions x 4 possible 5’ bases x 4 possible 3’ bases ) . The number of observations of each substitution type was normalised to the triplet frequency extracted from the canine mitochondrial genome . The following example illustrates how to calculate the observed/expected ratio for T[C>T]G occurring on the heavy strand . We observed a total of 835 substitutions occurring across the MT genome; given that the TCG triplet is observed 117 times in the dog mitochondrial reference genome heavy strand , the frequency of TCG triplets occurring in the dog mtDNA heavy strand is 117/16727 = 0 . 007 , where 16 , 727 bp is the length of the dog mitochondrial genome . Using the frequency of TCG triplets in the reference genome , we can calculate the expected number of T[C>T]G substitution types on the heavy strand as ( total number of mutations ) x ( TCG frequency on the heavy strand ) / 3 ( as there are 3 possible C>N substitutions ) i . e . expected number of T[C>T]G substitutions on the heavy strand = 835 x 0 . 007/ 3 ≈ 1 . 95 . As we observed 22 T[C>T]G mutations on the heavy strand , the observed/expected ratio for this mutation type was 22/1 . 95 = 11 . 28 . Triplets within region MT:16129–16430 inclusive ( 'Substitution calling-Extraction and filtering' ) , as well as a set of specific excluded sites ( 'Additional quality checks and validation' ) were excluded from our analysis . This was accounted for during the calculation of expected substitutions described above . | A unique cancer called canine transmissible venereal tumour ( CTVT ) causes ugly tumours to form on the genitals of dogs . Unlike most other cancers , CTVT is contagious: the cancer cells can be directly transferred from one dog to another when they mate . The disease originated from the cancer cells of one individual dog that lived approximately 11 , 000 years ago . CTVT now affects dogs all over the world , which makes it the oldest and most widespread cancer known in nature . Like healthy cells , cancer cells contain compartments known as mitochondria that produce the chemical energy needed to power vital processes . Inside the mitochondria , there is some DNA that encodes the proteins that mitochondria need to perform this role . Changes ( or mutations ) to this mitochondrial DNA ( mtDNA ) may stop the mitochondria from working properly . CTVT cells have previously been found to occasionally capture mtDNA from normal dog cells , which suggests that replenishing their mtDNA may help promote CTVT cell growth . Furthermore , these captured mtDNAs act as genetic "flags" that can help trace the spread of the disease . Here , Strakova , Ní Leathlobhair et al . analysed the mtDNA in CTVT tumours collected from over 400 dogs in 39 countries . The analysis shows that CTVT cells have captured mtDNA from normal dog cells on at least five occasions . Over the last 2 , 000 years , the disease appears to have spread rapidly around the world , perhaps transported by dogs travelling on ships along historic trade routes . CTVT may have only reached the Americas within the last 500 years , possibly carried there by dogs brought by Europeans . Likewise , CTVT probably only came to Australia after European contact . The experiments also revealed that the most damaging types of mutations were absent from the mtDNA of CTVT , which suggests that fully functioning mitochondria play an important role in CTVT . Unexpectedly , Strakova , Ní Leathlobhair et al . found evidence that certain sections of mtDNA in some CTVT cells have been exchanged , or shuffled , with the mtDNA captured from normal dog cells . This type of “recombination” is not usually thought to occur in mtDNA , and has not previously been detected in cancer . Future studies will determine if this process is widespread in other types of cancer , including in humans . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"cancer",
"biology"
] | 2016 | Mitochondrial genetic diversity, selection and recombination in a canine transmissible cancer |
Neurons are sensitive to low oxygen ( hypoxia ) and employ a conserved pathway to combat its effects . Here , we show that p38 MAP Kinase ( MAPK ) modulates this hypoxia response pathway in C . elegans . Mutants lacking p38 MAPK components pmk-1 or sek-1 resemble mutants lacking the hypoxia response component and prolyl hydroxylase egl-9 , with impaired subcellular localization of Mint orthologue LIN-10 , internalization of glutamate receptor GLR-1 , and depression of GLR-1-mediated behaviors . Loss of p38 MAPK impairs EGL-9 protein localization in neurons and activates the hypoxia-inducible transcription factor HIF-1 , suggesting that p38 MAPK inhibits the hypoxia response pathway through EGL-9 . As animals age , p38 MAPK levels decrease , resulting in GLR-1 internalization; this age-dependent downregulation can be prevented through either p38 MAPK overexpression or removal of CDK-5 , an antagonizing kinase . Our findings demonstrate that p38 MAPK inhibits the hypoxia response pathway and determines how aging neurons respond to hypoxia through a novel mechanism .
Whereas the brain comprises about 2% of human body weight , it consumes about 20% of oxygen intake , highlighting that neurons require robust aerobic energy production ( Clarke and Sokoloff , 1999 ) . Neurons do not have plentiful glycolytic reserves , yet must expend tremendous amounts of ATP to maintain their membrane potential . Failure to do so , as during the low oxygen conditions ( hypoxia ) that occur during ischemic stroke , can lead to the collapse of the membrane potential , large scale glutamate neurotransmitter release , and overactivation of glutamate receptors ( Takahashi et al . , 2002; Lees , 2000; Xue et al . , 1994; Li and Buchan , 1993; Gill , 1994; Sheardown et al . , 1993; Nellgard and Wieloch , 1992; Buchan et al . , 1991 ) . An understanding of how neurons respond to hypoxic stress is important for the development of new therapies to prevent and treat damage caused by ischemic stroke or traumatic injury . Multicellular organisms respond to hypoxic stress by activating a conserved hypoxia response pathway , which includes prolyl hydroxylase enzymes as oxygen sensors and hypoxia-inducible factor alpha ( HIFα ) transcription factors as effectors . When oxygen levels are sufficiently high ( i . e . , normoxia ) , the EGL-9/PHD family of prolyl hydroxylases employs molecular oxygen , 2-oxoglutarate , and iron to hydroxylate specific proline side chains on the HIFα transcription factors , resulting in the inactivation of these transcription factors by ubiquitin-mediated protein degradation ( Fong and Takeda , 2008; Aragones et al . , 2009; Wong et al . , 2013; Majmundar et al . , 2010 ) . When oxygen levels are not sufficiently high ( i . e . , hypoxia ) , the EGL-9/PHD prolyl hydroxylases become inactive , resulting in HIFα protein stabilization and thus a transcriptional change in gene expression ( Semenza , 2009; Fandrey and Gassmann , 2009 ) . Depending on the specific cells and tissues undergoing stress , HIFα can mediate adaptive responses to hypoxia that include increased erythropoiesis , increased angiogenesis , and reprogramming of metabolism away from oxidative phosphorylation and towards glycolysis and anaerobic fermentation . In addition to its role in response to acute hypoxic stress , the hypoxia response pathway also helps to maintain stem cell niches and tumor growth and metastasis in cancer ( Amelio and Melino , 2015; Gupta et al . , 2014; Ito and Suda , 2014 ) . Whereas the major target of oxygen regulation through EGL-9/PHD proteins is HIFα , several studies have shown that EGL-9/PHD oxygen sensors regulate additional proteins as part of the overall hypoxia response ( Lee et al . , 2005; Koditz et al . , 2007; Fu et al . , 2007; Fu and Taubman , 2010; Cummins et al . , 2006; Park et al . , 2012; Lee et al . , 2015 ) . Thus , while the core pathway of the hypoxia response is well established , it is less clear how the pathway uses alternative effectors and modulators in different tissues and contexts so as to tailor the specific physiological response to stress . Mammalian neurons are particularly sensitive to hypoxia , making in vivo studies challenging . The genetically tractable and hypoxia tolerant model organism C . elegans has allowed investigators to study how the hypoxia response pathway functions in multiple tissue types , developmental stages , and aging ( Rodriguez et al . , 2013; Leiser and Kaeberlein , 2010; Powell-Coffman , 2010 ) . C . elegans possess a single prolyl hydroxylase , called EGL-9 , and a single HIFα , called HIF-1 . These two proteins are expressed in essentially all tissues in C . elegans , where they mediate the primary response that allows nematodes to survive when they encounter hypoxic niches within their natural environment of the soil . HIF-1 also has a complex role in regulating aging and protein homeostasis in C . elegans ( Rodriguez et al . , 2013; Fawcett et al . , 2015 ) . Hypoxia modulates a specific nematode behavior through EGL-9 but independent of HIF-1 ( Park et al . , 2012 ) . Nematodes navigate their environment using a biased random walk comprised of long runs of forward locomotion and spontaneous reversals of locomotion followed by changes in direction ( Cohen and Sanders , 2014; Gray et al . , 2005 ) . The frequency of spontaneous reversals is determined by the activity of AMPA-type glutamate receptors ( AMPARs ) located in a small number of command interneurons ( Schaefer and Rongo , 2006; Mellem et al . , 2002; Zheng et al . , 1999 ) . C . elegans avoids zones of hypoxia using a combination of sensory neuron-mediated aerotaxis and command interneuron-mediated spontaneous reversals ( Chang et al . , 2006; Cheung et al . , 2005; Park et al . , 2012 ) . In a normoxic environment , C . elegans exhibits a relatively high frequency of spontaneous reversals , resulting in a bias towards local foraging behaviors . When exposed to hypoxia for long periods , C . elegans exhibits a depressed frequency of spontaneous reversals , resulting in a bias towards roaming behavior that allows the animal to potentially exit the hypoxic environment . The local foraging behavior in normoxic environments requires EGL-9 , as egl-9 mutants exhibit decreased reversals similar to those observed under hypoxia ( Park et al . , 2012 ) . Surprisingly , this behavioral phenotype does not require HIF-1 . Hypoxia and EGL-9 regulate C . elegans reversal behavior by regulating the membrane trafficking of the AMPAR subunit GLR-1 . GLR-1-containing AMPARs act in the command interneurons to receive synaptic input and direct overall locomotory reversal behavior ( Hart et al . , 1995; Mellem et al . , 2002; Maricq et al . , 1995; Chang and Rongo , 2005 ) . Mutants that lack GLR-1 have a depressed frequency of spontaneous reversals . The synaptic localization of GLR-1 can be detected in vivo using a functional GLR-1::GFP chimeric protein , and mutants that fail to localize GLR-1 to synapses also have a depressed frequency of reversals ( Burbea et al . , 2002; Glodowski et al . , 2005; Schaefer and Rongo , 2006; Shim et al . , 2004; Zheng et al . , 1999; Rongo et al . , 1998 ) . Wild-type animals exposed to hypoxia ( or egl-9 mutants under normoxia ) accumulate GLR-1 receptors in internal endosomal compartments ( Park et al . , 2012 ) . Under normoxia , oxygen promotes the interaction and endosomal recruitment of EGL-9 with LIN-10 , an ortholog of the Mint/X11 scaffolding molecules , and LIN-10 in turn promotes GLR-1 recycling to the plasma membrane ( Whitfield et al . , 1999; Glodowski et al . , 2005; Park et al . , 2009; 2012 ) . Under hypoxia , EGL-9 releases LIN-10 , allowing LIN-10 to be phosphorylated by the CDK-5 kinase . Phosphorylated LIN-10 is then released from endosomes , resulting in diminished GLR-1 recycling , depletion of synaptic GLR-1 by endocytosis without accompanying recycling , and decreased GLR-1-mediated reversal behavior ( Park et al . , 2012; Juo et al . , 2007 ) . EGL-9 and oxygen regulate GLR-1 recycling through a novel HIF-1-independent mechanism , suggesting that different tissues can employ parts of the hypoxia response pathway for specialized functions , and that additional modulators and mediators of the pathway remain to be discovered . Here , we show that signaling through the kinases SEK-1 ( p38 MAPKK ) and PMK-1 ( p38 MAPK ) regulate GLR-1 recycling and GLR-1-mediated reversal behavior by modulating the hypoxia response pathway . Loss of function mutations in either pmk-1 or sek-1 mimic the effects of hypoxia on GLR-1 trafficking and behavioral output . Wild-type SEK-1 and PMK-1 promote the endosomal localization of EGL-9 and LIN-10 in neurons under normoxia , and the effect of sek-1 or pmk-1 mutations on EGL-9/LIN-10 co-localization and GLR-1 recycling requires the activity of the CDK-5 kinase . Wild-type SEK-1 and PMK-1 also regulate HIF-1 throughout the organism . Older animals show reduced levels of activated PMK-1 , GLR-1 internalization , and decreased GLR-1-mediated behaviors . The reduction of functional GLR-1 in older animals can be prevented through either the overexpression of PMK-1 or the removal of CDK-5 . Our findings demonstrate that p38 MAPK is a modulator of the hypoxia response pathway through EGL-9 , and that this novel mechanism helps determine how aging neurons respond to hypoxia .
We previously showed that the hypoxia response pathway regulates GLR-1 recycling and function ( Park et al . , 2012 ) , and we therefore reasoned that other signaling pathways that respond to oxidative stress conditions might also contribute to GLR-1 regulation . One such signaling molecule is the p38 MAPK ortholog PMK-1 , which is involved in oxidative stress response and innate immunity ( Berman et al . , 2001; Kim et al . , 2002; Inoue et al . , 2005 ) . To determine if this p38 MAPK regulates GLR-1 trafficking , we obtained a viable mutant strain homozygous for a complete loss of function ( deletion ) allele in pmk-1 ( Mizuno et al . , 2004 ) . We introduced a transgene , nuIs25 , which expresses full length , functional GLR-1 receptors tagged with GFP ( GLR-1::GFP ) , into pmk-1 mutants . In wild-type nematodes , GLR-1::GFP is localized to discrete puncta ( mean diameter of 0 . 48 microns , SEM of 0 . 01 microns , 95% of puncta are between 0 . 32 and 0 . 73 microns in diameter ) along the ventral cord dendrites of interneurons ( Figure 1A ) , with 85% of such puncta colocalized with synaptic markers ( Rongo et al . , 1998; Burbea et al . , 2002 ) . We found that pmk-1 mutants accumulated GLR-1::GFP in elongated structures ( mean length of 2 . 30 microns , SEM of 0 . 10 microns , 95% of these accumulations are between 1 . 32 and 3 . 49 microns in length ) along the ventral cord ( Figure 1B ) , similar to the GLR-1::GFP accumulations in elongated endosomes observed in mutants for membrane recycling factors ( Shi et al . , 2010; Glodowski et al . , 2007; Kramer et al . , 2010; Park et al . , 2009; Rongo et al . , 1998 ) . GLR-1 puncta and GLR-1 accumulations are distinct enough in shape and size to allow easy quantification of their respective numbers along the ventral cord dendrites ( Figure 1C , D ) . We found a sizeable decrease in the number of GLR-1 puncta ( Figure 1C ) and a five-fold increase in the number of GLR-1 elongated accumulations ( Figure 1D ) in pmk-1 mutants relative to wild type . Expression of a wild-type pmk-1 cDNA from the glr-1 promoter , which drives expression specifically in the command interneurons , was sufficient to rescue pmk-1 mutants , indicating a cell autonomous requirement for PMK-1 function ( Figure 1C , D ) . The observed changes in GLR-1 were unlikely to be due to general defects in synapse formation or overall cell polarity , as the localization of a synaptobrevin-GFP reporter ( SNB-1::GFP ) , which decorates synaptic vesicles at interneuron presynaptic elements when expressed from a transgene , did not change in pmk-1 mutants relative to wild type ( Figure 1E , F ) . In addition , we detected similar levels of glr-1 mRNA in wild-type animals and pmk-1 mutants , indicating that PMK-1 regulates GLR-1 in a posttranscriptional fashion ( Figure 1G ) . 10 . 7554/eLife . 12010 . 003Figure 1 . Signaling through PMK-1 p38 MAPK regulates GLR-1 AMPAR trafficking . GLR-1::GFP fluorescence in ( A ) wild-type animals and ( B ) pmk-1 ( km25 ) mutants . GLR-1 is localized to elongated accumulations ( indicated by yellow arrows ) . Bar: 5 μm . Average GLR-1::GFP number is quantified as ( C , I ) puncta or ( D , J ) accumulations per length of ventral cord dendrites . Average SNB-1::GFP puncta are quantified based on ( E ) number per length of ventral cord and ( F ) puncta width . ( G ) Relative glr-1 mRNA levels quantified by qRT-PCR and normalized to the mean value for wild type . ( H ) Spontaneous reversal frequency ( number of reversals measured over a 5-min period ) represented as a percentage of the mean value for wild type . Graph bar columns labeled with asterisks indicate statistical difference by ANOVA followed by Dunnett’s multiple comparison to wild type ( ****p<0 . 0001 , ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 ) . Lines connecting specific columns indicate pairwise comparisons using the Holm-Šídák test . Error bars indicate SEM . N = 13–47 animals per genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 12010 . 00310 . 7554/eLife . 12010 . 004Figure 1—figure supplement 1 . Additional related factors that do not Alter GLR-1 localization In C . elegans neurons . ( A–D ) Average GLR-1::GFP number is quantified as ( A , C ) puncta or ( B , E ) accumulations per length of ventral cord dendrites . Graph bar columns labeled with asterisks indicate statistical difference by ANOVA followed by Dunnett’s multiple comparison to wild type ( ****p<0 . 0001 ) . Error bars indicate SEM . N = 8–15 animals per genotype . ( E–G ) GLR-1::GFP fluorescence in wild-type animals that have been exposed to ( A ) OP50 E . coli , ( B ) 5 hr of Pseudomonas aeruginosa strain PA14 , or ( C ) 8 hr of PA14 . Bar: 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12010 . 004 PMK-1 is part of a p38 MAPK pathway that responds to bacterial infection and promotes innate immunity via the MAPKKK NSY-1 and the MAPKK SEK-1 ( Mizuno et al . , 2004 ) . To determine if these signaling molecules are required to regulate GLR-1 trafficking , we examined GLR-1::GFP in loss of function mutations for both genes . A loss of function allele for sek-1 resulted in a similar GLR-1 localization phenotype to that observed in pmk-1 mutants ( Figure 1C , D ) . Expression of a wild-type sek-1 cDNA from the glr-1 promoter rescued sek-1 mutants , indicating a cell autonomous requirement for SEK-1 function ( Figure 1C , D ) . Double mutants for sek-1 and pmk-1 showed a similar phenotype compared to either single mutant ( Figure 1C , D ) , suggesting that these mutations do not yield an additive phenotype and are thus likely acting in the same pathway . By contrast , two loss of function , putative null alleles in nsy-1 ( one a nonsense mutation and the other an insertion resulting in a frameshift ) , did not result in abnormal GLR-1 localization ( Figure 1C , D ) . Thus , whereas SEK-1 appears to be the MAPKK for PMK-1 to regulate GLR-1 localization , NSY-1 is unlikely to be the p38 MAPKKK . We examined mutants for several additional MAPKKK genes in C . elegans , including TIR-1 , MLK-1 , and KIN-18; however , we did not observe changes in GLR-1::GFP localization ( Figure 1—figure supplement 1 ) . Mutants for MAPKKK DLK-1 also do not show the same GLR-1::GFP localization defect observed in pmk-1 mutants ( Park et al . , 2009 ) . Our results suggest that the upstream MAPKKK for this function of p38 MAPK signaling is likely to be a noncanonical kinase relative to traditional MAPK signaling . The C . elegans genome contains an additional p38 MAPK , called PMK-2 , that functions redundantly with PMK-1 in the nervous system to regulate behavioral responses to pathogenic bacteria ( Pagano et al . , 2015 ) . We examined GLR-1:GFP in pmk-2 ( qd284 ) mutants , which contain a deletion and frameshift at the beginning of the ORF , making this mutant allele a likely null; however , we did not observe a difference in GLR-1 puncta or accumulations compared to wild type ( Figure 1—figure supplement 1 ) . The gain of function mutation pmk-2 ( qd171qd279 ) , which can suppress other nervous system defects of pmk-1 ( km25 ) mutations ( Pagano et al . , 2015 ) , did not suppress the effects of pmk-1 ( km25 ) mutations on GLR-1 localization ( Figure 1—figure supplement 1 ) . Two independent pmk-2 loss of function mutations – qd279 and qd280 – did not enhance the GLR-1 localization phenotype caused by pmk-1 ( km25 ) mutations ( Figure 1—figure supplement 1 ) , although qd280 showed a mild suppression of the depressed GLR-1 puncta number phenotype caused by the pmk-1 ( km25 ) mutation . Taken together , our results indicate the PMK-2 does not regulate GLR-1 localization either by itself or redundantly with PMK-1 . PMK-1 signaling promotes innate immunity and the oxidative stress response by phosphorylating and activating the Nrf2 transcription factor ortholog SKN-1 ( Hoeven et al . , 2011; Papp et al . , 2012 ) . We therefore examined GLR-1::GFP in skn-1 deletion allele homozygotes; however , we did not detect a difference in GLR-1::GFP localization relative to wild type ( Figure 1C , D ) . PMK-1 also regulates the transcription factor ATF-7 ( Shivers et al . , 2010 ) . We examined atf-7 single mutants and pmk-1 atf-7 double mutants; however , we did not observe a change in GLR-1::GFP localization ( Figure 1—figure supplement 1 ) . These findings suggest that PMK-1 does not regulate GLR-1 through its most well established transcriptional outputs . We also examined GLR-1::GFP localization in wild-type animals raised on the pathogenic bacteria Pseudomonas aeruginosa ( strain PA14 ) , which is known to promote an innate immune response by activating the PMK-1 pathway ( Papp et al . , 2012; Shivers et al . , 2010 ) however , we did not observe any significant changes relative to wild type ( Figure 1—figure supplement 1 ) . These results indicate that the MAPKK SEK-1 and the MAPK PMK-1 act in a distinct and novel p38 MAPK signaling pathway to regulate GLR-1 subcellular localization . The elongated accumulations of GLR-1 in sek-1 and pmk-1 mutants are similar to those observed in mutants in which GLR-1 recycling is impaired , suggesting that they might represent internalized receptors . Internalization of GLR-1 AMPARs results in diminished interneuron synaptic function . We examined functional synaptic GLR-1 through a standardized measurement: the frequency of spontaneous reversals of locomotion in the brief absence of food ( Schaefer and Rongo , 2006; Mellem et al . , 2002; Zheng et al . , 1999 ) . Wild-type animals exhibited a robust frequency of reversals , whereas glr-1 null mutants showed a depressed reversal frequency ( Figure 1H ) . Similar to glr-1 null mutants , null mutants for sek-1 and pmk-1 showed a reduced frequency of reversals , whereas a null mutant for skn-1 showed a reversal frequency that was similar to that of wild type ( Figure 1H ) . Expression of either a wild-type sek-1 or pmk-1 cDNA from the glr-1 promoter was sufficient to rescue the reversal phenotype of the corresponding mutation ( Figure 1H ) . Thus , SEK-1 and PMK-1 are required to promote GLR-1 function . If the elongated accumulations containing GLR-1::GFP in pmk-1 and sek-1 mutants represent AMPARs trapped in endosomes following endocytosis , then a reduction in GLR-1 endocytosis should suppress the accumulation of GLR-1 in these structures . We previously showed that expression of a dominant negative RAB-5 , which contains a mutation that mimics the GDP-bound state of this GTPase , reduces GLR-1 endocytosis and suppresses internal accumulation of GLR-1 in membrane recycling mutants ( Park et al . , 2009; Bucci et al . , 1992; Li et al . , 1994 ) . We introduced a transgene that expresses RAB-5 ( GDP ) from the glr-1 promoter into either sek-1 or pmk-1 mutants . For both mutants , expression of RAB-5 ( GDP ) restored the GLR-1 synaptic puncta number to wild-type levels ( Figure 2A–D ) and suppressed the accumulation of GLR-1::GFP in elongated accumulations ( Figure 2C , E ) , consistent with such accumulations being post-endocytic . 10 . 7554/eLife . 12010 . 005Figure 2 . The p38 MAPK pathway promotes GLR-1 AMPAR function and recycling from endosomes . GLR-1::GFP fluorescence in ( A ) wild-type animals , ( B ) pmk-1 ( km25 ) mutants , and ( C ) pmk-1 ( km25 ) mutants containing a transgene that expresses dominant negative RAB-5 with a GDP-locked mutation . Yellow arrows indicate elongated accumulations . Bar: 5 μm . Average GLR-1::GFP number is quantified as ( D ) puncta or ( E ) accumulations per length of ventral cord dendrites . ( F , J , N ) GLR-1::GFP and ( G , K , O ) mRFP::SYX-7 fluorescence observed in the PVC neuron cell body of ( F , G , H , I ) wild type , ( J , K , L , M ) sek-1 mutants , and ( N , O , P , Q ) pmk-1 mutants . ( H , L , P ) Merged image of the red and green channels . ( I , M , Q ) Binary image with white indicating pixels with significant signal ( colocalization ) in both channels . ( R ) Fraction of GLR-1::GFP-labeled pixels that overlap with mRFP:SYX-7-labeled pixels . Graph bar columns labeled with asterisks indicate statistical difference by ANOVA followed by Dunnett’s multiple comparison to wild type ( ****p<0 . 0001 , ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 ) . Lines connecting specific columns indicate pairwise comparisons using the Holm-Šídák test . Error bars indicate SEM . N = 13–16 animals per genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 12010 . 005 While the small size of C . elegans neurites has precluded an analysis of dendritic endosomes in the command interneurons , accumulation of GLR-1 in endosomes can be directly visualized in C . elegans neuron soma by examining GLR-1::GFP colocalization with an mRFP-tagged syntaxin ( SYX-7 ) that resides at early endosomes ( Chun et al . , 2008; Park et al . , 2009 ) . We co-expressed GLR-1::GFP with mRFP::SYX-7 using the glr-1 promoter and collected single confocal optical sections of neuron cell bodies that express both chimeric proteins . We then quantified co-localization by measuring the fraction of pixels that contain fluorescent from both proteins when such fluorescent was above baseline ( Figure 2F–Q ) . Approximately 30% of GLR-1::GFP is co-localized with mRFP::SYX-7 in wild-type animals ( Figure 2F-I , R ) . By contrast , there is approximately a 50% increase in the portion of GLR-1::GFP that is co-localized with mRFP::SYX-7 in sek-1 ( Figure 2J-M , R ) and pmk-1 ( Figure 2N-Q , R ) mutants , indicating that GLR-1 accumulates at SYX-7-decorated endosomes in these mutants , at least in soma . Taken together , our results indicate that in the absence of SEK-1/PMK-1 signaling , GLR-1 receptors accumulate in elongated , internal , and post-endocytic compartments in the command interneurons , resulting in diminished GLR-1 function and behavioral output . The elongated structures containing GLR-1::GFP observed in sek-1 and pmk-1 mutants resemble similar structures observed in wild-type nematodes exposed to hypoxia , as well as egl-9 mutants under normoxia ( Park et al . , 2012 ) . Given the role of PMK-1 in stress response , we reasoned that PMK-1 signaling might regulate GLR-1 recycling through EGL-9 and the hypoxia response pathway . To explore this possibility , we examined GLR-1::GFP in wild-type and pmk-1 mutant animals exposed to hypoxia using a published nitrogen displacement approach ( Pocock and Hobert , 2008; Park et al . , 2012 ) . As previously described ( Park et al . , 2012 ) , wild-type animals exposed to 0 . 5% oxygen ( hypoxia ) localized GLR-1::GFP to elongated structures along dendrites ( Figure 3A , B , E , F ) , similar to those observed in pmk-1 and sek-1 mutants under normoxic conditions ( Figure 3C , E , F ) . These internalized structures can be enhanced in an additive effect when mutations in distinct membrane trafficking pathways are combined in double mutant combinations ( Park et al . , 2009; Glodowski et al . , 2007; Shi et al . , 2010 ) . However , we observed no statistically significant additive effect of hypoxia exposure on GLR-1 localization in either pmk-1 or sek-1 mutants ( Figure 3D , E , F ) . Mutations in egl-9 under normoxia cause a similar effect on GLR-1 trafficking to that in wild-type animals under hypoxia ( Figure 3E , F ) . We therefore examined GLR-1::GFP in double mutants between egl-9 and either sek-1 or pmk-1 . We found no statistically significant difference between the double mutants and either single mutant , under conditions of both normoxia and hypoxia ( Figure 3E , F ) . Our results indicate that there is no additive effect of combining mutations that impair p38 MAPK signaling with egl-9 mutations , hypoxia exposure , or both , suggesting that these factors work together in a single pathway to regulate GLR-1 . 10 . 7554/eLife . 12010 . 006Figure 3 . Loss of p38 MAPK signaling occludes the effects of hypoxia on GLR-1 AMPAR trafficking . GLR-1::GFP fluorescence in ( A , B ) wild-type animals or ( C , D ) pmk-1 ( km25 ) mutants under conditions of ( A , C ) normoxia or ( B , D ) hypoxia . Yellow arrows indicate elongated accumulations . Bar: 5 μm . Average GLR-1::GFP number is quantified as ( E ) puncta or ( F ) accumulations per length of ventral cord dendrites . ( G , H ) PMK-1::GFP fluorescence in wild-type animals under ( G ) normoxia or ( H ) hypoxia . Cell bodies for AVG and RIGL are indicated . Average nuclear PMK-1::GFP fluorescence intensity ( normalized to the average value in wild type ) is quantified in ( I ) . Red bar columns indicate animals under normoxia , whereas blue bar columns indicate animals exposed to hypoxia . Graph bar columns labeled with asterisks ( ****p<0 . 0001 , **p<0 . 01 , *p<0 . 05 ) indicate statistical difference by ( E , F ) ANOVA followed by Dunnett’s multiple comparison to wild type or Tukey’s multiple comparison indicated by the brackets , and ( I ) Student t test . Error bars indicate SEM . N = 11–24 animals per genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 12010 . 006 EGL-9 and hypoxia regulate multiple physiological processes through their regulation of HIF-1 function ( Shen et al . , 2006; Chang and Bargmann , 2008; Pocock and Hobert , 2008; Shao et al . , 2010; Gort et al . , 2008 ) . However , the regulation of GLR-1 trafficking by EGL-9 and hypoxia does not require HIF-1 ( Park et al . , 2012 ) . Using a hif-1 molecular null allele ( Jiang et al . , 2001 ) , we tested whether hif-1 mutations suppressed the accumulation of GLR-1 observed in sek-1 and pmk-1 mutants; however , we observed no difference in either sek-1 hif-1 or pmk-1 hif-1 double mutants compared to sek-1 and pmk-1 single mutants ( Figure 3E , F ) . Taken together , our results are consistent with the p38 MAPK components SEK-1 and PMK-1 regulating GLR-1 trafficking through an EGL-9-dependent , HIF-1-independent mechanism . To examine PMK-1 subcellular localization in these neurons , we generated a transgene containing the glr-1 promoter driving a full length PMK-1::GFP chimeric protein . We found that PMK-1::GFP was enriched in the nuclei of the command interneurons and distributed in a diffuse fashion throughout the ventral cord dendrites ( Figure 3G ) . We also examined animals carrying the same transgene under conditions of hypoxia . In response to hypoxia , we observed a decrease in PMK-1::GFP in both the cell bodies and the dendrites ( Figure 3H ) . PMK-1 was less enriched in the nuclei and more diffusely distributed in the cell body cytosol ( Figure 3H , I ) . Given that PMK-1 is under the control of the glr-1 promoter and the unc-54 3’UTR sequences in this experiment , and that these regulatory sequences have not shown oxygen-dependent regulation in previous experiments ( Park et al . , 2012; Ghose et al . , 2013 ) , our results suggest that oxygen elevates PMK-1 levels through a post-transcriptional mechanism . Under normoxic conditions , EGL-9 promotes GLR-1 recycling by binding to the N-terminus of LIN-10 , thereby preventing the kinase CDK-5 from phosphorylating LIN-10 and triggering its diffusion ( delocalization ) along dendrites ( Park et al . , 2012 ) . If PMK-1 and SEK-1 regulate GLR-1 in the same manner as does EGL-9 , then one would expect that ( 1 ) a cdk-5 mutation would suppress the accumulation of GLR-1 observed in sek-1 and pmk-1 mutants ( similar to how it suppresses accumulation in egl-9 mutants ) , and ( 2 ) LIN-10 would be diffusely distributed in sek-1 and pmk-1 mutants ( similar to how LIN-10 is diffusely distributed in egl-9 mutants ) ( Park et al . , 2012 ) . To test the first expectation , we examined GLR-1::GFP in double mutants between cdk-5 and either sek-1 or pmk-1 , and we found that GLR-1 did not accumulate in both double mutants ( Figure 4A–F ) . Consistent with the GLR-1 trafficking data , we observed that a cdk-5 mutation suppressed the spontaneous reversal defects caused by sek-1 and pmk-1 mutations ( Figure 4G ) . These findings place CDK-5 genetically downstream of SEK-1 and PMK-1 . 10 . 7554/eLife . 12010 . 007Figure 4 . CDK-5 is required for p38 MAPK to regulate GLR-1 AMPAR trafficking . GLR-1::GFP fluorescence in ( A ) wild-type animals , ( B ) pmk-1 ( km25 ) mutants , ( C ) cdk-5 ( ok626 ) mutants , and ( D ) pmk-1 ( km25 ) cdk-5 ( ok626 ) double mutants . Yellow arrows indicate elongated accumulations . Bar: 5 μm . Average GLR-1::GFP number is quantified as ( E ) puncta or ( F ) accumulations per length of ventral cord dendrites . ( G ) Spontaneous reversal frequency ( number of reversals measured over a 5-min period ) represented as a percentage of the mean value for wild type . Graph bar columns labeled with asterisks indicate statistical difference by ANOVA followed by Dunnett’s multiple comparison to wild type ( ****p<0 . 0001 ) . Lines connecting specific columns indicate pairwise comparisons using the Holm-Šídák test . Error bars indicate SEM . N = 15–28 animals per genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 12010 . 007 To test the second prediction , we examined LIN-10 localization in p38 MAPK signaling mutants by introducing a transgene that expresses a functional LIN-10::GFP chimeric protein solely in the GLR-1-expressing command interneurons ( Rongo et al . , 1998 ) . LIN-10::GFP is localized to small puncta along dendrites ( Figure 5A ) , and this localization requires EGL-9 and oxygen but is inhibited by CDK-5 ( Park et al . , 2012; Juo et al . , 2007 ) . We found that LIN-10::GFP was diffusely distributed throughout dendrites ( and with few puncta ) in sek-1 and pmk-1 mutants , similar to its distribution in egl-9 mutants and animals undergoing hypoxic stress ( Figure 5B , E ) . By contrast , mutations in cdk-5 result in more LIN-10::GFP puncta ( Figure 5C , E ) , and these puncta are larger and brighter , resulting in more total localized LIN-10 along dendrites , which can be quantified as integrated optical density ( IOD ) along the dendrites ( Figure 5F ) . Moreover , the observed delocalization of LIN-10 in sek-1 and pmk-1 mutants is completely blocked when a cdk-5 mutation is introduced into these genetic backgrounds ( Figure 5D , E , F ) . This suggests that p38 MAPK signaling promotes LIN-10 localization into puncta by antagonizing the diffuse distribution ( delocalization ) that would otherwise be promoted by CDK-5 . Taken together , our results are consistent with p38 MAPK signaling working together with EGL-9 to promote LIN-10 localization into puncta , and that the underlying mechanism is through the prevention of CDK-5 from opposing LIN-10 localization . 10 . 7554/eLife . 12010 . 008Figure 5 . The PMK-1 p38 MAPK regulates LIN-10 localization . LIN-10::GFP fluorescence in ( A ) wild-type animals , ( B ) pmk-1 ( km25 ) mutants , ( C ) cdk-5 ( ok626 ) mutants , and ( D ) pmk-1 ( km25 ) cdk-5 ( ok626 ) double mutants . Bar: 5 μm . ( E ) Average LIN-10::GFP puncta number is quantified per length of ventral cord dendrites . ( F ) Average integrated optical density ( IOD ) per puncta per animal as a measurement of total localized LIN-10::GFP . IOD is the sum of the pixel values for each puncta , reflecting both puncta size and fluorescence intensity . Graph bar columns labeled with asterisks indicate statistical difference by ANOVA followed by Dunnett’s multiple comparison to wild type ( ****p<0 . 0001 , *p<0 . 05 ) . Lines connecting specific columns indicate pairwise comparisons using the Holm-Šídák test . Error bars indicate SEM . N = 15 animals per genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 12010 . 008 EGL-9 and oxygen promote LIN-10 subcellular localization , and a specific splice isoform of EGL-9 , called EGL-9E , is colocalized with LIN-10 in dendrites ( Park et al . , 2012 ) . As the p38 MAPK pathway could regulate LIN-10 either directly or indirectly by regulating EGL-9 , we examined EGL-9E::GFP subcellular localization in p38 MAPK signaling mutants . Whereas EGL-9E::GFP is localized to puncta along the ventral cord dendrites ( Figure 6A , E ) , it was diffusely distributed in sek-1 and pmk-1 mutants , with few puncta and little total punctate EGL-9E::GFP along dendrites ( Figure 6B , E , F ) . One explanation for the impaired localization of EGL-9E in p38 MAPK signaling mutants is if EGL-9E were to depend on LIN-10 , the localization of which is also affected in these p38 MAPK mutants . The loss of EGL-9E would thus be a secondary consequence of impaired LIN-10 localization in these mutants . We examined EGL-9E::GFP in cdk-5 mutants ( which have augmented LIN-10 subcellular localization ) and double mutants between cdk-5 and either sek-1 or pmk-1 . EGL-9E is localized similar to wild type in cdk-5 single mutants ( Figure 6C , E , F ) , and mutations in cdk-5 do not suppress the effects on EGL-9E subcellular localization observed in sek-1 or pmk-1 mutants ( Figure 6D , E , F ) . Similarly , whereas LIN-10 subcellular localization depends on EGL-9E , the subcellular localization of EGL-9E does not require LIN-10 ( Figure 6E , F ) . Taken together , our data indicate that p38 MAPK signaling acts genetically upstream of EGL-9 , promoting the subcellular localization of the EGL-9E isoform in neurons independent from its effects on LIN-10 or CDK-5 . 10 . 7554/eLife . 12010 . 009Figure 6 . The PMK-1 p38 MAPK regulates EGL-9 localization . EGL-9E::GFP fluorescence in ( A ) wild-type animals , ( B ) pmk-1 ( km25 ) mutants , ( C ) cdk-5 ( ok626 ) mutants , and ( D ) pmk-1 ( km25 ) cdk-5 ( ok626 ) double mutants . Bar: 5 μm . ( E ) Average EGL-9E::GFP puncta number is quantified per length of ventral cord dendrites . ( F ) Average integrated optical density ( IOD ) per puncta per animal as a measurement of total localized EGL-9E::GFP . IOD is the sum of the pixel values for each puncta , reflecting both puncta size and fluorescence intensity . Graph bar columns labeled with asterisks indicate statistical difference by ANOVA followed by Dunnett’s multiple comparison to wild type ( ****p<0 . 0001 ) . Error bars indicate SEM . N = 13–20 animals per genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 12010 . 009 Given our finding that p38 MAPK signaling promotes EGL-9E subcellular localization and its non-canonical , HIF-1-independent function in the command interneurons , we reasoned that p38 MAPK signaling might also promote canonical EGL-9 function , including its ability to repress HIF-1 , throughout the organism . We tested this hypothesis several ways . First , we examined the expression of a transcriptional reporter for HIF-1 using nIs470 , a transgene containing the cysl-2 promoter and Venus ( Ma et al . , 2012 ) . Expression from the cysl-2 reporter is inactive under normoxia but activated by hypoxia ( Figure 7A , B ) . We introduced nIs470 into pmk-1 mutants and found that cysl-2 reporter expression under normoxia was elevated in these mutants , similar to the expression observed in wild-type animals under hypoxia ( Figure 7C ) . Exposure to hypoxia did not result in increased cysl-2 reporter expression in pmk-1 mutants , suggesting that pmk-1 mutations occlude any additional effects of hypoxia on cysl-2 transcription ( Figure 7D ) . 10 . 7554/eLife . 12010 . 010Figure 7 . The p38 MAPK pathway modulates the hypoxia response pathway . Fluorescence from Venus expressed from the cysl-2 promoter in animals carrying a Pcysl-2::Venus transgene . Either ( A , B ) wild-type animals or ( C , D ) pmk-1 ( km25 ) mutants under ( A , C ) normoxia or ( B , D ) hypoxia are shown . Note that pharyngeal fluorescence is detected from the Pmyo-2::mCherry injection marker even under normoxia . Bar: 100 μm . ( E ) Relative nhr-57 mRNA levels from the indicated genotypes ( under normoxia ) quantified by qRT-PCR and normalized to the mean value for wild type . ( F , G , H ) Fluorescence from a HIF-1::GFP chimeric protein expressed from the glr-1 promoter in animals under normoxia and carrying a Pglr-1::HIF-1::GFP transgene . The PVC neuron cell body from ( F ) wild type , ( G ) egl-9 ( sa307 ) mutants , and ( H ) pmk-1 ( km25 ) mutants is shown . ( I ) Average relative HIF-1::GFP fluorescence levels ( normalized to the mean value for wild type ) observed in PVC nuclei under normoxia . Graph bar columns labeled with asterisks indicate statistical difference by ANOVA followed by Dunnett’s multiple comparison to wild type ( ****p<0 . 0001 , ***p<0 . 001 , **p<0 . 01 ) . Error bars indicate SEM . N = 17–20 animals per genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 12010 . 010 We also measured the levels of a different HIF-1 target gene , called nhr-57 , using qRT-PCR to measure endogenous mRNA levels in total nematode lysates ( Shen et al . , 2005 ) . We found that mutations in sek-1 and pmk-1 resulted in a seven-fold increase in nhr-57 mRNA levels relative to a control transcript ( actin ) ( Figure 7E ) , similar to the increase observed in wild-type animals treated under hypoxia ( Bishop et al . , 2004 ) . HIF-1 is required for this increase , as mutations in hif-1 blocked the effect of mutations in sek-1 and pmk-1 ( Figure 7E ) . As a direct measure of EGL-9 activity , we also visualized its HIF-1 substrate using a transgenic HIF-1::GFP chimeric protein expressed in the command interneurons via the glr-1 promoter ( Park et al . , 2012 ) . Low levels of HIF-1::GFP are visible in wild-type neurons ( Figure 7F ) . By contrast , strong HIF-1::GFP foci are visible in the nuclei of egl-9 mutants ( Figure 7G ) as well as in pmk-1 mutants ( Figure 7H ) , suggesting that p38 MAPK signaling , like EGL-9 , promotes HIF-1 turnover ( quantified in Figure 7I ) . Taken together , our results suggest that p38 MAPK signaling modulates the canonical hypoxia response pathway as well as the non-canonical pathway that regulates AMPAR recycling in neurons . The levels of activated PMK-1 decrease as animals age , and this depression in p38 MAPK activity can be detected on Western blots using an anti-phospho-p38 MAPK antibody ( Youngman et al . , 2011 ) . To confirm this change in p38 MAPK signaling , we generated lysates ( in three separate biological replicates ) from wild-type animals and pmk-1 mutants either from young larvae ( stage L4 ) or older adults ( day 9 post-L4 stage ) , separated the proteins by SDS-PAGE , and probed them with an anti-phospho-p38 MAPK antibody and an anti-actin antibody as a loading control ( Figure 8A ) . We detected a 50% decrease in phospho-PMK-1 levels in older animals relative to young larvae , consistent with a decrease in p38 MAPK signaling during aging ( Figure 8A , B ) . We also detected a 30% decrease in pmk-1 mRNA levels in older animals relative to young larvae ( Figure 8C ) . Signaling through PMK-1 and its downstream transcriptional effector ATF-7 promotes the transcription of multiple genes , including that of T24B8 . 5 ( Shivers et al . , 2010 ) . Thus , a transgene ( agIs219 ) containing the T24B8 . 5 promoter driving GFP expression provides an additional means to monitor p38 MAPK signaling via PMK-1 ( Shivers et al . , 2009 ) . We examined GFP expression from the T24B8 . 5 promoter in L4 stage animals and day 9 adults , finding a baseline GFP expression level in L4 animals ( Figure 8D ) that disappeared in older animals ( Figure 8E ) . Expression of GFP from the T24B8 . 5 promoter was abolished in pmk-1 mutants at both stages of development ( Figure 8F , G ) . Taken together , these findings confirm that PMK-1 p38 MAPK activity wanes in aging animals . 10 . 7554/eLife . 12010 . 011Figure 8 . PMK-1 p38 MAPK activity declines with aAge . ( A ) Western blot of whole animal lysates from the indicated genotype and developmental stage ( either L4 larvae or adults aged 9 days past L4 ) . Top panels show signal from anti-phospho-p38 MAPK antibody , whereas the bottom panels show signal from an anti-actin antibody . Arrows point to bands corresponding to the indicated protein . Asterisks indicate additional bands that cross react with the anti-phospho-p38 MAPK antibody but are not actually PMK-1 ( i . e . , they are not present in the pmk-1 molecular null mutant ) . The panels on the left and the panels on the right are from the same Western blot , but from different regions of the SDS-PAGE gel . ( B ) Quantification of the ratio of anti-phospho-p38 MAPK antibody signal to anti-actin antibody signal , normalized to the value of wild-type L4 animals . Values indicate an average from three independent Western blots . ( C ) Quantification of the ratio of pmk-1 mRNA to actin mRNA , normalized to the value of wild-type L4 animals . Values indicate an average from three independent qRT-PCR reactions . Graph bar columns labeled with asterisks indicate statistical significance by Student t test ( *p<0 . 05 ) . Error bars indicate SEM . ( D , E , F , G ) Fluorescence from GFP expressed from the T24B8 . 5 promoter in animals carrying a PT24B8 . 5::GFP transgene . Either ( D , E ) wild-type animals or ( F , G ) pmk-1 ( km25 ) mutants as ( D , F ) L4 stage larvae or ( E , G ) adults 9 days following the L4 stage are shown . Note that GFP expression from AIY ( arrowheads ) is from a transgenic marker ( Pttx-3::GFP ) incorporated into the array . Bar: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12010 . 011 Given our finding that PMK-1 promotes GLR-1 recycling in young adult animals , could an age-dependent decrease in p38 MAPK activity result in GLR-1 internal accumulation in older animals ? We tested this possibility by examining GLR-1::GFP in both L4 larvae and in older animals . Whereas L4 larvae resembled young adults , with ventral cord dendrites containing punctate GLR-1::GFP ( Figure 9A , G ) , we began to observe GLR-1 in elongated structures along the ventral cord of many 4-day adult ( 4 days after L4 stage ) animals . By day 9 after L4 stage ( 9-day adults ) , all older adults had accumulated GLR-1::GFP in elongated accumulations and had fewer GLR-1 puncta , a phenotype similar to that observed in egl-9 , pmk-1 , and sek-1 mutants ( Figure 9B , G , H ) . Consistent with this change in GLR-1 localization , we observed a decrease in spontaneous reversal frequency each day wild-type animals grew older ( Figure 9I ) . We also examined 9-day adult pmk-1 mutants and found that they had similar GLR-1::GFP localization and spontaneous reversal phenotypes to those observed in wild-type adults at all ages ( Figure 9G , H , I ) , suggesting that pmk-1 mutations occlude any additional effect on GLR-1 trafficking and function due to aging . These findings are consistent with depressed p38 MAPK activity causing GLR-1 trafficking defects in older animals . 10 . 7554/eLife . 12010 . 012Figure 9 . Age-onset downregulation of GLR-1 AMPARs through p38 MAPK and CDK-5 signaling . GLR-1::GFP fluorescence in ( A , B ) wild-type animals , ( C , D ) cdk-5 ( ok626 ) mutants , and ( E , F ) wild-type animals expressing a wild-type pmk-1 cDNA from the glr-1 promoter ( from a Pglr-1::PMK-1 ( + ) transgene , labeled as pmk-1 ( OE ) to indicate PMK-1 overexpression ) . Animals are either ( A , C , E ) L4 stage larvae or ( B , D , F ) adults aged 9 days past the L4 stage . Yellow arrows indicate elongated accumulations . Bar: 5 μm . Average GLR-1::GFP number is quantified as ( G ) puncta or ( H ) accumulations per length of ventral cord dendrites . Gray bar columns indicate L4 stage animals , whereas purple bar columns indicate older animals that are 9 days past the L4 stage . Graph bar columns labeled with asterisks indicate statistical difference by ANOVA followed by Dunnett’s multiple comparison to wild type ( ****p<0 . 0001 , ***p<0 . 001 ) . Lines connecting specific columns indicate pairwise comparisons using the Holm-Šídák test . Error bars indicate SEM . N = 13–18 animals per genotype . ( I ) Spontaneous reversal frequency ( number of reversals measured over a 5-min period and normalized per minute ) as measured at different days after the L4 stage in aging animals of the indicated genotype . Asterisks indicate statistical difference by ANOVA followed by Dunnett’s multiple comparison to wild type ( *p<0 . 05 ) at the indicated time point . DOI: http://dx . doi . org/10 . 7554/eLife . 12010 . 012 If older animals accumulate endosomal GLR-1 because of depressed p38 MAPK signaling , then one might expect that ( 1 ) overexpressing wild-type PMK-1 , or ( 2 ) removing CDK-5 activity ( the inhibitory target of PMK-1/EGL-9 regulation ) might suppress this age-onset phenotype . We tested these possibilities by first generating a transgene that overexpresses a wild-type PMK-1 cDNA via the glr-1 promoter . Wild-type animals that overexpress PMK-1 from this transgene showed normal punctate GLR-1::GFP even in 9-day adults ( Figure 9E–H ) , indicating that simply elevating PMK-1 is sufficient to suppress the defects observed during aging ( when the levels of activated PMK-1 are observed to drop ) . Next , we examined older cdk-5 mutants and found that they also failed to accumulate endosomal GLR-1 ( Figure 9C , D , G , H ) , indicating that CDK-5 is required for the defect in GLR-1 recycling in older animals . Mutations in cdk-5 also partially restored GLR-1 function in aging animals , as cdk-5 adults as late as 6 days post L4 have elevated rates of spontaneous reversals relative to wild-type animals of the same age ( Figure 9I ) . Moreover , mutations in cdk-5 restore reversals to pmk-1 mutants in young and old animals to levels that were higher than those for wild-type animals ( Figure 9I ) . Whereas cdk-5 mutants showed more robust reversal behavior during aging relative to wild type , they nevertheless still showed a decline in reversals and GLR-1 puncta number over time , suggesting that additional factors contribute to age-associated decline in this behavioral modality . For example , muscle decline dramatically impairs locomotion in 9-day adults , making interpretation of more subtle behavioral phenotypes like reversal frequency challenging ( Herndon et al . , 2002 ) . Taken together , our results indicate that AMPAR GLR-1 recycling and function decline with age , and that this decline is due , at least in part , to depressed p38 MAPK signaling , subsequent activation of the HIF-1-independent hypoxia response pathway , and CDK-5 activity .
Here we have shown that p38 MAPK signaling modulates both the canonical hypoxia response pathway and a non-canonical hypoxia response pathway that regulates GLR-1 AMPAR trafficking and GLR-1-mediated behavior . The canonical hypoxia pathway senses oxygen via the prolyl hydroxylase EGL-9 , which uses dioxygen to hydroxylate a proline residue on the transcription factor HIF-1 , resulting in the ubiquitin-mediated degradation of HIF-1 ( Epstein et al . , 2001 ) . The non-canonical pathway senses oxygen via a specific isoform of EGL-9 , called EGL-9E , which , when activated by oxygen , binds to the scaffolding molecule LIN-10 , recruiting it to endosomes where it promotes the recycling of GLR-1 AMPARs to the synapse ( Park et al . , 2012 ) . We find that if animals are under normal oxygen conditions and are young , then the p38 MAPKK SEK-1 and the p38 MAPK PMK-1 promote EGL-9 activity ( Figure 10A–D ) . Active EGL-9 in turn triggers HIF-1 turnover , thereby preventing a HIF-1 transcriptional response . In addition , p38 MAPK signaling also promotes the association of EGL-9E with LIN-10 , in turn mediating steady GLR-1 recycling . By contrast , if animals are under conditions of hypoxia , then EGL-9 activity is depressed , resulting in HIF-1 stabilization and the activation of the HIF-1 transcriptional response . In addition , depressed EGL-9 activity exposes LIN-10 to phosphorylation by the CDK-5 kinase , resulting in LIN-10 delocalization in neurons and depressed GLR-1 recycling ( Figure 10E–H ) . This regulation results in long term changes in synaptic efficacy in the command circuit , which switches locomotion behavior from local foraging to long-distance roaming . This behavioral switch in locomotion allows the animal to escape the hypoxic environment . As animals grow older , p38 MAPK activity decreases despite the presence of ample oxygen , resulting in impaired GLR-1 recycling through the action of CDK-5 . Our findings suggest that p38 MAPK signaling is an important part of the hypoxia response pathway . 10 . 7554/eLife . 12010 . 013Figure 10 . Hypothetical model for p38 MAPK regulation of the hypoxia response pathway . A hypothetical , step-by-step model of hypoxia response pathway interactions in C . elegans neurons is shown for conditions of either ( A-D ) normoxia or ( E-H ) hypoxia . ( A ) Under normoxia , oxygen binds to and activates EGL-9 ( pink ovals ) . Oxygen also activates SEK-1 and PMK-1 ( p38 MAPK , green and yellow ovals , respectively ) through a mechanism that remains unknown . Activated p38 MAPK in turn phosphorylates one or more proteins ( possible EGL-9 itself , as speculated in this cartoon with a ‘P’ in a yellow circle ) that activate EGL-9 and trigger its recruitment to endosomes . Meanwhile , GLR-1 receptors ( red channels ) undergo continual endocytosis at the synapse . ( B ) Isoform EGL-9E , now bound to oxygen and possibly phosphorylated by p38 MAPK , becomes localized to endosomes , where it binds the PDZ-PTB domain protein LIN-10 ( orange oval ) and recruits it to endosomes by preventing its phosphorylation by the CDK-5 kinase ( purple oval ) . ( C ) Once at endosomes , LIN-10 promotes the recycling of endocytosed GLR-1 AMPARs ( red channels in the endosome ) back to the synapse . ( D ) The final outcome is that GLR-1 synaptic levels are maintained . ( E ) Under hypoxia , lack of oxygen results in lower SEK-1/PMK-1 p38 MAPK activity and inactive EGL-9 . ( F ) In the absence of oxygen , EGL-9E does not bind to LIN-10 . This exposes the LIN-10 N-terminus ( the localization domain of LIN-10 ) to CDK-5 , which phosphorylates it , thereby inhibiting LIN-10 recruitment to endosomes . ( G ) Without endosomal LIN-10 , GLR-1 AMPARs continue undergoing endocytosis from synapses but are not recycled , resulting in their accumulation in elongated endosomal compartments . ( H ) The final outcome is that GLR-1 synaptic levels become depleted . DOI: http://dx . doi . org/10 . 7554/eLife . 12010 . 013 C . elegans encounters hypoxic and anoxic environments in the soil , which contains bacteria and rotting material ( Anderson and Dusenbery , 1977; Van Voorhies and Ward , 2000 ) . Gas exchange occurs through the cuticle , and oxygen is sensed in the fluid of the pseudocoelomic body cavity by a set of sensory neurons expressing soluble guanylate cyclases; these neurons mediate a rapid aerotaxis response to changes in oxygen levels ( Cheung et al . , 2005; Gray et al . , 2004 ) . By contrast , the PMK-1/EGL-9/HIF-1 pathway is broadly expressed , activates a slower response to hypoxia that is tailored for cellular stress , modulates the aerotaxis circuit describe above , and modulates the locomotory reversal circuit that nematodes use to escape hypoxic environments ( Chang and Bargmann , 2008; Pocock and Hobert , 2008; Branicky and Schafer , 2008; Park et al . , 2012 ) . The two pathways are integrating environmental information on oxygen availability over different time scales: the former over a period of seconds to minutes whereas the latter is over a period of minutes to hours . There are precedents for gas messenger-dependent oxygen sensing mechanisms that act through protein kinases . In the mammalian carotid body , oxygen stimulates heme oxygenase-2 to generate carbon monoxide ( CO ) , which activates soluble guanylate kinase and protein kinase G ( PKG ) ( Yuan et al . , 2015; Peng et al . , 2014; Maines , 2004 ) . PKG inactivates cystathionine-γ-lyase ( CSE ) through direct phosphorylation , resulting in reduced hydrogen sulfide levels , reduced carotid body neural activity , and regular breathing ( Yuan et al . , 2015 ) . Under hypoxia , a drop in CO levels results in inactive PKG , CSE activation , increased hydrogen sulfide ( H2S ) levels , increased carotid body neural activity , and accelerated breathing . Changes in oxygen levels are known to regulate p38 MAPK itself . In mammals , hypoxic damage ( e . g . , triggered by ischemia during myocardial infarction or stroke ) can induce the activation of p38 MAPK , which is activated by stress and inflammation ( Cook et al . , 1999; Kawasaki et al . , 1997 ) . The activation of p38 MAPK contributes to ischemic injury , necrosis , and apoptosis , resulting in heart failure in the case of myocardial infarction and neurodegeneration in the case of ischemic stroke ( Marber et al . , 2011; Kumphune et al . , 2012; Barone et al . , 2001; Lai et al . , 2014 ) . The specific mechanism by which activated p38 MAPK contributes to ischemia-induced damage is unclear , but is thought to stem in part from its regulation of growth factor and apoptosis signal transduction pathways . Several groups have observed changes in HIF activity triggered by ERK and p38 MAPK ( possibly activated by ROS generated from mitochondria under hypoxia ) , yet it remains controversial whether the mechanism is direct phosphorylation of HIF-1 by p38 MAPK ( Conrad et al . , 1999; Bardos and Ashcroft , 2004; Minet et al . , 2001 ) . Our findings here would suggest a novel mechanism by which p38 MAPK signaling regulates HIF-1: the regulation of the PHD enzymes that act upstream of HIF-1 . This mechanism could provide an important link between growth factor signaling and the hypoxia response pathway in maintaining stem cell populations and promoting tumor growth and metastasis ( Wang et al . , 2013 ) . If p38 MAPK regulates HIF-1 activity , then what is the specific physiological role of this regulation ? To what is p38 MAPK responding ? While PHD proteins are well-established oxygen sensors in the hypoxia response pathway , there is also a likely role for the mitochondrial electron transport chain ( ETC ) in sensing oxygen during hypoxia . Even under conditions of normal oxygen , the ETC produces low levels of ROS ( Turrens , 2003 ) . Perhaps ROS activates a baseline level of p38 MAPK under normoxic conditions , acting as part of an additional oxygen sensing mechanism ( Figure 10A ) . Indeed , we did observe that normal oxygen levels promoted PMK-1 nuclear localization in the command interneurons , and hypoxia resulted in PMK-1 depletion from the nucleus , suggesting that oxygen ( or its byproducts ) could be activating PMK-1 ( Figure 3G , H , I ) . An alternative explanation for why we observe activation of the hypoxia response pathway in pmk-1 mutants is that the activation is due to an indirect effect of losing the baseline expression of the oxidative stress response in these mutants . In this scenario , pmk-1 mutants accumulate products of oxidative stress ( e . g . , ROS and oxidized macromolecules ) , which would mimic hypoxia , perhaps through the actions of ROS directly inactivating EGL-9 . We feel that this scenario is unlikely because skn-1 mutants , which are arguably more impaired for the oxidative stress response than are pmk-1 mutants , do not show induction of hypoxia response target genes like nhr-57 ( Oliveira et al . , 2009 ) . Nor did skn-1 mutants show the same GLR-1 localization defects that we observed in pmk-1 and egl-9 mutants . Instead , we favor a model in which PMK-1 directly silences the hypoxia response during normoxia , and that it does so independent of its role in the oxidative stress response . It remains a possibility that the additional ROS that is generated during long-term hypoxia and/or reoxygenation might further activate PMK-1 , providing a negative feedback that restores the hypoxia response back to a normoxia baseline , thereby minimizing the dangers of ROS production that occur during extended hypoxia and subsequent reoxygenation . Too much PMK-1 activation ( e . g . , during extreme anoxia ) might contribute to toxicity; indeed , mutations in pmk-1 increase the survival of animals exposed to long-term anoxia ( Hayakawa et al . , 2011 ) Consistent with our model , C . elegans PMK-1 is activated by oxidative stress in addition to being activated by pathogenic infection ( Berman et al . , 2001; Kim et al . , 2002; Inoue et al . , 2005 ) . Bacterial pathogens and anoxia exposure can both activate PMK-1 through the Toll/IL-1 resistance ( TIR ) domain protein TIR-1 , the ASK1 ortholog MAPKKK NSY-1 , and the MKK3 MAPKK SEK-1 ( Liberati et al . , 2004; Papp et al . , 2012; Kim et al . , 2002; Hayakawa et al . , 2011 ) . Whereas SEK-1 and PMK-1 are activated by oxidative stress , NSY-1 and TIR-1 do not appear to be critical components through which C . elegans respond to oxidative stress and presumably ROS ( Inoue et al . , 2005 ) . Our results clearly show that NSY-1 is not required for oxygen levels to modulate the hypoxia response , and given that NSY-1 is the sole C . elegans ASK1 ortholog , it seems likely that SEK-1 and PMK-1 are activated by a different MAPKKK under these conditions . Identifying the specific MAPKKK will be an important next step in determining how the p38 MAPK PMK-1 pathway senses hypoxia ( perhaps through an alternative ROS sensor ) and modulates the hypoxia response . In addition to its acute role in promoting survival during oxygen deprivation stress , the hypoxia response pathway also has a complex role in regulating aging and lifespan beyond simply maintaining stem cell populations ( Katschinski , 2006 ) . In C . elegans , hypoxia and limited stabilization of HIF-1 promote longevity ( Leiser and Kaeberlein , 2010 ) . However , loss of EGL-9 , which results in extreme HIF-1 stabilization , does not promote longevity and can be actually detrimental to lifespan ( Chen et al . , 2009; Bellier et al . , 2009 ) . Moreover , loss of HIF-1 can also promote longevity under conditions of elevated temperature via a separate mechanism ( Leiser et al . , 2011 ) . As nematodes grow older , the levels of active PMK-1 decrease ( Youngman et al . , 2011 ) , which might result in elevated HIF-1 activity . Whereas mutants for pmk-1 have a similar lifespan to that of wild type ( Troemel et al . , 2006; Alper et al . , 2010 ) , it is worth noting that the observed decrease in PMK-1 levels over time could impair nervous system function in a manner that would be missed by simple life span analysis . We find that GLR-1 accumulates in endosomes as nematodes grow older , similar to what occurs in young pmk-1 mutants , and that either the simple overexpression of PMK-1 or the removal of CDK-5 , an inhibitor of its downstream target LIN-10 , restores both GLR-1 localization and function to levels observed in younger animals . Our findings highlight the idea that changes in kinase signaling could explain aspects of age-associated physiological decline .
Transgenes generated in this study include ( 1 ) a wild-type pmk-1 cDNA fused to the glr-1 promoter , ( 2 ) a wild-type sek-1 cDNA fused to the glr-1 promoter , and ( 3 ) a wild-type pmk-1 cDNA fused in frame to GFP and placed behind the glr-1 promoter . Transgenic plasmids were generated using standard techniques . Transgenic strains were isolated after microinjecting plasmids ( 10 ng/μl ) with the transgenic marker ttx-3::rfp ( a gift from O . Hobert , Columbia Univ . ) into the germline to form extrachromosomal arrays . All other transgenes used in this study were as described in the publications cited in the text . Animals were grown at 20oC on standard NGM plates seeded with OP50 E . coli . For hypoxia , animals were incubated in a hypoxia chamber ( C-174 chamber , Biospherix ) for 24 hr at 20°C and recovered in ambient oxygen for 12 hr at 20°C . The oxygen level was automatically maintained with an oxygen controller ( ProOx P110 , Biospherix ) supplied with compressed nitrogen gas . GFP- and RFP-tagged fluorescent proteins were visualized in nematodes by mounting larvae on 2% agarose pads with levamisole . Fluorescent images were observed using a Zeiss Axioplan II . A 100X ( N . A . = 1 . 4 ) PlanApo objective was used to detect GFP and RFP signals . Imaging was done with an ORCA charge-coupled device ( CCD ) camera ( Hamamatsu , Bridgewater , NJ ) using IPLab software ( Scanalytics , Inc , Fairfax , VA ) or iVision v4 . 0 . 11 ( Biovision Technologies , Exton , PA ) software . Exposure times were chosen to fill the 12-bit dynamic range without saturation . Maximum intensity projections of z-series stacks were obtained and out-of-focus light was removed with a constrained iterative deconvolution algorithm ( iVision ) . For images , we captured the anterior ventral cord dendrites in the anterior region containing the RIG and AVG cell bodies . The quantification of ventral nerve cord fluorescent objects ( i . e . , puncta and elongated compartments ) was done using ImageJ ( Collins , 2007 ) to automatically threshold the images and then determine the outlines of fluorescent objects in ventral cord dendrites . ImageJ was used to quantify both the shape and the size of all individual fluorescent objects along the ventral cord . This allowed us to distinguish between the small GLR-1::GFP puncta in wild-type animals and the large , aberrant compartments ( which have an elongated shape rarely observed in wild type ) in hypoxic animals , as well as in thevariousindicated mutants . Object size was measured as the maximum diameter for each outlined puncta . Object number was calculated by counting the average number of puncta per 100 microns of dendrite length . The amount of a given fluorescent protein per puncta was calculated by summing all of the pixel values contained within each individual punctum to yield an integrated optical density ( IOD ) score for each punctum . Colocalization between GLR-1::GFP and mRFP::SYX-7 was performed as previously described ( Park et al . , 2009 ) . Single optical images for neuronal cell bodies expressing both reporters were collected using a confocal microscope equipped with the BD CARV II Confocal Imager and a Carl Zeiss 100× Plan-Apochroma objective ( N . A . = 1 . 4 ) . For quantitative colocalization analysis , all image manipulations were performed with iVision v4 . 0 . 11 ( Biovision Technologies , Exton , PA ) software using the FCV colocalization function . We applied an empirically derived threshold to all images for both the GLR-1::GFP channel and the mRFP::SYX-7 channel to eliminate background coverslip fluorescence and other noise ( typically 5% of pixels for each channel ) . The fluorescent intensity values for both the GLR-1::GFP and mRFP::SYX-7 channels were then scatter plotted for each pixel . Pixels with similar intensity values for both channels ( within an empirically-established tolerance factor ) were counted as colocalized . To acquire the fraction of GLR-1::GFP colocalized with mRFP::SYX-7 , the number of colocalized pixels was normalized to the number of GLR-1::GFP pixels under threshold . To maximize our resolving power while observing the relatively small C . elegans neuron cell bodies , we restricted our analysis to a single confocal optical section taken through the middle of each cell body . The reversal frequency of individual animals was assayed as previously described , but with some modifications ( Zheng et al . , 1999 ) . Single young adult hermaphrodites were placed on NGM plates in the absence of food . The animals were allowed to adjust to the plates for 5 min , and the number of spontaneous reversals for each animal was counted over a 5-min period . Twenty or more animals were tested for each genotype , and the reported scores reflect the mean number of reversals per minute , normalized as a percentage of the value of wild-type controls . To measure phospho-PMK-1 protein levels , 35 young adults of each genotype were dissolved in 1× Laemmli buffer by flash freezing and boiling for 10 min . Lysates were analyzed on 10% SDS-polyacrylamide gels . Western blotting was performed using rabbit anti-phospho-p38 MAPK ( 1:2000 , Promega ) and mouse anti-Actin ( 1:2000 , MP biomedicals ) , with detection through chemiluminescence . Total RNAs were extracted with Trizol ( Invitrogen Co . , Carlsbad , CA ) . Young adult or L4 stage larvae ( 10–15 animals each ) were resuspended in 250 µl of Trizol and lysed by one round of freezing ( by liquid nitrogen ) and thawing ( 60°C ) with subsequent vigorous vortexing in 4°C for 30 min . PCR was performed in an Eco real-time qPCR system ( Illumina , San Diego , CA ) using iScriptTM One-Step RT-PCR Kit With SYBR Green ( Bio-Rad Laboratories Inc . , Hercules , CA ) in 20 µL reactions with 20 ng of RNA template . For glr-1 , we used as forward ( 5’-TGATACAATGAAAGTTGGAGCAAATC-3’ ) and reverse ( 5’-CATCGCATTGTCCTCTATCATACCAC-3’ ) primers . For pmk-1 , we used as forward ( 5’- CTGATGAGCCAATTGCAGAAG-3’ ) and reverse ( 5’-TTTTCTCCTCATCTTCCTCTTCG-3’ ) primers . For nhr-57 , we used as forward ( 5′-CGTGATTGCAGACTTGAAAGC-3′ ) and reverse ( 5′-GCGTTTGACTTCCATCGTTTG-3′ ) primers . For act-1 , we used as forward ( 5′-ACCATGTACCCAGGAATTGC-3′ ) and reverse ( 5′-TGGAAGGTGGAGAGGGAAG-3′ ) primers . Samples were measured two to three times and average values were used for the calculation of relative fold changes . The relative levels of glr-1 , pmk-1 , and nhr-57 mRNA were normalized to the levels of act-1 mRNA in each preparation . For each experiment , the value for wild type was set to 1 and other values were normalized accordingly . | The brain accounts for 2% of our body weight , but consumes about 20% of our oxygen intake . This oxygen gluttony is due to the tremendous appetite of brain cells for energy , which neurons satisfy through oxygen-dependent ( aerobic ) metabolism . As a result , the loss of oxygen to the brain during a stroke , heart attack , or due to another medical condition can be very damaging to cells in the brain . Human and other animal cells use a communication system called the hypoxia response pathway to sense oxygen and trigger a protective response when oxygen is low . This pathway includes an enzyme called prolyl hydroxylase , which senses oxygen and modifies another protein in the pathway that regulates the production of enzymes involved in metabolism . This alters the balance of enzymes involved in aerobic and oxygen-independent ( anaerobic ) metabolism in the cell . However , it is not clear how the activity of the prolyl hydroxylase is regulated . Much of our knowledge about the hypoxia response pathway has been gained from studies using a small worm called C . elegans . This worm uses the pathway to cope with hypoxia in the harsh environment of the soil . Mutant worms that lack the prolyl hydroxylase have several abnormalities including higher levels of anaerobic metabolism even in the presence of oxygen , and defects in the connections between neurons . Park and Rongo used C . elegans to study the pathway in more detail . The experiments show that another enzyme called p38 MAPK activates the prolyl hydroxylase . Mutant worms that lack this enzyme have similar abnormalities in the hypoxia response pathway as animals that lack the prolyl hydroxylase . In normal worms , decreasing levels of p38 MAPK as the animals grow older contribute to the decline in the nervous system . The p38 MAPK enzyme appears to work by regulating the activity of the prolyl hydroxylase and its location inside neurons . These findings provide a new target for the development of drugs that may help to protect us from tissue damage caused by hypoxia . Future challenges are to find out what activates p38 MAPK , and how it influences the location of prolyl hydroxylase in neurons . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2016 | The p38 MAP kinase pathway modulates the hypoxia response and glutamate receptor trafficking in aging neurons |
Small GTPases of the Rab family not only regulate target recognition in membrane traffic but also control other cellular functions such as cytoskeletal transport and autophagy . Here we show that Rab26 is specifically associated with clusters of synaptic vesicles in neurites . Overexpression of active but not of GDP-preferring Rab26 enhances vesicle clustering , which is particularly conspicuous for the EGFP-tagged variant , resulting in a massive accumulation of synaptic vesicles in neuronal somata without altering the distribution of other organelles . Both endogenous and induced clusters co-localize with autophagy-related proteins such as Atg16L1 , LC3B and Rab33B but not with other organelles . Furthermore , Atg16L1 appears to be a direct effector of Rab26 and binds Rab26 in its GTP-bound form , albeit only with low affinity . We propose that Rab26 selectively directs synaptic and secretory vesicles into preautophagosomal structures , suggesting the presence of a novel pathway for degradation of synaptic vesicles .
Synapses are highly dynamic structures exhibiting frequent turnover . The most dramatic phase of synaptic remodeling occurs during development when the majority of initially formed synapses are eliminated while the final synaptic network is being generated . However , even in the adult brain there is persistent turnover of synapses , mostly in response to experience and learning ( Caroni et al . , 2012; Chung and Barres , 2012 ) . Formation of a new synapse involves the establishment of highly specialized structures containing arrays of unique membrane and scaffold proteins , which necessitates close coordination between the presynaptic axon and the postsynaptic dendrite . Components of these structures are delivered by microtubule-based transport , although some of the proteins are locally synthesized in dendrites ( Steward and Levy , 1982; Holt and Schuman , 2013 ) . Similarly , delivery of synaptic membranes ( such as synaptic vesicles ) and active zone precursors to the synapse relies on kinesin-mediated transport ( Hirokawa et al . , 2010 ) . A lot has been learned in recent years about the signaling events and the downstream effectors involved in synaptogenesis as well as the mechanisms by which individual components are recruited and maintained ( Caroni et al . , 2012 ) . Less is known , however , about the molecular cascades involved in synaptic elimination . In principle , two mechanisms of elimination co-exist that are not mutually exclusive and may indeed be coordinated with each other: ‘attack from the outside’ or ‘self-destruction from the inside’ . Elimination from the outside is usually executed by microglial cells but the underlying signaling network is complex , and astrocytes have also recently been appreciated to play a major role in this process ( Chung and Barres , 2012; Stephan et al . , 2012; Clarke and Barres , 2013; Maor-Nof and Yaron , 2013 ) . In cell autonomous elimination , synaptic components may either be ( i ) recycled , that is , being removed in a functionally intact form for the use at another site , or ( ii ) degraded . With the exception of some recent evidence showing that synaptic vesicles can be exchanged between neighboring synapses , whether synaptic components can be reused after having been operational in a functional synapse remains largely unexplored ( Darcy et al . , 2006; Kamin et al . , 2010 ) . Conversely , an increasing body of evidence supports the view that synaptic components are rapidly degraded once a synapse is earmarked for elimination . Unsurprisingly , the ubiquitin-proteasome system is emerging as one of the central players , both at presynaptic ( Yao et al . , 2007; Jiang et al . , 2010 ) and postsynaptic sites ( [Bedford et al . , 2001; Patrick et al . , 2003; Lee et al . , 2004] , for a review see Mabb and Ehlers ( 2010 ) ) . At the presynaptic site , the ubiquitin system is not only involved in synaptic elimination , but also in the general regulation of synaptic plasticity ( Muralidhar and Thomas , 1993; Campbell and Holt , 2001; DiAntonio et al . , 2001; Murphey et al . , 2003; Yao et al . , 2007; Yi and Ehlers , 2007; Lee et al . , 2008 ) . For instance the protein RIM , a crucial hub for organizing active zones that form the release site for synaptic vesicles , was recently shown to be rapidly turned over upon ubiquitination , resulting in loss of synaptic function ( Yao et al . , 2007 ) . Furthermore , an increasing number of ubiquitin-modifying enzymes have been described from synapses ( particularly E3-ligases ) ( Ding and Shen , 2008 ) . In contrast to the emerging role of the ubiquitin proteasome system , little information is currently available regarding the mechanisms by which synaptic membrane proteins are eliminated . At the postsynaptic site , ubiquitin-dependent pathways are clearly involved in the regulation of surface receptor density ( Patrick et al . , 2003; Kato et al . , 2005; Schwarz et al . , 2010 ) . However , only scant information is available about the turnover of membrane proteins at the presynaptic site where a complicated and autonomous vesicle recycling machinery needs to deal with many 100s of synaptic vesicles . Surprisingly , the mechanisms by which synaptic vesicles are eliminated have thus far received little attention . By analogy to non-neuronal cells , it is frequently assumed that synaptic vesicle membrane proteins follow the canonical endosomal-lysosomal route for degradation which involves ubiquitination and recognition by the ESCRT machinery after being delivered to endosomes , followed by the formation of multivesicular bodies , retrograde transport , and ultimately fusion with lysosomes ( Katzmann et al . , 2002; Raiborg and Stenmark , 2009 ) . However , aside from a few hints from a recent proteomic study , whether synaptic vesicle proteins are ubiquitinated remains unclear ( Na et al . , 2012 ) . Similarly , whether sequestration into the lumen of multivesicular bodies is involved and , if so , to what extent is unknown . Indeed , multivesicular bodies are infrequently observed in axons and typically appear in response to pathological ( dystrophic or toxic ) conditions ( for review see [Von Bartheld and Altick , 2011] ) . Furthermore , no information is currently available concerning the involvement of the ESCRT pathway in the elimination of presynaptic components . Intriguingly , recent studies implicate the involvement of clathrin-dependent pathways in targeting plasma membrane components to autophagosomes , hinting at the potential involvement of this mechanism in the turnover of synaptic vesicles recovered by endocytosis following the release of their neurotransmitter content ( Ravikumar et al . , 2010 ) . In this study we report about data suggesting the presence of a novel pathway for the degradation of synaptic and secretory vesicles , which involves selective sequestration of vesicle clusters into structures resembling early autophagosomes . This pathway is triggered by Rab26 , a little-studied member of the Rab-GTPase superfamily that is related to the exocytotic Rab3/Rab27 subgroup . We show that Rab26 selectively localizes to presynaptic membrane vesicles and recruits both Atg16L1 and Rab33B , two components of the pre-autophagosomal machinery . Remarkably , these autophagosomal structures are filled almost exclusively with synaptic vesicles and proteins typically associated with large dense-core vesicles . Overexpression of EGFP-tagged Rab26 , but not of FLAG-tagged or wild-type ( WT ) Rab26 , induces the formation of giant autophagosomes in the cell bodies of hippocampal neurons—a phenotype that is mirrored upon transfection in HeLa cells . Based on these findings , we conclude that Rab26 may selectively channel synaptic vesicles into pre-autophagosomes and , thus , may represent a new regulator of synapse turnover .
Previously we reported that synaptic vesicles highly purified from rat brain contain more than 30 different Rab-GTPases ( Takamori et al . , 2006 ) . Of these , a subgroup of Rabs including Rab3a , Rab3b , Rab3c , and Rab27b were highly enriched in the vesicle fraction ( Pavlos et al . , 2010 ) . These Rabs are known to function in the regulation of exocytosis and constitute part of the ‘secretory Rab subfamily’ ( Pereira-Leal and Seabra , 2001; Fukuda , 2008; Pavlos and Jahn , 2011 ) . Rab26 , a comparatively uncharacterized member of the Rab superfamily , is also closely related to this subgroup . Since we detected Rab26 on purified synaptic vesicles in two previous independent proteomic studies ( Takamori et al . , 2006; Pavlos et al . , 2010 ) , we decided to further explore its endogenous localization in neurons . To this end , we raised a mouse monoclonal antibody that is specific for Rab26 and does not cross-react with other related Rab proteins including Rab27 ( Figure 1—figure supplement 1 ) . First , we used immunoblotting to monitor the subcellular distribution of Rab26 during the purification of synaptic vesicles from the rat brain . As shown in Figure 1A , Rab26 co-purified with synaptic vesicle markers ( as indicated here using synaptophysin ) , with the highest enrichment being observed in the synaptic vesicle ( SV ) fraction obtained after purification using consecutive density gradients and size exclusion chromatography . This fraction has been previously shown to be comprised almost exclusively of synaptic vesicles ( at least 95% purity ) ( Huttner et al . , 1983 ) . For independent confirmation , we carried out immunoisolation of synaptic vesicles using beads ( Eupergit C1Z ) covalently coupled with monoclonal antibodies specific for Rab26 or synaptophysin . As shown in Figure 1B , both antibodies resulted in the isolation of membranes highly enriched in both synaptophysin and Rab26 . As a control , the membranes were solubilized with the detergent Triton X-100 prior to immunoisolation ( Tx-IP ) . In this case , only the respective antigens were isolated ( Figure 1B ) , thus validating the specificity of the isolation procedure . We also verified the nature of the immunoisolated vesicles by transmission electron microscopy ( TEM ) . As previously reported , synaptophysin-beads were densely covered by small vesicular profiles with a size distribution typical for synaptic vesicles ( i . e . , 40–45 nm ) ( Figure 1C ) ( Burger et al . , 1989; Takamori et al . , 2000 ) . Rab26 beads were similarly populated with these vesicles , albeit to a lesser extent ( Figure 1D ) . Nevertheless , quantitative assessment of the size distribution of the bead-bound vesicles revealed no distinguishable difference between the vesicles bound to synaptophysin and Rab26 beads , respectively ( Figure 1E ) . 10 . 7554/eLife . 05597 . 003Figure 1 . The small GTPase Rab26 co-purifies with synaptic vesicles . ( A ) Rab26 co-purifies with synaptic vesicles using conventional fractionation . Synaptic vesicles were purified from rat brain homogenate ( H ) by two consecutive differential centrifugation steps , yielding a low-speed pellet ( P1 ) and a supernatant ( S1 ) , followed by a second centrifugation yielding a pellet P2 ( containing synaptosomes and mitochondria ) and a supernatant ( S2 ) . P2 was then lysed by osmotic shock , followed by centrifugation to separate large particles including synaptic junctional complexes ( LP1 ) and a supernatant from which small membranes enriched in synaptic vesicles are collected by high-speed centrifugation ( LP2 , supernatant LS2 only contains soluble proteins ) . LP2 was further fractionated by sucrose density gradient centrifugation followed by chromatography on controlled pore glass beads where larger membrane fragments ( PK1 ) were separated from synaptic vesicles ( SV ) ( Huttner et al . , 1983 ) . 12 µg of proteins of each fraction was analyzed by SDS-PAGE and immunoblotting for Rab26 and the synaptic vesicle marker synaptophysin . Note that Rab26 copurifies with synaptophysin , displaying a pattern typical of synaptic vesicle proteins . ( B ) Synaptic vesicles were immunoisolated using Eupergit C1Z/methacrylate beads to which monoclonal antibodies specific for Rab26 or synaptophysin ( Syph ) respectively , were covalently coupled . The beads were incubated with a resuspended LP2 fraction and collected ( see [Burger et al . , 1989; Takamori et al . , 2000] for details ) . SN , supernatant obtained after bead incubation; Bound , immunoisolates; Tx-SN and Tx-IP , same as before but the Input sample was solubilized with Triton X-100 to a final concentration of 1% before immunoprecipitation . For immunoblotting , 4% of the input or supernatant samples and 33% of bound samples were loaded for analyses . Note that Rab26 and synaptophysin cofractionate with the immunobeads irrespective of the antibody employed . Incubation with synaptophysin beads quantitatively depleted Rab26 from the supernatant whereas depletion of synaptophysin by Rab26-beads was less complete . In contrast , only the respective antigen was recovered from the detergent-solubilized samples . Asterisks denote IgG heavy and the light chains of the antibodies used for isolation , respectively . ( C and D ) Electron microscopy showing ultrathin sections of methacrylate beads containing bound organelles that were captured with synaptophysin- ( C ) or Rab26-specific ( D ) antibodies , respectively . ( E ) Size distribution ( diameter ) of bead-associated vesicles . Note that both populations exhibit a very similar and homogeneous size distribution , with a peak between 40–45 nm as is characteristic for synaptic vesicles ( Takamori et al . , 2006 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 00310 . 7554/eLife . 05597 . 004Figure 1—figure supplement 1 . Monoclonal Rab26 antibody generated in this study specifically recognizes Rab26 . 10 mg of lysates from NIH 3T3 cells transiently transfected with GFP-tagged versions of Rab3d , Rab27a , Rab27b or Rab26 were immunoblotted and probed with anti-GFP ( upper panel ) or the monoclonal anti-Rab26 ( lower panel ) antibodies . The newly generated monoclonal antibody recognizes only Rab26 but does not react with the other closely related Rab proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 004 Like other GTPases , Rabs function as molecular switches which undergo conformational changes during GDP/GTP cycling ( Grosshans et al . , 2006; Stenmark , 2009 ) . This cycle is frequently paralleled by a membrane ‘on/off’ association—dissociation cycle . Membrane association is achieved following posttranslational modification of Rabs by geranyl-geranylation , a prerequisite for membrane insertion and Rab activation . Conversely , membrane dissociation is regulated by a specific adaptor protein , termed GDP dissociation inhibitor ( GDI ) , which sequesters GDP-bound Rabs from membranes to the cytosol following GTP-hydrolysis ( Araki et al . , 1990; Ullrich et al . , 1993; Goody et al . , 2005; Wu et al . , 2010 ) . Therefore , to gain initial insights into the functional membrane association/dissociation cycle of Rab26 , we assessed whether it could be extracted from synaptic vesicle membranes following GDI treatment . For this , we incubated a fraction enriched in synaptic vesicles ( LP2 ) with purified recombinant GDI in the presence of GDP or GTPγS . The latter analogue was used to overcome the intrinsically high GTPase activity of Rabs ( see ‘Discussion’ ) . Consistent with previous observations , Rab3 is rapidly retrieved from synaptic vesicles membranes by GDI in the presence of GDP ( Araki et al . , 1990; Chou and Jahn , 2000; Pavlos et al . , 2010 ) . By comparison , Rab26 is resistant to GDI-mediated membrane extraction , even when GDP is present in excess ( Figure 2A ) . This feature is reminiscent of the biochemical characteristics of Rab27b , which also fails to be retrieved from synaptic vesicles by GDI treatment in vitro ( Pavlos et al . , 2010 ) . Rather , Rab27b is known to dimerize and persist on synaptic vesicle membranes in its GDP-bound form ( Chavas et al . , 2007; Pavlos et al . , 2010 ) . Therefore , to assess whether GDP-bound Rab26 exhibits a similar tendency to oligomerize , we additionally performed co-immunoprecipitation experiments between FLAG- and EGFP-tagged wild-type ( WT ) Rab26 and its variants containing mutations that selectively interfere with the Rab26 GDP/GTP switch domain ( s ) . As shown in Figure 2B , co-precipitation of FLAG-tagged Rab26 was only observed when cells were transfected with either wild-type EGFP-Rab26 or with a GDP-preferring variant ( Rab26T77N , henceforth referred to as Rab26TN ) . By comparison , little to no co-precipitation was observable when the ‘nucleotide empty’ Rab26 ( Rab26N177I , Rab26NI ) or ‘GTP-locked’ ( Rab26Q123L , Rab26QL ) variants were employed . Together , these data indicate that Rab26 is a synaptic vesicle protein that oligomerizes preferentially in its GDP-bound form , thereby precluding GDI-mediated membrane extraction—a feature shared with its synaptic vesicle relative Rab27b . 10 . 7554/eLife . 05597 . 005Figure 2 . GDP-Rab26 cannot be extracted by GDI from membranes and forms oligomers . ( A ) Rab26 is resistant to extraction by GDI from synaptic vesicle membranes . An enriched synaptic vesicle fraction ( LP2 ) was incubated with GTPγS or GDP ( 500 µM ) for 15 min at 37°C . His-GDI ( 5 µM ) or PBS ( control , first lane ) was added and the samples were incubated for an additional 45 min at 37°C . The membranes were then separated from the soluble fraction by centrifugation and analyzed by immunoblotting . While Rab3a was efficiently depleted from synaptic vesicles , Rab26 persisted on membranes . IB , immunoblotting . ( B ) Rab26 oligomerizes in a GDP-dependent manner . HEK293 cells transiently co-expressing EGFP-Rab26 variants ( WT , QL , TN or NI ) with FLAG-Rab26WT were lysed in detergent containing buffer followed by immunoprecipitation of EGFP-Rabs . Co-precipitation of FLAG-Rab26WT was observed with EGFP-Rab26 WT and even more efficiently with the GDP-preferring variant Rab26TN whereas co-precipitation with the nucleotide-empty variant ( Rab26NI ) was reduced and binding to the GTP-preferring variant ( Rab26QL ) was abolished . IP , immunoprecipitation . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 005 Next , to study its subcellular localization in more detail , we immunostained primary cultures of rat hippocampal neurons for Rab26 . First , the distribution of endogenous Rab26 was compared with that of synaptotagmin-I , one of the major membrane constituents of synaptic vesicles . As shown in Figure 3A , Rab26 labeling resulted in a conspicuous punctate staining pattern that overlapped with , although was not identical to , the pattern obtained with synaptotagmin-I antibodies . Higher magnification of neurites revealed that most of the Rab26 positive puncta colocalized with synaptotagmin-I . In contrast , many puncta positive for synaptotagmin-I were not stained with the Rab26 antibody ( Figure 3B , arrows show colocalization ) . 10 . 7554/eLife . 05597 . 006Figure 3 . Endogenous and expressed GTP-Rab26 variants localize to a subset of synaptic vesicles in cultured hippocampal neurons . In ( A–G ) , representative line scans of the two channels are shown below each set . In the y-axis , F ( a . u . ) indicates fluorescence intensity ( arbitrary units ) . ( A and B ) Localization of endogenous Rab26 ( detected with the newly generated monoclonal anti Rab26 antibody ) and synaptotagmin-I ( Syt-I ) in neurites of dissociated hippocampal neurons ( DIV 15 ) reveals that Rab26 colocalizes with a subset of Syt-I positive puncta ( B , arrows ) . ( C–E ) Expression of FLAG-tagged Rab26 variants in neurites ( DIV 9 cultures 48hr after transfection ) . Both FLAG-Rab26WT ( C ) and QL ( D ) co-localize with a subset of synaptotagmin positive puncta ( Syt-I ) , whereas FLAG-Rab26TN ( E ) exhibited a more diffuse distribution . ( F ) Overexpression of EGFP-Rab26WT exhibits a distribution comparable to endogenous and FLAG-tagged Rab26 in neuritis where it colocalizes with synaptophysin ( Syph ) . ( G ) Rab26-positive clusters contain actively recycling vesicles . Hippocampal cultures were incubated overnight with a labeled antibody directed against a lumenal epitope of synaptotagmin ( Syt-I clone 604 . 2 conjugated to Oyster-550 ) , resulting in uptake during endocytosis . EGFP-Rab26WT is present in recycled synaptic vesicles as evident from the co-localization with synaptotagmin-I positive puncta . ( H ) Localization of YFP-tagged Rab26 variants at the Drosophila neuromuscular junction ( third instar larvae ) . The Rab26 variants ( Rab26wt , GTP-preferring Rab26Q250L , and GDP-preferring Rab26T204N ) , were expressed using elav-Gal4 . Neuromuscular junctions of third instar larvae expressing these Rab26 variants were stained with anti-GFP ( green ) and for endogenous cysteine string protein as vesicle marker ( anti-Csp , red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 00610 . 7554/eLife . 05597 . 007Figure 3—figure supplement 1 . EGFP-Rab26WT colocalizes with the secretory neuropeptide Y ( NPY ) but not with EEA1 in neurites . ( A ) Hippocampal neurons ( DIV 10 ) co-expressing EGFP-Rab26WT and RFP-NPY showed colocalisation of both proteins in neurites as indicated by the arrows . ( B ) No overlap was observed with the early endosomal marker EEA1 ( arrows ) . Line scans beneath each figure shows extent of colocalization . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 00710 . 7554/eLife . 05597 . 008Figure 3—figure supplement 2 . Rab26 does not colocalize with the presynaptic scaffold protein Brp in Drosophila neuromuscular junctions . YFP-tagged Rab26WT localizes to the presynaptic compartment of neuromuscular junctions ( NMJs ) in Drosophila melanogaster and forms huge structures that overlapped with the presynaptic membrane marker HRP , but not with the active zone marker Bruchpilot ( Brp ) . The YFP tagged wild type Rab26 was expressed using the elav-Gal4 promoter . NMJs of larvae expressing YFP-Rab26WT were stained with anti-GFP ( green ) , anti-HRP ( blue ) and anti-Brp ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 008 To shed more light on the intracellular distribution of Rab26 , neurons were transiently transfected with variants of FLAG-tagged Rab26 and then labeled for FLAG and synaptotagmin 1 ( Figure 3 ) . The staining pattern obtained with Rab26WT ( Figure 3C ) and with Rab26QL ( the GTP-preferring variant ) ( Figure 3D ) was very similar to that of endogenous Rab26 , showing a high degree of overlap with endogenous synaptotagmin 1 . In contrast , the GDP-preferring Rab26TN exhibited a more diffuse but still somewhat granular pattern that displayed no significant co-localization with synaptotagmin 1 ( Figure 3E ) . A similar staining pattern was also observed when EGFP-tagged wildtype Rab26 was used , with the fusion protein again colocalizing with synaptic vesicle markers ( in this case synaptophysin ) ( Figure 3F ) and with the neuropeptide RFP-NPY but not with the early endosomal marker/Rab5 effector EEA1 ( Figure 3—figure supplement 1A , B , respectively ) . This suggests that Rab26 is likely to traffic in a pathway distinct from Rab5 . Next , we tested whether the Rab26 positive puncta represent synapses undergoing exo-endocytosis . To monitor this , live hippocampal neuronal cultures expressing EGFP-Rab26 were pre-incubated overnight with a fluorescently-conjugated antibody specific for the luminal domain of synaptotagmin-I . This antibody can only bind to synaptotagmin when the luminal domain of the protein is exposed to the extracellular surface following synaptic vesicle exocytosis , and is thus used to conveniently identify active synapses as well as synaptic vesicles that have undergone exo-endocytosis ( Kraszewski et al . , 1995; Fernandez-Alfonso et al . , 2006 ) . As shown in Figure 3G , many of the EGFP-Rab26 puncta colocalized with vesicles labeled with this antibody , thereby confirming that these Rab26-positive vesicles originated from endocytosis of vesicles that previously had undergone at least one round of exocytosis . Rab26 is one of the Rab GTPases conserved between mammals and Drosophila , with the genome of latter encoding three alternatively spliced isoforms . According to a systematic analysis of all Drosophila Rab proteins , Rab26 is expressed specifically in neurons at all developmental stages ( larval and pupal development , adults flies ) ( Chan et al . , 2011 ) . To test whether the distribution of Rab26 in Drosophila resembles that in cultured hippocampal neurons , we expressed YFP-tagged versions of wild-type ( WT ) , GTP-preferring ( RabQ250L ) and GDP-preferring ( Rab26T204N ) Rab26 using the pan neuronal elav-Gal4 driver ( Zhang et al . , 2007 ) . Although the tagged Rab26 protein variants were expressed throughout development , no lethality or delayed development was observed ( data not shown ) . Consistent with previous observations , analysis of third instar larvae nerve-muscle preparations revealed an exclusive localization of Rab26 to presynaptic compartments of the neuromuscular junction without staining of axons and cell bodies ( Figure 3—figure supplement 2; see also [Chan et al . , 2011] ) . Remarkably , expression of wild-type and of the gain-of-function ( QL ) Rab26 resulted in the appearance of large vesicle clusters within the presynaptic boutons whereas the GDP-preferring form ( TN ) was diffusely localized . This indicated that , like in hippocampal neurons , formation of these clusters is dependent on the nucleotide-bound state of Rab26 ( Figure 3H ) . These Rab26-positive compartments are present in neuromuscular junction boutons as indicated by staining with anti-horseradish peroxidase ( HRP ) ( Figure 3—figure supplement 2 ) and showed a partial overlap with the synaptic vesicle protein cysteine string protein ( Csp ) ( Figure 3H ) ( Zinsmaier et al . , 1990 ) . Conspicuously , large Rab26-positive structures often showed intense Csp staining at their borders . Importantly , Rab26 is excluded from active zones immunostained for the Bruchpilot ( Brp ) , a scaffold protein specifically localized to active zones ( Figure 3—figure supplement 2 ) ( Wagh et al . , 2006 ) . Whereas the distribution of endogenous and tagged-expressing Rab26 variants was comparable in neurites , expression of EGFP-Rab26WT in cultured hippocampal neurons resulted in a unique and highly conspicuous phenotype that was not observed with endogenous Rab26 , untagged or FLAG-tagged Rab26 and that has not been previously reported for any other EGFP-tagged Rab GTPase . In the soma , EGFP-Rab26WT induced the formation of large vesicular structures that , in some instances , filled the major part of the neuronal cytoplasm ( Figure 4 ) . These structures were intensely positive for both synaptic vesicle ( Figure 4A , B ) and large dense core vesicle markers , that is , the two types of neuronal secretory vesicles that usually do not show significant overlap ( Figure 4C and Figure 4—figure supplement 1C ) . Intriguingly , the overlap with Rab3A , the major Rab-GTPase associated with synaptic vesicles , was less apparent . Double labeling with a variety of compartment-specific markers revealed no overlap with early endosomes or Golgi ( Figure 4—figure supplement 1A , B , respectively ) but some , albeit limited , overlap with lysosomes ( Figure 4D ) . 10 . 7554/eLife . 05597 . 009Figure 4 . Expression of EGFP-Rab26WT in cultured hippocampal neurons induces formation of large GFP-positive clusters in the neuronal cell bodies . ( A–C ) : GFP-positive clusters colocalize with markers for synaptic and large dense core vesicles . ( A and B ) EGFP-Rab26WT clusters contain synaptobrevin ( Sybv ) and Rab3a . ( C ) Co-expression of EGFP-Rab26WT with RFP-NPY results in almost complete overlap of both proteins in these clusters ( C ) . ( D ) Partial overlap is also observable with the lysosomal membrane protein LAMP2 . DIV 10 , scale bar , 5 µm . Line scans of select vesicle clusters ( denoted by solid white lines in merged panel ) signify the relative correlation between the individual fluorescent channels . F ( a . u . ) indicates fluorescence intensity ( arbitrary units ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 00910 . 7554/eLife . 05597 . 010Figure 4—figure supplement 1 . EGFP-Rab26WT colocalizes with the secretory protein secretogranin II but not with EEA1 and GM130 in neuronal somata . Hippocampal neurons ( DIV 10 ) co-expressing EGFP-Rab26WT show no overlap between Rab26 and early endosomes ( EEA-1 ) ( A ) or Golgi ( GM130 ) ( B ) whose endogenous staining patterns are not affected by overexpression of EGFP-Rab26WT . In contrast , massive clustering of Secretogranin , SgII , a marker for large dense core vesicles ( C ) is observed , with the clusters perfectly colocalizing with EGFP-Rab26WT . Line scans adjacent of each figure shows the extent of colocalization . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 010 To better understand the nature of these structures we carried out immunogold-TEM on ultrathin cryosections of transfected neurons . As shown in Figure 5A , the soma contained large clusters of numerous small vesicles that were positive for both EGFP-Rab26WT and synaptobrevin ( Sybv ) . In many cases , these clusters were rather homogenous , but occasionally also contained larger vesicles and mitochondria ( Figure 5B ) . Although no systematic quantification was performed , some of these clusters reached enormous dimensions containing possibly 1000s of small vesicles ( Figure 5—figure supplement 1 ) . In some cases , these clusters were surrounded by a single and/or double membrane ( Figure 5C , inset ) , although this was somewhat variable . We also assessed neuromuscular junctions of Drosophila strains overexpressing YFP-Rab26WT by TEM . Here , dense clusters of vesicles ( devoid of surrounding membranes ) were regularly observable that were clearly set apart from surrounding synaptic vesicles ( Figure 5D , indicated by an arrow ) but were clearly absent in controls . 10 . 7554/eLife . 05597 . 011Figure 5 . Ultrastructure of EGFP-Rab26 induced vesicle clusters in cultured hippocampal neurons and in neuromuscular junctions of Drosophila third instar larvae . ( A–C ) Ultrathin cryosections obtained from hippocampal neurons expressing EGFP-Rab26WT were immunogold labeled for EGFP and synaptobrevin ( Sybv , monoclonal antibody 69 . 1 specific for synaptobrevin 2 ) and analyzed by electron microscopy . In the soma of hippocampal neurons organelles surrounded by one or two ( C , inset , arrow ) membranes were densely packed with small vesicles and very occasionally other organelles ( e . g . , mitochondria , m ) . Immunogold labeling for both EGFP and synaptobrevin was concentrated both on vesicles present inside and the surrounding membrane . Scale bar in insert , 50 nm . ( D ) Ultrathin sections of neuromuscular junctions obtained from Drosophila third instar larvae . Control animals show scattered vesicles of a somewhat heterogeneous size , typical for this developmental stage ( Rasse et al . , 2005 ) . Synapses of a strain expressing YFP-Rab26WT using the elavG::UAS system ( elavG::UAS Rab26WT ) display frequent clusters of densely packed vesicles ( arrow ) that are separated from the surrounding vesicles but lack a surrounding membrane . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 01110 . 7554/eLife . 05597 . 012Figure 5—figure supplement 1 . EGFP-Rab26WT induces massive vesicle clustering in neuronal cell bodies . Ultrathin cryosections obtained from hippocampal neurons expressing EGFP-Rab26WT were immunogold labeled for EGFP and analyzed by electron microscopy . Transient expression of EGFP-Rab26WT results in the clustering of enormous numbers of vesicles positive for EGFP-Rab26 . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 012 As described above , in neurites Rab26 is associated with large clusters containing synaptic vesicle proteins regardless of whether endogenous Rab26 is visualized or whether Rab26 is overexpressed . Thus , it is conceivable that the induction of these vesicle clusters is an intrinsic property of GTP-Rab26 , which is enhanced by the weak homo-dimerization of EGFP and YFP ( Shaner et al . , 2005 ) . What could be the identity of these clusters ? Considering their striking resemblance of these clusters to autophagosomal precursors , we hypothesized that these clusters may represent autophagosomes at various stages of formation and/or maturation . Autophagy is a degradative pathway during which cellular contents are enclosed by a double-membrane ( i . e . , the isolation membrane ) and then delivered to lysosomes for disposal ( for review see e . g . , [Mizushima et al . , 2011; Lamb et al . , 2013] ) . The pathway is initiated by two ubiquitin-like conjugation systems that operate in a sequential manner . The first conjugates the ubiqutin-like protein Atg12 to Atg5 which is then recruited by Atg16L1 to the pre-autophagosome structure . This complex then recruits a LC3 family member , a second ubiquitin-like molecule , and attaches it covalently to phosphatidylethanolamine in an E3-ligase like reaction ( Klionsky and Schulman , 2014 ) . Since LC3 remains associated with the autophagosomal membrane until its delivery to the lysosome , it is considered to be the most reliable marker for autophagosomes ( Klionsky et al . , 2012 ) . Therefore , to test whether the Rab26 containing clusters are linked to the autophagy pathway , we next checked for association with autophagosome-related proteins . First , we assessed for colocalization between Rab26 and Atg16L1 , a component of pre-autophagosomes ( Mizushima et al . , 2003; Ravikumar et al . , 2010 ) . For this purpose , hippocampal neurons transiently expressing EGFP-Rab26WT were immunostained for endogenous Atg16L1 . Indeed , an almost perfect colocalization between Atg16L1 and Rab26-positive clusters was detected in neuronal cell bodies ( Figure 6A ) , thereby identifying these clusters as autophagosomal precursors . Next , we stained untransfected neurons for endogenous Rab26 and Atg16L1 . Again , a high degree of overlap was observed between Rab26 and Atg16L1 in clusters decorating neurites ( Figure 6B ) but not in cell bodies which remained largely unstained ( not shown ) . Strong overlap with Atg16L1 was also observed when neurons were transfected with FLAG-Rab26WT and/or FLAG-Rab26QL , but not with FLAG-Rab26TN ( Figure 6C ) . This indicated that the association of Rab26 with autophagosomes depends on the GTP-form of the protein . This GTP-dependency was similarly noted in HeLa cells following ectopic expression of Rab26 . In this instance , overexpression of GTP-bound forms ( WT and QL ) , but not GDP-bound ( TN ) form , of EGFP-Rab26 led to the formation of large Atg16L1-positive clusters ( Figure 6—figure supplement 1A–C ) . Analysis by immunogold-TEM again revealed that these clusters consisted of small but often heterogeneous vesicles , partially surrounded by membranes , with EGFP labeling detected both on vesicles within clusters as well as on their encapsulating membrane ( s ) ( Figure 6—figure supplement 1D ) . 10 . 7554/eLife . 05597 . 013Figure 6 . Clusters containing GTP-Rab26 colocalize with autophagosome-specific proteins both in cell bodies and dendrites of cultured hippocampal neurons . Arrows indicate co-localization . ( A ) Somatic clusters induced by expression of EGFP-Rab26WT co-localize with endogenous Atg16L1 . ( B and C ) In neurites , Atg16L1 co-localizes with clusters of endogenous Rab26 ( non-transfected , DIV 15 , panel B ) and with clusters containing FLAG-tagged Rab26WT and Rab26QL , but not with Rab26TN ( transfected , DIV 9 , panel C ) . ( D ) Similar colocalization patterns were obtained from neurites expressing FLAG-Rab26 variants and autophagosomes labeled by GFP-LC3B . Note that occasional puncta were observed for the GDP-preferring variant Rab26TN that , however , showed no overlap with LC3B ( arrowhead ) . DIV 9 . ( E and F ) Co-expression of FLAG-Rab26WT and EGFP-Rab33WT in hippocampal neurons . In the soma ( E ) , EGFP-Rab33 in primary restricted to a perinuclear structure reminiscent of the Golgi apparatus whereas an almost perfect overlap was observed between the Rab33 and Rab26 in peripheral puncta lining neurites ( F ) . DIV 8 . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 01310 . 7554/eLife . 05597 . 014Figure 6—figure supplement 1 . Rab26 overexpression causes vesicle clusters in HeLa . Endogenous ATG16L1 is recruited to vesicle clusters induced by transient overexpression of EGFP-tagged Rab26WT ( A ) or Rab26QL ( B ) in HeLa cells . Although GFP-Rab26TN forms small clusters in these cells , endogenous Atg16L1 is not recruited to these structures ( C ) . ( D ) Immunogold electron microscopy analysis revealed that HeLa cells transfected with EGFP-Rab26WT exhibited formation of huge clusters of small vesicle ( left panel , inset ) that are sometimes surrounded by external membranes ( right panel , inset ) and large vacuole autolysosome-like compartments whose membrane were labeled with EGFP-Rab26WT ( center and right panel ) . Rab26 was visualized by anti-GFP antibody coupled with 10 nm gold particles . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 014 Transient association of Atg16L1 to pre-autophagosomal structures enables the recruitment and membrane attachment of LC3 family members that persist on the autophagosomal membranes until degradation . Therefore , to assess whether the Rab26/Atg16L1 clusters recruit LC3 to membranes , we co-transfected cultured hippocampal neurons with GFP-tagged LC3B ( one of the eight known LC3 family members ) and FLAG-tagged Rab26 variants . Co-expression of GFP-LC3 and active forms FLAG-Rab26 ( WT and QL ) resulted in a localization pattern comparable to that observed for Atg16L1 ( Figure 6D ) , thereby verifying the nature of these compartments as autophagosomes . Interestingly , however , LC3 recruitment was not observed when the EGFP-tagged Rab26WT was overexpressed ( not shown , see ‘Discussion’ ) . Taken together , the above findings indicate a novel functional link between a Rab GTPase and a hitherto unknown autophagy pathway in neurons that appears selective for synaptic and/or secretory vesicles , tentatively termed ‘vesiculophagy’ . Indeed , there is now a growing body evidence implicating several Rabs ( Rab1 , Rab7 , Rab9 , Rab11 , Rab24 , Rab32 , and Rab33 , inclusive ) in canonical autophagy ( for a comprehensive review [Chua et al . , 2011] ) . Among these , Rab33 , a Golgi resident Rab , participates in the formation of autophagosome precursors by recruiting Atg16L1 ( a Rab33 effector ) to isolation membranes ( Itoh et al . , 2008 ) . Given that Rab26 colocalizes with Atg16L1 , we checked for potential cooperation between Rab26 and Rab33 in neurons . For this , hippocampal neurons were co-transfected with FLAG-Rab26WT and EGFP-Rab33BWT . As shown in Figure 6E , EGFP-Rab33BWT was largely restricted to the perinuclear/Golgi region in the soma which showed no appreciable overlap with FLAG-Rab26WT . On the other hand , significant overlap between Rab26 and Rab33 was observed in more peripheral puncta lining neurites ( Figure 6F ) . Together these data imply that the autophagy-pathway regulated by Rab26 may functionally intersect with Rab33 . The overlap between Rab26 and Rab33 prompted us to further investigate whether Atg16L1 may also be an effector of Rab26 . To explore this possibility , we performed co-immunoprepitation experiments between FLAG-tagged Rab26 ( WT , QL or TN ) and endogenous Atg16L1 in HeLa cells . As shown in Figure 7A , all three FLAG-tagged Rab26 variants were efficiently immunoprecipitated with the FLAG antibody . Immunoblotting for endogenous Atg16L1 from the same immunoprecipitates revealed co-precipitation between Atg16L1 and Rab26QL . By comparison , little to no Atg16L1 was detectable in the precipitates of Rab26-WT and Rab26TN , respectively , indicating that the interaction between Rab26 and Atg16L1 is GTP-dependent . 10 . 7554/eLife . 05597 . 015Figure 7 . Atg16 is an effector of GTP-Rab26 . ( A ) Co-Immunoprecipitation of FLAG-tagged Rab26 variants expressed in HeLa cells with endogenous Atg16L1 protein . Immunoprecipitation was carried out following lysis in detergent-containing buffer and clearance by centrifugation to remove cell debris . Note that only the GTP-preferring QL variant of Rab26 showed significant binding to Atg16L1 ( arrow ) . ( B ) GST pulldown of purified recombinantly expressed GST-Rab26 variants with a pre-formed complex of His-tagged versions of Atg5 and the N-terminal domain of Atg16L1 ( Atg16NT ) . Note that Atg16NT selectively interacted with the GTP-preferring QL-variant of Rab26 . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 01510 . 7554/eLife . 05597 . 016Figure 7—figure supplement 1 . Analyses of Rab26 and ATG16L1 interaction by analytical gel filtration . To analyze complexes formed between Atg16L1 and Rab26 or Rab33 , samples containing either Rab protein were pre-incubated with Atg16NT/Atg5 and loaded on a Superdex 200 10/30 GL column . As controls , Rab26 , Rab33 , or Atg16NT/Atg5 alone were separately loaded and analyzed . Eluted fractions were collected , analyzed by SDS-PAGE and protein bands visualized using Coomassie staining . ( A ) The elution profiles for various samples analyzed . ( B–F ) Representative fractions were separated on SDS PAGE gels and stained with Coomassie to visualize the protein bands . ( B ) Co-migration of Rab26QL with Atg16NT/Atg5 was not detected when pre-incubated samples of these proteins were tested . ( C ) Complexes of Rab33QL with Atg16NT/Atg5 can be observed to co-migrated ( fractions 5–7 , arrows ) . ( D–F ) Control runs were performed using only Rab26QL , Rab33QL or Atg16NT/Atg5 . DOI: http://dx . doi . org/10 . 7554/eLife . 05597 . 016 In parallel , we performed GST-pulldown assays to verify the results from the coIP experiments . For this , purified bacterially expressed recombinant Rab26 variants ( QL or TN ) , tagged with GST were incubated with a preassembled complex of Atg5 and the N-terminal fragment of Atg16L1 containing its coiled coil domain ( Atg16NT ) . In agreement with our immunoprecipitation studies , GST-pulldown revealed an interaction between Atg16L1 and Rab26 , with Atg16L1 binding to the QL and to a lesser extent the TN-variant of Rab26 ( Figure 7B ) , with the latter being further reduced upon repetitive washings ( not shown ) . Atg5 remains bound in this complex . To further examine the interaction , we analyzed the binding between Rab26 and Atg16L1 using analytical gel filtration . Surprisingly , formation of Rab26 ( QL ) -ATG16L1 complexes were not detectable with this approach ( Figure 7—figure supplement 1 ) . As a positive control , we carried out the same experiment using Rab33 ( QL ) and ATG16L1 . Here , complex formation was detectable with this approach . Thus , while both IP and pull-down experiments show that RAB26 binds ATG16L1 in a GTP-dependent manner , this binding appears to be weaker than the interaction between Rab33 and ATG16L1 .
Rab26 is most closely related to the secretory GTPases Rab3 and Rab27 , which led to the conclusion that it may perform similar functions in membrane traffic ( Fukuda , 2008 ) . This view is supported by reports showing association of Rab26 with zymogen granules in exocrine cells ( Nashida et al . , 2006 ) and by the observation that overexpression of the dominant-negative form abolishes the formation of zymogen granules ( Tian et al . , 2010; Li et al . , 2012 ) . More recently , Rab26 has been found to be associated with lysosomes in zymogen-secreting cells ( Jin and Mills , 2014 ) implying that its functions in secretory cells extend beyond that of exocytosis . In our previous work ( Takamori et al . , 2006; Pavlos et al . , 2010 ) , Rab26 was identified among the list of Rab proteins ( up-to-30 ) enriched on highly purified synaptic vesicles . Our present data now show that this association is exclusive , with Rab26 being absent from other organelles such as early endosomes , paralleling the distribution of other secretory Rabs . On the other hand , the preferential association of Rab26 with large clusters of synaptic vesicles and its conspicuous absence from smaller boutons positive for synaptic vesicle markers is clearly distinct from Rab3 and Rab27b and indicates that Rab26 may not be contributing to the canonical function of these Rabs in regulated exocytosis . Intriguingly , in contrast to for example , Rab3 and Rab5 , Rab26 cannot be extracted from synaptic vesicle membranes by GDI in its GDP-form—a feature it shares with Rab27b . Rather , Rab26 exhibits a tendency to oligomerize in the GDP-form , again a feature shared with Rab27b and perhaps with some others such as Rab11 and Rab9 , which crystallize as dimers in the GDP-state ( Pasqualato et al . , 2004; Wittmann and Rudolph , 2004; Chavas et al . , 2007 ) . It is somewhat surprising that , along with the GDP-bound variant , wild-type Rab26 also appears to oligomerize ( albeit to a lesser extent ) . However , this might be explained by the known high intrinsic GTP-hydrolysis rates of wild-type and native Rab proteins which would render a percentage of wild-type Rab26 inactive ( approximately 40% by co-immunoprecipitation ) , thereby prompting its oligomerization on membranes and hence preclude its GDI-extraction from synaptic vesicles . By comparison , the propensity of the ‘GDP-locked’ variant ( Rab26TN ) to rapidly oligomerize would likely result a steric clash with the GDI-related Rab escort protein ( REP ) , which is required for prenylation and delivery of new synthesized Rabs to their cognate membranes ( Andres et al . , 1993 ) . Such a scenario would be in keeping with the predominantly cytosolic distribution pattern observed in neurons expressing FLAG/EGFP-tagged Rab26TN . Perhaps the most conspicuous feature of Rab26 is that it is not only preferentially associated with secretory vesicle clusters but actually induces their formation in a GTP-dependent manner as becomes apparent upon the expression of exogenous Rab26 variants in both neurons and non-neuronal cells . This is most dramatically observed with the EGFP-tagged variant suggesting that the weak homodimerization tendency of EGFP enhances the phenotype ( note that no other EGFP-tagged Rab exhibits similar features including the most abundant secretory GTPase , Rab3a ) . At present , the exact mechanism underlying this clustering phenotype is unclear . Nevertheless , since the GTP-form of Rab26 does not oligomerize , it is unlikely that clustering is effected by homophilic Rab26 interactions . Rather , it possible that clustering is mediated by a hitherto unknown effector protein . This effector is probably distinct from Atg16L1 as overexpression of EGFP-Rab33B ( that also recruits Atg16L1 ) does not induce such clusters ( Figure 6 , and data not shown ) . However , given that the central terminal region of Atg16L1 has a tendency for homo-multimerization , this possibility cannot be excluded ( Mizushima et al . , 2003; Ishibashi et al . , 2011 ) . Intriguingly , our findings agree with a recent report according to which overexpression of Rab26 in exocrine cell lines induces clustering of lysosomes , reminiscent of the partial co-localization of the EGFP-induced Rab26 clusters with lysosomes in neuronal cell bodies ( Jin and Mills , 2014 ) . Our results indicate that the core autophagy protein Atg16L1 is an effector of Rab26 that binds to the GTPase exclusively in the GTP-form , paralleling previous findings on the Golgi-resident Rab33B ( Itoh et al . , 2008 ) . Interestingly , binding of Rab26 to Atg16L1 appears to be weaker than that between Rab33 and Atg16L1 , which plays a role in canonical autophagy , probably explaining why Itoh et al . ( 2008 ) did not observe binding of GST-Rab26 to a full-length Atg16L1 protein using stringent conditions . It is conceivable that the interaction is more transient , or else , that it requires additional factors for stabilization , thus allowing for fine-tuning the flow of synaptic vesicles targeted for selective autophagy . How does recruitment of Atg16L1 to synaptic vesicle clusters relate to the established steps of autophagosome formation ? First of all , it cannot yet be excluded with certainty that upon recruitment to these vesicles Atg16L1 performs a non-canonical function that is not related to autophagosome formation ( see e . g . , [Pimentel-Muinos and Boada-Romero , 2014] for review of such functions ) . In particular , Atg16L1 and Rab33A have recently been found to be associated with secretory vesicles in neuroendocrine PC12 cells , with the data suggesting a role for Atg16L1 in regulating exocytosis independent of autophagy ( Ishibashi et al . , 2012 ) . On the other hand , based on our extensive morphological assessment using double immunolabeling microscopy , we strongly favor that the Rab26-Atg16L1 complexes in neurons represent pre-autophagosomal structures because ( i ) Rab26 is not present on all synaptic vesicles but rather confined to vesicle aggregates that may be functionally impaired , and ( ii ) LC3 is recruited to these clusters suggesting that the formation of an autophagosomal membrane is , at least in part , initiated . Our data indicates that the vesicle clusters containing Rab26 and Atg16L1 have undergone exo-endocytotic cycling . Intriguingly , clathrin has recently been shown to interact with Atg16L1 , thus targeting plasma membrane constituents towards autophagosome precursors via clathrin-mediated endocytosis ( Ravikumar et al . , 2010 ) . Since clathrin-mediated endocytosis constitutes the main endocytotic pathway for synaptic vesicles , it is conceivable that there is a synergy between Rab26- and clathrin-induced autophagocytosis in nerve terminals that further fine-tunes the targeting of synaptic vesicles to preautophagosomal structures . Taken together , the data support the view that Rab26 is member of a signaling cascade that selectively targets synaptic and secretory vesicles towards autophagocytosis , which may represent a novel and highly specific form of autophagy ( ‘vesiculophagy’ ) . In recent years , it has become apparent that in addition to the classical , mTOR/ULK kinase-induced pathway of macroautophagy there is a panoply of diverse and highly regulated pathways ( e . g . , pexophagy and mitophagy ) that all converge on the common pathway but differ in the mechanism of initiation and cargo recruitment . In many of these cases the pathway is initiated by ubiquitination of target proteins . While we do not know whether this is also the case here , it is conceivable that the initiation event may indeed be the recruitment of active Rab26 to the membrane of subsets of synaptic vesicles that then interacts with other factors to form clusters and to recruit an isolation membrane , the origin of which remains to be identified . Following the classical work in the early 70s of last century demonstrating that synaptic vesicles undergo multiple rounds of recycling in the synapse , ( Atwood et al . , 1972; Ceccarelli et al . , 1973; Heuser and Reese , 1973 ) , recent attention has focused primarily on unraveling the mechanisms of endocytosis and vesicle re-formation ( Saheki and De Camilli , 2012 ) . However , all membrane constituents age and accumulate structural defects requiring sorting out of damaged constituents . Although no increase in the number of late endosomes , lysosomes or autophagosomes was observed following even massive stimulation , it was hypothesized as early as 1971 that newly reformed synaptic vesicles could either be actively re-used as functional synaptic vesicles or re-directed to a pathway ultimately leading to lysosomes as the final destination for degradation ( Holtzman et al . , 1971 ) . Our discovery of vesiculophagy as a pathway initiated in presynaptic boutons that directs entire synaptic vesicle pools towards autophagosomes provides a previously uncharacterized link towards lysosomal degradation of trafficking organelles which is distinct from the classical endosomal route . Indeed , recent data suggest that presynaptic neurotransmission is functionally modulated by macroautophagy . Induction of autophagy in neurons increased the amount of autophagic vacuoles in presynaptic terminals and with an accompanying reduction in synaptic vesicle number and decrease in evoked neurotransmitter release ( Hernandez et al . , 2012 ) . Furthermore , two groups have recently suggested that in axons autophagosomes originate distally and then are transported by retrograde motors towards the cell body . During their travel they undergo fusion with acidic compartments and finally with the lysosomes ( Lee et al . , 2011; Maday et al . , 2012 ) . It is therefore conceivable that Rab26 feeds vesicle membranes into autophagosomes that may form and mature during retrograde transport . How this novel pathway is initiated and regulated will be the subject of future studies .
Mouse monoclonal and rabbit polyclonal antibodies specific for synaptophysin , synaptotagmin , synaptobrevin , Rab3a , GDI ( Cl 81 . 2 ) and GFP were provided by Synaptic Systems ( Göttingen , Germany ) . Mouse anti-LAMP2 antibody was purchased from the Developmental Studies Hybridoma Bank ( DSHB , University of Iowa , IA ) . Antibodies against EEA1 and GM130 were purchased from BD Bioscience ( Franklin Lakes , NJ ) . The antibody against the FLAG epitope was obtained from Stratagene ( La Jolla , CA ) . Antibodies specific for Atg16L1 were purchased from CosmoBio ( Tokyo ) and MBL ( Nagoya ) . Anti-Atg5 antibody was from Novus Biological ( Littleton , Colorado ) . The antibody against secretogranin II was kindly provided by Sharon Tooze ( Cancer Research UK ) . The monoclonal antibody against Rab26 used in this study was raised by immunizing 8- to 10-week-old female BALB/c mice over a period of 17 days with full length recombinant human Rab 26 . Cells from knee lymph nodes were fused with the mouse myeloma cell line P3X63Ag . 653 ( ATCC CRL-1580 ) . Cell culture supernatants obtained from individual clones were then screened using enzyme-linked immunosorbent assay ( ELISA ) , immunoblot assays and immunoflourescence . The final hybridoma used in this study was cloned two times by limiting dilution . The monoclonal antibody produced from this clone was determined to be of the IgG2a subclass and is specific for Rab26 ( Figure 1—figure supplement 1 ) . The antibody is commercially available from Synaptic Systems ( Göttingen , Germany ) . Cy3-labeled goat anti-mouse or anti-rabbit and Alexa 488-labeled goat anti-mouse secondary antibodies were purchased from Dianova ( Hamburg , Germany ) and used at a dilution of 1:400 . Horseradish peroxidase-conjugated anti-mouse and anti-rabbit secondary antibodies were obtained from Bio-Rad ( Hercules , CA ) and used at a dilution of 1:2000 or 1:4000 . The coding sequence of human Rab26 ( NM_014353 . 4 ) was amplified by PCR and inserted into pEGFP-C1 ( Clontech , Palo Alto , CA ) using BglII and BamHI restriction sites or into pCMV-Tag2a ( Agilent Technologies , La Jolla , CA ) using BamHI and XhoI restriction sites for expression in mammalian cells . Likewise , inserts encoding Rab26 Q123L , T77N or N177I mutants were generated by recombinant PCR and similarly inserted into these vectors . For recombinant protein expression in bacteria , inserts for the Rab26 variants were inserted into pGEX-KG using EcoRI and BamHI while the insert encoding alpha-GDI was sub-cloned into pET-28a ( Novagen , Madison , WI ) . The sequence corresponding to murine Atg16L1 ( 1–265 ) ( BC049122 ) was cloned into pET-28a ( Novagen ) using NdeI and NotI restriction sites . Full-length murine Atg5 ( 1–275 ) ( BC002166 ) was cloned with an N-terminal thrombin cleavage site into the multiple cloning site 1 of pETDuet-1 ( Novagen ) using the SalI and NotI sites . The vector expressing neuropeptide Y ( NPY ) was generated by inserting the sequence encoding human pro-NPY into the pmRFP vector . Cloning was performed according to standard procedures ( Janssen , 2001 ) . The plasmid expressing GFP-tagged human LC3B was a kind gift from Dr Zvulun Elazar ( Weizmann Institute , Israel ) . Culturing of the HEK 293 and HeLa SS6 cell lines and the preparation of high density primary rat hippocampal neurons have been previously described ( Rosenmund and Stevens , 1997; Chua et al . , 2012 ) . Neurons were transfected between 7 to 12 days after seeding or , in the case of the cell lines , 1 day after seeding using Lipofectamine 2000 ( Invitrogen , Carlsbad , CA ) according to the manufacturer's protocol . Neurons in Figures 3F , 4 , Figure 3—figure supplement 1 and Figure 4—figure supplement 1 were transfected using calcium phosphate as previously described ( Pavlos et al . , 2010 ) . Immunostaining was then performed as described in Chua et al . ( 2012 ) . Briefly , cells were washed once with PBS to remove serum and then fixed using 4% paraformaldehyde . Afterwards , cells were permeabilized with 0 . 3% Triton-X-100 in PBS , rinsed with PBS and then blocked with 10% normal goat serum in PBS . Incubation with primary antibodies diluted in blocking solution was then carried out for 1 hr at room temperatures or overnight at 4°C . Subsequently , cells were exposed to secondary Cy3 or Alexafluor 488 conjugated goat anti-rabbit and anti-mouse antibodies , respectively , for 1 hr at room temperature . After washing , cells were mounted on slides ( SuperFrost Plus , VWR International bvba , Leuven , Belgium ) and then imaged using a confocal microscope ( LSM 780 , Zeiss , Germany ) or an epifluorescence microscope ( Axiovert 200M , Zeiss , Germany ) . Linescan analyses were performed using ImageJ or LAS AF Lite software . To visualize synaptic vesicles that have undergone recycling , live neurons transfected with EGFP-Rab26WT were incubated in culture for 24 hr with Oyster 550-labeled anti-synaptotagmin-I antibodies ( Synaptic Systems ) that recognize its luminal domain ( Willig et al . , 2006 ) . The neurons were then washed twice with PBS , fixed with 4% paraformaldehyde and analyzed under the microscope . The UAST-YFP . Rab26 , UAST-YFP . Rab26Q250L , UAST-YFP . Rab26T204N ( Zhang et al . , 2007 ) and elav-Gal4 transgenic fly stocks were obtained from the Bloomington stock collection . Dissection and immunostaining of neuromuscular junctions from third instar larvae were performed as described ( Schmid and Sigrist , 2008 ) using the following antibodies: mouse Anti-Brp ( hybridoma clone nc82 , DSHB; 1:50 dilution ) , anti-Csp antibody ( hybridoma clone ab49 , DSHB; 1:100 dilution ) , the chicken anti-GFP antibody ( Abcam; 1:1000 dilution ) and the goat anti-HRP ( Sigma; 1:400 dilution ) . Dylight-649 labeled anti-goat and Alexa-488 labeled anti-chicken secondary antibodies were purchased from Jackson ImmunoResearch Laboratories ( West Grove , PA ) . Alexa-568 conjugated anti-mouse secondary antibodies were purchased from Invitrogen ( Carlsbad , CA ) . Images were acquired with a microscope ( DMR-E; Leica , Germany ) equipped with a scan head ( TCS SP2 AOBS; Leica , Germany ) and an oil-immersion 63 × 1 . 4 NA objective . Biochemical isolation of synaptic vesicles from the rat brain was performed as described previously ( Huttner et al . , 1983; Takamori et al . , 2006 ) . 12 µg of each brain fraction were loaded for analysis by immunoblotting to monitor the protein enrichment profile . Purified monoclonal antibodies directed against Rab26 ( described above ) and synaptophysin ( clone 7 . 2; [Jahn et al . , 1985] ) were coupled to epoxy-activated methacrylate microbeads ( Eupergit C1Z , Röhm Pharmaceutical; note that these beads are no longer commercially available ) and used for immunoisolation as described previously ( Burger et al . , 1989; Takamori et al . , 2000 ) . The bound vesicles were subsequently analyzed by electron microscopy or eluted with 40 µl 2 × SDS sample buffer for immunoblots analysis . The RabGDI assay was performed as described in Pavlos et al . ( 2010 ) . Briefly , crude synaptic vesicles ( LP2 ) were used as the starting material . 50 µg of LP2 were pre-incubated with 500 µM GDP or 500 µM GTPγS for 15 min at 37°C in 200 µl of extraction buffer containing 100 mM KCl , 5 mM MgCl2 , 10 mM EDTA , 50 mM HEPES-KOH pH 7 . 4 supplemented with a protease inhibitor cocktail ( cOmplete EDTA-free , Roche , Mannheim , Germany ) . 5 µM of purified His-GDI were then added to each sample and further incubated for 45 min at 37°C . The samples were then kept on ice and subsequently centrifuged for 20 min at 200 , 000×g , 4°C using a Beckman S100 AT3 rotor . The resulting pellet was re-suspended in 50 µl 2 × lithium dodecyl sulfate ( LDS ) sample buffer ( Invitrogen ) , boiled at 95°C for 5 min and analyzed by immunoblotting . For morphological analyses of immunoisolated vesicles , synaptophysin- or Rab26-conjugated microbeads containing the immunoisolated vesicles were first pelleted and subsequently immobilized using 2% agarose in 0 . 1 M cacodylate buffer at pH 7 . 4 . Small agarose cubes containing the immobilized beads were fixed overnight at 4°C using 2% glutaraldehyde in 0 . 1 M cacodylate buffer at pH 7 . 4 . After post-fixation in 1% osmium tetroxide and pre-embedding staining with 1% uranyl acetate , tissue samples were dehydrated and embedded in Agar 100 . Thin sections ( 80 nm ) were examined using a Philips CM 120 BioTwin transmission electron microscope ( Philips Inc . Eindhoven , The Netherlands ) . Images were taken with a TemCam F224A slow scan CCD camera ( TVIPS , Gauting , Germany ) . The evaluation of the samples was done using the iTEM software ( Olympus Soft Imaging Solutions GmbH , Münster , Germany ) . For immunogold electron microscopy , ultrathin cryosections of neuronal cultures ( Figure 5A and Figure 5—figure supplement 1 ) and HeLa cells ( Figure 6—figure supplement 1 ) transfected with EGFP-Rab26WT , were prepared as described previously ( Tokuyasu , 1973 , 1980; Zink et al . , 2009 ) . For the ultrastructural analyses of the Drosophila neuromuscular junction ( Figure 5D ) , a standard protocol was used . Briefly Drosophila filets were fixed by immersion using 2% glutaraldehyde in 0 . 1 M cacodylate buffer at pH 7 . 4 overnight at 4°C . After post-fixation in 1% osmium tetroxide and pre-embedding staining with 1% uranyl acetate , tissue samples were dehydrated and embedded in Agar 100 . Thin sections ( 80 nm ) were examined using a Philips CM 120 BioTwin transmission electron microscope ( Philips Inc . Eindhoven , The Netherlands ) . Images were taken with a TemCam F416 slow scan CMOS camera ( TVIPS , Gauting , Germany ) . Human GST-tagged Rab26WT , Q123L , T77N were expressed in Escherichia coli BL21 ( D3 ) . 200 ml of pre-culture were grown at 37°C overnight . 10 ml of the pre-culture were then inoculated into 1 l of fresh LB medium supplemented with 100 µg/ml ampicillin and incubated for 2 . 5 hr at 37°C until the OD600 reached a value of 0 . 6–0 . 8 . The cultures were then incubated for 1 hr at 16°C . Induction was initiated by adding 1 mM IPTG to the cultures and the expression was carried out overnight at 16°C . Thereafter , cells were harvested by centrifugation at 4000 rpm for 10 min using a Beckman centrifuge . Pellets obtained from each 1 l culture flask were re-suspended in 25 ml of protein buffer containing 50 mM HEPES pH 7 . 4 , 500 mM NaCl , 5 mM DTT , 5 mM MgCl2 , 100 µM GTPγS/GDP , supplemented with protease inhibitor cocktail and 1 mg/l of DNase . The samples were left for 10–15 min at 4°C and subsequently sonicated four times for 30 s each , separated by a 1 min incubation on ice , using a Branson Sonifier 450 . The lysate was then cleared at 13 , 000 rpm using a SLA 1500 rotor for 40 min at 4°C . The resulting supernatant was collected and filtered using a 0 . 45 µm Whatman filter . The filtrate was then loaded onto a GST-column ( GST Trap4B GE Healthcare , Germany ) and eluted using 30 mM glutathione in protein buffer . The eluted fractions were collected and dialyzed three times for 3 hr each using protein buffer to remove glutathione . The purified proteins were then used for GST pulldown assays . His-tagged murine Atg16L1 ( 1–265 ) -pET-28a and His-tagged murine Atg5-pETDuet-1 were co-transformed into E . coli Rosetta 2 cells ( Merck Millipore , Germany ) . Proteins were co-expressed in 3 l of ZYM-5052 autoinducing medium ( Studier , 2005 ) supplemented with 100 mg/l ampicillin and 30 mg/l kanamycin for 3 hr at 37°C and followed by an overnight incubation at 18°C . Cells were harvested by centrifugation at 4500×g for 20 min . Pellets were resuspended in 100 ml buffer A ( 0 . 3 M NaCl , 1 mM MgCl2 , 35 mM imidazole , 50 mM NaH2PO4 , pH 7 . 5 ) . Cells were lysed by sonication and centrifuged for 1 hr at 25 , 000×g . The resulting supernatant was loaded onto two in line connected 1 ml HisTrap columns ( GE Healthcare , Germany ) , washed with 80 ml buffer A and then eluted with a gradient from 0 to 100% over 80 ml of buffer B ( 0 . 3 M NaCl , 1 mM MgCl2 , 0 . 5 M imidazole , 50 mM NaH2PO4 , pH 7 . 5 ) . Fractions containing the purified proteins were pooled and dialyzed overnight at 4°C in gel filtration buffer ( 0 . 2 M NaCl , 1 mM MgCl2 , 25 mM HEPES , pH 7 . 5 ) . The complex was concentrated and loaded onto a Superdex 200 16/60 column ( GE Healthcare , Germany ) . Fractions were pooled and concentrated to 5 mg/ml , divided into aliquots , flash cooled in liquid nitrogen , and stored at −80°C . Co-immunoprecipitation assays were performed as described in Chua et al . ( 2012 ) . Briefly , transiently transfected cells were washed once with ice-cold PBS and then lysed using lysis buffer ( 50 mM HEPES , 150 mM NaCl , 1 mM MgCl2 , 1% Tx-100 supplemented with a protease inhibitor cocktail ) for 30 min . The lysate was then clarified by centrifugation at 10 , 000×g for 10 min . The resulting supernatant was incubated for 2 hr with anti-Flag or anti-GFP antibodies . Subsequently , 30 µl of protein G-Sepharose beads ( GE Healthcare , Sweden ) were added to each sample and further incubated for 1 hr under constant rotation . The samples were then washed thrice with lysis buffer . Finally , 25–30 µl of 2 × LDS sample buffer were then added to the beads and the mixture was boiled at 95°C for 5 min . 15 µl of the immunoprecipitated samples and 5 µl of the input were analyzed by immunoblotting . For GST pulldown assays , 10 µg of each purified GST-Rab26 variant were first immobilized to 15 µl of Glutathione Sepharose 4 Fast Flow beads ( GE Healthcare , Sweden ) in a buffer containing 200 mM NaCl , 5 mM DTT , 5 mM MgCl2 , 10 µM GTPγS/GDP , 1% NP40 , 50 mM HEPES , pH 7 . 4 for 30 min at 4°C with constant rotation . Subsequently , 10 µg of the pre-formed His-Atg16L1NT/His-Atg5FL complex was added to the mixture and further incubated for 1 more hr . The beads were then washed three times with buffer . 30 µl of 2 × LDS sample loading buffer was used to elute the proteins from the beads and 5 µl of each sample were subjected to SDS-PAGE and Western blotting . Human Rab26 ( 54–233 ) Q123L was cloned into pET-28a using NdeI and XhoI cleavage sites . Murine Rab33B ( 30–202 ) Q92L ( BC065076 ) was cloned into pETDuet-1 with BamHI and NotI restriction sites . Plasmids were transformed into E . coli BL21 ( DE3 ) . Bacteria were grown in 3 l ZYM-5052 autoinducing medium supplemented with the appropriate antibiotic for 8 hr at 37°C . Cells were harvested by centrifugation and resuspended in 100 ml buffer A ( 30 mM imidazole , 0 . 2 M NaCl , 5 mM MgCl2 , 30 mM HEPES , pH 7 . 5 ) . Bacteria were lysed with a microfluidizer and centrifuged for 1 hr at 25 , 000×g . Supernatant was loaded on a 1 ml HisTrap column ( GE Healthcare , Germany ) and washed with 25 ml buffer A and then eluted with a gradient from 0 to 100 % B over 20 ml ( 0 . 4 M imidazole , 0 . 2 M NaCl , 5 mM MgCl2 , 30 mM HEPES , pH 7 . 5 ) . Fractions containing protein were pooled and diluted 1:1 with gel filtration buffer ( 0 . 15 M NaCl , 5 mM MgCl2 , 20 mM Hepes pH 7 . 5 ) and 250 µM GTP was added . Proteins were kept at 4°C overnight . Proteins were concentrated and loaded onto a Superdex 75 16/60 column ( GE Healthcare , Germany ) . Fractions were then pooled and concentrated to at least 30 mg/ml , aliquoted and then flash cooled in liquid nitrogen and stored at −80°C . For analytical gel filtration experiments 500 µl samples were loaded onto a Superdex 200 10/30 GL column ( GE Healthcare , Germany ) using an ÄktaPurifier ( GE Healthcare , Germany ) . The gel filtration buffer was 0 . 15 M NaCl , 5 mM MgCl2 , 20 mM HEPES , pH 7 . 5 . 80 µM Rab26 ( 54–233 ) Q123L , 8 µM Atg16NT/Atg5 and 1 mM GTP were mixed , incubated for 20 min at RT and then loaded to the column . For comparison , 80 µM Rab33B ( 30–202 ) Q92L was similarly mixed with 8 µM Atg16NT/Atg5 and 1 mM GTP and analyzed . As controls , 8 µM Atg16NT/Atg5 alone , 80 µM Rab26 ( 54–233 ) Q123L containing 1 mM GTP or 80 µM Rab33B ( 30–202 ) Q92L with 1 mM GTP were ran separately . Samples were analyzed on Coomassie stained 15% SDS PAGE gels . | Our brain contains billions of cells called neurons that form an extensive network through which information is readily exchanged . These cells connect to each other via junctions called synapses . Our developing brain starts off with far more synapses than it needs , but the excess synapses are destroyed as the brain matures . Even in adults , synapses are continuously made and destroyed in response to experiences and learning . Inside neurons there are tiny bubble-like compartments called vesicles that supply the synapses with many of the proteins and other components that they need . There is a growing body of evidence that suggests these vesicles are rapidly destroyed once a synapse is earmarked for destruction , but it is not clear how this may occur . Here , Binotti , Pavlos et al . found that a protein called Rab26 sits on the surface of the vesicles near synapses . This protein promotes the formation of clusters of vesicles , and a membrane sometimes surrounds these clusters . Further experiments indicate that several proteins involved in a process called autophagy—where unwanted proteins and debris are destroyed—may also be found around the clusters of vesicles . Autophagy starts with the formation of a membrane around the material that needs to be destroyed . This seals the material off from rest of the cell so that enzymes can safely break it down . Binotti , Pavlos et al . found that one of the autophagy proteins—called Atg16L—can bind directly to Rab26 , but only when Rab26 is in an ‘active’ state . These findings suggest that excess vesicles at synapses may be destroyed by autophagy . Further work will be required to establish how this process is controlled and how it is involved in the loss of synapses . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | The GTPase Rab26 links synaptic vesicles to the autophagy pathway |
The population genetics of most range expansions is thought to be shaped by the competition between Darwinian selection and random genetic drift at the range margins . Here , we show that the evolutionary dynamics during range expansions is highly sensitive to additional fluctuations induced by environmental heterogeneities . Tracking mutant clones with a tunable fitness effect in bacterial colonies grown on randomly patterned surfaces we found that environmental heterogeneity can dramatically reduce the efficacy of selection . Time-lapse microscopy and computer simulations suggest that this effect arises generically from a local 'pinning’ of the expansion front , whereby stretches of the front are slowed down on a length scale that depends on the structure of the environmental heterogeneity . This pinning focuses the range expansion into a small number of 'lucky’ individuals with access to expansion paths , altering the neutral evolutionary dynamics and increasing the importance of chance relative to selection .
Stochasticity and its competition with deterministic forces plays an integral role in biology , such as in stochastic gene expression , cellular decision making , and cell differentiation ( Balázsi et al . , 2011 ) . Stochastic processes are also at the heart of evolutionary dynamics: not only do the mutations entering a population occur at random times in random individuals and at random positions in their genome , but in addition the fate of a mutation and its clonal lineage is largely stochastic and only partly determined by its effect on the individual’s fitness . The random fluctuations in the frequency of a mutant allele due to the stochasticity associated with reproduction are called genetic drift . Genetic drift is particularly strong at the front of range expansions , where only a relatively small number of individuals at the front of the expansion contributes to future growth and thus has any influence on the future genotypic composition of the population . The neutral diversity and adaptation in spatially expanding populations has been studied in computer simulations ( Edmonds et al . , 2004; Klopfstein et al . , 2006; Kuhr et al . , 2011; Kuhr and Stark , 2015; Lavrentovich and Nelson , 2014; Otwinowski and Krug , 2014 ) , in the field ( Ramachandran et al . , 2005; White et al . , 2013; Louppe et al . , 2017 ) , and in microbial colonies ( Hallatschek et al . , 2007; Fusco et al . , 2016; Gralka et al . , 2016b; Korolev et al . , 2011 ) , which can serve as a useful model system because short generation times and ease of handling allow for quantitative investigations of the evolutionary dynamics of range expansions . In microbial colonies , nutrient gradients and mechanical effects limit the number of proliferating individuals to a small region close to the colony perimeter called the growth layer ( Grant et al . , 2014; Gralka et al . , 2016b; Warren et al . , 2019 ) . For mutations occurring inside the growth layer , most mutant offspring are concentrated in a relatively small number of enormously successful lineages that manage to remain at the front and 'surf’ on the expanding population wave ( Excoffier and Ray , 2008 ) . As a consequence , the evolutionary dynamics is different in microbial colonies compared with well-mixed populations . For instance , the clones ( that is the collection of the initial mutant’s offspring ) of spontaneous neutral mutations often reach much larger sizes ( Fusco et al . , 2016 ) , and existing beneficial variants can sweep to high frequency much faster in microbial colonies than in well-mixed populations ( Gralka et al . , 2016b ) . Conversely , deleterious mutations are predicted to remain at the population frontier for extended periods because genetic drift is strong at the front , which prevents selection from efficiently weeding out deleterious alleles ( Travis et al . , 2007; Burton and Travis , 2008; Lavrentovich et al . , 2016; Peischl et al . , 2013; Gralka et al . , 2016a ) . The quantitative outcome of the competition of selection and genetic drift in microbial colonies is determined by the local shape and roughness of the front ( Gralka et al . , 2016b; Farrell et al . , 2017 ) , which in turn is determined by microscopic details , such as cell-cell adhesion or cell morphology shaping the mechanical interactions between cells ( Kayser et al . , 2018b; Kayser et al . , 2018a; Giometto et al . , 2018 ) , although the direct mapping is typically unknown ( Farrell et al . , 2017 ) . The evolutionary effects of fluctuations at expanding microbial population fronts have been studied in depth , but these studies have focused only on fluctuations associated with the growth , division , and random motion of cells , whose strength may depend on intrinsic properties of the microbial species , in homogeneous environments . However , any range expansion will experience varying degrees of environmental heterogeneity , which can be viewed as a source of extrinsic noise . In macro-ecology , it has long been realized that environmental heterogeneity can dramatically alter the invasion dynamics of invasive species ( With , 1997; With , 2002 ) or the genetic diversity after macroscopic range expansion ( Wegmann et al . , 2006 ) . By contrast , how the evolutionary dynamics in microbial populations is affected by environmental heterogeneity in the form of , for example locally varying nutrient availability , temperature gradients , or imperfections in the growth substrate , has received much less attention . Experimental efforts have concentrated mostly on simple temporal and spatial gradients in antibiotic concentration , which have been shown to facilitate the emergence of resistance in shaken cultures ( Lindsey et al . , 2013 ) , microfluidic devices ( Zhang et al . , 2011 ) and on agar plates ( Baym et al . , 2016 ) , as predicted by theory ( Greulich et al . , 2012; Gralka et al . , 2017; Hermsen et al . , 2012; Hermsen , 2016; Hermsen and Hwa , 2010 ) . The effects of spatial heterogeneity on evolutionary dynamics in expanding microbial populations have been studied in experiments only with neutral alleles in fixed geometries , such as isolated obstacles creating 'geometry-enhanced’ genetic drift ( Möbius et al . , 2015; Beller et al . , 2018 ) . These studies have shown that obstacles obstructing locally the advance of the front can doom lineages that happen to lie on the blocked part of the expanding population , whereas unobstructed lineages close to the edge of the obstacles obtain a boost as they fill the vacant area behind the obstacle . Even for neutral alleles , however , not much is known about the evolutionary dynamics in more complex heterogeneous environments . Moreover , selection during range expansions can dramatically alter the population structure over just a few generations ( Gralka et al . , 2016b ) , but how the action of selection is affected by environmental heterogeneity has so far remained completely unexplored . Here , we study the impact of complex environmental heterogeneities on the growth and evolutionary dynamics of microbial colonies . To this end , we introduce plasmid loss in E . coli as a model system for spontaneous mutations with tunable growth rate effects whose clones can be tracked under the microscope . By growing colonies on solid substrates with a weakly patterned surface with random microscopic features much bigger than individual cells , but much smaller than the whole colony , we find that environmental heterogeneity can overpower selection such that even strongly beneficial mutations are unable to establish at rates higher than expected for neutral mutations . Using a minimal computer model of populations expanding in randomly disordered environments , we show that dramatic changes in the efficacy of selection can arise from a local 'pinning’ of the expansion front , whereby stretches of the front are slowed down on a length scale that depends on the structure of the environmental heterogeneity . This pinning focuses the range expansion into a small number of individuals with access to expansion paths , increasing the importance of chance and thus limiting the efficacy of selection . We expect these results to generalize to other spatially growing populations , such as biofilms , tumors , and invasive species , when the growing population front is transiently hindered by the local environment .
We grew colonies from single cells of a strain of E . coli MG1655 carrying a plasmid that is costly to produce , resulting in a 20% growth rate disadvantage s in plasmid-bearing cells compared to their plasmid-less ( but otherwise isogenic ) conspecifics ( Figure 1 ) . This strain loses the plasmid stochastically at a rate of about 5×10-3 per cell division ( approximately independent of antibiotic concentration , see Figure 1—figure supplement 1 ) . The plasmid codes for a fluorescence gene and confers resistance to the antibiotic doxycycline ( dox , a tetracycline analog ) such that when grown at increasing antibiotic concentrations , cells missing the plasmid grow more and more slowly , until they eventually grow more slowly than cells harboring the plasmid ( for [dox]>0 . 3μg/ml ) , despite the inherent cost associated with bearing the plasmid . This allowed us to treat plasmid loss effectively as a spontaneous mutation whose fitness effect s , that is the relative growth rate difference between the plasmid-bearing ( 'wild type’ ) and non-bearing ( 'mutant’ ) cells , we can finely tune from +20% to -15% by varying the amount of doxycycline in the growth media ( see Figure 2—figure supplement 1 ) , thus making the mutation either beneficial , neutral or deleterious . Since plasmid loss in our system is also coupled with a loss in fluorescence , we can easily detect mutant clones , that is the individual cell that incurred the mutation originally and its offspring , under the microscope as dark patches in the colonies ( see Figure 1 ) , allowing us to observe the evolutionary dynamics in real time . Our approach extends previous experimental model systems for evolutionary dynamics during microbial range expansion that employed either an initial mixture of wild-type and mutant cells ( Hallatschek et al . , 2007; Van Dyken et al . , 2013; Müller et al . , 2014; Gralka et al . , 2016b; Korolev et al . , 2012; Kayser et al . , 2018a ) or were confined to spontaneous neutral ( Fusco et al . , 2016 ) or deleterious ( Lavrentovich et al . , 2016 ) mutations . The ability to track spontaneous mutations in colonies grown from single cells is essential to ensure identical starting conditions in our experiments , allowing a quantitative comparison of the evolutionary outcomes between the two growth conditions . To investigate the impact of environmental heterogeneity on colony growth and adaptation dynamics , we grew the colonies on two different substrates ( Figure 1 ) : standard , 'smooth’ , agar plates as well as agar surfaces with random microscopic features , created by depositing filter paper onto melted agar and removing it after cooling and drying ( see Materials and methods ) . The resulting substrate had an average roughness ( i . e . , standard deviation of the substrate height ) of 10 μm with ridges and valleys much wider than individual cells ( about 15–30 μm ) , but small compared to whole colonies ( see Figure 1—figure supplement 2 for a detailed characterization of the rough substrates ) . Notably , the width of the valleys is comparable to the growth layer width of E . coli colonies ( Gralka et al . , 2016b ) , which is the fundamental length scale characterizing the range of mechanical interactions within microbial colonies ( Kayser et al . , 2018a ) . Colonies grew more slowly on these rough substrates compared to smooth substrates ( Figure 2a ) , but this disorder-induced reduction in radial growth rate was consistent between wild-type and mutant cells , such that their selective difference s , defined as the difference between colony expansion rates , normalized by the wild-type colony expansion rate , was independent of surface structure ( Figure 2b–c ) . Colonies grown on rough substrates ( hereafter called 'rough’ colonies ) also had a rougher front line ( see Figure 1d and Figure 1—figure supplement 3 ) than those grown on smooth substrates ( 'smooth’ colonies ) , and displayed branch-like outgrowths where the bacteria tended to colonize grooves in the agar surface much faster than the surrounding areas ( Figure 1e , arrows ) . These branches grew far ahead of the rest of the population , becoming visible at a width of about 20 μm ( consistent with the width of the valleys in the substrate ) , and broadened as they were incorporated into the bulk of the colony . This kind of growth pattern is reminiscent of the 'pinning’ phenomenon observed in the study of interfaces in heterogeneous media , such as the capillary rise of water or autocatalytic fluid interfaces in porous media ( Delker et al . , 1996; Atis et al . , 2015 ) , macro-ecological species invasions ( Keitt et al . , 2001 ) , or magnetic domains ( Lemerle et al . , 1998 ) . Pinning refers to the effect whereby certain regions of an expanding interface are slowed or even stopped entirely in their advance by heterogeneities , whereas other regions can advance unimpeded . Given the importance of the front morphology for evolutionary dynamics ( Gralka et al . , 2016b; Farrell et al . , 2017 ) , we hypothesized that by changing the growth patterns of rough colonies , the structured agar surface should also impact the dynamics of spontaneous mutations . The primary readout of our experiments is the final mutant frequency fMT in the colony and the number of surviving mutant clones ( sectors ) as a function of the selective advantage s that the mutation confers . These measures are proxies for the degree of adaptation of the population during the expansion process and the success probability of spontaneous mutations in shaping the composition of the population , respectively . Thus , our system gives us direct access to population genetic measures of interest . Alternatively , one can measure the frequency of mutants at the colony perimeter , which has a more direct influence on the future genetic composition of the population at the front . However , this measure can quickly become uninformative for beneficial mutations as mutants overtake the whole perimeter , and it is often difficult to measure accurately because of the low fluorescence signal at the front . The mutant frequency averaged over the whole population still gives a good , if conservative , measure of the mutant frequency at the perimeter . On smooth substrates , in accord with previous experimental results ( Gralka et al . , 2016b; Korolev et al . , 2012 ) , advantageous mutants increased in frequency fMT rapidly as the colony grew ( see Figure 2—figure supplement 2 for an analysis of the mutant dynamics over time ) : for s=0 . 2 , mutants made up roughly half of the total population ( Figure 2d ) after 72 hrs . At higher antibiotic concentration , mutants became first neutral and eventually deleterious ( Figure 2—figure supplement 1 ) , and accordingly , the final mutant frequency was lower , decreasing approximately exponentially with the fitness cost s ( Figure 2—figure supplement 4 ) such that strongly deleterious mutants made up only a small fraction of the final population . On rough substrates , deleterious and neutral mutants remained at frequencies comparable to those observed in smooth colonies at the corresponding values of the fitness ( dis ) advantage s . However , in contrast to smooth colonies , beneficial mutants in rough colonies did not increase in frequency with s relative to the neutral case . This finding is surprising , given that the growth rate advantages of mutant over wild-type colonies were the same on rough and smooth substrates ( Figure 2c ) . Thus , this apparent inefficacy of selection in affecting evolutionary outcomes was not caused by an altogether elimination of growth rate differences . Instead , a closer look at the colony growth dynamics on rough substrates , shown in Figure 1e , suggests a different mechanism: the surface structure of the rough substrate constrains and guides growth along predetermined paths where growth proceeds faster than in the immediate surroundings , such that any mutation can only be successful , that is establish a sector and thus rise to high frequency in the population , if it happens to arise in one of the branch-like regions of accelerated growth . Conversely , a beneficial mutant clone will be unable expand even if its per capita growth rate is higher than its wild-type neighbors if the mutation occurs in a portion of the front that is slowed down by the environmental heterogeneity . If this proposed mechanism is indeed the root cause for the apparent inefficacy of selection on rough substrates , then one would expect the number of successful mutants , that is those that manage to establish sectors , to be independent of the mutant’s selective advantage s . Indeed , this is what we observed: the probability of forming a sector , which quantifies the evolutionary success probability of individual mutations , increased with s in smooth colonies ( Figure 2e ) , but was constant in rough colonies as long as s>0 . Notably , the establishment probability was extremely low in both scenarios: Even for the most advantageous mutants in smooth colonies , we estimate u∼10-7 per mutation making evolutionary success an extremely rare event . The low success probability is a consequence of two processes: firstly , the mutation must occur in a favorable location , namely in the first layer of cells at the front of the population ( Gralka et al . , 2016b ) , which reduces the number of mutations eligible for sector formation by a factor of about 1000 ( assuming a growth layer width of 10 cells and a colony height of about 100 cells , see Figure 2—figure supplement 3 ) . We estimate that about 2000 mutations per colony arose in favorable positions , each of which had an establishment probability of order 10-3 . Secondly , each eligible mutation has to survive genetic drift , which in microbial colonies is manifest in the random fluctuations in the sector boundaries as a consequence of stochastic cell growth and division , and subsequent cell motion due to mechanical pushing of cells on each other ( Hallatschek et al . , 2007; Farrell et al . , 2017; Kayser et al . , 2018a ) . The low establishment probability means that most mutations will not manage to create sectors , but instead they will form so-called bubbles , that is individual mutant clones that have lost contact with the front . We extracted the size of mutant clones , both bubbles and sectors , by measuring the individual areas of non-fluorescent patches in the colony micrographs . The resulting clone size distribution P ( X>x ) ( clone areas normalized by colony area , shown in Figure 3 ) is related to the site frequency spectrum in population genetics , where it can be used to predict rare evolutionary outcomes such as fitness valley crossing ( Weissman et al . , 2009 ) and evolutionary rescue ( Fusco et al . , 2016 ) , and is well understood for toy models of microbial colonies ( Fusco et al . , 2016; Otwinowski and Krug , 2014 ) . For neutral mutations , the clone size distribution is expected to be broad up to a shoulder indicating the typical size of the largest expected bubble . In our experiments , we indeed observed a broad shoulder-like distribution for neutral mutations , consistent with earlier experiments using population sequencing ( Fusco et al . , 2016 ) . In smooth colonies , beneficial mutations created a larger number of bulging sectors , leading to an even broader distribution with maximum clone sizes of almost half the population , while the distribution for strongly deleterious mutations was cut off at small clone sizes . This clone size distribution is consistent with our initial observation that a larger selective advantage s gave rise to a larger overall mutant frequency , but it also shows that even at the largest s≈0 . 2 , most mutant clones remained small , with more than half of the visible clones reaching frequencies of at most 1% . By contrast , the clone size distributions obtained from rough colonies were virtually indistinguishable for all s>0 , whereas we observed the same cut-off for large clones for deleterious mutations . In summary , a microscopically randomly patterned growth surface had several effects on the population and evolutionary dynamics of our colonies . The heterogeneity of the substrate decreased the radial growth rate during early colony growth and gave rise to colonies with an overall rougher morphology . In terms of evolutionary dynamics , heterogeneity decreased the dependence of the final mutant frequency ( or , equivalently , the rate of adaptation ) on the selective effect of mutations . These effects are large , even though the perturbation we impose seems relatively weak . After all , the rough substrate is only distinguished from the smooth substrate by troughs and elevations of order ten micrometers , and colony growth rate differences are consistent for both substrate types ( Figure 2c ) . How can such a relatively small change in environmental conditions have such a dramatic effect on the evolutionary dynamics ? Above , we have proposed that the transient colony pinning seen in Figure 1e may be responsible , by giving a boost to certain regions irrespective of whether that region harbors beneficial mutants or not . However , the growth of the colony and the mutational dynamics within it are highly complex processes affected by the mechanical properties of growing cells and their interactions with each other and the growth substrate ( Grant et al . , 2014; Boyer et al . , 2011; Kayser et al . , 2018a; Giometto et al . , 2018; Farrell et al . , 2017 ) . In addition , the substrate heterogeneity in our experiments is complex and characterized by long-range correlations ( see Figure 1—figure supplement 2 ) . This raises the question whether our key findings may hinge on these complexities , or whether much simpler uncorrelated heterogeneities in growth rates can also have comparable effects on population genetics . To answer this question , we have devised a minimal simulation model for populations expanding in heterogeneous environments . Briefly , colonies grow from single cells on a square lattice , only cells with empty neighbors can divide , and a wild type can mutate upon cell division with probability μ to the mutant type carrying a fitness advantage or disadvantage s ( Materials and methods ) . Disorder sites ( density ρ ) confer a reduced growth rate k ( 0≤k<1 ) to any individual growing on it . We call k the transparency of the disorder sites; for k=0 , we refer to the disorder sites as ( impassable ) obstacles . The simplicity of the model allows us to explore exhaustively the whole parameter space in k and ρ . Our model is based on the classical Eden lattice model ( Eden , 1961 ) that is commonly used to model growing microbial colonies ( Gralka et al . , 2016b; Fusco et al . , 2016; Ben-Zion et al . , 2019 ) . The Eden model without environmental heterogeneity is in the so-called KPZ universality class , that is its statistical properties are described the KPZ equation , a classical model of interface growth ( Kardar et al . , 1986 ) . The KPZ equation has been extended to include environmental heterogeneity ( discussed in detail in the Materials and methods ) which was shown to induce a pinning transition: the environmental heterogeneity induces a characteristic length scale on which the interface cannot advance ( is pinned by the heterogeneities ) . Thus , adding environmental heterogeneity to the Eden model may likewise introduce a pinning transition under certain conditions , making our generalized Eden model a natural candidate for a minimal model of range expansions in heterogeneous environments . Indeed , in agreement with our experiments , increasing the density ρ of disorder sites leads to a decrease in the radial colony expansion speed in our simulations ( Figure 4b ) that becomes more extreme as the obstacle transparency k goes to 0 . For small k≪1 , the expansion speed decreases first slowly and then rapidly as the density reaches a critical value ρc≈0 . 4 . For impassable obstacles ( k=0 ) at densities ρ>ρc , the obstacles form a closed ring around the incipient colony and prevented further growth ( Figure 4b , black line ) . This is the anticipated pinning transition , which occurs in our model at a critical density ρc≈0 . 4 . This critical density corresponds to the scenario where the colony can no longer percolate through the network of obstacles , suggesting that ρc is equivalent to the site percolation threshold 1-0 . 592…=0 . 407… ( Bunde et al . , 1985; Barabási and Stanley , 1995 ) . Notably , while the percolation transition only occurs at the critical point k=0 , non-zero but small values of k give rise to transient pinning near the critical density that is essentially indistinguishable from the k=0 case over time scales shorter than 1/k , while still allowing the expansion to progress indefinitely , albeit slowly . Close to the pinning transition , small changes in obstacle density can have dramatic effects: not only does the colony expansion speed decrease rapidly , but the colony morphology changes drastically and , as we show below , so do the evolutionary dynamics . As a result of the local pinning of the colony interface , the colony morphology depends on the density of obstacles ( Figure 4c ) , most drastically for impassable obstacles ( k=0 ) on which we concentrate for now . Without obstacles , the colonies are compact and relatively smooth . At intermediate obstacle densities ( Figure 4c , middle ) , colonies are punctured by small holes and the overall density of the colony decreases . At the critical density ρc the colony is characterized by the fragmented morphology of percolation clusters with a large number of holes and a very rough exterior ( see Figure 4—figure supplement 1 and Materials and methods for a quantitative analysis of the colony interfaces ) . Below the critical obstacle density , the interface is pinned locally over a length scale that depends on the proximity to the pinning transition; the whole interface becomes pinned when this length scale reaches the system size , whereas the interface is unaffected on length scales much larger than this pinning length ( see Theory ) . An example of local pinning is shown in Figure 4d , where the interface can only advance when the individuals located in unpinned portions of the front grow into the pinned areas ( indicated by arrows ) . This process is equivalent to the branch-like outgrowths in the experiments ( Figure 1e ) which correspond to unpinned portions of the front outgrowing the pinned areas . In the following , we show how the changes in colony morphology produced by the environmental heterogeneity affect the evolutionary dynamics . We begin by replicating the experimental situation to assess the efficacy of selection in the presence of environmental heterogeneity . We simulated mutations conferring a selective advantage s ( that is increasing the growth rate by a factor 1+s ) , shown in Figure 5 . Transparent obstacles ( k=0 . 1 , Figure 5a ) only have a relatively mild effect on the mutant frequency fMT . For any value of the obstacle density ρ , fMT increases roughly exponentially with s , but the dependence on obstacle density is non-monotonic . This is easiest to see for the most beneficial mutations: the final mutant frequency fMT at s=0 . 2 is lower at intermediate ρ≈0 . 5 than at the extremes ρ≈0 or ρ≈1 . Intuitively , this non-monotonicity results from a symmetry between high and low ρ: in both cases , there is only a small fraction of sites of the 'other’ type ( i . e . , disorder sites at low ρ or regular sites at high ρ ) , and their density is too small to effect a strong change in the population genetics . The reduction in the sensitivity of fMT to s at intermediate ρ becomes much more dramatic as the obstacle transparency k is decreased ( see Figure 5b for the extreme case k=0 ) . A similar pattern is found for the final frequency of neutral mutants ( s=0 ) , which is largest at intermediate ρ . To quantify the effects of varying k and ρ and summarize the simulation results over many parameter combinations , we introduce the selection efficacy ks and the neutral diversity f0 by parametrizing the mutant frequency with an exponential function fMT ( s ) =f0ekss . Although this choice is merely a heuristic , rescaling the mutant frequency curves for a range of values of ρ and k by the fitted values for the neutral diversity f0 and the selection efficacy ks , all points fall close to a single master curve given by a simple exponential ( Figure 5f ) . The selection efficacy ks and the neutral diversity f0 have a minimum and maximum near ρc , respectively , which is increasingly sharp as k approaches the critical point k=0 . At the critical point , the selection efficacy vanishes entirely as ρ approaches ρc ( Figure 5c , d ) . Thus , selection is completely unable to affect the final mutation frequency as the critical point is approached . The virtual independence of the evolutionary dynamics of the per-capita fitness s holds even at the scale of individual clones , whose size distributions for different values of s are practically indistinguishable for obstacle densities near ρc ( Figure 5—figure supplement 1 ) . Importantly , while we find a proper phase transition in the evolutionary dynamics only at the critical point ( k=0 ) , the percolation transition is also manifest in populations grown in generic heterogeneous environments with k>0 which do not give rise to a percolation transition . As a consequence , tiny changes in environmental parameters near a non-trivial critical obstacle density ρc can have a dramatic effect on the population growth dynamics and colony morphology as well as its evolutionary dynamics . The close connection between colony morphology and evolutionary dynamics is underscored by the empirical observation that the two descriptive parameters , the selection efficacy ks and the neutral diversity f0 , introduced as independent parameters measured directly from the simulations , are not independent in practice ( Figure 5e ) . Plotting ks vs . f0 for various choices of k and ρ reveals that the two parameters represent two sides of the same coin: environmental heterogeneity alters the growth pattern of the colony , which in turn affects both the neutral diversity and the selection efficacy . Our minimal model has shown that simple uncorrelated heterogeneities can have a strong impact on the evolutionary dynamics and morphology of the range expansions . Remarkably , even incomplete obstacles that merely slow down growth can generate large roughness and quasi-neutral population genetics when they are at intermediate density . The reason is that , because the positioning and dynamics at the front are so important in range expansions , a slowdown has similar effects to a complete halt . Regarding environmental heterogeneity as simply another source of ( extrinsic ) noise , it is perhaps not surprising from a classical population genetics perspective that the addition of noise effectively weakens selection , as other sources of noise , such as small population sizes , are known to push evolutionary dynamics towards the neutral limit ( Gillespie , 2004 ) . However , as we show in the following , the environmental heterogeneity in our simulations changes the evolutionary dynamics on a fundamental level that is not consistent with a mere increase in total noise level . Consider the neutral diversity f0 in Figure 5e , which corresponds to the rate at which neutral mutations accumulate in the population . On average and in the absence of environmental heterogeneity , this rate is μ ( N/π ) 1/2 since a fraction μ of cells at the population front acquire new neutral mutations in every generation , and the front size scales as the square root of the population size N ( Fusco et al . , 2016 ) . Importantly , this result is independent of the level of noise in the system since it concerns only the average over many populations . Since the population size is the same across all our simulations , we would expect the same neutral diversity for all parameter values ρ and k . By contrast , our simulations show clear systematic deviations from the expected value ( dotted line in Figure 5e ) . To further characterize the qualitative changes on the neutral dynamics induced by environmental heterogeneity , we computed the spatially resolved phylogenetic tree of the population , obtained by tracing the lineages of all individuals at the population front back to the origin , focusing on the extreme cases of no ( ρ=0 ) and critical ( ρ=ρc ) obstacles ( i . e . , k=0 ) . For intermediate ρ , there is a crossover length set by the obstacle density . On length scales much shorter than this crossover length , the dynamics correspond to that in heterogeneous environments , while the homogeneous dynamics are recovered on length scales much larger than the crossover length ( see Theory for details and Figure 4—figure supplement 1b for an illustration of the crossover length at intermediate ρ ) . As shown in Figure 6a , b , the lineage tree has a vastly different appearance depending on the environmental heterogeneity . Without obstacles ( panel a ) the lineages are relatively straight and roughly aligned with the radial direction . By contrast , at the critical obstacle density , where the colony has a rough exterior ( Figure 4c ) , lineages are erratic and often have segments oriented perpendicular to the radial direction . To quantify the differences in the lineage structure between the two scenarios , we focused on the pair coalescence 'time’ T2 , that is the time when two individuals , sampled a distance Δx apart , had their last common ancestor ( measured in lattice sites , see Figure 6c ) , as well as the strength of lineage fluctuations , that is how much lineages deviate from straight lines over time . The strength of lineage fluctuations determines how likely two randomly lineages are to meet ( 'coalesce’ ) and thus directly shapes the coalescence structure of the population ( Korolev et al . , 2010; Chu et al . , 2018 ) . The lateral lineage fluctuations l⟂ ( see Materials and methods for details ) scale with distance t from the origin as l⟂∼tξ , where a larger value of ξ indicates rougher boundaries . We find that lineages in rough colonies are not only rougher in absolute value , but also in terms of their scaling in rough colonies . Whereas in the standard Eden model we recover the known ( super-diffusive ) scaling ξ=0 . 66±0 . 006 ( Kardar et al . , 1986 ) , we find a larger scaling exponent ξ=0 . 86±0 . 006 in rough colonies ( ρ=0 . 4 ) . This is consistent with the corresponding change of the statistical properties of the colony interface , which transitions from the Kardar-Parisi-Zhang ( KPZ ) universality class to the quenched Edwards-Wilkinson ( QEW ) universality class ( see Materials and methods and Figure 4—figure supplement 1 ) . The changes in lineage fluctuations in the presence of environmental heterogeneity are reflected in the shape and orientation of individual neutral clones ( Figure 6—figure supplement 1 ) . Mutant clones have an approximately ellipsoidal shape oriented preferentially along the radial direction in the absence of heterogeneity , whereas they have essentially random orientations in rough colonies ( Figure 6—figure supplement 1c ) , in agreement with the observation that lineages lose their radial orientation as the number of obstacles increased . Similarly , we measured the scaling of clone widths l⟂ with its length l∥ as l⟂∼l∥ζ , where the exponent ζ quantifies the anisotropy of the clones ( ζ=0 corresponding to clones whose width is independent of their length , ζ=1 corresponding to isotropic clones ) . In our simulation , ζ changed from ζ=2/3 for ρ=0 ( consistent with KPZ interface statistics [Fusco et al . , 2016] ) to ζ≈0 . 95 for ρ=ρc ( Figure 6—figure supplement 1d ) , indicating roughly isotropic neutral clones . The increased roughness of lineages is also reflected in the number of successful lineages emanating from the initial population founder . We quantify this by computing the pairwise coalescence time T2 ( Figure 6e , f ) over all pairs of cells at the front of the population . We find that , for a given distance Δx between the sample pairs , the relative coalescence time and persistence probability ( i . e . , the probability of not having a common ancestor until time T2 , shown in panel f ) is always smaller in the presence of obstacles . This indicates that fewer lineages reach the population edge in the presence of environmental heterogeneity . This makes intuitive sense from the phylogenetic trees shown in Figure 6a , b , where in rough colonies all individuals at the front coalesce quickly into a small number of large lineages . Thus , for any mutation to be successful and grow into a large clone , it has to belong to one of those large lineages . Since the number of those lineages is small , such mutations are extraordinarily unlikely; at the same time , if they occur , they can grow to large size simply by virtue of having occurred in a fortuitous location .
Using a combination of bacterial colony experiment and population genetics theory , we examined how environmental heterogeneity can shape the population genetics of range expansions . By growing colonies on heterogeneous substrates , we found that microscale ridges and troughs in the growth substrate were enough to dramatically alter the growth dynamics and morphology of the colonies as well as reduce the ability of beneficial mutations to establish and expand . Time-lapse microscopy and minimal model simulations showed that this reduction in selection efficacy on heterogeneous substrates can be explained by a local pinning of the colony front . Since mutations occur only within the growing population at the front , the properties of the front dictate the evolutionary dynamics , including the strength of selection and the size of individual clones . Local pinning impacts the dynamics at the front by reducing the expansion speed of some parts of the population , leading to an effective reduction in the number of expansion paths that can actively contribute successful mutations . Thus , most individuals and their clones will get stuck in dead-ends , and only a few lucky individuals will be able to find the paths along unpinned front positions and be able to establish a large clone . Once established , though , the size of a mutant clone is roughly independent of its fitness because it is constrained by the network of obstacles . Thus , the evolutionary success of a mutation , that is whether a sector can form or not , and how large a mutant clone can get , depends entirely on where it arises and not at all on its fitness . In this sense , locally pinned expansions bear little resemblance to unconstrained radial expansions ( see Figure 7 ) . Rather , expansions along each available path more closely correspond to linear expansions , for example along a coastline , where mutations spread deterministically after local establishment ( Fisher , 1937 ) . While we have focused here on microbial populations , we expect these results to hold more generally . This is one of the conclusions of our deliberately minimal model , which showed that simple uncorrelated heterogeneities are enough to create a strong impact on the population genetics and morphology of the range expansions . These effects persist even when the heterogeneity does not present as impassable obstacles but merely slows down growth . This is because the properties of the front dictate the evolutionary dynamics , since mutations occur only within the growing population at the front , such that a slowdown have comparable effects on the evolutionary dynamics as a complete halt . Since our model does not include any particular biological details , its results may apply generally in systems with front-limited growth and environmental heterogeneity . Thus , we would expect a reduced selection efficacy to generalize to other dense cellular populations in disordered environments , such as tumors and biofilms , but also to macroscopic range expansions of invasive species ( With , 2002 ) . When a population undergoes a range expansion , it will arguably not experience a completely homogeneous environments: at the very least , some areas will be more hospitable than others , but other parts of the environment may also be entirely inaccessible to the population because of , for example rivers and lakes , a strong local competitor or predator , or lack of resources . Environmental heterogeneity is thus arguably the rule rather than the exception . Despite its simplicity , our minimal model reproduces many experimental findings qualitatively , such as a rougher colony morphology and a reduced efficacy of selection in the presence of environmental heterogeneity ( in experiments , by a factor of 2 . 5 - 4 depending on fitting strategy ) . However , the model cannot quantitatively account for all our experimental results . For instance , the model predicts that the frequency of neutral and deleterious mutants should be greater in heterogeneous than in homogeneous environments ( Figure 5d ) . By contrast , in our experiments , we find only about half as many neutral mutants in rough colonies as in smooth colonies and comparable mutant frequencies for deleterious mutations ( Figure 2d ) . A potential reason for this discrepancy may be that the random pattern imposed in our experiments is not correlation-free as in the simulations ( Figure 1—figure supplement 3 ) , which may impact the dynamics of mutant clones as follows . A beneficial mutation has to overcome genetic drift , and to do so , it must grow to a lateral size large enough for selection to take over ( Gralka et al . , 2016b ) . However , if the characteristic length scale of the environmental heterogeneity is smaller than this 'establishment size’ , then the evolutionary dynamics is effectively neutral . On the other hand , a deleterious mutation born on a ridge or in a trough never grows to large enough size to 'see’ the disorder in the first place and thus its dynamics are largely unaffected by the environmental heterogeneity . In the simulations , however , the heterogeneity can be felt on all length scales , such that it affects mutant clones of all sizes the same way . We expect environmental heterogeneity to impact not only the fate of beneficial mutations . Since deleterious mutations are typically more numerous than beneficial ones , environmental heterogeneity may also increase the chances of an overall decrease in population fitness through the accumulation of deleterious mutations . The accumulation of deleterious mutation , called the expansion load , is already more likely in range expansions than in well-mixed populations ( Peischl et al . , 2013; Lavrentovich et al . , 2016; Gralka et al . , 2016a ) . By altering the efficacy of selection in depressing deleterious mutations and elevating beneficial mutations , heterogeneities in the environment may further facilitate the accumulation of deleterious mutations . Depending on the mutational supply , that is the relative rate and effect of deleterious and beneficial mutations , environmental heterogeneity may not only slow down the process of adaptation but also lead to entirely different long-term evolutionary outcomes . As an example , consider Figure 8 , where we compute the rate of adaptation in range expansions in various habitats for a given distribution of fitness effects ( DFE ) . We find that the rate of adaptation can transition from positive ( adaptation over time ) to negative ( accumulation of deleterious mutations ) depending only on the degree of environmental heterogeneity . Thus , environmental heterogeneity can fundamentally alter the evolutionary dynamics of range expansions . | Throughout evolution , countless populations have expanded their territories by invading new areas . This invasion can happen on the scale of kilometers and millennia – such as when humans migrated out of Africa – or millimeters and months , such as the growth of a population of cells in a solid tumor . During this expansion , mutations can occur that can either increase or decrease fitness in the new territory . If a favorable mutation occurs at the edge of the population , then it has plenty of room to expand . Such a mutation has a high chance of becoming established , and so it can have a very strong impact on the genetic makeup of the population . This increase in evolutionary advantage in individuals at the front is called “gene surfing” . This phenomenon is well known in populations living in “homogeneous” territories , where the new space a population is invading is more or less the same as the one they already occupy – think of the endless flat grasslands of the Siberian steppes . But in reality , many territories are not like that . What happens if the new territory is not completely homogeneous ? For instance , if a species’ expansion is impeded by a mountain range with a series of valleys . Gralka and Hallatschek investigated how such changes in landscape could affect phenomena like gene surfing . Experiments using E . coli as a model system and computer simulations showed that a varied environment – such as roughness akin to a mountain range and valleys , but on a bacterial scale – had a strong influence on the fate of mutations arising in a population . Whether the environment is favorable for expansion or not in the place where the mutation happens becomes much more important than how the mutation itself affects fitness . So , if a beneficial mutation occurs at a cliff-edge , it is not likely to get far . But if it happens at a population edge by the bacterial equivalent of gently rolling hills , there is a much better chance of the mutation taking hold . The findings suggest that the amount a population can adapt during expansion is limited , and it can even lead to the spread of harmful mutations in a population if they occur in just the right spot . Piecing together these scenarios is important in order to accurately infer the evolutionary history of a species based on mutations present in its genome now . This type of knowledge can also be useful in developing new treatments for cancers , making use of these evolutionary processes to slow or halt a tumor’s expansion . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"evolutionary",
"biology",
"physics",
"of",
"living",
"systems"
] | 2019 | Environmental heterogeneity can tip the population genetics of range expansions |
Liver metabolism follows diurnal fluctuations through the modulation of molecular clock genes . Disruption of this molecular clock can result in metabolic disease but its potential regulation by immune cells remains unexplored . Here , we demonstrated that in steady state , neutrophils infiltrated the mouse liver following a circadian pattern and regulated hepatocyte clock-genes by neutrophil elastase ( NE ) secretion . NE signals through c-Jun NH2-terminal kinase ( JNK ) inhibiting fibroblast growth factor 21 ( FGF21 ) and activating Bmal1 expression in the hepatocyte . Interestingly , mice with neutropenia , defective neutrophil infiltration or lacking elastase were protected against steatosis correlating with lower JNK activation , reduced Bmal1 and increased FGF21 expression , together with decreased lipogenesis in the liver . Lastly , using a cohort of human samples we found a direct correlation between JNK activation , NE levels and Bmal1 expression in the liver . This study demonstrates that neutrophils contribute to the maintenance of daily hepatic homeostasis through the regulation of the NE/JNK/Bmal1 axis .
Circadian rhythms regulate several biological processes through internal molecular mechanisms ( Dibner et al . , 2010 ) and the chronic perturbation of circadian rhythms is associated with the appearance of metabolic syndrome ( Kolla and Auger , 2011 ) . This homeostasis is closely dependent on the circadian system in the liver , which shows rhythmic expression of enzymes associated with glucose and lipid metabolism ( Haus and Halberg , 1966; North et al . , 1981; Tahara and Shibata , 2016 ) . Moreover , mice with mutations in clock genes encoding nuclear receptors have impaired glucose and lipid metabolism and are susceptible to diet-induced obesity and metabolic dysfunction , consistent with the idea that these genes control hepatic metabolic homeostasis ( Delezie et al . , 2012; Kudo et al . , 2008; Lamia et al . , 2008; Rey et al . , 2011; Tong and Yin , 2013; Turek et al . , 2005; Yang et al . , 2006 ) . Besides , recent reports have shown that hepatic physiology follows a diurnal rhythm driven by clock genes , with expression of proteins involved in fatty acid synthesis higher in the morning while those controlling fatty acid oxidation are higher at sunset ( Toledo et al . , 2018; Zhou et al . , 2015 ) . Blood leukocyte levels also oscillate diurnally , as does the release of hematopoietic stem cells and progenitor cells from the bone marrow ( BM ) ( Haus and Smolensky , 1999; Lucas et al . , 2008; Méndez-Ferrer et al . , 2008 ) and their recruitment into tissues ( Adrover et al . , 2019; He et al . , 2018; Scheiermann et al . , 2012 ) . Oscillatory expression of clock genes in peripheral tissues is largely tuned by the suprachiasmatic nucleus ( Dibner et al . , 2010; Druzd and Scheiermann , 2013; Huang et al . , 2011; Reppert and Weaver , 2002 ) ; however , the potential regulation of daily rhythms of specific tissues by immune cells remains largely unexplored , both in steady state and during inflammation . Although the molecular mechanisms linking circadian rhythms and metabolic disease are largely unknown , several studies have demonstrated a strong association between leukocyte activation and metabolic diseases ( McNelis and Olefsky , 2014 ) . A prime example is the BM , where engulfment of infiltrating neutrophils by tissue-resident macrophages modulates the hematopoietic niche ( Casanova-Acebes et al . , 2013 ) . The circadian clock is dysregulated by obesity ( Kohsaka et al . , 2007; Xu et al . , 2014 ) , and recent studies suggest that liver leukocyte recruitment and migration show a circadian rhythm ( Scheiermann et al . , 2012; Solt et al . , 2012 ) whose alteration can result in steatosis ( Solt et al . , 2012; Xu et al . , 2014 ) . Neutrophils are key factors in steatosis development ( González-Terán et al . , 2016; Keller et al . , 2009; Mansuy-Aubert et al . , 2013; Nathan , 2006 ) and show diurnal oscillations in their recruitment and migration to multiple tissues ( Scheiermann et al . , 2012; Solt et al . , 2012 ) . Here , we demonstrate that circadian neutrophil infiltration into the liver controls the expression of clock genes through the regulation of c-Jun NH2-terminal kinase ( JNK ) and the hepatokine fibroblast growth factor 21 ( FGF21 ) , driving adaptation to daily metabolic rhythm .
Virtually all cell types have an internal clock that controls their rhythmicity through the periodic expression of clock genes ( Robles et al . , 2014; Tahara and Shibata , 2016 ) . However , it is unknown how these multiple cell rhythms are integrated . The liver is an essential metabolic organ that controls body glucose and lipid homeostasis ( Manieri and Sabio , 2015 ) , and neutrophil infiltration alters its function ( González-Terán et al . , 2016 ) . We hypothesized that the metabolic cycles in the liver might be entrained by rhythmic neutrophil infiltration . To test this , we harvested liver , BM , and blood from C57BL6J mice at 4 hr intervals over a 24 hr period . Liver neutrophil infiltration showed a clear diurnal pattern , with a peak at ZT2 , coinciding with liver-driven lipogenesis in mice ( Zhou et al . , 2015 ) , and a nadir during the night , at ZT14 ( Figure 1A ) , correlating with lipolysis ( Zhou et al . , 2015 ) . These oscillations corresponded directly to changes in neutrophil numbers in blood ( Figure 1—figure supplement 1A ) , suggesting that liver infiltration might result from higher neutrophil migration to the liver . We first confirmed that neutrophils were infiltrated in the liver using 3D microscopy . According to published data ( Casanova-Acebes et al . , 2018 ) , infiltrated neutrophils presented an intrasinusoidal distribution in the liver , different to that observed in the Kupffer cells population ( Figure 1B and Figure 1—figure supplement 1B ) . Then we evaluated whether myeloid chemokines could be involved in circadian neutrophil recruitment into the liver . Analysis of liver lysates indicated that the expression of the hepatocyte-derived neutrophil chemoattractant Cxcl1 ( Su et al . , 2018 ) was higher at ZT2 than a ZT14 . Moreover , mRNA of Cxcl1 in liver samples showed the same oscillation pattern than infiltrated neutrophils , suggesting that this chemokine may be important in the regulation of the neutrophil diurnal cycle ( Figure 1—figure supplement 1C ) . The infiltration pattern correlated with liver expression levels of the clock-gene Bmal1 , peaking at ZT2 and bottoming at ZT14 ( Figure 1C ) . Infiltration also correlated inversely with the expression of Nr1d2 ( encoding Rev-erb β ) , Per2 , and Cry2 ( Figure 1C ) , which are important proteins in the control of circadian rhythms ( Reppert and Weaver , 2002 ) , consistent with the feedback loop that controls their expression . Bmal1 is thought to induce lipogenesis ( Zhang et al . , 2014 ) , whereas Nr1d2 controls lipid metabolism and its reduced expression promotes lipogenesis and steatosis ( Delezie et al . , 2012; Solt et al . , 2012 ) . In agreement with these studies , liver triglycerides were higher at ZT2 than at ZT14 ( Figure 1D ) . Our results show a correlation between neutrophil infiltration , hepatocyte Bmal1 expression , and lipid metabolism regulation , raising the possibility that neutrophils signal to hepatocytes to modulate the expression of circadian genes . Exposure of mouse hepatocytes in vitro to freshly isolated neutrophils increased hepatocyte expression of the clock genes Bmal1 and Clock . In contrast , no effect was observed upon exposure to T or B lymphocytes , or macrophages , suggesting the existence of a neutrophil-to-hepatocyte communication that controls hepatocyte clock-gene expression ( Figure 1E and Figure 1—figure supplement 1D ) . We then investigated whether neutrophil elastase ( NE ) , a proteolytic enzyme reported to regulate liver metabolism , could regulate hepatocyte clock genes ( Mansuy-Aubert et al . , 2013; Talukdar et al . , 2012 ) . Exposure to elastase reproduced the same increase in hepatocyte Bmal1 and Clock expression in contrast with another protease that did not affect Bmal1 expression ( Figure 1F and Figure 1—figure supplement 1D ) . Next , neutrophil-mediated regulation of liver clock-gene expression in vivo was investigated using a previously characterized genetic model of neutrophil deficiency ( Dzhagalov et al . , 2007; Steimer et al . , 2009; Figure 1—figure supplement 1E , F and Figure 1—figure supplement 2A–C ) . Low hepatic neutrophil infiltration in neutropenic mice correlated with reduced expression of Bmal1 and Clock ( Figure 1G ) and increased expression of Cry2 and Per2 at ZT2 ( Figure 1G ) . These changes in clock-gene expression were accompanied by lower liver triglyceride levels ( Figure 1H ) . Furthermore , lack of neutrophils perturbed the diurnal rhythmicity in Bmal1 , Clock , and Per2 expression in the liver without affecting clock genes in other organs such as the lung , in which there is no correlation between the peak of neutrophil infiltration and Bmal1 expression ( Figure 1—figure supplement 2D , E ) . Our results thus indicate that neutrophils might specifically control the expression of hepatocyte circadian clock genes in steady state . Chronic jet lag alters liver circadian genes and disrupts liver metabolism ( Kettner et al . , 2016 ) . Analysis of a mouse model of jet lag revealed complete disruption of the circadian liver neutrophil infiltration with increased hepatic neutrophil infiltration even at ZT14 ( Figure 2A ) . Abolition of rhythmic neutrophil hepatic infiltration under jet lag correlated with increased steatosis and high levels of liver triglycerides ( Figure 2B ) . To evaluate whether the metabolic effect of circadian perturbation was caused by the increased neutrophil infiltration , we exposed neutropenic and control mice to the jet lag protocol ( Figure 2—figure supplement 1A , B ) . Jet lag-induced steatosis was less severe in neutropenic mice ( Figure 2C ) , and disruption of diurnal liver expression of Bmal1 detected in control jet-lagged mice was partially ablated in neutropenic mice ( Figure 2D ) . Similar results were also observed in mice with impaired neutrophil migration such as Cxcr2MRP8-KO BM transplanted mice ( Eash et al . , 2010; Mei et al . , 2012 ) and p38γ/δLyzs-KO mice ( González-Terán et al . , 2016 ) . In both models , the reduction of neutrophil infiltration correlated with decreased levels of liver Bmal1 expression and protection from jet lag-induced steatosis ( Figure 2—figure supplement 1C–G ) . These results are consistent with the role of neutrophils in the control of liver clock genes . Inflammation plays a key role in the pathogenesis of non-alcoholic fatty liver disease ( Tiniakos et al . , 2010 ) and the development of hepatic steatosis is associated with increased liver infiltration by myeloid cells , particularly neutrophils ( González-Terán et al . , 2016; Mansuy-Aubert et al . , 2013; Talukdar et al . , 2012; Tiniakos et al . , 2010 ) . Two widely used mouse models of hepatic steatosis , high-fat diet ( HFD ) and methionine-choline-deficient ( MCD ) diet , increased liver neutrophil infiltration in WT mice at ZT2 , ZT14 , and ZT18 ( Figure 2E , F ) . Consistent with a neutrophil-to-hepatocyte communication in the regulation of hepatocyte clock genes , the MCD diet enhanced Bmal1 expression and inhibited Cry2 and Per2 expression in control mice , but not in neutropenic mice at ZT2 ( Figure 2G ) . Altered liver clock-gene regulation in neutropenic mice was associated with protection against steatosis and lower liver triglycerides ( Figure 2H ) . To confirm the role of neutrophils in modulating liver clock genes , we depleted neutrophils by injecting anti-Ly6G antibody into MCD diet-fed mice ( González-Terán et al . , 2016 ) . Anti-Ly6G administration for 7 days reduced circulating neutrophil levels without affecting monocytes ( Figure 2—figure supplement 2A , B ) , and treatment for 21 days markedly decreased hepatic diurnal Bmal1 and Clock expression , increased expression of Cry2 , and Per2 ( Figure 2—figure supplement 2C ) and consequently reduced steatosis ( González-Terán et al . , 2016 ) . To further support the role of neutrophil liver infiltration in the regulation of liver clock genes and hepatic lipogenesis during diet-induced steatosis , we leveraged a mouse model ( p38γ/δLyzs-KO ) that exhibits deficient neutrophil migration and subsequently , reduced liver neutrophil infiltration after MCD diet ( González-Terán et al . , 2016 ) . Compared with diet-matched control ( Lyzs-Cre ) mice , MCD-diet-fed p38γ/δLyzs-KO mice showed hepatic down-regulation of Bmal1 , which was associated with higher expression of Cry2 , and Per2 ( Figure 2I ) . These results suggest that the reduced neutrophil infiltration in mice lacking myeloid p38γ/δ expression is responsible for the altered expression of circadian clock genes . Overall , these findings strongly support that neutrophil infiltration modulates clock-gene expression in the liver , with downstream effects on liver metabolism . It has been suggested that JNK activation in the liver may be regulated in a circadian manner with a peak at noon ( Robles et al . , 2014 ) . To evaluate whether neutrophils might mediate this diurnal regulation of JNK , we analyzed JNK activation in neutropenic mice . Lack of neutrophils was associated with lower liver expression and activation of JNK , lower activation of the JNK downstream effector c-Jun , and lower expression of acetyl-CoA carboxylase ( Acaca ) , a key enzyme in metabolic regulation ( acetyl-CoA carboxylase; ACC ) that mediates inhibition of beta-oxidation and activation of lipid biosynthesis ( Figure 3A and Figure 3—figure supplement 1A ) . Similar results were found in p38γ/δLyzs-KO mice , in which reduced liver neutrophil infiltration was associated with decreased JNK phosphorylation and ACC protein levels ( Figure 3B and Figure 3—figure supplement 1B ) . Moreover , neutrophil-treated hepatocytes showed increased JNK activation together with increased levels of ACC expression ( Figure 3—figure supplement 1C ) . NE represents a potential mediator of this neutrophil function because elastase-treated hepatocytes also showed higher JNK activation , suggesting that this protease modulates the expression of the clock genes through the JNK signaling pathways ( Figure 3C and Figure 3—figure supplement 1D ) . This JNK activation was accompanied by increased Bmal1 expression ( Figure 3D ) , indicating that neutrophils altered liver clock-gene expression through the elastase-JNK pathway . Our results suggest that neutrophil-mediated JNK activation might modulate hepatocyte clock genes and metabolism through the regulation of ACC . Supporting this hypothesis , specific JNK depletion in hepatocytes downregulated Bmal1 , Clock , and Acaca compared to Alb-Cre ( Figure 3E and Figure 3—figure supplement 1E ) . According to these results , JNK inhibition reduced the expression of Bmal1 , Clock and Acaca in WT liver but not in neutropenic mice ( Figure 3—figure supplement 1F , G ) . These data strongly suggest that JNK activation caused by neutrophil infiltration modulates clock genes and daily metabolism in hepatocytes . JNK is an important modulator of the expression of the hepatokine circadian regulator FGF21 ( Vernia et al . , 2014 ) , which controls glucose and lipid metabolism ( Fisher and Maratos-Flier , 2013; Li et al . , 2013; Potthoff et al . , 2012 ) . Mice lacking JNK in hepatocytes had higher FGF21 mRNA expression ( Figure 3E ) . In concordance with high JNK activation , FGF21 expression was reduced in neutrophil-exposed hepatocytes ( Figure 3—figure supplement 1H ) . Moreover , neutropenic and p38γ/δLyzs-KO mice showed increased FGF21 expression ( Figure 3F and Figure 3—figure supplement 1I , J ) , which was consistent with the reduced hepatocyte JNK activation in these mice . To further define the role of FGF21 in the neutrophil-mediated regulation of liver metabolism , we suppressed FGF21 expression using two independent lentiviral shRNA vectors ( Figure 3G and Figure 3—figure supplement 1K ) . The protection of p38γ/δLyzs-KO mice against MCD-diet-induced alterations was abrogated by shFGF21 and these mice developed steatosis with an elevated hepatic triglyceride content ( Figure 3H , I ) . These data further supported the idea that neutrophil infiltration controls liver metabolism through the regulation of FGF21 expression . To formally confirm the involvement of NE in circadian clock alteration , we first evaluated the diurnal oscillation of NE levels in liver from WT mice fed a normal diet ( ND ) . According to infiltration pattern of neutrophils in the liver ( Figure 1A ) , we found higher NE levels at ZT2 than at ZT14 . ( Figure 4A ) . Next , circadian clock-gene expression in NE-/- mice revealed lower Bmal1 and elevated Per2 and Cry2 expression , compared to control mice ( Figure 4B ) , which mimicked the behavior of neutropenic mice . In addition , NE-/- mice presented lower respiratory quotient during the lights-on period than WT mice , indicating that these mice have increased fat utilization as a source of energy ( Figure 4C ) , supporting the data that reduced liver-neutrophil infiltration results in higher lipid oxidation . Interestingly , when fed MCD or HFD diet , NE-/- mice were protected against steatosis ( Figure 4D , E and Figure 4—figure supplement 1A , B ) , presented lower JNK activation , and expressed less ACC than control mice ( Figure 4F , G and Figure 4—figure supplement 1D ) . Besides , NE-/- mice were protected against alterations in clock-gene expression induced by MCD diet , presenting lower expression of Bmal1 and higher of Cry2 and Per2 comparing to control mice at ZT2 ( Figure 4H ) . Furthermore , under HFD , NE-/- mice were also refractory to these changes as these mice maintained a pattern of clock-gene expression similar to control mice in ND ( Figure 4—figure supplement 1E ) . To formally test a direct contribution of NE in the regulation of hepatic clock-gene expression and liver metabolism , we infused WT or NE-/- neutrophils into neutropenic mice under the jet lag protocol ( Figure 5A ) . The infusion of WT neutrophils was able to increase Bmal1 expression in the liver after jet lag , while neutropenic mice infused with NE-/- neutrophils presented the same levels of Bmal1 than non-infused neutropenic mice ( Figure 5B ) . In addition , while infusion of neutropenic mice with WT neutrophils increased steatosis , neutropenic mice infused with NE-/- neutrophils presented the same levels of steatosis than control neutropenic mice ( Figure 5C , D ) . All these data indicate that diet or jet-lag -induced hepatic infiltration of neutrophils results in dysregulation of the liver clock , and the lack of NE is enough to protect mice against these alterations . Finally , to evaluate the translational relevance of these findings for human physiology we quantified in human livers the expression levels for the genes encoding NE , JUN ( as an indicator of JNK activation ) and Bmal . Our results suggest that the levels of ELANE expression directly correlate with BMAL1 and JUN mRNA in livers from a human cohort ( Figure 5E ) . These correlations reinforce the idea that a rhythmic neutrophil infiltration in the liver controls the expression of clock genes through the JNK pathway activation and could be a target for therapeutic intervention during non-alcoholic fatty liver disease .
Our analysis demonstrates that neutrophils control clock genes in the liver and that reduced neutrophil infiltration protects against jet lag and diet-induced liver steatosis by altering the expression of these temporal regulators . These findings establish neutrophils as unexpected players in the regulation of daily hepatic metabolism . Our results also demonstrate that at least part of this neutrophil-induced clock modulation is mediated by elastase . These results agree with previous data showing that NE mediates the deleterious effects of neutrophils on liver metabolism and that mice lacking NE are protected against diet-induced steatosis ( Mansuy-Aubert et al . , 2013; Talukdar et al . , 2012 ) . The molecular mechanism underlying this regulation involves neutrophil NE that induces activation of JNK and consequently inhibits the production of the hepatokine FGF21 . The JNK pathway is an important modulator of liver metabolism , and lack of JNK1 and JNK2 in hepatocytes protects against steatosis ( Manieri and Sabio , 2015 ) . Here , we also demonstrate that JNK also regulates hepatocyte clock genes and , therefore , modulates diurnal adaptation of liver metabolism . Recently published data have demonstrated that lipogenesis is increased in the light phase , in agreement with our analysis ( Guan et al . , 2018 ) . We show that neutrophil infiltration causes JNK activation down-stream of elastase secretion , a time-dependent process . Indeed , phosphoproteomic analysis of the hepatic phosphorylation network identifies JNK as a key signaling enzyme with peak activation at ZT6 ( Robles et al . , 2017 ) immediately prior to the peak of lipogenic gene expression ( Guan et al . , 2018 ) . Our results suggest that neutrophils induce an accumulative activation of JNK with a peak during the day that would control the lipogenic program . Recent evidence established that the metabolic effects of JNK in the liver are mediated by FGF21 ( Vernia et al . , 2016; Vernia et al . , 2014 ) . Our results now show that liver FGF21 expression can be modulated through the control of JNK by neutrophils . Reduction of FGF21 by shRNA reverted the protective effect and metabolic changes induced by reduced neutrophil infiltration . In conclusion , our results show that the diurnal oscillating migratory properties of neutrophils regulate liver function in a manner that preserves daily metabolic rhythms , and that disturbance of this rhythmicity can cause disease . These results might imply a novel mechanism of action for the potential use of clock-modulating small molecules in liver health .
For the analysis of human liver mRNA levels , individuals were recruited among patients who underwent laparoscopic cholecystectomy for gallstone disease . The study was approved by the Ethics Committee of the University Hospital of Salamanca ( Spain ) , and all subjects provided written informed consent to participate . Patients were excluded if they had a history of alcohol use disorders or excessive alcohol consumption , chronic hepatitis C or B , or body mass index ≥35 . Baseline characteristics of these groups are listed in Figure 5—source data 1 . Neutropenic mice were generated with MCL1 ( B6 . 129-Mcl1tm3Sjk/J ) crossed with B6 . Cg-Tg ( S100A8-Cre , -EGFP ) 1Ilw/J mice or B6 . 129P2-Lyz2tm1 ( cre ) Ifo/J mice . Mice deficient in NE , with compound JNK1/2 deficiency in hepatocytes , with Cxcr2 deficiency in neutrophils or with p38γ/δ deficiency in myeloid compartment have been described ( Belaaouaj et al . , 1998; Das et al . , 2011; Das et al . , 2009; González-Terán et al . , 2016 ) All mice were backcrossed for 10 generations to the C57BL/6J background ( Jackson Laboratory ) . Genotypes were confirmed by PCR analysis of genomic DNA . Mice were housed under a 12 hr light:12 hr dark cycle ( Light is on at Zeitgeber Time ZT0 and off at ZT12 ) . For jet lag experiments , the 12 hr:12 hr dark/light cycle was disrupted by extending the dark cycle 12 hr every 5 days over 3 weeks ( Kettner et al . , 2016 ) . Cxcr2MRP8-KO chimeras were generated by exposing WT recipient mice to 2 doses of ionizing radiation ( 625 Gy ) and reconstituting them with 5 × 106 donor BM ( Cxcr2MRP8-KO ) cells injected into the tail vein . Mice were fed a methionine-choline-deficient ( MCD ) diet for 3 weeks or a high-fat diet ( HFD ) for 8 weeks ( Research Diets Inc ) . For neutrophil depletion , mice mini-osmotic pumps ( Alzet ) were implanted with anti-Ly6G antibody or saline ( 0 . 4 mg/kg per day , 21 days ) . For JNK inhibition experiments , mice were intraperitoneally injected with SP600125 ( 15 mg/kg ) ( Santa Cruz Biotechnology ) at ZT0 . For neutrophil infusion experiments , mice were intravenously injected with 3 × 106 WT or NE-/- purified neutrophils each 3–4 days . Neutrophils were isolated from BM using biotinylated anti-Ly6G antibody ( Clone:1A8 ) and streptavidin-labeled magnetic microbeads ( Miltenyi Biotec ) . All animal procedures conformed to EU Directive 86/609/EEC and Recommendation 2007/526/EC regarding the protection of animals used for experimental and other scientific purposes , enacted under Spanish law 1201/2005 . Hepatocytes were isolated from adult females by collagenase liver perfusion and cells were filtered through a 70 μm strainer . Hepatocytes pelleted from centrifuged Percoll gradients were plated at 4 × 105 cells/well on 6-well plates coated with collagen type one and incubated at 37°C . After 24 hr , cells were treated with 0 . 5 mM palmitate ( Sigma-Aldrich ) for 6 hr and then exposed for 1 hr to freshly neutrophils ( 2 × 106 cells/well ) in the presence of 1 µM FMLP ( Sigma-Aldrich ) . Neutrophils were isolated from BM as described above . For some experiments , neutrophils were sorted purified form the BM using an anti-Ly6G antibody ( Clone: 1A8 ) . T and B lymphocytes were sorted purified from spleens using anti-CD3 ( Clone: 145–2 C11 ) and anti-B220 ( Clone: RA3-6B2 ) , and bone marrow macrophages ( BMDM ) were differentiated as previously described ( González-Terán et al . , 2013 ) . All antibodies were purchased from BD Pharmingen . Alternatively , hepatocytes were exposed 2 hr to 5 nM NE ( R and D Systems ) or 0 . 5 mg/mL of collagenase A ( Roche ) after palmitate treatment . Mice were perfused with 20 mL of PBS and livers were collected and dissociated . Cell suspension was passed through a 70 μm strainer and centrifuged twice at 50 xg for 2 min to discard the liver parenchyma . For some experiments , livers were incubated for 15 min with 1 mg/mL Collagenase A ( Roche ) and 2 U/mL DNase ( Sigma ) at 37°C , and lungs were incubated for 25 min with 0 , 25 mg/ml Liberase TL ( Sigma ) and 5 U/mL DNase ( Sigma ) at 37°C Leukocyte fraction was collected and stained with anti-CD45 ( Clone: 30-F11 ) , from Invitrogen , anti-CD11b ( Clone: M1/70 ) , anti-Ly6G ( Clone: 1A8 ) or anti-Ly6C/G ( Clone: RB6-8C5 ) , from BD Pharmingen , and alternatively , with anti-F4/80 ( Clone: BM8 ) , from Invitrogen , and Goat anti-Clec4F from R and D Systems and conjugated with anti-goat Alexa 647 . Cells were sorted on a FACSAria to >95% purity . Flow cytometry experiments were performed with a FACScan cytofluorometer ( FACS Canto BD ) , and data were analyzed with FlowJo software . Transient calcium phosphate transfection of HEK-293 cells ( #CRL-1573 , ATCC ) was performed with the pGIPZ empty or pGIPZ . shFGF21 vector ( V3LMM_430499 and V3LMM_430501 , from Dharmacon ) together with pΔ8 . 9 and pVSV-G . The supernatants were collected , centrifuged ( 700 xg , 4°C , 10 min ) and concentrated ( 165x ) by ultracentrifugation for 2 hr at 121 , 986 xg at 4°C ( Ultraclear Tubes , SW28 rotor and Optima L-100 XP Ultracentrifuge; Beckman ) . Mice received tail-vein injections of 200 μl of lentiviral particles . Expression of mRNA was examined by qRT-PCR using a 7900 Fast Real Time thermocycler and Fast Sybr Green assays ( Applied Biosystems ) . Relative mRNA expression was normalized to Gapdh and Actb mRNA . The primers used were as follows: Actb ( F: GGCTGTATTCCCCTCCATCG; R: CCAGTTGGTAACAATGCCATGT ) ; Gapdh ( F: TGAAGCAGGCATCTGAGGG; R: CGAAGGTGGAAGAGTGGGA ) ; Clock ( F: AGAACTTGGCATTGAAGAGTCTC; R: GTCAGACCCAGAATCTTGGCT ) ; Bmal1 ( F: TGACCCTCATGGAAGGTTAGAA; R: GGACATTGCATTGCATGTTGG ) ; Nr1d2 ( F: CAGACACTTCTTAAAGCGGCACTG; R: GGAGTTCATGCTTGTGAAGGCTGT ) ; Cry2 ( F: CACTGGTTCCGCAAAGGACTA; R: CCACGGGTCGAGGATGTAG ) ; Per2 ( F: GAAAGCTGTCACCACCATAGAA; R: AACTCGCACTTCCTTTTCAGG ) ; Acaca ( F: GATGAACCATCTCCGTTGGC; R: GACCCAATTATGAATCGGGAGTG ) ; Fgf21 ( F: CTGCTGGGGGTCTACCAAG; R: CTGCGCCTACCACTGTTCC ) ; Mip1a ( F: TTCTCTGTACCATGACACTCTGC; R: CGTGGAATCTTCCGGCTGTAG ) ; Mip2 ( F: CCAACCACCAGGCTACAGG; R: GCGTCACACTCAAGCTCTG ) ; KC ( F: CTGGGATTCACCTCAAGAACATC; R: CAGGGTCAAGGCAAGCCTC ) ; Sdf-1 ( F: GCTCTGCATCAGTGACGGTA; R: ATCTGAAGGGCACAGTTTGG ) ; Elane ( F: ATTTCCGGTCAGTGCAGGTAGT; R: GGTCAAAGCCATTCTCGAAGAT ) ; GAPDH ( F: CCATGAGAAGTATGACAACAGCC; R: GGGTGCTAAGCAGTTGGTG ) ; ELANE ( F: TCCACGGAATTGCCTCCTTC; R: CCTCGGAGCGTTGGATGATA ) ; BMAL1 ( F: GCCGAATGATTGCTGAGG; R: CACTGGAAGGAATGTCTGG ) ; JUN ( F: GGATCAAGGCGGAGAGGAAG; R: GCGTTAGCATGAGTTGGCAC ) . Lipids were extracted from 25 mg of liver in isopropanol ( 50 mg/mL ) and centrifuged ( 15 min 9500 xg 4°C ) . Triglycerides were detected in the supernatant ( Sigma-Aldrich ) . Tissue samples were fixed in 10% formalin for 48 hr , dehydrated , and embedded in paraffin . Sections ( 5 μm ) were cut and stained with hematoxylin and eosin ( Sigma-Aldrich and Thermo Scientific ) . Sections ( 8 µm ) from frozen tissue and embedded in OCT compound ( Tissue-Tek ) were stained with Oil Red O ( American Master Tech Scientific ) . Sections were examined in Leica DM2500 microscope using 20x objective . Tissue extracts were prepared in Triton lysis buffer [20 mM Tris ( pH 7 . 4 ) , 1% Triton X-100 , 10% glycerol , 137 mM NaCl , 2 mM EDTA , 25 mM β-glycerophosphate , 1 mM sodium orthovanadate , 1 mM phenylmethylsulfonyl fluoride , and 10 µg/mL aprotinin and leupeptin] . Extracts ( 20–50 µg protein ) were examined by immunoblot . The antibodies employed were anti-FGF21 ( 1/1000 , #RD281108100 , BioVendor ) , anti-phospho JNK ( 1/1000 , #4668S , Cell Signaling ) , anti-JNK ( 1/1000 , #9252S , Cell Signaling ) , anti-phospho c-Jun ( 1/1000 , #9164L , Cell Signaling ) , anti-c-Jun ( 1/1000 , #9165S , Cell Signaling ) , anti-ACC ( 1/1000 , #3676S , Cell Signaling ) , and anti-vinculin ( 1/5000 , #V9131 , Sigma ) . Anti-phospho JNK and anti-JNK antibodies recognize the two different JNK isoform ( JNK1 and JNK2 ) and their two spliced variants ( JNK1 ( 46 kDa ) , JNK1 ( 54 kDa ) and JNK2 ( 46 kDa ) and JNK2 ( 54 kDa ) ) . Immunocomplexes were detected by enhanced chemiluminescence ( Amersham ) . For 3-D imaging , livers were fixed in a solution of paraformaldehyde 4% in PBS at 4°C . After washing in PBS , tissues were stored overnight in 30% sucrose ( Sigma ) with PBS . Then , livers were embedded in OCT compound ( Tissue-Tek ) and frozen at −80°C . Cryosections of organs ( 70 µm ) were washed in PBS and blocked/permeabilized in PBS with 10% donkey serum ( Millipore ) and 1% Triton . Primary antibodies diluted in blocking/permeabilization buffer were incubated overnight at 4°C , followed by three washes in PBS and 2 hr incubation with secondary antibodies and DAPI at room temperature . After three washes in PBS , cells were mounted with Fluoromount-G ( SouthernBiotech ) . The following primary and secondary antibodies were used: rat anti-CD31 ( 1:200 , #553370 BD Pharmingen , ) , rabbit anti-S100A9 ( mrp14 ) ( 1:100 , #AB242945 , Abcam , ) , goat anti-Clec4f ( 1:100 , #AF2784 , RD System ) , Alexa 488 donkey anti rat IgG ( 1:200 , #A-21208 , ThermoFisher ) , Cy3 AffiniPure Fab Fragment Donkey Anti-Rabbit IgG ( 1:200 , #711-167-003 , Jackson Laboratories ) , Alexa Fluor 633 donkey anti goat IgG ( H+L ) ( 1:200 , #A21082 , ThermoFisher ) . Immunostaining were imaged with a SP8 confocal microscope using 40x objectives . Individual fields or tiles of large areas were acquired every 2 . 5 µm for a total of 30 µm in depth . 3D images were obtained with Fiji/ImageJ 3D Viewer plugging . For 2-D imaging , liver sections ( 12 µm ) prepared from frozen tissue and embedded in OCT compound were fixed with 2% paraformaldehyde and permeabilized with PBS 0 . 1% Triton . After blocking with PBS 5% BSA 0 . 1% Triton and washing , tissues were incubated overnight at 4°C with primary antibody . Then , sections were washed and incubated with conjugated secondary antibodies for 1 hr at room temperature and nuclei were stained with Sytox Green ( Invitrogen ) after washing . The following primary and secondary antibodies were used: rat anti-mouse S100A9 ( Mrp-14 ) antibody ( 1:200 , #AB105472 , Abcam ) , rabbit anti-Neutrophil Elastase antibody ( 1:200 , #AB68672 , Abcam ) , goat Alexa Fluor 405 anti-rabbit ( 1:200 ) and goat Alexa Fluor 568 anti-rat IgG ( 1:500 ) . Sections were mounted in Vectashield mounting medium ( Vector , H-1000 ) and examined using a Leica SP5 multi-line inverted confocal microscope and 20x objectives . 20 mL of PBS prefunded livers were crushed with a syringe plunger , resuspended in 4 mL of PBS/EDTA 5 mM/0 . 5% FBS and filtered ( 70 µm ) . Cell suspension was centrifuged at 1800 rpm 5 min and the supernatant was filtered ( 22 µm ) . Supernatants were concentrated using Amicon Ultra centrifugal filters ( Sigma-Aldrich ) . NE levels were determined with Mouse Neutrophil Elastase ELISA kit ( R and D system ) . All data are expressed as means ± SEM . For comparisons between two groups , the Student’s t-test was applied . For data with more than two data sets , we used one-way ANOVA coupled with Turkey’s multigroup test . When variances were unequal , Welch’s test or Kruskal-Wallis test coupled with Dunn’s multiple comparison test were applied , respectively . Multiple group comparisons in the rhythmicity of neutrophil infiltration were analyzed with two-way ANOVA followed by Fisher’s post hoc test . Significance was determined as a 2-sided p < 0 . 05 . All statistical analyses were conducted in GraphPad Prism software . Statistical details were indicated in the figure legends . | Every day , the body's biological processes work to an internal clock known as the circadian rhythm . This rhythm is controlled by ‘clock genes’ that are switched on or off by daily physical and environmental cues , such as changes in light levels . These daily rhythms are very finely tuned , and disturbances can lead to serious health problems , such as diabetes or high blood pressure . The ability of the body to cycle through the circadian rhythm each day is heavily influenced by the clock of one key organ: the liver . This organ plays a critical role in converting food and drink into energy . There is evidence that neutrophils – white blood cells that protect the body by being the first response to inflammation – can influence how the liver performs its role in obese people , by for example , releasing a protein called elastase . Additionally , the levels of neutrophils circulating in the blood change following a daily pattern . Crespo , González-Terán et al . wondered whether neutrophils enter the liver at specific times of the day to control liver’s daily rhythm . Crespo , González-Terán et al . revealed that neutrophils visit the liver in a pattern that peaks when it gets light and dips when it gets dark by counting the number of neutrophils in the livers of mice at different times of the day . During these visits , neutrophils secreted elastase , which activated a protein called JNK in the cells of the mice’s liver . This subsequently blocked the activity of another protein , FGF21 , which led to the activation of the genes that allow cells to make fat molecules for storage . JNK activation also switched on the clock gene , Bmal1 , ultimately causing fat to build up in the mice’s liver . Crespo , González-Terán et al . also found that , in samples from human livers , the levels of elastase , the activity of JNK , and whether the Bmal1 gene was switched on were tightly linked . This suggests that neutrophils may be controlling the liver’s rhythm in humans the same way they do in mice . Overall , this research shows that neutrophils can control and reset the liver's daily rhythm using a precisely co-ordinated series of molecular changes . These insights into the liver's molecular clock suggest that elastase , JNK and BmaI1 may represent new therapeutic targets for drugs or smart medicines to treat metabolic diseases such as diabetes or high blood pressure . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"immunology",
"and",
"inflammation"
] | 2020 | Neutrophil infiltration regulates clock-gene expression to organize daily hepatic metabolism |
CRISPR-Cas systems provide sequence-specific immunity against phages and mobile genetic elements using CRISPR-associated nucleases guided by short CRISPR RNAs ( crRNAs ) . Type III systems exhibit a robust immune response that can lead to the extinction of a phage population , a feat coordinated by a multi-subunit effector complex that destroys invading DNA and RNA . Here , we demonstrate that a model type III system in Staphylococcus epidermidis relies upon the activities of two degradosome-associated nucleases , PNPase and RNase J2 , to mount a successful defense . Genetic , molecular , and biochemical analyses reveal that PNPase promotes crRNA maturation , and both nucleases are required for efficient clearance of phage-derived nucleic acids . Furthermore , functional assays show that RNase J2 is essential for immunity against diverse mobile genetic elements originating from plasmid and phage . Altogether , our observations reveal the evolution of a critical collaboration between two nucleic acid degrading machines which ensures cell survival when faced with phage attack .
Nearly all known archaea and about half of bacteria possess adaptive immune systems composed of clusters of regularly-interspaced short palindromic repeats ( CRISPRs ) and CRISPR-associated ( Cas ) proteins ( Makarova et al . , 2015 ) . CRISPR-Cas systems utilize small CRISPR RNAs ( crRNAs ) to recognize and destroy prokaryotic viruses ( phages ) ( Barrangou et al . , 2007; Brouns et al . , 2008 ) and other mobile genetic elements ( Bikard et al . , 2012; Marraffini and Sontheimer , 2008 ) . The CRISPR-Cas immune response occurs in three steps: adaptation , crRNA biogenesis , and interference ( reviewed in Klompe and Sternberg , 2018 ) . During adaptation , Cas proteins capture short segments of foreign nucleic acids ( 24–40 nucleotides ( nts ) in length ) and integrate them as ‘spacers’ into the CRISPR locus in between repeat sequences of similar lengths ( Mojica et al . , 2005; Yosef et al . , 2012 ) in order to record a molecular memory of past invaders . During crRNA biogenesis , the repeats and spacers are transcribed into a long precursor crRNA , which is subsequently processed to generate mature crRNAs that each contain a single spacer sequence . Mature crRNAs combine with one or more Cas proteins to form an effector complex , which during interference , senses and degrades ‘protospacers’ , invading nucleic acids complementary to the crRNA . Although all CRISPR-Cas systems adhere to this general pathway , they exhibit striking diversity in their cas gene composition and corresponding mechanisms of action . Accordingly , the current classification scheme divides these systems into two classes , six types ( I-VI ) , and dozens of subtypes ( Koonin et al . , 2017; Makarova et al . , 2015 ) . Class one systems ( types I , III , and IV ) encode multi-subunit effector complexes , while Class two systems ( types II , V , and VI ) possess single subunit effectors . Although Class two systems have garnered significant attention over the past decade due to their versatile genetic applications ( Klompe and Sternberg , 2018 ) , Class one systems are more widespread in nature , and of these , type III systems are believed to be the most closely related to the common ancestor from which all other CRISPR types have evolved ( Mohanraju et al . , 2016 ) . Staphylococcus epidermidis RP62a harbors a well-established model type III-A CRISPR-Cas system ( Figure 1A ) , here onward referred to as CRISPR-Cas10 . This system encodes three spacers ( spc ) and nine CRISPR-associated proteins ( Cas and Csm ) that can prevent the transfer of a conjugative plasmid ( Marraffini and Sontheimer , 2010 ) and stave off phage infection ( Maniv et al . , 2016 ) . This system employs an elaborate transcription-dependent mechanism of defense that exerts exquisite spatial and temporal control over the immune response . Defense begins with crRNA biogenesis ( Figure 1B ) , during which the Cas6 endoribonuclease cleaves the precursor crRNA within repeat sequences to generate 71 nt intermediate crRNAs ( Hatoum-Aslan et al . , 2011; Hatoum-Aslan et al . , 2014 ) . These intermediates are subsequently trimmed on their 3’-ends to generate mature species that range in length from 31 to 43 nts ( Hatoum-Aslan et al . , 2013 ) . Mature crRNAs combine with Cas10 , Csm2 , Csm3 , Csm4 , and Csm5 to form the Cas10-Csm effector complex ( Hatoum-Aslan et al . , 2013 ) ( Figure 1C ) . During interference , this complex , in conjunction with the accessory ribonuclease Csm6 , wage a two-pronged attack against foreign DNA and RNA . Interference is triggered by the binding of the crRNA to the targeted transcript ( Goldberg et al . , 2014 ) , an event that leads to the activation of at least three nucleases: Cas10 cleaves the non-template ( coding ) DNA strand ( Samai et al . , 2015 ) , each Csm3 subunit slices RNA within the protospacer region ( Samai et al . , 2015; Staals et al . , 2014; Tamulaitis et al . , 2014 ) , and Csm6 degrades nonspecific transcripts in the vicinity ( Foster et al . , 2018; Jiang et al . , 2016 ) . In related type III-A systems , target RNA binding has also been shown to trigger two additional functions of Cas10: the cleavage of non-specific single-stranded DNA by its HD domain ( Kazlauskiene et al . , 2016; Liu et al . , 2017 ) , and the generation of cyclic oligoadenylates by its Palm polymerase domain ( Kazlauskiene et al . , 2017; Niewoehner et al . , 2017 ) . The cyclic oligoadenylates act as second messengers which bind and further stimulate Csm6 , thus accelerating its activity near the site of foreign nucleic acid detection . The combined activities of the Cas and Csm nucleases , together with the high tolerance for mismatches between the crRNA and protospacer pairing ( Pyenson et al . , 2017 ) , contribute to a particularly powerful immune response that can lead to phage extinction with a single targeting spacer ( Bari et al . , 2017; Pyenson et al . , 2017 ) . The robust protection conferred by type III CRISPR-Cas systems motivated us to pose the question: Do these systems conspire with other cellular pathways to ensure a successful defense ? Indeed , our previous work showed that the Cas10-Csm complex , when expressed in its native S . epidermidis host , co-purifies with several cellular nucleases in trace amounts ( Walker et al . , 2017 ) . Furthermore , in a reconstituted system , Csm5 can bind and stimulate one of these nucleases , PNPase ( polynucleotide phosphorylase ) , a highly-conserved enzyme that is responsible for the processing and degradation of cellular RNAs in bacteria and eukaryotes ( Cameron et al . , 2018 ) . This observation lead us to propose a model wherein PNPase and other non-Cas nucleases are likely responsible for crRNA maturation ( Walker et al . , 2017 ) . However , the extent to which the success of CRISPR-Cas10 defense relies upon the activities of these non-Cas nucleases remains unknown . In bacteria , PNPase can act independently , or associate with other enzymes ( including nucleases and helicases ) to form an RNA degrading machine called the degradosome ( Cameron et al . , 2018 ) . In this study , we posited the hypothesis that PNPase and other degradosome-associated nucleases are essential for CRISPR-Cas10 immunity . To test this , we deleted the genes encoding PNPase and Ribonuclease ( RNase ) J2 , a second degradosome nuclease that co-purifies with the Cas10-Csm complex ( Walker et al . , 2017 ) . The single and double mutants were then assayed for defects in crRNA biogenesis and interference . We discovered that PNPase is indeed required for efficient crRNA maturation , but RNase J2 is not . In contrast , while PNPase appears to be dispensable for CRISPR-Cas10 function , RNase J2 is essential to block the transfer of a conjugative plasmid and defend against two unrelated phages , therefore establishing a general requirement for RNase J2 in CRISPR-Cas10 immunity . We also used quantitative PCR to track the accumulation of nucleic acids during an active phage infection , and found that both PNPase and RNase J2 are required for the efficient clearance of phage-derived DNA and transcripts . It has been well documented by others that RNase J2 promotes cellular RNA degradation by binding and stimulating the ribonuclease activity of its paralog RNase J1 ( Hausmann et al . , 2017; Linder et al . , 2014; Mathy et al . , 2007; Mathy et al . , 2010; Raj et al . , 2018 ) . Here , we show that S . epidermidis RNase J2 also promotes robust DNA degradation through this same mechanism . Altogether , our results support a model for CRISPR-Cas10 immunity in which two steps of the pathway , crRNA biogenesis and interference , rely upon the activities of degradosome-associated nucleases , thus revealing the evolution of a critical collaboration that ensures cell survival when faced with phage infection .
PNPase is a processive exonuclease that degrades RNA and DNA in the 3’−5’ direction , and also catalyzes the reverse reaction in which nucleotide diphosphates are added back onto the 3’-end of the substrate ( Cameron et al . , 2018; Cardenas et al . , 2009; Walker et al . , 2017 ) . Our previous work showed that in a purified system , Csm5 can bind to PNPase and stimulate its nucleolytic activity while repressing its polymerization function ( Walker et al . , 2017 ) . Since the deletion of Csm5 leads to loss of crRNA maturation in vivo ( Hatoum-Aslan et al . , 2011; Hatoum-Aslan et al . , 2014 ) , and since we were unable to locate a nucleolytic active site in Csm5 after extensive mutagenic and biochemical analyses , the observed physical and functional interactions between Csm5 and PNPase lead us to propose a model in which Csm5 recruits and stimulates the catalytic activity of PNPase to carry out crRNA maturation ( Walker et al . , 2017 ) . Here , to lend further support to this model , we created in-frame deletions of pnp in two S . epidermidis strains ( Figure 2—figure supplement 1 ) : RP62a , the native CRISPR-Cas10 containing strain , and LM1680 ( Jiang et al . , 2013 ) , a CRISPR-less derivative of RP62a that is more receptive to transformation and conducive to protein purification . To determine the impact of this mutation on crRNA maturation , the entire CRISPR-Cas system was introduced back into LM1680/Δpnp on plasmid pcrispr-cas ( Hatoum-Aslan et al . , 2013 ) , which encodes a 6-His tag on the N-terminus of Csm2 . We then purified Cas10-Csm complexes from the mutant and wild-type strains using Ni2+ affinity chromatography , and extracted crRNAs bound to the complexes . While complex formation remained unaffected in the Δpnp mutant ( Figure 2A ) , we observed that the crRNA sizes associated with these complexes differed dramatically from the wild-type ( Figure 2B ) —when purified from wild-type background , 6 . 2 ( ±2 . 4 ) % of complex-associated crRNAs exists in the intermediate state , whereas in the Δpnp mutant , intermediates represent 50 . 6 ( ±3 . 0 ) % of all crRNAs in the complex ( Figure 2C and Figure 2—source data 1 ) , indicating that PNPase is indeed required for efficient crRNA maturation . To rule out the possibility of an unintentional second-site mutation causing this phenotype , we created a complementation strain in which pnp was re-introduced into the genome of the knockout strain , along with several silent mutations in the coding region ( denoted as pnp* ) to differentiate between the original wild-type and knock-in strains ( Figure 2—figure supplement 2 ) . As expected , the crRNA size distribution returned to the wild-type phenotype in the complementation strain ( Figure 2B and C and Figure 2—source data 1 ) , providing confirmation that efficient crRNA maturation relies upon the activity of PNPase . The persistence of residual maturation in the Δpnp mutant suggests there are probably additional nuclease ( s ) contributing to this process . In order to identify other maturation nuclease candidates , we consulted the short list of cellular nucleases that were previously found to co-purify with the Cas10-Csm complex ( Walker et al . , 2017 ) . Now , after having established a bona fide function for PNPase in crRNA maturation , two other co-purifying nucleases emerged as promising candidates: Ribonucleases J1 and J2 . Since PNPase , RNase J1 , and RNase J2 have all been found to work together as components of the degradosome in gram positive organisms ( Cho , 2017; Redder , 2018; Roux et al . , 2011 ) , the possibility exists that they might also be working together in the CRISPR-Cas10 immunity pathway . Additionally , the gene encoding RNase J2 , rnjB , is located directly downstream of pnp in the S . epidermidis genome , further hinting at a functional link . Ribonucleases J1 and J2 are paralogous enzymes that have been shown to form a complex in the cell and catalyze endonucleolytic and 5’−3’ exonucleolytic cleavage of RNA substrates ( Hausmann et al . , 2017; Linder et al . , 2014; Mathy et al . , 2007; Mathy et al . , 2010; Raj et al . , 2018 ) . Although the dominant active site in the J1/J2 complex resides in RNase J1 , RNase J2 has been shown to cause a synergistic stimulation of RNase J1’s nuclease activity through an allosteric mechanism . In order to determine the contribution of these ribonucleases toward CRISPR-Cas10 function , we sought to delete the genes that encode both enzymes . However , despite multiple attempts , we were unable to delete rnjA , the gene encoding RNase J1 , suggesting this enzyme might be essential for S . epidermidis survival . Alternatively , our method for mutagenesis ( Bae and Schneewind , 2006 ) , which requires cell growth at a range of temperatures , might be incompatible with the extreme temperature sensitivity that has been observed in a S . aureus ΔrnjA mutant ( Hausmann et al . , 2017 ) . Nonetheless , we were able to create an in-frame deletion of rnjB , the gene that encodes RNase J2 , in both S . epidermidis RP62a and LM1680 strains ( Figure 2—figure supplement 1 ) . To test for defects in crRNA maturation in the LM1680/ΔrnjB mutant , we overexpressed and purified Cas10-Csm complexes from this strain and extracted and visualized the complex-associated crRNAs ( Figure 2 ) . We observed that Cas10-Csm remains intact when expressed in the ΔrnjB background ( Figure 2A ) , and the crRNA size distribution is similar to that observed in the wild-type control ( Figure 2B and C and Figure 2—source data 1 ) , indicating that RNase J2 is unlikely to play a significant role in crRNA maturation . To confirm , we also created the ΔpnpΔrnjB double-mutant ( Figure 2—figure supplement 1 ) and found that it exhibits a maturation defect similar to that observed in the Δpnp single mutant ( Figure 2B and C ) . We noted that longer RNAs running between 43 and 71 nucleotides appear more prominently in the double mutant when compared to either single mutant—these species likely represent crRNAs that have undergone incomplete maturation . The more prominent appearance of these bands in the double mutant when compared to the ∆rnjB single mutant may imply that RNase J2 plays a modest role in crRNA maturation , and its loss can be compensated by the presence of PNPase . However , since this partial defect in the double mutant does not lead to additional accumulation of the 71 nucleotide intermediate species , we hesitate to speculate further on the significance of these longer crRNAs . From these observations , it can be concluded that PNPase , but not RNase J2 , is essential for efficient crRNA maturation , and that there exist other maturation nuclease ( s ) that have yet to be identified . We next wondered whether the activities of PNPase and/or RNase J2 are essential for CRISPR-Cas10 function . To test this , three independent interference assays against diverse mobile genetic elements were conducted ( Figure 3 ) . The first is a plasmid challenge assay that measures the transfer efficiency of the conjugative plasmid pG0400 from a S . aureus donor into S . epidermidis LM1680 recipients bearing pcrispr-cas ( Figure 3A ) . In this system , spc1 ( the first spacer ) targets the nickase gene in pG0400 , therefore , a functional CRISPR-Cas10 system in the recipient strain is expected to lower the conjugation efficiency . Indeed , the controls behaved consistently with previous results ( Hatoum-Aslan et al . , 2013 ) , where the wild-type LM1680 strain bearing the empty vector ( pC194 ) showed a conjugation efficiency in the 10−3 range , while this same strain harboring pcrispr-cas exhibited an efficiency in the 10−7 range ( Figure 3B and Figure 3—source data 1 ) . Interestingly , despite the maturation defect in Δpnp/pcrispr-cas , this strain had a conjugation efficiency similar to that of the wild-type strain expressing the CRISPR-Cas10 system , suggesting that PNPase is dispensable for anti-plasmid immunity . In contrast , ΔrnjB/pcrispr-cas exhibited a significant defect in immunity , with a conjugation efficiency similar to that of the negative control . Consistent with this result , ΔpnpΔrnjB/pcrispr-cas also showed a significant defect in immunity . Interestingly , CRISPR-Cas10 appears to retain a modest level of anti-plasmid immunity in the double mutant , resulting in a conjugation efficiency that is an order of magnitude lower than that observed in the absence of pcrispr-cas ( compare 2 . 2 × 10−4 to 2 . 6 × 10−3 ) . Importantly , complementation strains containing a copy of rnjB on pcrispr-cas ( a plasmid called pcrispr-cas-rnjB ) exhibited conjugation efficiencies similar to those of the wild-type strain bearing pcrispr-cas ( Figure 3B and Figure 3—source data 1 ) . These observations suggest that RNase J2 is essential for CRISPR-Cas10 anti-plasmid immunity . To determine if this defect applies specifically to plasmid transfer , or if there is a more general phenotype , we tested CRISPR immunity against two unrelated phages . The first , CNPx , is a variant of CNPH82 , a temperate phage belonging to the morphological family Siphoviridae ( Daniel et al . , 2007; Maniv et al . , 2016 ) . CNPx can infect S . epidermidis LM1680 , and spc2 in pcrispr-cas bears complementarity to its cn20 gene ( Figure 3C ) . When dilutions of this phage are plated with the LM1680/pC194 negative control , many plaques , or zones of clearing in the bacterial lawn can be observed , approximately 107 plaque-forming units per milliliter ( pfu/ml ) ( Figure 3D and Figure 3—source data 2 ) . However , when the LM1680/pcrispr-cas strain is challenged with CNPx , not a single plaque is visible , in agreement with the robust anti-phage immunity previously reported for this system ( Bari et al . , 2017; Pyenson et al . , 2017 ) . Importantly , when we challenged the mutant strains with CNPx , we observed a significant defect , even more pronounced than that seen for anti-plasmid immunity: While the Δpnp strain appeared to perform like wild-type , the ΔrnjB and ΔpnpΔrnjB mutations abolished CRISPR-Cas10 function altogether . In addition , returning rnjB to these defective mutants on pcrispr-cas-rnjB restored full immunity against CNPx . Finally , we wanted to determine if the phenotypes observed thus far are specific to LM1680 , or if they can also be seen in the wild-type RP62a strain . S . epidermidis LM1680 was originally derived from RP62a as a CRISPR ‘escape’ mutant which harbors a ~ 258 , 000 nt deletion encompassing the CRISPR locus and flanking regions ( Jiang et al . , 2013 ) . We wondered if the immunity defect observed in LM1680 might be conditional upon the loss of the regions adjacent to the CRISPR-Cas system . To rule out this possibility , the RP62a Δpnp , ΔrnjB , and ΔpnpΔrnjB mutants were challenged with phage Andhra ( Cater et al . , 2017 ) in the presence of the plasmid pcrispr-spcA2 ( Bari et al . , 2017 ) ( Figure 3E ) . Andhra is a lytic phage belonging to the morphological family Podoviridae , and spcA2 targets its DNA polymerase ( gp09 ) gene . It is important to note that this system relies upon the RP62a genome-encoded cas and csm genes , with only the Andhra-targeting spacer being overexpressed on the plasmid . As expected , without pcrispr-spcA2 , all strains are equally susceptible to Andhra , with plaque counts ranging between 108–109 pfu/ml ( Figure 3F and Figure 3—source data 3 ) . However , when transformed with pcrispr-spcA2 , only the wild-type and Δpnp strains gain complete immunity to Andhra , while the ΔrnjB and ΔpnpΔrnjB strains remain as sensitive to Andhra as the corresponding negative controls without the pcrispr-spcA2 plasmid . From these collective observations , we can conclude that while PNPase appears dispensable , RNase J2 is essential for CRISPR immunity against diverse mobile genetic elements originating from plasmid and phage . In addition , RNase J2 most likely acts in the CRISPR pathway at a step downstream of crRNA maturation . Although CRISPR immunity appeared unperturbed in the Δpnp mutant , we wondered if a partial defect might be present which cannot be resolved by the functional assays . For example , a partial defect in anti-plasmid immunity might lead to a decrease in plasmid copy number instead of complete elimination , but the copy number reduction might be sufficient to prevent growth on the media that selects for transconjugants . Similarly , a partial defect in anti-phage immunity might result in the survival of a small number of phages , but too few to create a visible plaque in the lawn of bacteria . In order to detect such partial defects in immunity , we used qPCR and qRT-PCR ( quantitative PCR and reverse-transcriptase PCR , respectively ) to measure DNA and RNA accumulation during an active phage infection . Of the three systems for testing CRISPR function , the RP62a-Andhra system ( Figure 3E ) was selected for this assay for two reasons: First , RP62a is the original clinical isolate containing a single genomic copy of the CRISPR-Cas10 system , thus providing the closest approximation to what might naturally occur . Second , Andhra is strictly lytic and therefore expected to express all genes in a single infection cycle ( Cater et al . , 2017 ) . The latter feature eliminates any gene-specific variations in CRISPR-Cas10 function that would otherwise occur using a temperate phage ( Goldberg et al . , 2014; Jiang et al . , 2016 ) . In the assay ( Figure 4A ) , early-log cells were challenged with Andhra in a 2:1 ( bacteria:phage ) ratio , and phages were allowed to adsorb to cells for 10 min . Infected cells were then washed with fresh media to remove any phages in suspension . A fraction of the cells was harvested immediately after adsorption ( time = 0 min ) and every ten minutes thereafter for up to 30 min . Importantly , the latent period for Andhra was previously determined to be 35 min ( Cater et al . , 2017 ) , therefore the selected time points all occur within the timeframe of a single infection cycle before the first burst . Immediately after each harvest , cells were heat-killed to prevent any further changes in nucleic acid content , and then subjected to nucleic acid extraction . The RNA and DNA extracts were analyzed by qRT-PCR and qPCR , respectively , using two pairs of primers: Andhra-specific primers that flank the protospacer region within gene gp09 ( Figure 4—figure supplement 1 ) , and genome-specific primers that bind the gap gene ( encodes glyceraldehyde-3-phosphate dehydrogenase ) for normalization . For quantification , the relative abundance of the phage-derived PCR product detected at zero minutes post-infection in RP62a cells was set to one as the reference point . As expected , in wild-type RP62a cells lacking spcA2 , phage RNA and DNA levels increased significantly in a time-dependent manner ( Figure 4B and C and Figure 4—source datas 1 and 2 ) ; however , the same strain bearing pcrispr-spcA2 exhibited a ten-fold depletion in phage RNA levels , and no increase in phage DNA content by 30 min post-infection . Interestingly , the Δpnp/pcrispr-spcA2 strain showed a significant ( ~10 fold ) accumulation of phage RNA and DNA , but not as high as that observed in the negative control strain devoid of pcrispr-spcA2 . These results suggest that PNPase is indeed required to promote efficient clearance of phage-derived nucleic acids . We also tested the ΔrnjB/pcrispr-spcA2 and ΔpnpΔrnjB/pcrispr-spcA2 strains , and in agreement with their impaired CRISPR-Cas10 function , both strains showed significant accumulation of phage-derived DNA and transcripts when compared to the wild-type strain bearing pcrispr-spcA2 . Notably , the double mutant showed a greater defect in the clearance of phage nucleic acids than the ΔrnjB single mutant , thus confirming the contribution of PNPase towards phage nucleic acid degradation . One additional observation in this data was striking: The relative amounts of phage nucleic acids in the ΔrnjB/pcrispr-spcA2 and ΔpnpΔrnjB/pcrispr-spcA2 strains significantly exceeded those seen in the wild-type strain lacking pcrispr-spcA2 . This data implies that PNPase and/or RNase J2 may be acting against phage nucleic acids even in the absence of CRISPR targeting . To test this possibility , we conducted the same time course infection assay in the Δpnp and ΔrnjB mutants lacking pcrispr-spcA2 and tracked DNA accumulation ( Figure 4—figure supplement 1 and Figure 4—figure supplement 1—source data 1 ) . As expected , the RP62a wild-type control showed that by 30 min post-infection , phage DNA levels rose to ~35 times greater than that seen immediately following adsorption . Strikingly , Δpnp and ΔrnjB mutants accumulated over 400 and 1300 times more phage DNA , respectively , by this same time point . These data indicate that PNPase and RNase J2 facilitate the restriction of phage DNA amplification within the cell , even in the absence of a targeting CRISPR system . The contributions of PNPase and RNase J2 toward the elimination of phage-derived nucleic acids can occur through multiple mechanisms . One possibility is that these enzymes indirectly prevent phage DNA accumulation by acting strictly as ribonucleases—their swift degradation of phage transcripts might deter the expression of the phage-encoded DNA polymerase , and thus prevent phage DNA replication . Another noncompeting possibility is that phage DNA is degraded directly by one or both of these enzymes . Although PNPase is most well-known for its ribonuclease activity ( Cameron et al . , 2018; Hui et al . , 2014 ) , we and others have recently shown that bacterial PNPase homologs can also degrade DNA substrates ( Cardenas et al . , 2009; Walker et al . , 2017 ) . Therefore , it is plausible that the efficient CRISPR-mediated clearance of phage nucleic acids is assisted by PNPase’s dual-cleavage of phage DNA and RNA . In contrast , studies on bacterial RNase J homologs have strictly reported ribonuclease activities in these enzymes ( Hausmann et al . , 2017; Linder et al . , 2014; Mathy et al . , 2007; Mathy et al . , 2010 ) . One exception seems to be in two archaeal RNase J homologs , which were shown to catalyze the cleavage of single-stranded DNA ( Levy et al . , 2011 ) . We therefore wondered whether one or both RNase J homologs in S . epidermidis might act similarly against DNA substrates . To test the nucleolytic activities of the S . epidermidis RNases J homologs , we overexpressed and purified recombinant RNase J1 and RNase J2 from E . coli ( Figure 5A ) . In addition , since previous reports have indicated that the dominant active site in an RNase J1/J2 complex resides within RNase J1 ( Mathy et al . , 2007; Mathy et al . , 2010 ) , we also purified a catalytically-dead variant of RNase J1 ( denoted as dJ1 ) bearing H74A and H76A mutations . We first tested for metal-dependent RNase activity in these enzymes by combining them with a 31 nt RNA substrate ( Figure 5B ) in the presence of various metals ( Figure 5—figure supplement 1 ) . Consistent with previous reports ( Hausmann et al . , 2017; Mathy et al . , 2010 ) , we observed insignificant nuclease activity in RNase J2 , and Mn2+-dependent ribonuclease activity in RNase J1 that leaves behind a single small cleavage product , implying 5’−3’ exonuclease activity . Furthermore , this function appeared more pronounced when RNase J1 was combined with RNase J2 in a 1:1 ratio . To examine the kinetics in more detail , we tracked substrate degradation over time ( Figure 5C and D ) , and confirmed that when combined , RNases J1 and J2 act synergistically—whereas RNase J1 alone could only degrade 23 ( ±14 ) % of the substrate during a 20 min incubation period , the combined proteins eliminated 95 ( ±1 ) % of the substrate in the same amount of time ( Figure 5D and Figure 5—source data 1 ) . Furthermore , the catalytically-dead RNase J1 mutant combined with RNase J2 showed no nucleolytic activity , confirming that the nuclease active site in the J1/J2 complex originates from RNase J1 . We next tested for metal-dependent DNase activity in these enzymes by combining them with a 60 nt DNA substrate ( Figure 5B ) in the presence of various metals ( Figure 5—figure supplement 1 ) . While we detected little/no nuclease activity from each enzyme alone , a striking Mn2+-dependent DNase activity was observed when the enzymes were combined , which again resulted in a single small cleavage product indicative of a 5’−3’ exonuclease function . Additional experiments measuring DNA degradation over time revealed a synergistic activation of DNA cleavage even more pronounced than that observed for RNA cleavage ( Figure 5E and F and Figure 5—source data 2 ) . For example , while RNase J1 alone cleaved 17 ( ±8 ) % of the DNA substrate after 5 min , the J1/J2 complex had already degraded 99 ( ±1 ) % of the substrate in the same amount of time . In addition to 5’−3’ exonuclease activity , RNase J homologs have also been shown to catalyze endoribonuclease cleavage ( Hausmann et al . , 2017; Linder et al . , 2014; Mathy et al . , 2007; Mathy et al . , 2010 ) . Therefore , we also tested for endonuclease activity against DNA by combining the enzymes alone and in combination with circular DNA substrates , both single-stranded and double-stranded ( Figure 5B ) . While cleavage of double-stranded circular or linear DNA could not be detected ( Figure 5G ) , we observed degradation of the single-stranded circular DNA by RNase J1 , an activity that is significantly stimulated when combined with RNase J2 ( Figure 5H ) . Again , the catalytically-dead RNase J1 completely lost this function , indicating that the DNA endonuclease active site resides within RNase J1 . Altogether , these observations demonstrate that RNase J2 stimulates both RNase and DNase activities of RNase J1 , and therefore lend support to the possibility that these enzymes assist with CRISPR-Cas10 immunity by directly degrading phage DNA and RNA .
Type III CRISPR-Cas systems rely upon multiple Cas proteins which interact with each other , and also have the potential to recruit and regulate other enzymes with relevant functions to ensure a successful defense . Here , we show that the degradosome-associated nuclease PNPase promotes crRNA maturation ( Figure 2 ) and is required for the efficient CRISPR-mediated clearance of invading DNA and RNA ( Figure 4 ) . We also show that a second degradosome nuclease , RNase J2 , is essential for CRISPR immunity against diverse mobile genetic elements originating from plasmid and phage ( Figure 3 ) . Although bacterial RNase J homologs have been strictly reported as ribonucleases , we present , to our knowledge , the first demonstration of DNA degradation by these enzymes ( Figure 5 ) . Altogether , our results support a model for CRISPR-Cas10 immunity in which two of the three steps of the pathway , crRNA biogenesis and interference , rely upon the activities of these degradosome-associated nucleases . The possibility remains that PNPase and/or RNase J2 might play indirect roles in CRISPR-Cas10 function , wherein deletion of these degradosome subunits may impact the expression of other host factors that are required for CRISPR function . However , the direct physical interactions that we and others have observed between degradosome components with each other and with the CRISPR machinery ( discussed in details below ) help to support direct roles for these nucleases in CRISPR immunity ( Figure 6 ) . We previously showed that Csm5 binds directly to PNPase and stimulates its nucleolytic activity in a purified system ( Walker et al . , 2017 ) , thus providing mechanisms for PNPase’s recruitment and regulation during crRNA maturation and interference in vivo . Since PNPase has been shown to bind RNase J1 ( Raj et al . , 2018; Roux et al . , 2011 ) , and since RNases J1 and J2 exist as a complex in the cell ( Mathy et al . , 2010 ) , one possible mechanism for the recruitment of RNase J2 to the site of CRISPR interference could be through its connection with PNPase . However , if such an interaction were essential for RNase J2’s participation in immunity , then a pnp knockout should phenocopy the rnjB mutant—this was not the case . Therefore , there may be an additional , more direct interaction between RNase J2 and one or more of the CRISPR-associated proteins that has yet to be identified . Another compatible alternative is that RNase J2 and/or PNPase might locate and degrade phage-derived nucleic acids through a CRISPR-independent mechanism . In support of this notion , we observed that in the absence of a targeting CRISPR-Cas system , the deletion of pnp and rnjB enables significant phage DNA accumulation which exceeds phage DNA levels in the wild-type strain by 10–40 fold , respectively ( Figure 4—figure supplement 1 ) . The mechanism by which these degradosome-associated nucleases repress phage DNA replication in the absence of CRISPR interference remains unclear . Interestingly , there have been two reports of phage-encoded proteins that specifically inhibit RNase E , a central subunit of the degradosome in gram negative organisms ( Van den Bossche et al . , 2016; Hui et al . , 2014; Marchand et al . , 2001 ) . These observations support the possible existence of a more direct antagonistic relationship between phages and the degradosome that undoubtedly deserves further investigation . DNA endonuclease activity is particularly important for CRISPR function . In order for phage DNA to get degraded , there must be an initiating endonucleolytic cleavage event that exposes free ends , which are subsequently accessed and processively degraded by exonucleases . Thus far , of the CRISPR-associated nucleases characterized in type III systems , only one has been shown to degrade DNA: Cas10 ( Kazlauskiene et al . , 2016; Liu et al . , 2017; Samai et al . , 2015 ) . Furthermore , S . epidermidis Cas10 ( SeCas10 ) cleaves only the coding DNA strand in the protospacer region , leaving the non-coding strand unharmed ( Samai et al . , 2015; Wang et al . , 2019 ) . Although Cas10 homologs in related type III-A systems exhibit nonspecific DNA endonuclease activity ( Kazlauskiene et al . , 2016; Liu et al . , 2017 ) , this function in SeCas10 has yet to be demonstrated . Therefore , our discovery that RNase J2 is essential for CRISPR-Cas10 immunity , combined with the fact that this enzyme is capable of promoting robust DNA endonuclease cleavage ( Figure 5H ) , opens the possibility that the RNase J homologs assist with CRISPR-Cas10 interference by catalyzing the initiating endonucleolytic cleavage on the non-coding DNA strand . It is worthwhile mentioning that while PNPase exhibits 3’−5’ exonuclease activity on DNA and RNA substrates ( Cameron et al . , 2018; Cardenas et al . , 2009; Walker et al . , 2017 ) , RNase J homologs catalyze endonuclease activity , as well as 5’−3’ exonucleolytic activity , a function only recently thought to be non-existent in bacteria ( Hausmann et al . , 2017; Linder et al . , 2014; Mathy et al . , 2007; Mathy et al . , 2010 ) . Therefore , the combined capabilities of these enzymes provide the full spectrum of nucleolytic activities that are necessary to shred DNA and RNA down to their base parts . The reliance on non-Cas nucleases by a type III CRISPR-Cas system described herein adds to the sparse collection of similar observations that have been reported for diverse CRISPR-Cas types . The earliest example of such a collaboration was seen in a type II CRISPR-Cas system in Streptococcus pyogenes , in which both the crRNA and the tracrRNA are processed by the host-encoded RNase III ( Deltcheva et al . , 2011 ) . In a more recent example , a type III-Bv ( variant ) system in Synechocystis 6803 , which lacks a Cas6 homolog , was found to utilize the host-encoded RNase E to catalyze crRNA processing ( Behler et al . , 2018 ) . In yet a third example , a CRISPR-like element in Listeria monocytogenes devoid of its own cas genes , was shown to rely upon a type I CRISPR-Cas system and host-encoded PNPase for small RNA processing and interference ( Sesto et al . , 2014 ) . From these seemingly disparate observations , perhaps a common theme is beginning to emerge in which diverse CRISPR-Cas types have independently evolved different mechanisms to tap into the arsenal of nucleolytic enzymes that bacteria utilize for routine functions , and channel their activities towards defense in the event of a phage infection . Indeed , such a strategy would help to minimize the energetic costs associated with maintaining dedicated nucleases , while also maximizing the likelihood of a successful defense .
S . aureus RN4220 was grown in Tryptic Soy Broth ( TSB ) medium ( BD Diagnostics , NJ , USA ) . S . epidermidis RP62a and LM1680 were grown in Brain Heart Infusion ( BHI ) medium ( BD Diagnostics , NJ , USA ) . E . coli DH5α was grown in Luria Bertani ( LB ) broth ( VWR , PA , USA ) , and E . coli BL21 ( DE3 ) was grown in Terrific broth ( VWR , PA , USA ) for protein purification . Growth media was supplemented with the following: 10 µg/ml chloramphenicol ( to select for pcrispr-based and pKOR1-based plasmids ) , 15 µg/ml neomycin ( to select for S . epidermidis cells ) , 5 µg/ml mupirocin ( to select for pG0400 ) , 30 µg/ml chloramphenicol ( to select for E . coli BL21 ( DE3 ) ) and 50 µg/ml kanamycin ( to select for pET28b-His10Smt3-based plasmids ) . Phages CNPx and Andhra were grown on S . epidermidis LM1680 and S . epidermidis RP62a host cells , respectively . To propagate phages , overnight cultures of each host were diluted 1:100 in BHI supplemented with 5 mM CaCl2 and phage ( 1 × 107 pfu ) . Cultures were incubated at 37°C with agitation for 5 hr . One-fifth the volume of fresh mid-log host cells was then added into the bacteria:phage culture , and the culture was incubated for an additional 2 hr at 37°C with agitation . Cells were pelleted at 5000 x g for 5 min at 4°C , and the supernatant containing phage was passed through a 0 . 45 µM filter . Phage titer was determined by spotting ten-fold dilutions of the supernatant atop a semisolid layer of 0 . 5 X heart infusion agar ( HIA ) medium ( Hardy Diagnostics , CA , USA ) containing 5 mM CaCl2 and a 1:100 dilution of overnight host culture . Spots were allowed to air dry , and plates were incubated overnight at 37°C . On the following day , plaques were enumerated and phage titers in plaque-forming units/ml ( pfu/ml ) were determined . The pKOR1 system ( Bae and Schneewind , 2006 ) was used to create in-frame deletions of pnp and/or rnjB , and to re-insert a pnp variant ( pnp* , which has five silent mutations , Figure 2—figure supplement 2 ) into S . epidermidis LM1680 and/or RP62a strains . pKOR1_Δpnp , pKOR1_ΔrnjB , and pKOR1_ΔpnpΔrnjB , which were used to delete indicated genes , were created using a 3-piece Gibson assembly strategy ( Gibson et al . , 2009 ) . Briefly , for pKOR1_Δpnp and pKOR1_ΔrnjB , 1 . 0 kb DNA fragments flanking the gene ( s ) of interest ( upstream and downstream ) were obtained via PCR amplification using the S . epidermidis RP62a WT genome as template and primers F179 , F180 , and F331-334 ( for pKOR1_Δpnp ) and F452-F457 ( for pKOR1_ΔrnjB ) ( Supplementary file 1 ) . For pKOR1_ΔpnpΔrnjB , 1 . 0 kb DNA fragments flanking both genes were obtained via PCR amplification using the S . epidermidis RP62a Δpnp strain as template and primers F452 , F459 , F455-F457 , and F460 . For all constructs , the pKOR1 plasmid was used as a template to amplify the backbone with indicated primers . The three PCR products generated for each construct were purified using the EZNA Cycle Pure Kit ( Omega Bio-tek , CA , USA ) and Gibson assembled . To create the complementation plasmid pKOR1_pnp* , a two-piece Gibson assembly was used , wherein the pnp gene was amplified from S . epidermidis RP62a and assembled with the pKOR1_Δpnp backbone generated with primers L047-L050 . All assembled constructs were transformed via electroporation into S . aureus RN4220 . Four transformants were selected for each construct and confirmed to harbor the plasmid using PCR and DNA sequencing with primers F016 , F181 , F182 , F335 , and L009 ( Supplementary file 1 ) . Confirmed plasmids were purified using the EZNA Plasmid DNA Mini Kit ( Omega Bio-tek , CA , USA ) and transformed into S . epidermidis LM1680 via electroporation . Confirmed S . epidermidis LM1680 transformants were used to transfer the pKOR1-based plasmids into S . epidermidis RP62a using phage-mediated transduction . The temperate phage CNPx was used to transfer pKOR1-based plasmids from S . epidermidis LM1680 into S . epidermidis RP62a . Briefly , overnight cultures of S . epidermidis LM1680 strains harboring the plasmids were used to propagate phage CNPx as described above . Filtered phage lysates were then diluted 1:10 in mid-log S . epidermidis RP62a cells . The phage:cell mixture was incubated at 37°C for 20 min , and pelleted at 5000 x g for 5 min at 4°C . Cell pellets were resuspended in fresh BHI and plated onto BHI agar containing neomycin and chloramphenicol . Plates were incubated overnight at 30°C . Four colonies were picked and confirmed to harbor the plasmid by subjecting them to PCR and sequencing using the appropriate sequencing primers listed in Supplementary file 1 . Allelic replacement was used to generate all mutants ( Bae and Schneewind , 2006 ) . Briefly , S . epidermidis strains bearing pKOR1_Δpnp , pKOR1_ΔrnjB , pKOR1_ΔpnpΔrnjB , or pKOR1_pnp* were grown overnight at 30°C . Overnight cultures were streaked onto BHI agar plates containing neomycin and chloramphenicol , and incubated overnight at 43°C to force plasmid integration into the genome . Several colonies were then inoculated into separate tubes containing fresh BHI broth supplemented with neomycin , and incubated overnight at 30°C with agitation in order to force plasmid excision from the genome . The overnight cultures were diluted 100 , 000 times with sterile water , and 100 µL of diluted culture was plated onto BHI agar plates containing neomycin and anhydrotetracycline ( 50 ng/µl ) and incubated overnight at 37°C to force plasmid loss . Several colonies of different sizes were replica-plated on BHI agar containing neomycin alone and on plates containing neomycin plus chloramphenicol to select against colonies that have lost the plasmid . Only colonies that lost the plasmid ( i . e . did not grow on the latter plate ) were subjected to PCR amplification and sequencing of the PCR product to confirm the presence of the intended mutation ( Figure 2—figure supplement 1 and Supplementary file 1 ) . Confirmed mutants were then purified by streaking and selecting single colonies over three consecutive nights . The complementation plasmid pcrispr-cas-rnjB was constructed using a two-piece Gibson assembly ( Gibson et al . , 2009 ) . Briefly , rnjB was amplified from the genome of S . epidermidis RP62a using primers L035 and L038 ( Supplementary file 1 ) and the backbone was generated in a PCR reaction with pcrispr-cas ( Hatoum-Aslan et al . , 2013 ) as template and primers L036 and L039 . PCR products were purified using the EZNA Cycle Pure Kit and Gibson assembled . The assembled construct was transformed via electroporation into S . aureus RN4220 . Four transformants were selected and confirmed to harbor the plasmid using PCR and DNA sequencing with primers L035 and L038 . Confirmed plasmids were purified using the EZNA Plasmid DNA Mini Kit and transformed into S . epidermidis LM1680 strains via electroporation . Conjugation was carried out exactly as described in Walker and Hatoum-Aslan ( 2017 ) . Phage challenge assays were carried out by spotting ten-fold dilutions of phage lysate atop a semisolid layer of 0 . 5 X HIA containing 5 mM CaCl2 and a 1:100 dilution of overnight host culture ( wild-type and mutant variants ) . Spots were air-dried , and plates were incubated overnight at 37°C . On the following day , pfu/ml values were determined . The conjugation and phage challenge data reported represents mean values ( ±S . D . ) of 3–4 technical replicates as a representative of several independent trials ( see Figure 3—source datas 1–3 for exact n values ) . Cas10-Csm complexes containing a 6-His tag on the N-terminus of Csm2 were purified from S . epidermidis LM1680/pcrispr-cas strains using Ni2+ affinity chromatography exactly as described in Chou-Zheng and Hatoum-Aslan ( 2017 ) . Complexes were resolved on a 15% SDS PAGE and visualized with Coomassie G-250 . Gels were imaged using the FluorChemR system ( Protein Simple , CA , USA ) . Total crRNAs were extracted from purified Cas10-Csm complexes using the TRIzol Reagent ( Invitrogen , NY , USA ) according to Hatoum-Aslan et al . ( 2014 ) . Extracted crRNAs were phosphorylated with T4 Polynucleotide Kinase ( New England Biolabs , MA , USA ) , radiolabeled with γ-[32P]-ATP ( PerkinElmer , MA , USA ) , and resolved on a 15% Urea PAGE gel . The gel was exposed for 10 min to a storage phosphor screen and visualized using a Typhoon FLA 7000 phosphor imager ( GE Healthcare Bio-Sciences , PA , USA ) . For densitometric analysis , the ImageQuant software was used . Percent of intermediate crRNAs was determined with the following equation: ( ( intensity of intermediate crRNA signal ( 71 nt ) ÷ sum of signal intensities for the dominant crRNA species ( 71 nt + 43 nt+37 nt+31 nt ) ) ×100% ) . The reported values represent an average of 3–5 replicates ( ±S . D . ) , as indicated in the Figure 2 legend and Figure 2—source data 1 . S . epidermidis RP62a cells were grown in TSB supplemented with 5 mM CaCl2 to an OD600 of 0 . 3–0 . 4 and infected with phage Andhra at a multiplicity of infection ( MOI ) of 0 . 5 . A control culture was also prepared by omitting Andhra . Control and bacteria:phage mixtures were incubated at 37°C for 10 min ( without agitation ) to allow phage absorption , and cells were pelleted at 5000 x g for 5 min . The supernatant was discarded , pellets were resuspended in 40 ml of fresh TSB , and pelleted again as above . Cells pellets were resuspended in 40 ml of fresh TSB . 10 ml of this culture was harvested immediately ( time = 0 ) , and the remaining culture was returned to 37°C with agitation . Additional 10 ml aliquots of culture were harvested in a similar manner at time = 10 , 20 , and 30 min post-infection . Harvested cultures were immediately heat-killed at 95°C for 10 min , pelleted , and washed with 10 ml of sterile water . Final cell pellets were stored at −80°C for nucleic acid extraction . To extract DNA , cell pellets were resuspended in 100 µl of sterile water , followed by 10 µg of lysostaphin ( AmbiProducts , NY , USA ) and 5 mM MgCl2 . The cell suspension was digested for 2 hr at 37°C . For DNA purification , The Wizard Genomic DNA Purification Kit ( Promega Corporation , WI , USA ) was used according to the manufacturer’s instructions . DNA was stored at 4°C . For RNA extraction , cell pellets were washed twice with 1 ml of TSM Buffer ( 50 mM Tris-HCl pH 7 . 5 , 0 . 5 M Sucrose , 10 mM MgCl2 ) and pelleted at 16 , 000 x g for 1 min . Cells were resuspended in 500 µL of TSM buffer supplemented with 10 µg of lysostaphin and incubated for 20 min at 37°C . Digested cells were pelleted and resuspended in 750 µL of TRI Reagent RT ( Molecular Research Center , OH , USA ) and 37 . 6 µL of 4-bromoanisole ( Molecular Research Center , OH , USA ) , and subsequent steps were completed as indicated in the manufacturer’s protocol . To eliminate trace genomic DNA , 10 µg of RNA was incubated with 3 units of DNase I ( New England Biolabs , MA , USA ) for 30 min at 37°C . RNA was extracted with an equal volume of phenol: chloroform: isoamyl alcohol ( 25:24:1 ) , vortexed for 15 s , and centrifuged at 10 , 000 x g for 2 min . The resulting aqueous phase was mixed with an equal volume of chloroform , vortexed and centrifuged as earlier . The aqueous phase was then combined with 1/10 vol of 3 M NaOAc , pH 5 . 2 , and 3 volumes of ethanol . Mixtures were incubated at −80°C for 30 min to precipitate the RNA . The RNA precipitate was pelleted and washed with 1 ml of 75% ethanol . RNA pellets were then air-dried for 10 min and resuspended in 10 µL of sterile water to obtain a final concentration of 1 µg/µL of RNA . cDNA was synthesized in 10 µL reactions containing 6 µg of RNA , annealing buffer ( 5 mM Tris-HCl pH 8 . 3 , 75 mM KCl , 1 mM EDTA ) , and 1 µM of reverse primer ( A475 and S002 ) . Mixtures were incubated at 95°C for 1 min and 48°C for 45 min . The mixtures were then combined with 4 µL of 5X synthesis Buffer ( 250 mM Tris-HCl pH 8 . 3 , 375 mM KCl , 22 . 5 mM MgCl2 , 75 mM DTT ) and 16 µL of Reverse Transcriptase Synthesis Mix ( 10 mM each of dATP , dCTP , dGTP , dTTP , and 100 U M-MuLV reverse transcriptase ) , and incubated at 37°C for 30 min . The resulting mix was used as template for qRT-PCR reactions as described below . PCR reactions ( 25 µL ) contained 1X PerfeCTa SYBR Green SuperMix ( Quanta Biosciences ) , 0 . 4 nM of phage-specific primers ( A474/A475 ) or genome-specific primers ( S001/S002 ) ( Supplementary file 1 ) , and 500 ng of total DNA ( for qPCR ) or 1 µL of prepared cDNA ( for qRT-PCR ) as template . Separate standard reactions containing 102–109 DNA molecules were also prepared using purified phage DNA extract or bacterial genomic DNA extract . A CFX Connect Real Time PCR Detection System ( Bio-Rad , CA , USA ) was used to amplify the DNA templates according to the following program: one cycle , 95°C for 3 min; 40 cycles , 95°C for 10 s and 55°C for 30 s . Melt curves were also generated to confirm homogenous product by exposing samples to a final temperature gradient of 65°C to 95°C . Relative fold difference was determined using the equation from the Real-Time PCR Handbook ( ThermoFisher Scientific , MA , USA ) : Fold Difference = ( Etarget ) Ct_target / ( Enormalizer ) ΔCt_normalizer , where E = 10 ( -1/slope ) , Ct_target = Ct_targetcalibratior – Ct_targetsamples , and ΔCt_normalizer = Ct_ normalizercalibratior – Ct_ normalizersamples . The glyceraldehyde-3-phosphate dehydrogenase ( gap ) gene was used for normalization . The Ct value of S . epidermidis RP62a at 0 min post-infection was used to calibrate the remaining Ct values . Specific numbers of replicates are found in appropriate figure legends . pET28b-His10Smt3-rnjA and pET28b-His10Smt3-rnjB were constructed using a two-piece Gibson assembly . Briefly , inserts were obtained by amplifying rnjA and rnjB from the genome of S . epidermidis RP62a and amplifying the backbone from a pET28b-His10Smt3 template using primers L030-L033 ( for the rnjB construct ) and L042-L045 ( for the rnjA construct ) ( Supplementary file 1 ) . PCR products were purified using the EZNA Cycle Pure Kit and Gibson assembled . pET28b-His10Smt3-drnjA , encoding a catalytically-dead variant of RNase J1 , was constructed via inverse PCR using primers L055 and L056 . PCR products were digested with DpnI ( New England Biolabs , MA , USA ) and purified using the EZNA Cycle Pure Kit . Purified PCR products were 5’- phosphorylated by incubating with T4 Polynucleotide Kinase at 37°C for 30 min , and then circularized by incubating with T4 DNA Ligase ( New England Biolabs , MA , USA ) overnight at room temperature . Assembled/ligated constructs were introduced into E . coli DH5α by chemical transformation . Four transformants for each construct were selected and confirmed to have the plasmid using PCR and DNA sequencing with primers T7P and T7T . Confirmed plasmids were purified using the EZNA Plasmid DNA Mini Kit and introduced into E . coli BL21 ( DE3 ) for protein purification . E . coli BL21 ( DE3 ) cells harboring the pET28b-His10Smt3-based plasmids were grown and induced exactly as previously described ( Walker et al . , 2017 ) . The downstream purification of recombinant proteins from cell pellets had slight modifications . Briefly , cell pellets were resuspended in 30 ml of Buffer A ( 50 mM Tris-HCl pH 7 . 0 , 1 . 25 M NaCl , 200 mM Li2SO4 , 10% sucrose , 25 mM Imidazole ) supplemented with one complete EDTA-free protease inhibitor tablet ( Roche ) , 0 . 1 mg/ml lysozyme , and 0 . 1% Triton X-100 . Cells were incubated at 37°C for 1 hr and sonicated . Insoluble material was removed via centrifugation and filtration . Cleared lysates were incubated with 3 ml of Ni2+-NTA agarose resin ( ThermoFisher Scientific , MA , USA ) pre-equilibrated with Buffer A , and incubated for 1 hr with constant rotation . The resin was pelleted and washed with 40 ml of Buffer A , followed by a 5 ml wash of 3 M KCl . The resin was then resuspended in 5 ml of Buffer A and transferred to a 5 ml gravity column ( G-Biosciences , MO , USA ) . The resin was further washed with 20 ml of Buffer A . Proteins were eluted stepwise with three aliquots of 1 ml each of IMAC buffer ( 50 mM Tris-HCl pH 7 . 0 , 250 mM NaCl , 10% glycerol ) containing 50 , 100 , 200 and 500 mM imidazole . Eluted protein fractions were resolved in a 15% SDS PAGE and visualized with Coomassie G-250 . The most concentrated fractions ( 3 ml total ) were pooled and mixed with SUMO Protease ( MCLAB , CA , USA ) and provided SUMO buffer ( salt-free ) . The mixtures were dialyzed for 3 hr against IMAC buffer containing 25 mM imidazole . The dialysate was mixed with 1 ml of Ni2+-NTA agarose resin ( pre-equilibrated with dialysis buffer ) , and incubated for 1 hr with constant rotation to collect the His10Smt3 tag . The digested dialysate was transferred into a 5 ml gravity column , and the untagged protein was collected in the flow-through . Additional untagged protein was collected by flowing through the column two 500 µL aliquots each of IMAC buffer containing 50 , 75 , 100 and 500 mM imidazole . Proteins were resolved and visualized as described above . Protein concentrations were determined using absorbance measurements at 280 nm with a NanoDrop2000 spectrophotometer ( ThermoFisher Scientific ) . Single stranded RNA or DNA substrates were labeled on their 5’-ends by incubating with T4 polynucleotide kinase and γ-[32P]-ATP , and purified over a G25 column ( IBI Scientific , IA , USA ) . Labelled substrates were combined with 0 . 5 pmol of enzymes alone , and in combination , in nuclease buffer ( 25 mM Tris-HCl pH 7 . 5 , 2 mM DTT ) supplemented with 10 mM MnCl2 ( unless indicated otherwise ) . For metal-dependent nuclease activity assays , nuclease reactions were carried out at 37°C for 10 min ( ssDNA ) , or 20 min ( ssRNA ) using 10 mM of EDTA , MgCl2 , MnCl2 , NiCl2 and ZnCl2 . For time course assays , nuclease reactions were carried out at 37°C for 0 , 2 , 5 and 10 min ( ssDNA ) , or 0 , 5 , 10 and 20 min ( ssRNA ) . Reactions were stopped by adding an equal volume of 95% formamide loading buffer and resolved on a 15% Urea PAGE . Gels were exposed to a storage phosphor screen and visualized using a Typhoon FLA 7000 phosphor imager . ImageQuant software was used for densitometric analysis . The fraction of cleaved substrate was determined using the following equation: intensity of cleaved substrate signal ÷ sum of intensities of cleaved plus uncut substrate signals . Values are representative of at least three independent trials . The double-stranded DNA substrate pUC19 ( New England Biolabs , MA , USA ) was linearized by digesting with PstI ( New England Biolabs , MA , USA ) and purified using the EZNA Cycle Pure Kit . Circular or linear pUC19 ( 0 . 1 µg ) was combined in nuclease buffer with 2 pmol of each nuclease alone and in combination . Circular ssDNA ( M13MP18 ) substrate ( 0 . 5 µg , New England Biolabs , MA , USA ) was combined with 4 pmol of each cellular nuclease alone and in combination , in nuclease buffer . Reactions were incubated at 37°C for 1 hr ( dsDNA ) or 15 , 30 and 60 min ( circular ssDNA ) . Reactions were stopped by incubating on ice for 5 min and digesting with 1 µg of Proteinase K Solution ( VWR , PA , USA ) for 15 min at room temperature . DNA was resolved on a 0 . 7% agarose gel containing ethidium bromide , and visualized under UV light using the FluorChemR system . Refer to the appropriate figure legends for specific replicate values . All graphed data represent the mean ( ±S . D . ) of n replicates , where n is indicated in figure legends and source data files . Average values were analyzed using a one-tailed t-test , and p values below 0 . 05 were considered statistically significant . No statistical methods were used to predetermine sample size . The following terms are used to describe the types of repetitions where appropriate in figure legends and source data files: technical replicate , independent trial , and biological replicate . Technical replicates refer to multiple measurements taken on a single cell line in a single experiment; independent trials refer to repetitions of the same experiment conducted at different times using the same cell lines; and biological replicates refer to independent trials conducted on different lines of cells in which each line comprises a transformant or mutant that was independently generated . | Just as humans are susceptible to viruses , bacteria have their own viruses to contend with . These viruses – known as phages – attach to the surface of bacterial cells , inject their genetic material , and use the cells’ enzymes to multiply while destroying their hosts . To defend against a phage attack , bacteria have evolved a variety of immune systems . For example , when a bacterium with an immune system known as CRISPR-Cas encounters a phage , the system creates a ‘memory’ of the invader by capturing a small snippet of the phage’s genetic material . The pieces of phage DNA are copied into small molecules known as CRISPR RNAs , which then combine with one or more Cas proteins to form a group called a Cas complex . This complex patrols the inside of the cell , carrying the CRISPR RNA for comparison , similar to the way a detective uses a fingerprint to identify a criminal . Once a match is found , the Cas proteins chop up the invading genetic material and destroy the phage . There are several different types of CRISPR-Cas systems . Type III systems are among the most widespread in nature and are unique in that they provide a nearly impenetrable barrier to phages attempting to infect bacterial cells . Medical researchers are exploring the use of phages as alternatives to conventional antibiotics and so it is important to find ways to overcome these immune responses in bacteria . However , it remains unclear precisely how Type III CRISPR-Cas systems are able to mount such an effective defense . Chou-Zheng and Hatoum-Aslan used genetic and biochemical approaches to study the Type III CRISPR-Cas system in a bacterium called Staphylococcus epidermidis . The experiments showed that two enzymes called PNPase and RNase J2 played crucial roles in the defense response triggered by the system . PNPase helped to generate CRISPR RNAs and both enzymes were required to help to destroy genetic material from invading phages . Previous studies have shown that PNPase and RNase J2 are part of a machine in bacterial cells that usually degrades damaged genetic material . Therefore , these findings show that the Type III CRISPR-Cas system in S . epidermidis has evolved to coordinate with another pathway to help the bacteria survive attack from phages . CRISPR-Cas immune systems have formed the basis for a variety of technologies that continue to revolutionize genetics and biomedical research . Therefore , along with aiding the search for alternatives to antibiotics , this work may potentially inspire the development of new genetic technologies in the future . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2019 | A type III-A CRISPR-Cas system employs degradosome nucleases to ensure robust immunity |
Crimean-Congo hemorrhagic fever ( CCHF ) is the most widely distributed tick-borne viral infection in the world . Strikingly , reported mortality rates for CCHF are extremely variable , ranging from 5% to 80% ( Whitehouse , 2004 ) . CCHF virus ( CCHFV , Nairoviridae ) exhibits extensive genomic sequence diversity across strains ( Deyde et al . , 2006; Sherifi et al . , 2014 ) . It is currently unknown if genomic diversity is a factor contributing to variation in its pathogenicity . We obtained complete genome sequences of CCHFV directly from the tick reservoir . These new strains belong to a solitary lineage named Europe 2 that is circumstantially reputed to be less pathogenic than the epidemic strains from Europe 1 lineage . We identified a single tick-specific amino acid variant in the viral glycoprotein region that dramatically reduces its fusion activity in human cells , providing evidence that a glycoprotein precursor variant , present in ticks , has severely impaired function in human cells .
Crimean-Congo hemorrhagic fever ( CCHF ) is severe human disease present in an increasing number of regions of Europe , Africa , and Asia ( Bente et al . , 2013 ) . In Turkey alone , more than one thousand endemic CCHF cases are reported annually . Outbreaks of CCHF are sporadic , and reported mortality rates are extremely variable ( 5–80% ) ( Whitehouse , 2004 ) . The CCHF etiological agent , CCHF virus ( CCHFV ) , is the most widespread tick-borne virus of medical importance and is primarily maintained in and transmitted by hard ticks of the Hyalomma genus ( Gargili et al . , 2017 ) . Human infections occur through tick bites or exposure to blood or other bodily fluids of infected animals or CCHF patients . CCHFV is classified in the family Nairoviridae in the genus Orthonairoviridae . CCHFV is a negative-sense , single-stranded RNA virus with a genome consisting of three segments called large ( L ) , medium ( M ) , and small ( S ) , which encode for the L RNA-dependent RNA polymerase , the glycoprotein precursor ( GPC ) , and the nucleoprotein ( NP ) , respectively . The GPC is post-translationally processed into several non-structural proteins and the structural glycoproteins , Gn and Gc , that mediate attachment of the virion to the host cell and fusion of the virion and host cell membranes ( Zivcec et al . , 2016 ) . To date , the cellular receptor ( s ) that mediate CCHFV entry are unknown . While a functional interaction between cell surface nucleolin and CCHFV Gc glycoprotein has been suggested ( Xiao et al . , 2011 ) , further studies are required to test the role of this interaction in the context of CCHFV cellular entry and infection . CCHFV strains exhibit great diversity at the RNA and protein sequence levels , and are divided into seven genetic lineages ( Africa 1 , 2 , and 3; Asia 1 and 2; Europe 1 and 2 ) . All lineages except Europe 2 are believed to be transmitted by Hyalomma spp . ticks and cause severe disease in humans . In contrast , the Europe 2 lineage is a phylogenetic outlier that is not usually associated with severe disease . The prototype strain of Europe 2 , AP92 , was isolated from Rhipicephalus bursa ticks collected in Greece in 1975 ( Deyde et al . , 2006; Papadopoulos and Koptopoulos , 1980 ) . Since then , Europe 2 strains have only been associated with three documented human cases , including one fatality in Iran and two mild cases in Turkey ( Midilli et al . , 2009; Salehi-Vaziri et al . , 2016; Elevli et al . , 2010 ) . Although CCHFV seroprevalence is relatively high in both humans and livestock in the Balkans ( Papa et al . , 2011; Sidira et al . , 2012; Sargianou et al . , 2013; Papa et al . , 2014; Papa et al . , 2016 ) , almost all clinical cases are caused by Europe 1 strains despite the circulation of Europe 2 strains in ticks and livestock . Thus , given the high CCHFV seroprevalence in areas of Europe 2 circulation and the low number of disease cases associated with these strains , the Europe 2 lineage may contain CCHFV strains that are less pathogenic in humans than strains of other lineages ( Papa et al . , 2011; Sidira et al . , 2012; Sargianou et al . , 2013; Papa et al . , 2016; Papa et al . , 2013; Sidira et al . , 2013 ) . The genomic and molecular properties of CCHFV that directly contribute to its transmission from the tick vector to the human host are largely unknown , partly because only one complete CCHFV genome sequence has been derived directly from a tick ( Cajimat et al . , 2017 ) . All other reported CCHFV genomes are derived from human patients . Thus , studies of the changes in CCHFV genomic signatures upon transmission from ticks to humans or other susceptible species have been precluded . While Europe 2 CCHFV strains have often been detected directly in ticks collected in the Balkans and in Turkey ( Sherifi et al . , 2014; Papa et al . , 2014; Panayotova et al . , 2016; Papa et al . , 2017; Dinçer et al . , 2017 ) , no full-genome of a Europe 2 strain has been derived directly from a tick to date . Only two Europe 2 CCHFV full-genome sequences have been described , the AP92 and Pentalofos strains , both detected in Greece ( Deyde et al . , 2006; Papa et al . , 2018 ) , but they were determined after passaging the virus in vertebrate culture systems . Thus , we sought to better address viral gene functions from tick-derived Europe 2 lineage CCHFV in human cells and in the context of virus transmission from ticks to humans by obtaining a Europe 2 CCHFV sequence directly from ticks .
To obtain a CCHFV sequence directly from ticks , ticks were collected in Strandja Nature Park , a unique pristine forest in Bulgaria . A total of 1541 ticks were collected from vegetation , humans , tortoises , and livestock . CCHFV was detected in three R . bursa ticks feeding on one cow from Malko Tarnovo village; the cow showed no obvious signs of disease . Similar CCHFV genome quantities ( 1 × 105 to 1 × 106 genome copies/ml ) were detected in all CCHFV-positive ticks , indicating that the three CCHFV strains replicated similarly in ticks . Complete CCHFV genomic sequencing directly from tick homogenates and phylogenetic analyses revealed that the CCHFV strains obtained from these three ticks , tentatively named Malko Tarnovo-BG2012-T1302 ( MT-1302 ) , Malko Tarnovo-BG2012-T1303 ( MT-1303 ) , and Malko Tarnovo-BG2012-T1362 ( MT-1362 ) , grouped with other strains of the Europe 2 lineage ( Figure 1A , Figure 1—figure supplement 1 ) . These Malko Tarnovo strains establish a novel clade and share a common ancestor with strains recently detected in Bulgaria and Greece . Compared to the AP92 and Pentalofos strains , the Malko Tarnovo S segment nucleotide sequences showed 91 . 9% and 96 . 5% , the M segment , 86 . 9% and 96 . 3% , and the L segment , 96 . 2% and 96 . 2% identity , respectively . A detailed comparison between genomes of the Malko Tarnovo strains and prototypes of all lineages , as well as amino acid changes between AP92 and Pentalofos strains , are listed in Supplementary files 1 and 2 . The Malko Tarnovo sequences are the first full-genome sequences of Europe 2 CCHFV strains obtained from a tick without passaging the virus in a vertebrate culture system ( Supplementary file 3 ) . In order to analyze the phenotypic properties of the newly identified Malko Tarnovo strains , we attempted virus isolation experiments in different tick and vertebrate cell lines . Unfortunately , we failed to isolate any of the three strains as virus could not be maintained through serial passaging in any of the tested cell lines ( Figure 1—figure supplement 2 ) . Next , we examined the ability of Malko Tarnovo strain proteins to support viral replication in human cells . We generated constructs expressing MT-1303 nucleoprotein ( NP ) , glycoprotein precursor ( GPC ) , and L protein , and tested these in a transcription- and entry-competent virus-like particle ( VLP ) assay . When cells are transfected with these expression constructs , the CCHFV proteins are expressed , processed , and assembled into virus-like particles ( VLPs ) that can bud from the host cell ( Zivcec et al . , 2015 ) . These particles can package a minigenome reporter RNA consisting of viral 5′ and 3′ untranslated regions ( UTRs ) flanking a reporter gene ( NanoLuc ) that is transcribed by the viral polymerase upon VLP entry . The UTRs serve as transcription and replication signals specific for CCHFV L-RNA polymerase ( Bergeron et al . , 2010 ) . In addition , NP encapsidation of the CCHFV minigenome RNA is required for its efficient transcription and replication . The GPC is required to generate VLPs morphologically similar to authentic CCHFV ( Zivcec et al . , 2015 ) . VLPs can be transferred to fresh recipient cells , and reporter activity can be read in the recipient cell as a proxy for VLP infectivity . Thus , the VLP assay tests the contribution of NP and L to transcription and replication of viral RNA and the role of GPC in assembly , release , and entry of virus particles . To assess the relative contribution of the individual MT-1303 proteins to VLP activity , we generated VLPs with different combinations of viral proteins from the MT-1303 and prototypic IbAr10200 strains . Surprisingly , VLPs generated with all three MT-1303 proteins failed to generate robust reporter activity when passaged onto naive human cells ( Figure 1B ) . VLPs containing MT-1303 NP or L protein individually yielded high reporter activity when complemented with IbAr10200 proteins . These results indicate that the MT-1303 L and NP proteins can recognize their own UTRs and those from the cell culture-adapted IbAr10200 strain to effectively drive transcription . Indeed , MT-1303 L protein was capable of driving transcription from minigenomes derived from other divergent CCHFV strains , though less efficiently than the IbAr10200 L protein ( Figure 1—figure supplement 3 ) . Strikingly , MT-1303 GPC failed to produce robust VLP reporter activity with any combination of NP and L protein ( Figure 1B ) . In contrast , GPCs from 11 other CCHFV strains representative of all other phylogenetic clades produced high VLP reporter activity ( Zivcec et al . , 2017 ) , and thus attenuation of VLP activity was mainly attributed to the MT-1303 GPC ( Figure 1—figure supplement 4 ) . Expression of GPC constructs harboring an N-terminal FLAG-tag were comparable between IbAr10200 and MT-1303 strains , indicating that entry deficiency was not due to differences in cellular GPC protein levels ( Figure 1—figure supplement 5A ) . We thus concluded that GPC of MT-1303 was expressed normally yet failed to yield robust VLP activity in human cells . To test which region of the MT-1303 GPC is responsible for poor VLP activity , we generated chimeric constructs of MT-1303 and IbAr10200 GPCs by exchanging the PreGc region ( Figure 2A ) . IbAr10200 GPC could not support robust VLP reporter activity when it contained the MT-1303 PreGc region ( Figure 2B; two-tailed Student’s t-test , p=0 . 0489 ) . Conversely , reporter activity levels from VLPs generated with the MT-1303 GPC harboring the IbAr10200 PreGc region were comparable to those from complete IbAr10200 GPC ( Figure 2B; two-tailed Student’s t-test , p=0 . 7902 ) . These results indicate that the MT-1303 PreGc region precludes robust VLP activity . Next , we investigated whether specific amino acid variations in the MT-1303 PreGc protein sequence correlated with poor VLP reporter activity . We compared the amino acid sequences just following the PreGc cleavage site ( RKPL↓ ) from the MT-1303 strain to other Europe 2 lineage CCHFV strains and 11 other diverse strains that yield robust VLP activity ( Zivcec et al . , 2017; Supplementary file 2 ) . We identified two amino acids in the PreGc region , Gly1116 and Met1129 , that were unique to MT-1303 . The MT-1303 strain contains a glycine ( G ) residue at position 1116 , while all other publicly available sequences of strains obtained from human samples or mammalian cell culture virus isolates have an arginine ( R ) residue at the corresponding position , indicating a high degree of conservation at this position . MT-1303 also contains a methionine ( M ) residue at position 1129 , whereas other strains possess either an isoleucine ( I ) or valine ( V ) ( Figure 2C ) . To test the potential functional effects of these unique MT-1303 amino acid variants on viral activity , we generated constructs expressing the MT-1303 GPC harboring single point mutations at positions 1116 or 1129 , changing the amino acid at either position to the consensus residue ( G1116R and M1129I ) . Strikingly , G1116R completely restored VLP reporter activity of MT-1303 GPC to levels similar to those of wild-type IbAr10200 GPC ( Figure 2D; two-tailed Student’s t-test , p=0 . 8397 ) . Consistently , mutating the IbAr10200 GPC residue 1105 , which corresponds to MT-1303 residue 1116 , from R to G reduced reporter activity to levels comparable to those generated by wild-type MT-1303 GPC . Thus , reduced VLP activity of the MT-1303 GPC can be mainly attributed to a single amino acid variant in the Gc region . The MT-1303 GPC M1129I mutation only modestly increased VLP reporter activity compared to wild-type MT-1303 GPC , and the inverse mutation did not affect IbAr10200 GPC VLP activity ( Figure 2D ) . We next searched all CCHFV PreGc sequences deposited in GenBank to discover any other variants at the position corresponding to MT-1303 GPC 1116 . Strikingly , all other sequences had an arginine at this position except for one other strain , MG48 , which contained a lysine ( K ) ( Figure 2C ) . The CCHFV-MG48 partial sequence also belongs to the Europe 2 lineage and was derived directly from a tick sample recently collected in the neighboring country of Turkey ( Dinçer et al . , 2017 ) . Importantly , this finding shows that variation at this position is detectable in ticks , whereas CCHFV isolated from mammalian cells or human samples does not contain this variation , possibly because variants with R at this position are preferentially amplified . To test whether substituting K at this position could support VLP activity in human cells , we assessed reporter activity of VLPs generated with the MT-1303 GPC with the G1116K mutation . Like G1116R , the G1116K mutation allowed high levels of reporter activity comparable to those produced by wild-type IbAr10200 GPC ( Figure 2D; two-tailed Student’s t-test , p=0 . 9028 ) , suggesting that a positively charged residue at this position is important for CCHFV GPC function in human cells . To confirm these VLP results in the context of replicating virus , we attempted to generate recombinant fluorescent reporter CCHFV using the wild-type MT-1303 M segment or mutant M segment with the consensus R residue using a CCHFV reverse genetics system ( Bergeron et al . , 2015 ) . In this system , cells are transfected with several plasmid constructs that encode the individual CCHFV genome segments under control of the T7 promoter , as well as helper plasmids . To track viral replication , the CCHFV S genome segment is replaced with a variant that also encodes the ZsGreen1 fluorescent reporter ( Welch et al . , 2017 ) . Thus , the CCHFV reverse genetics systems allows functional testing of individual genome segments as well as discrete molecular variants in genome or protein sequence in the context of infectious CCHFV . When complemented with IbAr10200 S and L genome segments , rescue attempts in Huh7 cells were only successful with the MT-1303 M segment encoding GPC-G1116R . Consistent with the VLP results , these data indicate that the MT-1303 M genome segment supports viral replication in human cells when R is present at position 1116 , but not when the tick-associated G1116 variant is used . We next attempted to rescue MT-1303 or reassortant CCHFV with IbAr10200 segments . Viruses containing either MT-1303 S , L or mutated M-G1116R could be obtained , emphasizing the importance of R1116 presence in Gc to recover MT-1303 in mammalian cells . MT-1303/IbAr10200 reassortant viruses sequences were confirmed to match the DNA templates used in the virus rescue transfections . The growth kinetics of the recombinant MT-1303 M-G1116R reassortant virus was analyzed in human ( Huh7 , A549 ) , hamster ( BSR-T7/5 , CHO ) , and tick ( ISE6 ) cell lines . Virus generated with MT-1303 M-G1116R or IbAr10200 M displayed similar growth kinetics in hamster cells ( Figure 2E ) . In contrast , MT-1303 M-G1116R exhibited reduced viral replication in human cells compared to IbAr10200 M . In tick cells , replicating MT-1303 M-G1116R virus was detected in all tested wells ( 3/3 ) , while IbAr10200 M virus infected only one out three biological replicates . The single well of tick cells infected with IbAr10200 M virus nevertheless showed delayed growth and reduced viral titers . These results suggest that the M segment of the MT-1303 strain is more successful at infecting tick cells as opposed to IbAr10200 whose replication was more robust in human cells . Given that the G1116 GPC variant is specific to the MT-1303 sequence obtained directly from a tick source , we next tested whether this variant mediated entry in tick cells . In Huh7 cells , we generated VLPs containing either the IbAr10200 , IbAr10200 R1105G ( corresponding amino acid to M-G1116 in MT-1303 ) , MT-1303 , or MT-1303 G1116R GPC . These VLPs were then transferred to fresh Huh7s or ISE6 cells , and luciferase signal in the recipient cells was measured to assess VLP entry . VLPs containing the MT-1303 GPC did not exhibit higher activity in ISE6 cells compared to Huh7 cells ( Figure 2—figure supplement 1 ) . One possible explanation for this result is that the G1116 variant renders the GPC temperature-sensitive , functioning more efficiently at lower temperatures required for maintenance in tick cells . To test this possibility , we performed VLP production and VLP incubation with Huh7 or ISE6 recipient cells at lower temperatures to mimic tick cell temperatures . VLPs containing MT-1303 GPC produced and incubated at 28°C did not restore VLP activity relative to the IbAr10200 GPC VLPs ( Figure 2—figure supplement 2 ) . These results suggest that impaired entry activity of the MT-1303 GPC is not due to temperature sensitivity . We next sought to identify the replication step ( s ) associated with poor infectivity of MT-1303 strain with residue G1116 in human cells . CCHFV Gc must be exported out of the endoplasmic reticulum to ultimately mature in the Golgi apparatus , the site of CCHFV assembly ( Zivcec et al . , 2016 ) . First , we examined the subcellular localization of Gc and observed that both IbAr10200 and MT-1303 Gc proteins co-localized with giantin , a marker of the Golgi apparatus ( Figure 1—figure supplement 4 ) , suggesting proper folding and normal trafficking of MT-1303 Gc to the site of CCHFV assembly . We next investigated whether MT-1303 Gc was being incorporated into membrane enveloped particles . To assess viral particle formation , cells were transfected with constructs expressing C-terminally V5-tagged GPC proteins ( Figure 3A ) . The C-terminal tail of the Gc glycoprotein is located within the CCHFV particle ( Figure 3B ) . Cell supernatants were ultracentrifuged through a sucrose cushion and the pellets were treated with trypsin to assess the incorporation of V5-tagged Gc into lipid-membrane-protected particles . The V5 epitope was detected in the pelleted fraction from either IbAr10200 or MT-1303 GPC and corresponded to the mature Gc glycoprotein ( Figure 3C ) . V5-containing fragments in these pellet fractions were protected from proteolytic treatment , and this protection was lost after solubilization of the VLP envelope with detergent ( Figure 3C ) . Together , these data indicate that wild-type MT-1303 GPC is able to promote formation of enveloped VLPs similarly to IbAr10200 GPC . Since Gc is believed to mediate entry by fusion of the viral envelope with the host membrane , we tested whether MT-1303 Gc could promote membrane fusion . Huh7 cells were transfected with an expression plasmid encoding IbAr10200 or MT-1303 GPC along with a plasmid encoding T7 polymerase . These cells were then incubated at acidic pH before co-culture with cells transfected with a T7 promoter-driven GFP reporter plasmid . In this assay , GPC-mediated cell-cell fusion results in the formation of fluorescent syncytia as transcription of GFP reporter only occurs after cell fusion . Wild-type and N-terminally FLAG/HA-tagged IbAr10200 GPC efficiently promoted the formation of large syncytia when GPC-expressing cells were treated with pH 6 media ( Figure 4A and B ) . Closer examination of individual GFP foci revealed that IbAr10200 GPC-induced GFP foci represented large cytoplasmic areas containing multiple nuclei , indicative of membrane fusion of individual cells ( Figure 4A ) . Large syncytia were largely absent when MT-1303 GPC was tested compared to IbAr10200 GPC ( Figure 4A and B; two-tailed Student’s t-test , p<0 . 0001 ) . However , cells expressing the MT-1303 G1116R GPC mutant promoted the formation of large GFP-positive syncytia comparably to IbAr10200 GPC ( Figure 4A and B ) , although these syncytia were slightly smaller in area ( Figure 4C ) . These data indicate that GPC R1116 is critical for membrane fusion activity , and the G1116 variant severely impairs this activity . Consistent with this , modifying the corresponding position in IbAr10200 GPC to a G residue ( R1105G ) severely impaired membrane fusion activity ( Figure 4A and B; two-tailed Student’s t-test , p<0 . 0001 ) . To test whether a lower pH treatment could rescue fusion activity of the MT-1303 GPC , we performed additional fusion assays at pH 4 and pH 5 . Fusion events mediated by the MT-1303 GPC were not restored at these pH levels ( Figure 4—figure supplement 1 ) . Interestingly , we observed a marked decrease in MT-1303 G1116R GPC-mediated fusion activity at pH 5 compared to pH 6 , suggesting that while this mutation can restore fusion activity in the MT-1303 strain , it does not exhibit the same range of pH tolerance as the IbAr10200 GPC . Together , these results indicate that wild-type MT-1303 Gc severely attenuates membrane fusion activity in human cells , leading to poor infectivity . Therefore , we conclude that the tick-associated MT-1303 GPC variant G1116 expressed in human cells poorly supports entry of cells in general .
Most complete genome sequences of CCHFV are derived from human virus isolates typically performed by passaging in mammalian cell lines or in newborn suckling mice . Virus isolation may result in the selection of virus variants that might affect the virus biology . To circumvent potential biases , we obtained three independent CCHFV sequences from ticks collected in South-Eastern Bulgaria . Although this study does not directly address the pathogenicity of CCHFV Malko Tarnovo , it revealed that certain genetic variants found in ticks have impaired glycoprotein activity in human cells compared to tick cells . Even more dramatic was the discovery that strain MT-1303 GPC fusion activity is severely impaired by the absence of a positively charged amino acid at position 1116 . Unlike mammalian species , CCHFV infection in ticks can be lifelong and results in expansion of the intra-host viral genetic diversity ( Xia et al . , 2016 ) . This could lead to the generation of CCHFV variants like the MT-1303 GPC variant G1116 , which would likely be poorly transmitted and amplified by vertebrate host species or would not cause severe disease in humans . Vertebrates can serve as amplification hosts , supporting CCHFV transmission to naive ticks feeding on the same animal . Interestingly , all the CCHFV-infected ticks in this study were collected from the same animal , which might indicate transmission of CCHFV between the ticks during feeding . G1116 may limit CCHFV amplification by vertebrate hosts and consequently reduce transmission to uninfected ticks feeding on the same animal . To gain insights into possible Malko Tarnovo strain transmission between ticks , we analyzed single-nucleotide variations between the consensus genome sequences of the collected CCHFV-positive ticks . Sequences from ticks MT-1302 and MT-1362 were identical , but they differed from the sequence derived from MT-1303 by three nucleotides located in the coding region of the M and L segments; MT-1303 sequences included G1116R in the GPC and A1288T and I1502V variations in the L protein . Reconstruction of the chain of infection is difficult . All three ticks contained similar amounts of CCHFV genome copies ranging from 1 . 93 × 105 – 3 . 17 × 106 copies per mL indicating viral amplification in all ticks . The ticks may either have become infected independently of one another from other sources or from the cow or one of the ticks infected the others during co-feeding on the same host . Since all other ticks sampled in this study were negative for CCHFV , the most parsimonious model would be that the three ticks were infected from the cow or during co-feeding on the cow . Since no blood was collected from the cow upon which the three CCHFV-positive ticks were feeding , whether any of these Malko Tarnovo variants were amplified by the vertebrate host is unknown . However , the strong requirement in mammalian host systems ( recombinant CCHFV and VLPs ) for a basic amino acid at position 1116 argues in favor of preferential amplification of the M segment from MT-1302 and MT-1362 over that from MT-1303 , which has the G1116 . Thus , one explanation could be that MT-1303 was infected first and that the R1116 variant was positively selected and enabled replication in the cow . The two ticks MT-1302 and MT-1362 may then have become infected afterwards with the positively selected variant R1116 . Of note , the glycoproteins of MT-1302 and MT-1362 were sequenced by next-generation sequencing and position 1116 did not show any nucleotide variations . Unfortunately , MT-1303 could not be re-sequenced by next-generation sequencing due to paucity of sampling material . More studies will be required to fully appreciate the dynamics and importance of CCHFV genetic variations in viral amplification and transmission in vertebrate and tick hosts and the extent to which this influences CCHFV genome evolution and pathogenicity . Further , characterization of MT-1303 sequence revealed that the wild-type S and L segment sequence could support IbAr10200 strain replication and efficient VLP production , suggesting that MT-1303 M encoding a GPC with G1116 would likely be under more stringent selective constraints to change into an arginine or a lysine as opposed to S and L which supported virus replication in mammalian cells . One limitation of this study resides in the difficulty of addressing MT-1303 GPC function in a tick cell environment due to the technical hurdle of manipulating tick cells to express exogenous proteins . Future studies will require the development of improved methods to transfect and express proteins in tick cells . In summary , we generated and characterized the full genomic sequences of tick-derived CCHFV strains belonging to the Europe 2 genetic lineage . The M segment was associated with poor CCHFV replication in human cells , contrasting efficient infection of tick cells . The reduced infectivity of the MT-1303 strain in human cells was attributed to reduced Gc-meditated membrane fusion activity . This defect is likely associated with the incompatible cellular environment of human cells as VLPs derived from Huh7 cells were not more efficient at infecting tick cells . Together , our study identifies the presence of CCHFV variants produced by ticks that are likely associated with reduced infectivity in humans . We propose that the detection of CCHFV variants lacking fitness in human cells might contribute to asymptomatic CCHFV infection .
The Strandja Nature Park represents the last remaining temperate forest with evergreen plants in Europe that was not reached by the land-ice during the last ice ages in the Pleistocene and Holocene epochs . It has primeval flora from the Paleogene and Neogene periods , including pontic rhododendron ( Rhododendron ponticum ) , oriental beech ( Fagus orientalis ) , and various oak species . 1541 ticks were collected by flagging from the vegetation and sampling from livestock , tortoises , and humans in the area of Strandja Nature Park from May to August 2012 . Sampling sites were located in the towns of Stoilovo , Silkosiya , Sredoka , Kosti , Bulgari , Sinemorets , Zvezdets , and Malko Tarnovo . Collected ticks were individually cryoconserved . Sex and species were morphologically identified using taxonomic keys ( Babos , 1964; Walker , 2014 ) . Adult ticks were individually homogenized in 500 µL L-15 medium without additives using six steel beads and a SpeedMill PLUS homogenizer ( Analytik Jena AG , Germany ) . For nymphs and larvae , 200 μL media and 10 ceramic beads were used . Homogenization was performed in two to three cycles for 2 min at a frequency of 30 pulses/sec . The suspension was cleared by centrifugation at 2500 rpm for 10 min at 4°C . Pools were generated by combining 100 µL supernatants of 10 homogenized ticks each according to species , life stage , and sampling site . Viral RNA was extracted using the QIAamp Viral RNA Mini Kit following the manufacturer’s instructions . RNA was transcribed in cDNA using Superscript III reverse transcriptase and random hexamer primers ( Invitrogen GmbH , Karlsruhe , Germany ) . Pools were screened by PCR using primers based on a fragment of the S-segment of CCHFV as described before ( Bergeron et al . , 2015 ) . Subsequently , supernatants of homogenates of individual ticks composing the PCR-positive pools were tested by PCR as described above . PCR products were sequenced by Seqlab ( Göttingen , Germany ) . Sequences were analysed using Geneious Pro v9 ( https://www . geneious . com ) and compared to other sequences using the NCBI Basic Local Alignment Tool ( Altschul et al . , 1990 ) . For Sanger sequencing , genomes were amplified from tick homogenate using a combination of fragment-specific primers and primers based on the sequence of CCHFV strain AP92 ( accession numbers DQ211638 , DQ211625 , DQ211612 ) . The 3′ and 5′ genomic termini were amplified by rapid amplification of cDNA ends-PCR ( RACE-PCR; Roche , Mannheim , Germany ) . Sequencing was performed by SeqLab . Next-generation sequencing based on long-range PCR products was used for in-depth genomic analyses , using specific primers , the Phusion PCR kit ( Thermo-Fisher ) , and a MiSeq platform ( Illumina ) . Genomes were assembled by reference mapping to the Sanger genomes using Geneious Pro v9 . Sequences were analysed in Geneious Pro v9 . Signal sequence cleavage sites were identified using SignalP version 4 . 1 ( Petersen et al . , 2011 ) . N , GPC , and RdRp genes were aligned using MAFFT , and maximum likelihood ( ML ) trees were inferred based on 1000 bootstrap iterations in RAxML . All cell cultures were checked for mycoplasma every 20 passages . Freshly seeded 4 × 104 Vero E6/7 ( African green monkey kidney , ATCC , cell identity has been authenticated by STR profiling by ATCC ) , 4 × 104 SW13 ( Human Adrenal gland cortical small cell carcinoma , gift from DA Bente , University of Texas Medical Branch , originally obtained from ATCC , cell identity has been authenticated by STR profiling by ATCC ) , 2 . 5 × 105 HAE/CTVM8 ( Hyalomma anatolicum embryo , Tick Cell Biobank , University of Liverpool , cell identity has been authenticated by Tick Cell Biobank [Bell-Sakyi , 1991] ) and 1 . 8 × 105 BDE/CTVM16 ( Rhipicephalus ( Boophilus ) decoloratus embryo , Tick Cell Biobank , University of Liverpool , cell identity has been authenticated by Tick Cell Biobank [Bell-Sakyi , 2004] ) cells in 48-well-plates were inoculated with the supernatant of CCHFV-positive ticks and CCHFV-positive pools filtrated through a 0 . 45 μm filter under S3 laboratory conditions . Each cell line was inoculated with 20 µL , 4 µl and 0 . 4 µL of the respective supernatant and L-15 media without additives was added up to a final volume of 200 µL . After 1 hr of incubation at 37°C and 5% CO2 ( VeroE6/7 and SW13 cells ) or 28°C ( HAE/CTVM8 and BDE/CTVM16 cells ) 300 µL of cell line specific medium was added . Cells were observed daily for signs of a cytopathic effect ( CPE ) . Eight days post-infection , 100 µL of supernatants of cell lines SW13 and VeroE6/7 were passaged onto fresh cells . This was repeated five times . The remaining supernatant was centrifuged at 1000 × g for 5 min and stored at −80°C . Infected tick cells were observed for a period of 70 days . The sequences were deposited under the GenBank accession numbers: MK299338 - CCHFV strain Malko Tarnovo-BG2012-T1302 ( MT-1032 ) segment S; MK299339 - Malko Tarnovo-BG2012-T1302 ( MT-1302 ) segment M; MK299340 - Malko Tarnovo-BG2012-T1302 ( MT-1302 ) segment L; MK299341 - Malko Tarnovo-BG2012-T1303 ( MT-1303 ) segment S; MK299342 - Malko Tarnovo-BG2012-T1303 ( MT-1303 ) segment M; MK299343 - Malko Tarnovo-BG2012-T1303 ( MT-1303 ) segment L; MK299344 - Malko Tarnovo-BG2012-T1362 ( MT-1362 ) segment S; MK299345 - Malko Tarnovo-BG2012-T1362 ( MT-1362 ) segment M; MK299346 - Malko Tarnovo-BG2012-T1362 ( MT-1362 ) segment L . Huh7 ( Apath , LLC , cell identity has been authenticated by Apath , LCC ) , BSR-T7/5 ( gift from KK Conzelmann , Ludwig-Maximilians-Universität , Munich , Germany ) , and A549 ( ATCC , cell identity has been authenticated by STR profiling by ATCC ) cells were cultured in Dulbecco’s modified eagle media ( DMEM ) supplemented with 5–10% ( v/v ) fetal bovine serum , 1% ( v/v ) non-essential amino acids , 1 mM sodium pyruvate , 2 mM GlutaMAX , and 100 U/mL penicillin/streptomycin at 37°C and 5% CO2 . CHO-K1 cells ( ATCC , cell identity has been authenticated by STR profiling by ATCC ) were cultured in Ham’s F12 media supplemented with 10% fetal bovine serum , sodium pyruvate and antibiotics . ISE6 cells ( gift from U Munderloh , University of Minnesota ) were cultured in L-15B300 supplemented with 10% fetal bovine serum and 5% tryptose phosphate broth at 34°C and 0% CO2 ( Munderloh et al . , 1994 ) . Cell lines were occasionally checked to confirm absence of mycoplasma . Minigenome assays were performed in Huh7 cells as previously described ( Zivcec et al . , 2015 ) . Briefly , Huh7 cells were transfected with plasmids expressing CCHFV NP and L proteins and the T7 polymerase along with a minigenome reporter . For the MT-1303 minigenome , a DNA sequence containing the nanoluciferase ( nLuc ) coding sequence flanked by the UTRs from the MT-1303 L segment was synthesized ( IDT DNA ) and cloned into the V ( 0 . 0 ) plasmid backbone ( Bergeron et al . , 2010 ) . The NP open reading frame and human codon-optimized L coding sequence of the MT-1303 strain were synthesized ( Genscript ) , amplified by PCR , and inserted into the mammalian expression vector pCAGGS . All transfections were performed with LT1 ( Mirus Bio ) according to the manufacturer’s recommendations . nLuc luminescence was determined in technical triplicate on a Synergy 4 plate reader ( BioTek ) as a measure of genome amplification 2 days post-transfection; firefly luciferase ( pGL3; Promega ) was used as a transfection control . VLPs were generated following the steps above with the addition of a pCAGGS expression plasmid harboring a codon-optimized coding sequence for the GPC during the transfection step . Three days post-transfection , supernatants were harvested and clarified by spinning at 1500 × g for 5 min at room temperature . Clarified supernatants were passaged onto fresh Huh7 or ISE6 cells . nLuc activity was read the following day and normalized to firefly luciferase signal from the transfected cells . Recombinant IbAr10200 CCHFV was generated in the biosafety level four facilities at the Centers for Disease Control and Prevention ( Atlanta , GA , USA ) as described previously ( Bergeron et al . , 2015 ) . Briefly , Huh7 cells were transfected with plasmids encoding the S , M and L genome segments under control of the T7 promoter , and helper plasmids encoding T7 polymerase , CCHFV NP , and human codon-optimized L . Supernatants were harvested 4–7 days post-transfection . ZsGreen1-expressing reporter viruses were generated by replacing the S segment with a variant that also contains the ZsGreen1 gene ( Welch et al . , 2017 ) . The MT-1303 M genomic segment was synthesized by GenScript , and the M-G1116R variant was generated using site-directed mutagenesis . Sequences of all recombinant viruses were confirmed by next-generation sequencing ( Illumina ) . Signal peptides of the human codon-optimized pCAGGS-GPC coding sequence of the IbAr10200 or MT-1303 CCHFV strains were replaced with the Gaussia luciferase signal peptide ( MGVKVLFALICIAVAEAK ) followed by the FLAG epitope , a tobacco etch virus cleavage site , and 3 HA epitope repeats . PreGc chimeric constructs were generated by amplifying GPC regions from the parent IbAr10200 and MT-1303 strains by PCR and assembling regions using InFusion cloning . The GPC breakpoint for the PreGc chimeras was after V996 in the IbAr10200 strain ( corresponding to V1007 in MT-1303 strain ) , which lies 41 amino acids upstream of the RPKL site-1 protease recognition sequence . GPC point mutant constructs were generated from parent GPC vectors using site-directed mutagenesis primers coupled with InFusion assembly . Primer sequences are listed in Supplementary file 4 . Huh7 cells were transfected with pCAGGS plasmids expressing IbAr10200 or MT-1303 strain GPC and fixed with 1:10 buffered formalin 24 hr post transfection . Cells were permeabilized with 1 × PBS with 0 . 1% Triton-X100 and blocked with 1 × PBS and 3% BSA . Mouse monoclonal 11E7 was obtained from the Joel M . Dalrymple-Clarence J . Peters USAMRIID Antibody Collection through BEI Resources ( National Institute of Allergy and Infectious Diseases , National Institutes of Health ) , and was used to detect Gc . A rabbit anti-giantin polyclonal antibody ( Covance ) was used to detect the cis and median Golgi cisternae . Primary antibodies were detected with Alexa Fluor 488- and 594-conjugated secondary antibodies , and cells were imaged on a Cytation5 imager ( Biotek ) at constant power , integration , and camera gain within each channel . Huh7 cell lysates were prepared in Passive Lysis Buffer ( Promega ) and proteins were separated on 4–12% bis-tris SDS-PAGE gels . Proteins were transferred to nitrocellulose membranes and blocked in PBS-Tween and 5% milk . Mouse monoclonal anti-FLAG M2 ( Sigma-Aldrich ) and anti-V5 ( Thermo-Fisher ) antibodies were used to detect N-terminal FLAG-tagged and C-terminal V5-tagged GPC . Anti-mouse HRP-conjugated secondary antibody ( Sigma-Aldrich ) was used to label primary antibodies , and protein bands were detected using SuperSignal West Dura Fast Western blotting reagent on a Bio-Rad ChemiDoc MP imaging system . An α-tubulin ( Sigma-Aldrich ) antibody was used as a loading control marker . Huh7 cells were transfected with expression constructs encoding C-terminal V5-tagged IbAr10200 or MT-1303 GPC proteins or with empty vector control . After 4 days , supernatants were harvested and clarified by centrifugation at 1500 × g for 10 min . Clarified supernatants were layered onto a 20% sucrose cushion and centrifuged at 96 , 800 × g for 2 hr at 4°C using the SW-41 Ti rotor ( Beckman Coulter ) . Pellets were dried at room temperature for 10 min and resuspended in 1 × PBS . Aliquots of the pellet sample were either left untreated or were treated with 1 mg/mL trypsin in the presence or absence of 1% Triton X-100 for 1 hr at 37°C . Digestion was quenched by the addition of complete protease inhibitor ( Roche ) to 1 × final concentration for 2 min at room temperature . Samples were mixed with Laemmli sample buffer with 2-mercaptoethanol ( Bio-Rad ) to 1 × final concentration , boiled for 5 min at 98°C , and analyzed by Western blotting as described above . Huh7 cells were transfected with pCAGGS plasmids expressing IbAr10200 or MT-1303 strain CCHFV GPC or with an empty plasmid control , along with an expression plasmid encoding T7 polymerase . Transfected cells were then incubated for 15 min at 37°C in DMEM acidified to pH 6 , 5 , or 4 with 2N HCl or in neutral pH DMEM ( pH 7 . 4 ) . Huh7 cells transfected with a reporter plasmid encoding GFP under the control of the T7 promoter were dissociated with 1 × PBS and 0 . 5 mM EDTA and resuspended in DMEM at neutral pH . Resuspended Huh7 cells were added to GPC-expressing Huh7 cells and co-cultured in DMEM at neutral pH for 24 hr . Cells were fixed with 1:10 buffered formalin and counterstained with 1 µg/µL 4′ , 6-diamidino-2-phenylindole ( DAPI ) to visualize nuclei . Cells were imaged on a Cytation5 imager ( Biotek ) . GFP foci were measured for each pH treatment condition , and the average areas of the foci were determined using ImageJ . Syncytia in experimental wells were called using an area threshold of two standard deviations above the mean area of the foci in the mock-transfected wells at the same pH . A biological replicate is defined as an additional data point acquired by repeating all steps of an experimental protocol from sample generation to data processing . A technical replicate is defined as an additional data point in which the same sample from an experimental protocol is measured additional times for precision . Means from biological replicate data were compared using unpaired , two-tailed Student’s t-tests for normally distributed data sets and Mann-Whitney U-tests for non-normally distributed data sets . p-values<0 . 05 were considered statistically significant . | Crimean-Congo hemorrhagic fever ( CCHF ) is caused by infection with a virus spread by ticks in Europe , Africa and Asia . It can cause severe disease in humans , including high fevers and bleeding . How deadly CCHF is varies with between 5% to 80% of those infected dying . Scientists suspect genetic differences in various strains of the virus may account for the differences in death rates , but they do not know the exact mutations that make the CCHF virus more or less deadly . To learn more , scientists have sorted strains of CCHF virus into different groups based on how similar they are genetically . One group called Europe 2 infects many people in the Balkans , but it rarely causes illness . In fact , only two mild cases of illness have been associated with Europe 2 strains , while other CCHF virus strains circulating in this region have caused thousands of more severe illnesses . Now , Hua et al . identified a mutation in one Europe 2 strain of the CCHF virus that may explain why this subgroup of viruses rarely causes severe human disease . The researchers collected a strain of CCHF virus from infected ticks found in Bulgaria and sequenced its genome . They named the virus strain Malko Tarnovo . Through a series of experiments , Hua et al . showed that the Malko Tarnovo strain very efficiently infects tick cells but not human cells . A single amino acid change in the genetic sequence of the virus appears to make the virus less able to infect human cells . The mutation prevents a protein on the surface of the virus from fusing with human cells , an essential step in infection . This may explain why this strain and others in the Europe 2 group do not cause severe human disease . Hua et al . also demonstrate the importance of studying viruses in the animals that spread them . By studying the CCHF virus in ticks , scientists may be able to learn more about how viruses evolve to infect new species , which may help scientists prevent future pandemics . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2020 | A single mutation in Crimean-Congo hemorrhagic fever virus discovered in ticks impairs infectivity in human cells |
In our dynamic world , decisions about noisy stimuli can require temporal accumulation of evidence to identify steady signals , differentiation to detect unpredictable changes in those signals , or both . Normative models can account for learning in these environments but have not yet been applied to faster decision processes . We present a novel , normative formulation of adaptive learning models that forms decisions by acting as a leaky accumulator with non-absorbing bounds . These dynamics , derived for both discrete and continuous cases , depend on the expected rate of change of the statistics of the evidence and balance signal identification and change detection . We found that , for two different tasks , human subjects learned these expectations , albeit imperfectly , then used them to make decisions in accordance with the normative model . The results represent a unified , empirically supported account of decision-making in unpredictable environments that provides new insights into the expectation-driven dynamics of the underlying neural signals .
Even the simplest perceptual judgments , like detecting the presence of a dim light , take time for the brain to process ( Luce , 1986 ) . Some of this time reflects sensory and motor processing , but a considerable fraction is dedicated to the decision process that converts the incoming sensory information into a categorical judgment that guides behavior ( Sternberg , 2001 ) . Under certain conditions , this temporally unfolding process serves a normative purpose: improving the accuracy of the decision by reducing uncertainty about the source or identity of noisy inputs . The sequential probability ratio test ( SPRT ) , drift-diffusion model , and related sequential-sampling models are forms of ‘belief-updating’ rules for this normative process , based on perfect integration over time of the logarithm of the likelihood ratio ( LLR ) associated with each data point ( Barnard , 1946; Wald , 1947; Good , 1979; Link , 1992; Gold and Shadlen , 2001; Smith and Ratcliff , 2004; Bogacz et al . , 2006 ) . These models have been useful for studying neural mechanisms of decision-making ( Gold and Shadlen , 2007 ) but are normative for only a restricted set of conditions in which: ( 1 ) the ideal starting time-point for accumulation is known ( e . g . , given by the onset of an experimental trial ) ; and ( 2 ) the statistics of the incoming information are perfectly stable throughout the entire sequence , with no change in the underlying signal and all noise coming from the same probability distribution . Perfect integration can be particularly problematic for tasks that require the detection of signal changes ( Clifford and Ibbotson , 2002 ) . When there is certainty about when the change might occur , integrated signals from before vs after that time can be compared to detect the change ( Green and Swets , 1966; Macmillan and Creelman , 2004 ) . However , when there is temporal uncertainty about the change , integrating evidence at the wrong time might miss the signal or add unnecessary noise , resulting in a loss of sensitivity to the change ( Lasley and Cohn , 1981 ) . Several possible solutions to this problem have been proposed , including using a leaky integrator , taking a time derivative of the evidence to identify changes , or using knowledge of the spatial and temporal structure of the stimulus to guide a more directed search for the evidence ( Henning et al . , 1975; Nachmias and Rogowitz , 1983; Smith , 1995 , 1998; Verghese et al . , 1999; Schrater et al . , 2000 ) . However , none of these solutions provide more general insights into how to balance the operations used to identify both steady , noisy , signals and unpredictable changes in those signals . Here we present a normative model of decisions between two alternatives that provides such an account . In a variety of learning and other tasks , the tradeoff between signal identification and change detection has been related to inference algorithms in hidden Markov models and other Bayesian algorithms . These algorithms estimate statistical parameters in the presence of abrupt and unpredictable change-points in the otherwise stable statistics of a data-generating process ( Zakai , 1965; Liptser and Shiryaev , 1977; Rabiner , 1989; Yu and Dayan , 2005; Adams and MacKay , 2007; Behrens et al . , 2007; Fearnhead and Liu , 2007; Wilson et al . , 2010; McGuire et al . , 2014; Sato and Kording , 2014 ) . Here we express these algorithms in a novel form that , unlike previous change-point models , is based on the LLR and thus can be compared directly to standard decision models based on evidence accumulation ( Gold and Shadlen , 2001; Usher and McClelland , 2001; Smith and Ratcliff , 2004; Bogacz et al . , 2006 ) . The form thus yields quantitative predictions of both choice behavior and the underlying neural signals for decisions about unstable , noisy stimuli ( Gold and Shadlen , 2007 ) . A key feature of the model is that the expected amount of instability in the environment governs the temporal dynamics of the decision process . When perfect stability is expected , evidence is accumulated perfectly . Otherwise , evidence is accumulated with a leak ( Usher and McClelland , 2001 ) to a non-absorbing boundary that expedites the identification of unexpected changes that should re-start the accumulation process , where both the leak and the boundary depend on the level of expected instability in the environment . These expectation-dependent dynamics represent a novel view of leaky , saturating , or otherwise imperfect evidence accumulation , which here may be understood as facilitating , rather than hindering , statistical inference . We show that human decision-makers can use these dynamics to solve two different tasks on different timescales ( tens of seconds vs hundreds of milliseconds ) that each requires information accumulation in the presence of unpredictable change-points occurring at different rates .
Consider a decision about which of two alternatives is the present source of a sequence of noisy data arriving over time . We derived a belief-update rule for these kinds of decisions based on Bayesian principles that have typically been used to understand learning processes in dynamic environments on relatively slow timescales ( Figure 1A ) ( Yu and Dayan , 2005; Adams and MacKay , 2007; Behrens et al . , 2007; Fearnhead and Liu , 2007; Wilson et al . , 2010; McGuire et al . , 2014; Sato and Kording , 2014 ) . This rule both accounts for environmental instability and relates directly to models of perfect , leaky , and bounded accumulation that have been used to understand decision processes in stable environments ( Link , 1992; Gold and Shadlen , 2001; Usher and McClelland , 2001; Smith and Ratcliff , 2004; Bogacz et al . , 2006 ) . We define belief as the logarithm of the posterior odds of the alternative sources of information ( L ) given all information collected until a given time point . The sign of L indicates which source is currently believed to be generating the information , and the magnitude of L indicates how certain that belief is . The update rule is optimal when there is a fixed probability that the source could switch to the alternative at any time ( i . e . , according to a Bernoulli process ) . Specifically , ( 1 ) Ln=ψ ( Ln−1 , H ) +LLRn , where Ln is the belief at time step n , LLRn is the sensory evidence ( the log likelihood ratio ) at step n , H ( the ‘hazard rate’ ) is the expected probability at each time step that the source will switch from one alternative to the other , and ψ is the time-varying prior expectation ( the logarithm of the prior odds ) about the source before observing the new evidence: ( 2 ) ψ ( Ln−1 , H ) =Ln−1+log[1−HH+exp ( −Ln−1 ) ]−log[1−HH+exp ( Ln−1 ) ] . 10 . 7554/eLife . 08825 . 003Figure 1 . Normative model . ( A ) Illustration of the belief-updating process . ( B ) Discrete-time log-prior odds at a given moment as a function of the belief at the prior moment , plotted as Equation 2 for different values of H . ( C ) Continuous-time version of the model , with log-prior odds plotted as a function of belief , computed by numerically integrating Equation 4 with dx ( t ) = 0 over a 16 ms interval . Here expected instability ( λ ) has units of number of changes per s . Thus , λ → ∞ is analogous to discrete-time H → 0 . 5 . ( D–F ) Examples of how the normative model ( solid lines ) and perfect accumulation ( dashed gray lines ) process a time-dependent stimulus ( light vs dark grey for the two alternatives , shown at the top ) for different hazard rates ( H ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08825 . 00310 . 7554/eLife . 08825 . 004Figure 1—figure supplement 1 . Dynamics of the continuous-time model . ( A ) Example of the temporal derivative of belief magnitude plotted as a function of belief magnitude ( Equation 4 ) , given an average sensory evidence of 7 units of LLR per unit time . The point of intersection with the dotted line ( i . e . , derivative = 0 ) corresponds to the approximate steady-state value of the belief ( the mode of the probability distribution ) . ( B ) Histogram of belief values at steady state for hazard rate λ = 0 . 1 Hz from panel A , with the approximated solution in Equation 14a shown in magenta . Note the heavy tail , which is typical for moderate-to-strong input to the model . ( C ) Asymmetric effects of transient ‘perturbations’ of evidence of equal magnitude ( 35 units of LLR per s ) but opposite signs . Belief magnitude re-converges to steady state faster after the positive perturbation ( i . e . , favoring the current belief ) than the negative perturbation ( i . e . , opposing the current belief ) , which accounts for the heavy tail pointed towards zero in B . ( D–I ) Simulated temporal evolution of the probability distribution of belief value , given an average sensory evidence of 7 units of LLR per s , with an abrupt sign reversal at t = 5 s . ( D–F ) Histograms of belief at the time-points indicated by arrows . ( G ) Pseudocolor plot of the full probability distributions normalized by the peak probability over the entire time series . Hot/cold colors indicate high/low probabilities . ( H ) Expected value of the belief variable . ( I ) Standard deviation of the belief variable . Note that after a change-point , leaky integration is re-started and beliefs pass through the ‘low-confidence’ regime of the model , resulting in a transient increase of variability that is approximately Gaussian . All simulations used 10 , 000 iterations . DOI: http://dx . doi . org/10 . 7554/eLife . 08825 . 004 The prior expectation ψ is the key feature of the model , balancing integration to identify steady signals and differentiation to detect changes by dynamically filtering sensory information in a way that depends on both L and H ( Figures 1 , 2 ) . For the special case of H = 0 ( perfect stability ) , the two rightmost terms in Equation 2 cancel . In this case , the update Equation 1 reduces to perfect accumulation as in random-walk and related decision models used to identify steady , but noisy , signals ( Figure 1D ) ( Smith and Ratcliff , 2004; Bogacz et al . , 2006 ) . In contrast , when H is high and changes are expected , accumulation over time is severely limited to facilitate change detection ( Figures 1F , 2G ) . For intermediate values of H , these operations trade-off to emphasize change detection at the expense of steady signal identification ( for higher H ) or vice versa ( for lower H; Figures 1E , 2G ) . Finally , in the special case of H = 0 . 5 , the history of evidence is irrelevant at all times and all three terms in Equation 2 cancel , so ψ = 0 and Ln = LLRn . 10 . 7554/eLife . 08825 . 005Figure 2 . Features of the discrete-time ( A , C , E , G ) and continuous-time ( B , D , F ) normative models . ( A , B ) Leak rate as a function belief state and hazard rate . Blues are the least leaky and correspond to longer temporal accumulation; reds are the most leaky and correspond to a sign reversal in the change in current belief , resulting in damped oscillations in choice behavior . For the continuous-time model ( B ) , there are no leak rates analogous to discrete-time H > 0 . 5 . ( C , D ) Bias as a function belief state and hazard rate . Dark greens are the most biased in favor of the alternative associated with negative log-odds; yellows are the most biased in favor of the other alternative . ( E , F ) Predicted choice accuracy one sample ( E ) or <300 ms ( F ) after a change-point vs during steady-state conditions , at different expected hazard rates , as shown , for two difference strengths of evidence ( E: |LLR| = 0 . 5 for leftmost curves and 5 for rightmost; F: |LLR| = 4/s for leftmost curves and 80/s for rightmost ) . ( G ) Average belief from the discrete-time model over 1000 simulations for each condition shown , each with a single change-point at trial 20 . DOI: http://dx . doi . org/10 . 7554/eLife . 08825 . 005 To gain further insight into the dynamics of the model and how it controls this trade-off , we made approximations of the nonlinearity in Equation 2: ( 3 ) ψ ( Ln−1 , H ) ≈ ( 1−Kn ) ×Ln−1+θn , ( 3a ) ≈ ( 1−2H ) ×Ln−1 when Ln−1≈0 , ( 3b ) ≈log[ ( 1−H ) /H]when Ln−1≫0 , ( 3c ) ≈−log[ ( 1−H ) /H] when Ln−1≪0 . Here Kn governs the leakiness of the accumulation process , and θn , governs a bias . Both parameters are adaptive , depending on both H and Ln , with dynamics that jointly establish a boundary on the prior and thus limit subsequent belief strength . The dynamics include two regimes , as follows . First , when beliefs are uncertain ( i . e . , regimes around Ln − 1 = 0 in Figures 1B , 2A , C; Equation 3a , in which Kn predominates over θn ) , the model acts like a leaky accumulator , in which the prior expectation is a fraction of the previous belief ( Busemeyer and Townsend , 1993; Usher and McClelland , 2001; Bogacz et al . , 2006; Tsetsos et al . , 2012 ) . Thus , the dynamics of a leaky accumulator can , in principle , act like the normative model , but only in the low-certainty regime ( Figure 3 ) . In this regime , the normative leak is adaptive , which has been demonstrated previously ( Ossmy et al . , 2013 ) , and is directly dependent on H , which has not been described previously . For low H and thus relative stability , a small leak provides long integration times . For H ≈ 0 . 5 ( the correct answer is equally likely to stay or switch after each sample ) , the model discards all historical information and L depends only on LLR . For H > 0 . 5 ( the correct answer is more likely to switch after each sample ) , the prior expectation undergoes damped oscillations ( Figure 2G ) , even when the source of evidence is transiently stable . These oscillations repeatedly switch the direction of existing beliefs because of the high expected probability of change on each discrete time step . 10 . 7554/eLife . 08825 . 006Figure 3 . Two ways of approximating the discrete-time normative model , accounting separately for its dynamics when the sensory evidence is consistently weak ( A , average |LLR| ≈ 0 . 25 ) or strong ( B , average |LLR| ≈ 10 ) . As in Figure 1B , each panel has discrete-time log-prior odds as a function of the belief at the previous moment . Dark blue lines correspond to the normative model for H = 0 . 1 . Light blue lines correspond to a leaky accumulator with no bias , related to the linear approximation in Equation 3a but optimized to best approximate the normative model separately for each average evidence strength in A and B . Magenta lines correspond to perfect accumulation ( no leak ) to a stabilizing boundary related to Equation 3b , c , also optimized for each evidence strength . In general , the leaky accumulator is better at approximating the normative solution for weak sensory evidence ( A ) , whereas the bounded accumulator is better at approximating the solution for strong sensory evidence ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08825 . 006 Second , as the magnitude of Ln − 1 increases and belief certainty becomes high ( i . e . , regimes around Ln − 1 far from zero in Figures 1B , 2A , C; Equation 3b , c , in which θn predominates over Kn ) , such as when the incoming evidence is strong or during periods of stability in the source , the prior expectation approaches a ‘stabilizing boundary’ whose height directly depends on H . Thus , the dynamics of a model that stabilizes the decision process at a hazard-dependent value can , in principle , act like the normative model , but only in the high-certainty regime ( Figure 3 ) . This boundary represents a suspension of the accumulation process but , unlike the decision bound in the SPRT and related models ( Barnard , 1946; Wald , 1947; Good , 1979; Link , 1992; Smith and Ratcliff , 2004; Bogacz et al . , 2006; Gold and Shadlen , 2007 ) , does not terminate the decision process . Instead , it stabilizes Ln when no changes occur ( i . e . , temporarily ending further evidence accumulation ) while still allowing for the sampling of new evidence that might lead to changes in belief and a re-start of the accumulation process ( Resulaj et al . , 2009 ) . The stabilizing boundary is also in contrast to the asymptote in leaky accumulation , which increases linearly with the strength of evidence ( Busemeyer and Townsend , 1993; Usher and McClelland , 2001; Bogacz et al . , 2006; Tsetsos et al . , 2012 ) . Together these properties navigate an inherent trade-off between identification of steady signals and change detection . This trade-off depends on both evidence strength and expected H ( Figure 2E , G ) . For weak evidence , the trade-off is most severe , as the model uses expected H to err on the side of either detecting changes quickly when H is high or identifying stable signals when H is low . As the strength of evidence increases , performance improves steadily for both conditions and the trade-off diminishes . We used two separate tasks to investigate if and how human subjects could use these dynamics to adapt to different rates of change and find the optimal trade-off between stable signal identification and change detection . For both tasks , we found that: ( 1 ) subjects adapted , albeit imperfectly , to different hazard rates ( via comparisons to a suboptimal model , which ignored block-wise changes in H ) and used their subjective estimates of hazard rate in a manner consistent with the normative model; and ( 2 ) their choice dynamics were better described by the normative model than two other adaptive , but suboptimal , alternatives inspired by the approximations to the normative model ( one was an accumulator with a leak that could vary as a free parameter for each hazard-specific block of trials but no stabilizing boundary; the other was a perfect accumulator with a stabilizing boundary that could vary as a free parameter for each hazard-specific block of trials; see Figure 3 ) .
We derived a normative model of evidence accumulation for decision tasks that is based on Bayesian principles for inferring changes in the statistics of a generative process ( Rabiner , 1989; Adams and MacKay , 2007; Behrens et al . , 2007; Fearnhead and Liu , 2007; Brown and Steyvers , 2009; Wilson and Finkel , 2009; Nassar et al . , 2010 , 2012; Wilson et al . , 2010; Boerlin et al . , 2013; Wilson et al . , 2013; Gonzalez Castro et al . , 2014; McGuire et al . , 2014; Sato and Kording , 2014 ) . Our model incorporates change detection into sequential-sampling decision models and is related to other , modified versions of these models that have been used to combine multiple sensory cues of different but known reliabilities or infer unknown sensory reliability assumed to be stable during the course of decision-making ( Hanks et al . , 2011; Deneve , 2012; Drugowitsch et al . , 2014 ) . However , unlike those models , which invoked a separate learning-rate term or had other , more complex forms , our model casts adaptation directly in the context of the evidence-accumulation process that is a key focus of studies of decision-making ( Usher and McClelland , 2001; Roitman and Shadlen , 2002; Huk and Shadlen , 2005; Uchida et al . , 2006; Brunton et al . , 2013; Hanks et al . , 2015 ) . This formulation allowed us to identify , for the first time , features of evidence accumulation that can underlie normative , adaptive decision-making , including expectation-dependent changes in leaky accumulation when beliefs are weak and saturating accumulation when beliefs are stronger . We showed that human subjects made decisions on two separate tasks , requiring evidence accumulation either across or within trials , that were consistent with the adaptive , hazard-dependent accumulation process prescribed by the model . Our findings substantially extend previous studies that similarly suggested that human decision-making behavior can reflect adaptations to the rate of environmental changes ( Behrens et al . , 2007; Brown and Steyvers , 2009; Gonzalez Castro et al . , 2014 ) . Specifically , we showed that subjects could both learn a range of hazard rates and then use those learned rates in a normative manner to interpret sequences of evidence to make decisions . However , they tended to learn imperfectly , over-estimating low hazard rates and under-estimating high hazard rates . Thus , although their use of these imperfectly learned hazard rates was consistent with the normative model , their overall decisions in some cases fell short of the ideal observer . Our framework provides a new way to interpret these deviations from optimality: not simply as poor performance , but rather as different , hazard-dependent set-points of an inherent trade-off . This tradeoff balances sensitivity to change during periods of expected instability , and sensitivity to steady-state signals during periods of expected stability . These different set points may have reflected certain prior expectations about the improbability of either perfect stability or excessive instability that could constrain performance when those conditions occur . Such prior expectations about a lack of perfect environmental stability interpreted in the context of our framework might also provide new insights into previous studies of the temporal dynamics of evidence accumulation . In some cases , decisions about perfectly stable stimuli appear to involve perfect accumulation , as described by drift-diffusion and related models ( Gold and Shadlen , 2000; Roitman and Shadlen , 2002; Brunton et al . , 2013; Hanks et al . , 2015 ) . Under those conditions , deviations from perfect accumulation in the brain may be considered as inefficient , operating under other constraints ( e . g . , computational costs ) , or of uncertain relevance to decision-making ( Usher and McClelland , 2001; Drugowitsch et al . , 2012 ) . In contrast , our results imply that at least some deviations from perfect accumulation might reflect normative adjustments to expected instabilities , even under the nominally stable conditions used for many tasks . For example , leaky accumulation that places more emphasis on recent vs past information or rates of accumulation that vary as a function of time , which can account for the temporal dynamics of certain decisions about stimuli that are presented with stable statistics for 100's of ms or more , might reflect prior expectations that instabilities are likely to occur within that time frame ( Usher and McClelland , 2001; Eckhoff et al . , 2008 ) . Likewise , reports of an ‘urgency’ signal that limits temporal integration based on a drive to respond quickly might reflect similar expectations of impending instabilities ( Reddi and Carpenter , 2000; Ditterich , 2006; Cisek et al . , 2009; Drugowitsch et al . , 2012; Thura et al . , 2012 ) . More extreme expectations of instabilities might relate to other tasks that appear not to require temporal integration at all and instead show little dependence of performance on stimulus duration beyond what is needed to activate the sensory detectors ( Ludwig et al . , 2005; Uchida et al . , 2006 ) . These interpretations are consistent with the idea that the temporal integration window for many kinds of decisions might be highly flexible and adapt to the temporal dynamics of the environment ( Ossmy et al . , 2013; Gonzalez Castro et al . , 2014 ) . Insofar as the accumulated evidence that serves as the decision variable governing choice behavior can also be thought of as a confidence signal , such adaptive dynamics might also pertain to confidence judgments associated with certain decision tasks ( Kepecs et al . , 2008; Kiani and Shadlen , 2009; Ma and Jazayeri , 2014 ) . Further work is needed to understand if and how these findings can be understood in the context of a common set of normative principles that balance the identification of steady signals with change detection . Our results might also have implications for understanding the trade-off between speed and accuracy inherent to many tasks ( Gold and Shadlen , 2007; Bogacz et al . , 2010 ) . Sequential-sampling models like drift-diffusion typically account for this trade-off in terms of an absorbing decision boundary . This boundary can be set to a pre-defined value to terminate the decision process while emphasizing either speed or accuracy at the expense of the other , or possibly balancing the two in the service of maximizing related quantities like reward rate ( Gold and Shadlen , 2002; Palmer et al . , 2005; Bogacz et al . , 2006 , 2010; Simen et al . , 2009 ) . Alternatively , in our model the adaptive accumulation process can be suspended , at least temporarily , not by an extrinsically imposed decision rule like an absorbing decision boundary but rather by the non-linear dynamics of the accumulation process itself . In principle , certain decisions might be made by committing to an alternative once this asymptotic regime is reached . This regime represents an upper limit on the expected level of confidence and thus precludes the need for either additional data for that alternative or for an additional boundary . In this case , the resulting speed-accuracy trade-off would not necessarily reflect a pre-defined attempt to control those factors explicitly but rather expectations about the rate at which the evidence-generating process is changing . Future work is needed to investigate how key features of our model might be implemented in the nervous system for different tasks and different timescales . Previous studies using tasks that required information accumulation on the timescale of the triangles task ( e . g . , over many seconds to minutes ) have similarly suggested that humans can approximate optimal change detection , which in some cases includes a sensitivity to different hazard rates ( Behrens et al . , 2007; Brown and Steyvers , 2009; Nassar et al . , 2010 ) . The neural mechanisms of these abilities are not yet known , but fMRI and pupillometry data suggest possible roles for the arousal system including the anterior cingulate cortex and the noradrenergic system , and genotype data imply possible contributions of the dopamine system ( Yu and Dayan , 2005; Nassar et al . , 2012; Behrens et al . , 2007; Krugel et al . , 2009; O'Reilly et al . , 2013; McGuire et al . , 2014 ) . Conversely , evidence-accumulation processes that operate over shorter timescales , like for various versions of the random-dot motion task , have focused on dynamic neural signals in other parts of cortex , the basal ganglia , and the superior colliculus that can reflect the rapid build-up of evidence to select a particular motor response ( in these cases involving eye movements ) ( Gold and Shadlen , 2007; Ding and Gold , 2013 ) . There are some suggestions that these systems may interact under certain conditions ( O'Reilly et al . , 2013 ) , but much more work is needed to understand the brain mechanisms responsible for the kinds of normative , scale-invariant dynamics of evidence accumulation we characterized in this study . Extending our framework to more than two alternatives and to conditions in which the statistics of the evidence changes gradually , as opposed to abruptly , would also be an important step towards better understanding how the brain accumulates and interprets dynamic evidence to solve complex , real-world problems .
The normative model is based on the posterior probability each of option ( z1 or z2 ) given all of the evidence collected so far ( x1:n ) , q ( zin ) ≡p ( zin|x1:n ) . We assume that at each time step , there is a probability ( H , for ‘hazard rate’ ) that there will be a switch in the correct option . Beginning with Bayes' Rule , and using the sum and product rules of probability , it can be shown that: ( 5 ) q ( z1n ) ∝p ( xn|z1 ) [ ( 1−H ) q ( z1 , n−1 ) +Hq ( z2 , n−1 ) ] , q ( z2n ) ∝p ( xn|z2 ) [Hq ( z1 , n−1 ) + ( 1−H ) q ( z2 , n−1 ) ] , where p ( xn|zi ) is the likelihood of observing the evidence from source i . This relationship is the forward recursion for the Baum-Welch algorithm in Hidden Markov Models and has been proven elsewhere ( Bishop , 2006 ) . We derived the model ( Equations 1 , 2 ) by taking the logarithm of the ratio of the two equations; that is , defining Ln≡log ( q ( z1n ) /q ( z2n ) ) and expanding the logarithm , giving:Ln=log[p ( xn|z1 ) /p ( xn|z2 ) ]+log[ ( ( 1−H ) q ( z1 , n−1 ) +Hq ( z2 , n−1 ) ) / ( Hq ( z1 , n−1 ) + ( 1−H ) q ( z2 , n−1 ) ) ] , where the first term of the RHS is the LLR in Equation 1 by definition . The second term of the RHS can be manipulated to yield ψ ( Equation 2 ) first by dividing both the numerator and denominator by Hq ( z2 , n − 1 ) , then expanding the expression while using q ( z1 , n−1 ) q ( z2 , n−1 ) =exp ( Ln−1 ) by definition , giving ψ ( Ln−1 , H ) =log[1−HHexp ( Ln−1 ) +1]−log[exp ( Ln−1 ) +1−HH] . Factoring out exp ( Ln − 1 ) from the first term of the RHS yields Equation 2 . The special cases of H = 0 and H = 0 . 5 are most straightforward to see from Equation 5 . When H = 0:q ( z1n ) ∝p ( xn|z1 ) q ( z1 , n−1 ) , q ( z2n ) ∝p ( xn|z2 ) q ( z2 , n−1 ) , andLn=log ( p ( xn|z1 ) /p ( xn|z2 ) ) +log ( q ( z1 , n−1 ) /q ( z2 , n−1 ) ) =LLRn+Ln−1 , which is perfect integration of the log likelihood ratios . When H = 0 . 5:q ( z1n ) ∝p ( xn|z1 ) [0 . 5q ( z1 , n−1 ) +0 . 5q ( z2 , n−1 ) ] , q ( z2n ) ∝p ( xn|z2 ) [0 . 5q ( z1 , n−1 ) +0 . 5q ( z2 , n−1 ) ] , andLn=log ( p ( xn|z1 ) /p ( xn|z2 ) ) =LLRn . Akin to the discrete-time model , the continuous-time version is based on the posterior probabilities of each option given all evidence collected until a given time point t . It has been shown previously that the non-normalized posterior probabilities of each of two states in a Markov jump process dx ( t ) , with average values ±μ and noise magnitude σ , can be written as a system of stochastic differential equations ( Zakai , 1965; Liptser and Shiryaev , 1977 ) : ( 6 ) dq1 ( t ) =[−λq1 ( t ) +λq2 ( t ) ]dt+q1 ( t ) μσ2dx ( t ) , dq2 ( t ) =[λq1 ( t ) −λq2 ( t ) ]dt−q2 ( t ) μσ2dx ( t ) . We used this result to write the log-odds ratio signal as L ( t ) , seeking the derivative dL ( t ) ≡dlog ( q1 ( t ) /q2 ( t ) ) , by beginning with Equation 6 , separating out the deterministic and stochastic components of the incoming evidence , and rewriting Equation 6 in vector form: ( 7 ) dq ( t ) = ( L+μσ2Dh ( t ) ) q ( t ) dt+μσDq ( t ) dW , q ( t ) ≡ ( q1 ( t ) q2 ( t ) ) TL≡ ( −λλλ−λ ) D≡ ( 100−1 ) . . Applying Itō's Lemma: ( 8 ) dL ( t ) =df ( q ( t ) ) =[ ( ∇f ) T ( L+μσ2Dh ( t ) ) q ( t ) +12Tr[ ( μσDq ( t ) ) T ( ∇2f ) ( μσDq ( t ) ) ]]dt+ . . . ( ∇f ) T ( μσDq ( t ) ) dW , f=log ( q1 ( t ) ) −log ( q2 ( t ) ) ∇f= ( 1/q1 ( t ) −1/q2 ( t ) ) T∇2f= ( −1/ ( q1 ( t ) ) 2001/ ( q2 ( t ) ) 2 ) . We now expand each component of Equation 8 , beginning with those in the deterministic expression: ( 9 ) ( ∇f ) T ( L+μσ2Dh ( t ) ) q ( t ) = ( 1/q1 ( t ) −1/q2 ( t ) ) [−λ+h ( t ) μ/σ2λλ−λ−h ( t ) μ/σ2] ( q1 ( t ) q2 ( t ) ) T=1/q1 ( t ) × ( −λ+h ( t ) μ/σ2 ) ×q1 ( t ) +1/q1 ( t ) ×λ×q2 ( t ) + . . . −1/q2 ( t ) ×λ×q1 ( t ) −1/q2 ( t ) × ( −λ−h ( t ) μ/σ2 ) ×q2 ( t ) =2h ( t ) μ/σ2+λ ( q2 ( t ) /q1 ( t ) −q1 ( t ) /q2 ( t ) ) , and ( 10 ) 12Tr[ ( μσDq ( t ) ) T ( ∇2f ) ( μσDq ( t ) ) ]=12 ( μσ ) 2 ( q1 ( t ) −q2 ( t ) ) ( −1/ ( q1 ( t ) ) 2001/ ( q2 ( t ) ) 2 ) ( q1 ( t ) −q2 ( t ) ) T=12 ( μσ ) 2[− ( q1 ( t ) ) 2/ ( q1 ( t ) ) 2+ ( q2 ( t ) ) 2/ ( q2 ( t ) ) 2]=0 . Turning to the stochastic component: ( 11 ) ( ∇f ) T ( μσDq ( t ) ) =μσ[1/q1 ( t ) −1/q2 ( t ) ][q1 ( t ) 00−q2 ( t ) ]=μσ[q1 ( t ) /q1 ( t ) +q2 ( t ) /q2 ( t ) ]=2μσ . Substituting Equations 9–11 into Equation 8 yields: ( 12 ) dL ( t ) =[−λ× ( q1 ( t ) /q2 ( t ) −q2 ( t ) /q1 ( t ) ) +2μσ2h ( t ) ]dt+2μσdW . Letting A=2μ/σ2 , and using the hyperbolic sine function , we have dL ( t ) =[−2λsinhL ( t ) +Ah ( t ) ]dt+AσdW , which can be rewritten as Equation 4 using dx ( t ) =h ( t ) dt+σdW . Simulations in Figure 1—figure supplement 1 show examples of time-evolution of the belief variable by approximating Equation 4 with the Euler-Maruyama method . We made first-order Taylor approximations of the deterministic terms in each model ( Figure 2A–D , Equation 3 ) . 48 subjects ( 29 female , 19 male; age range = 19–45 years ) participated in the triangles task , and 13 subjects ( 7 female , 6 male; age range = 19–38 years ) participated in the dots-reversal task after providing informed consent . Human subject protocols were approved by the University of Pennsylvania Internal Review Board . Both tasks were performed on an iMac with a 27′′ ( 68 . 5 cm ) screen . All models were fit to choice data using Matlab's optimization toolbox by minimizing the cross-entropy error function ( Bishop , 2006 ) : ( 15 ) e=−∑n ( 1−ρn ) log ( 1−ρ^n ) +ρnlog ( ρ^n ) , where ρn is a binary variable indicating which alternative was chosen on trial n ( arbitrarily defined as 0 for the right and 1 for the left ) and ρ^n is the choice probability predicted by the given model . All models assumed the choices were based on the sign of the subjective log-odds . Choice data shown in Figure 5A were fit by a two-parameter logistic function: ρ^n=1/[1+exp ( ( LLRn−ϕ ) /β ) ] , where ϕ represents the LLR for which subjects have a 50% chance of choosing either side , and β is the slope of the function around that point . Average differences in parameter ϕ by hazard-rate condition are reflected in the horizontal shift of the psychometric function , representing a bias towards ( leftward shift ) or away from ( rightward shift ) repeating the same choice . Like for the fits of the main normative model , here the first 200–400 trials of each block were excluded to allow for a period of learning . We report the correlation between ϕ and the prediction from the asymptotic approximation of the fit normative model: log ( ( 1−Hsubj ) /Hsubj ) . The subjective mappings of ψn to Ln − 1 in Figure 6 were estimated as a non-parametric function , fit to choice data based on Ln=ψ^ ( Ln−1 ) +LLRn where here the LLR was based on the generative variance used for the task . We expressed ψn as an interpolated function of Ln − 1 , with values spread evenly between −10 and 10 in steps of one log-odds ratio and interpolation performed with cubic splines . We fit the mapping with the same objective function as for the parametric models ( Equation 10 ) , using the entire data set within a given block to estimate the ψ that best fit choice data . We used Tikhonov regularization of the derivative ( for smoothness ) using the added penalty term: γ∑i ( Δψi ) 2 , where i indexes the value of L for which ψ was estimated and γ = 1/20 was determined through ad hoc methods . Standard errors and statistical tests on performance measured as a function of viewing duration following the final change-point on the dots-reversal task were based on bootstrapped samples of the behavioral data or model predictions ( Figure 8A–F , Figure 9H–K ) . Viewing duration bins were 0–200 , 200–500 , 500–1000 , 1000–1500 and 1500–3000 ms . These analyses only included trials for which there was at least one change-point and the duration of the second-to-last direction was at least 300 ms to avoid immediately sequential change-points . The mean ± SEM durations of the second-to-last direction for trials used in this analysis were 2945 ± 55 ms for 0 . 1 Hz and 789 ± 13 ms for 2 Hz . At these durations , discrimination accuracy was likely to be at nearly asymptotic levels at the time of the final change-point . For a given duration bin specific to coherence and condition ( indexed jointly as k ) , a single bootstrapped sample m of performance was calculated as ρ¯km= ( ∑nρ¯kmn ) /Nkm , where m indexes subject , generated as a random integer between zero and 14; n indexes trial; and Nkm is the total number of trials within the duration bin for that subject and trial type ( e . g . , 0 . 1 Hz , low coherence ) . Means and standard errors were calculated as means and standard deviations of all bootstrapped samples ( 1000 samples per comparison ) . Statistical tests between condition-by-coherence trials i and j were based on paired differences between the same bootstrapped samples . The probability that ρ¯i>ρ¯j was calculated as 1M∑m ( ρ¯im−ρ¯jm>0 ) , ( where M was the total number of bootstrapped samples for that comparison ) , likewise for ρ¯i<ρ¯j , and statistical significance indicated when one of these was less than the desired confidence level ( 0 . 05 ) . Fit leaks for the dots-reversal task ( Figure 9 ) used the same algorithm as in the leaky-accumulator model described above but with separate leak terms for both hazard rate and low vs high coherence . Specifically , for subject i , session s , time-step m , trial n , and coherence level c , we defined the leak Lismn= ( 1−Kisc ) Lism−1 , n+kiCismn+2ki〈|Cisn|〉ηismn . The noise was based on the stationary standard derivation of the log-odds , given the current coherence and model parameters: σisn=ki〈|Cisn|〉/Kisc . We then fit the choice data to the predicted probability of a given choice for each trial , p^i , s , n=12+12erf ( Li , s , M , n/2ki〈|Cisn|〉/Kisc ) , where M indicates the final sample of trial n . We used a beta distribution prior on leak rate , letting the maximized model log probability for each subject be based on the sum of log likelihoods and this log prior probability , similar to what we did for the triangles task . We were interested in the dependence of leak on coherence level and comparing these dependencies between the choice data and model predictions . To control for overall level of leak by session ( and subject ) , we computed the dependence as a normalized quantity: dis= ( Kis , high−Kis , low ) / ( Kis , high+Kis , low ) where the ‘high’ and ‘low’ subscripts denote leaks for high and low coherences . The comparison between the data and normative model prediction of the dependence on hazard rate ( session ) used an analogous normalized measure: dis= ( Kis , fast−Kis , slow ) / ( Kis , fast+Kis , slow ) where ‘fast’ and ‘slow’ indicate the 2 Hz and 0 . 1 Hz sessions respectively . Statistics reported were based on these difference measures . | Organisms gather information from their surroundings to make decisions . Traditionally , neuroscientists have investigated decision-making by first asking what would be optimal for the animal , and then seeing whether and how the brain implements the optimal process . This approach has assumed that the environment consists of noisy , but stable , signals that the brain must decipher by accumulating information over time and ‘averaging out’ the noise . Previous research had suggested that most animals can accumulate information . However , these studies also showed that animals , including humans , often fall short of the optimal solution by being overly sensitive to noise and failing to completely average it out . Of course , in real life , the signals themselves can change abruptly and unpredictably , challenging us to distinguish noise from changes in the underlying signals . If a moving target suddenly jolts to the right , is that change part of the normal jitter that should be ignored , or does it predict where the target will be next ? How do we know when to keep old information that is still relevant to the decision , and when to discard the old information because a change might have occurred that renders it irrelevant ? Glaze et al . have addressed this question by building optimal change detection into the traditional ‘information-accumulation’ framework . The model suggests that what researchers previously thought was an over-sensitivity to noise might actually be optimal for the real-life challenge of detecting change . In two different tasks , Glaze et al . tested human volunteers to see if they could make decisions in ways predicted by the model . One task involved the volunteers making decisions about which one of two possible sources of noisy signals generated a given piece of information , with the correct answer changing unpredictably every 1–20 trials . The other task involved looking at a crowd of moving dots , which jolted and wobbled as they changed direction , and the volunteers had to decide which direction the dots were moving at the end of each trial . Both experiments showed that the volunteers were remarkably good at making decisions in the ways predicted by the new model , and incorporated learned expectations about the rate of change in underlying signals . The results suggest that humans , and potentially other organisms , are capable of detecting changes in the optimal ways suggested by the decision-making model . The study also makes predictions about what kinds of neural patterns neuroscientists might find when measuring brain activity while organisms do similar tasks . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Normative evidence accumulation in unpredictable environments |
Elucidating cardiac evolution has been frustrated by lack of fossils . One celebrated enigma in cardiac evolution involves the transition from a cardiac outflow tract dominated by a multi-valved conus arteriosus in basal actinopterygians , to an outflow tract commanded by the non-valved , elastic , bulbus arteriosus in higher actinopterygians . We demonstrate that cardiac preservation is possible in the extinct fish Rhacolepis buccalis from the Brazilian Cretaceous . Using X-ray synchrotron microtomography , we show that Rhacolepis fossils display hearts with a conus arteriosus containing at least five valve rows . This represents a transitional morphology between the primitive , multivalvar , conal condition and the derived , monovalvar , bulbar state of the outflow tract in modern actinopterygians . Our data rescue a long-lost cardiac phenotype ( 119-113 Ma ) and suggest that outflow tract simplification in actinopterygians is compatible with a gradual , rather than a drastic saltation event . Overall , our results demonstrate the feasibility of studying cardiac evolution in fossils .
The hearts of ray-finned fishes ( actinopterygians ) are presently described as a succession of four muscular chambers that perform inflow ( sinus venosus and atrium ) and outflow ( ventricle and conus arteriosus ) roles , followed by the bulbus arteriosus , a terminal , non-chambered , elastic cardiac segment ( Simões-Costa et al . , 2005; Grimes et al . , 2006; Durán et al . , 2008 ) . In basal actinopterygians , the conus arteriosus dominates the cardiac outflow , while in teleosts , it is the bulbus arteriosus that prevails , a notion that harks back to Gegenbaur ( 1866 ) and before . The conus arteriosus displays multiple fibrous valve rows , a character state that represents the general gnathostome condition , primitively retained in basal actinopterygian groups ( Durán et al . , 2008; Boas , 1880; Boas , 1901; Schib et al . , 2002; Xavier-Neto et al . , 2010; Parsons , 1929; Icardo et al . , 2002a; Durán et al . , 2014; Icardo et al . , 2002b ) . The multiple conal valve rows of basal actinopterygians prevent backflow and protect the delicate gill vessels from the elevated pulsations generated by the ventricle ( Satchell and Jones , 1967 ) . In contrast , in derived actinopterygians such as the teleost zebrafish , the valveless bulbus arteriosus protects the gills through its prominent elastic properties ( i . e . functioning as a windkessel [Farrell , 1979] ) . Thus , teleost hearts display only one valve row at the bulbo-ventricular transition , which is now regarded as an evolutionary remnant of the conus arteriosus ( Grimes et al . , 2006 ) . The transition from a heart packed with dozens of outflow tract valves in basal actinopterygians , such as in the genus Polypterus ( Durán et al . , 2014 ) , to the single valve row in the cardiac outflow tract of derived actinopterygians , such as in the cypriniform teleost Danio rerio ( the zebrafish ) ( Grimes and Kirby , 2009 ) represents a celebrated , hundred-year-old , case of secondary cardiac simplification . The emphasis on the bulbus arteriosus , rather than on the conus arteriosus in teleosts , and the concurrent reduction in the number of outflow valve rows are presently almost completely unconstrained in evolutionary and developmental times . We know that the primitive actinopterygian Polypterus diverged from other actinopterygian lineages ( including the zebrafish ) by about 390 Mya ( Takeuchi et al . , 2009 ) and that the elastic bulbus arteriosus of teleosts represents a very late ontogenetic addition , being added to the heart only after cardiac chambers are formed ( Grimes et al . , 2006; Grimes and Kirby , 2009 ) . With such limited information , it is impossible to answer whether outflow tract simplification in teleosts represented another case of phyletic gradualism , or resulted from drastic developmental effects . Significant morphological changes are sometimes associated with major genetic changes , such as large-scale gene duplications and/or changes in the function of genes with major developmental effects , both known to have taken place in teleost evolution ( Shapiro et al . , 2004; Shubin et al . , 1997 ) . Knowledge of morphological transitions between character states is critical to the construction of any evolutionary hypothesis . Thus , the first steps toward the understanding of any evolutionary modification ideally involve the discovery of intermediate morphologies . Unfortunately , there are no universally recognized descriptions of fossilized vertebrate chambered hearts ( Xavier-Neto et al . , 2010; Janvier , 1996; Rowe et al . , 2001; Fisher et al . , 2000; Cleland et al . , 2011 ) . Although provocative clues accumulate ( Janvier , 1996; Rowe et al . , 2001; Fisher et al . , 2000; Cleland et al . , 2011; Shu et al . , 2003; Carvalho and Maisey , 1996 , Janvier et al . , 1991 ) , none of the specimens described so far retained enough original attributes to establish beyond dispute that cardiac preservation is possible . Part of the problem is that the heart is formed by soft tissues , which fossilize only under special conditions ( Martill , 1988 ) . However , other soft organs have been described in the Cretaceous of Araripe , Brazil ( Martill , 1988; 1990; Pradel et al . , 2009; Brito et al . , 2010 ) and even in Paleozoic fishes ( 380 million-years old ) from the Gogo Formation in Australia ( Trinajstic et al . , 2007; 2013; Long et al . , 2008 ) and Antarctica ( Young et al . , 2010 ) , which suggests that the difficulty lies not with cardiac preservation , but with the lack of a systematic search .
In the course of a wider search for fossil hearts , we fortuitously found evidence for a long and gradual evolutionary reduction of the conus arteriosus and of its multiple fibrous valve rows in teleosts . The relevant fossils are from the extinct pachyrhizodontid fish Rhacolepis buccalis ( Agassiz , 1841 ) , known from fossils of remarkable three-dimensional ( 3D ) preservation ( Maisey , 1994 ) . The fossils were collected from the Romualdo Member of the Santana Formation , a Cretaceous Konservat Lagerstätte in the Araripe Basin in the Northeast of Brazil . A pollen and spore-based biostratigraphical analysis indicates a temporal range from 119 to 113 Ma for the strata in which the fossils are found ( de Moraes Rios-Netto et al . , 2012 ) . Rhacolepis buccalis is one of the most abundant fishes in the Santana Formation ( Maisey , 1991 ) and belongs to the extinct Mesozoic clade Pachyrhizodontoidei ( Arratia , 2008; 2010 ) . The relationships of Pachyrhizodontoidei among teleosts have been disputed ( Taverne , 1974; Taverne , 1976; Forey , 1977 ) . There is , however , mounting evidence that some features unite Pachyrhizodontoidei with Elopomorpha ( Maisey , 1991 ) and , recently , Pachyrhizodontoidei , among Crossognathiformes , were placed as a group closely related to Elopomorpha in a basal position among all living teleosts ( Arratia , 2008; 2010 ) . Thus , because of its basal phylogenetic position , R . buccalis is well suited for studies on the evolution of morphological characters in teleosts . Here , we report complete fossil hearts from two R . buccalis specimens ( Figure 1 ) . The fossils were scanned with propagation phase contrast synchrotron radiation microtomography ( PPC-SR-µCT ) at 6 µm resolution The remains of the R . buccalis heart are compressed along the latero-medial axis and were found in an orthotopic position , that is , posterior to gills and between the bones of the pectoral girdle ( Figure 1a–b , Video 1 ) . Their cardiac affinity is inferred on the basis of an S-shaped configuration , four chambers ( conus arteriosus , ventricle , atrium and sinus venosus ) , typical ventricular ( thick , arrowheads ) and atrial ( thin , arrows ) muscular trabeculae ( Figure 1c–f; Video 2 ) , as well as paired Cuvier ducts that join the sinus venosus in the posterior-most region of the heart ( not shown ) . 10 . 7554/eLife . 14698 . 003Figure 1 . Phase contrast synchrotron micro tomography of teleost fossil hearts . ( a , b ) 3D reconstructions of specimen CNPEM 27P obtained from PPC-SR-μCT . ( a ) , Left lateral view . ( b ) , Ventral view . ( c , d ) , ( e , f ) Sagittal sections of specimens CNPEM 01P and CNPEM 17P , respectively . Blue masks in ( d ) and ( f ) highlight fossil cardiac chambers and pericardium in the specimens CNPEM 01P and CNPEM 17P , respectively . Note that thin trabeculae are associated to the atrium ( arrows ) and that thick trabeculae are typical of the ventricle ( arrowheads ) Abbreviations: A , atrium; C . A . , conus arteriosus; P , pericardium; S . V . , sinus venosus; V , ventricle . DOI: http://dx . doi . org/10 . 7554/eLife . 14698 . 00310 . 7554/eLife . 14698 . 004Video 1 . 3D reconstruction of Rhacolepis buccalis CNPEM 27P PPC-SR-μCT . Animated rotation of the whole specimen zooming at heart position . DOI: http://dx . doi . org/10 . 7554/eLife . 14698 . 00410 . 7554/eLife . 14698 . 005Video 2 . Rhacolepis buccalis PPC-SR-μCT . Details of tomography at the heart region and 3D reconstruction of the conal valves . DOI: http://dx . doi . org/10 . 7554/eLife . 14698 . 005 The outflow tract in R . buccalis displays a well-defined conus arteriosus encased by pericardium . Our observations indicate that the R . buccalis conus arteriosus is formed by a thick muscular wall , displays the morphology of a cylinder , which eventually tapers off before joining the aorta at its cranial end and is endowed with multiple valve rows ( Figure 1c–d , Video 2 , Figure 2 ) . In the region immediately apposed to the heart , the pericardial sac assumes the shape of a pyramid wedged between posterior bilateral gill regions ( Video 2 ) . In one of the specimens ( CNPEM 17P ) , the pericardial layer is easily identified near the conal myocardial wall ( Figure 2p ) , while in the other specimen ( CNPEM 01P ) , the limits between the conus arteriosus muscular wall and the pericardium are less marked ( Figure 2q ) . In summary , R . buccalis heart is unique among teleosts in that it displays a large , dominant , conus arteriosus , rather than a predominant bulbus arteriosus in its outflow tract . 10 . 7554/eLife . 14698 . 006Figure 2 . The fossil conus arteriosus of Rhacolepis buccalis . ( a-c ) Coronal , transversal and sagittal sections of the conus arteriosus of specimen CNPEM 17P taken by Phase contrast synchrotron microtomography ( PPC-SR-µCT ) , respectively . Arrowheads in ( c ) indicate five conal valve rows in sagittal perspective . ( d-f ) , Drawings of sections in ( a-c ) highlight conal valve rows ( gray ) . ( g-i ) Didactic scheme to indicate the orientation of individual valve rows along the three orthogonal body planes ( a-c ) and ( j-l ) , ( j-l ) Coronal , transversal and sagittal sections of the conus arteriosus of specimen CNPEM 01P taken by PPC-SR-µCT . Arrowheads in ( l ) indicate five individual conal valves in sagittal perspective . ( m-o ) Drawings of sections in ( j-l ) represent conal valves ( gray ) . ( p-q ) , 3D reconstruction and segmentation of conal valves from specimens CNPEM 01P and 17P , respectively . Note that the pericardium ( pink ) outlines the conus arteriosus ( p ) . Each individual conal valve is represented by a specific spectral color . DOI: http://dx . doi . org/10 . 7554/eLife . 14698 . 006 Inside the fossilized conus arteriosus it is possible to discern multiple , nearly parallel layers ( Figure 2 ) , which , upon 3D reconstruction , appear disposed as helicoid rings along the cranio-caudal axis of the chamber ( Figures 2b , e , h , k , n , Video 3 ) . These structures are interpreted as the fossilized remnants of the fibrous component of individual conal valves , presumably valve leaflets . For comparison , we depict the fibrous components ( valve leaflets ) of the two conal valves of Megalops atlanticus ( Figure 3 ) , a living basal teleost , related to R . buccalis . 10 . 7554/eLife . 14698 . 007Video 3 . Rhacolepis buccalis PPC-SR-μCT . Sections of conal valve . DOI: http://dx . doi . org/10 . 7554/eLife . 14698 . 007 Sagittal and coronal tomographic sections and 3D reconstructions in the two fossil specimens are consistent with the presence of at least five valve rows per conus arteriosus ( Figures 2c , f , l , o , p , q ) . Because of post-mortem changes , of the imperfect alignment of the conus arteriosus to the body axes , and of the semi-lunar character of conal valves ( Figure 2 ) , the transverse sections shown in Figure 2 actually represent shallow oblique sections that allow the depiction of more than one valve row per transverse plane ( Figure 2b , e , h , k , n ) , although it is difficult to describe with precision the exact number of valves in each valve row due to the incomplete state of preservation . 10 . 7554/eLife . 14698 . 008Figure 3 . The heart of the extant elopiform Megalops atlanticus with a focus on its outflow tract . ( a ) Dissected heart of M . atlanticus . ( b ) The M . atlanticus heart was cut open along the sagittal plane to expose right and left components of the two conus arteriosus valves . ( c ) Magnification of the conus arteriosus in ( b ) showing valve leaflets from the two valve rows ( white arrowheads ) and the endocardial surface overlying conus arteriosus muscles ( black arrows ) . ( d ) Scheme representing the right valve leaflets from the conus arteriosus of M . atlanticus as displayed in ( c ) . ( e ) 3D reconstruction and segmentation of conal valves ( blue ) superimposed on a M . atlanticus Magnetic Resonance Imaging ( MRI ) . ( f ) Detail of ( e ) . ( g ) MRI of the M . atlanticus outflow tract , highlighting two conal valves ( arrowheads ) . Abbreviations: A , atrium; AO . , aorta; B , bulbus arteriosus; C . A . ; conus arteriosus; L , left side; R , right side; S . V . , sinus venosus; V , ventricle . DOI: http://dx . doi . org/10 . 7554/eLife . 14698 . 008
One important issue in the study of evolution is the idea of direction , that is , whether natural selection intrinsically favors the emergence of more complex forms or not ( McShea , 1996 ) . However , an unequivocal association of evolution with complexity is not a requirement of evolutionary theory ( Darwin and Wallace , 1858 ) . Moreover , such a view is at odds with biological evidence of frequent secondary simplifications in the evolution of microorganisms , parasites and in miniaturized/cave/fossorial fishes ( Lwoff , 1943; Brusca and Brusca , 2003; Britz et al . , 2014 ) . Indeed , cases of morphological simplification reported in the literature most likely represent only a fraction of the examples that falsify the notion that evolution must lead to increased complexity . Many other examples of simplification are found in the evolution of animal circulatory systems ( Xavier-Neto et al . , 2010; Brusca and Brusca , 2003 ) . Was this the result of traditional phyletic gradualism , or of a saltation event in the wake of large-scale gene duplication ( Amores et al . , 1998 ) ? The five rows of conal valves of R . buccalis contrast to the nine valvar rows , each containing three to six individual valves in the basal actinopterygian Polypterus ( Durán et al . , 2014 ) . Rhacolepis buccalis valves also stand out when compared to the very limited number of conal valves in living teleosts: two valve rows in Elopomorpha ( excepting Elops , with one ) and a single valve row at the bulbo-ventricular transition in remaining teleosts . Taken together , these two characters , valvar content and overall composition of the R . buccalis heart ( i . e . chamber vs . elastic segment ) , suggest that the outflow tract of this extinct fish represents an intermediate morphology between basal and higher actinopterygians , frozen in time by fossilization as an evolutionary picture taken at the Aptian/Albian boundary , 119–113 Ma ( Rios-Netto et al . , 2012 ) ( Figures 2 and 3 ) . Valvar reduction in Actinopterygii was neither seamless , nor restricted to the teleost clade . In fact , acipenseriforms and amiiforms display independent evolutionary tendencies for conus arteriosus simplification and valve reduction when compared to Polypterus . Moreover , among chondrosteans , only acipenseriforms display a clear phenotype of valve reduction , while within holosteans only amiiforms show a reduced number of valves . This suggests that valve reduction happened at least three times independently in actinopterygians ( Figure 3 ) . The five fossil valve rows we describe in R . buccalis indicate that the process of outflow tract simplification involved at least three major transitions at the base of the teleost radiation ( around 284 Ma ( Betancur et al . , 2013; Broughton et al . , 2013 ) ) : one from nine valve rows to five valve rows ( e . g . from Polypterus to R . buccalis ) ; another from five to two valve rows ( e . g . from R . buccalis to living elopomorphs ) ; a third to the single outflow valve row retained in all other teleosts ( Figure 3b ) . It is important to recognize that events other than simplification are also involved in the evolution of the outflow tract in vertebrates . For instance , there is evidence for increase in the number of valves occurring independently in basal Actinopterygii clades , explicitly in Polypteriformes and Lepisosteiformes , which is not illuminated by our present findings . 10 . 7554/eLife . 14698 . 009Figure 4 . The Rhacolepis buccalis conus arteriosus is morphologically intermediate in actinopterygian cardiac outflow tract evolution . ( a ) Hypothetical transition from a character state composed by an array of multiple valve rows in the conus arteriosus of basal actinopterygians , such as Polypteriformes ( top ) , to a derived state characterized by the dominance of the valveless bulbus arteriosus , in living teleosts ( here represented by a generalized elopomorph at the bottom ) , through an intermediate state represented in the conus arteriosus of fossilized R . buccalis hearts ( middle ) . Anterior to left . ( b ) Cladogram depicting phylogenetic relationships among early and derived gnathostomes and their corresponding morphologies of the cardiac outflow region . Drawings represent either the inner sides of right ( R ) and left ( L ) counterparts , or only the inner right side of the cardiac outflow tract . Drawings were modified from classic illustrations ( Parsons , 1929; Danforth , 1912; Senior , 1907 ) ( not to scale ) . Blue and pink coloring highlight , respectively , bulbus and conus arteriosus ( and respective valves ) in extant species . Valvar arrangement in Rhacolepis is suggested by data in Figure 2 . A parsimony ancestral character state reconstruction was made for the number of conal valves , following the color code in terminals . General relationships of Teleostei were based on Arratia , 2010 . Genera illustrating the conal condition in each Actionopterygian branches are: Squalus for Chondrichthyes; Neoceratodus for Sarcopterygii; Polypterus for Polypteriformes; Lepisosteus for Lepisosteiformes; Amia for Amiiformes; Pterothrissus for Albuliformes; Gadus for Clupeocephala . Abbreviations: B , bulbus; C . A . , conus arteriosus; L , left side; P , pericardium; R , right side; VE , ventricle . DOI: http://dx . doi . org/10 . 7554/eLife . 14698 . 009 Currently , it is not possible to ascertain genetic correlates for the valve reduction event in R . buccalis . However , this does not prevent informed speculation that may set parameters for investigation in extant species with longer , or shorter , divergence times from the exuberantly valved Polypterus . In this sense , it is useful to observe that valve reductions in acipenseriforms and amiiforms are uncoupled from the teleost extra round of large-scale genome duplication that may , or may not , have affected R . buccalis . This suggests the possibility that slow , smaller scale , mutational events produced incremental phenotypic changes , which may have been gradually selected for outflow tract simplification in teleosts ( Shapiro et al . , 2004 ) . What developmental mechanisms could underlie the transition from conal to bulbal dominance and from valve-rich to single-valved outflow tracts ? Although we deal here with outflow composition and number of valve rows as independent characters , it is possible that these traits are not completely independent , and that the relevant parameter is simply the relative extent of outflow tract occupied by the conus arteriosus and its valves , or by the bulbus arteriosus and its valveless , elastic , character ( Munoz-Chapuli et al . , 1997 ) . Outflow tract variability among actinopterygians was modeled according to Turing’s reaction-diffusion paradigm ( Munoz-Chapuli et al . , 1997 ) . The major conclusion of this exercise was that the various rows of valves distributed across the cranio-caudal extent of the conus arteriosus and the interspersed valveless spaces can be described as an ensemble of multiple positive and negative domains of endothelial to mesenchymal transformation ( Runyan and Markwald , 1983 ) set up by the interaction between diffusible molecules playing activator and inhibitor roles ( Munoz-Chapuli et al . , 1997 ) . The data now available suggest a case of phyletic gradualism , rather than an abrupt saltation-like event for actinopterygian outflow tract simplification . Three lines of evidence support this speculation: evidence for three independent ( i . e . convergent ) events of valve reduction in Actinopterygii ( in acipenseriforms , amiiforms and teleosts ) ; the three valvar simplification steps in teleosts ( multiple to five , five to two , and two to one; Figure 4 ) and the inferred simplicity of developmental mechanisms capable of producing these phenotypes . The discovery of a fossil heart in R . buccalis demonstrates that systematic , non-destructive approaches can be employed to study cardiovascular evolution and suggests that these sensitive techniques can be utilized not only in the context of species associated with abundant fossils , but also with rare fossils of animals at key phylogenetic positions . Regardless of these specific questions , we hope our results will open exciting new possibilities for research in cardiovascular paleontology and evolution .
The Rhacolepis buccalis fossils used in this study were collected from the Romualdo Member of the Santana Formation , in the Cretaceous of Araripe Basin in the Northeast of Brazil . They are deposited in the Exceptional Preservation Collection at the Brazilian Biosciences National Laboratory ( LNBio , Campinas , Brazil ) and Brazilian Center for Research in Energy and Materials ( CNPEM ) under the following accession numbers: CNPEM 01P; CNPEM 17P; CNPEM 27P . Carbonatic nodules were scanned at the ID17 and ID19 beamlines of the European Synchrotron Radiation Facility ( ESRF , Grenoble , France ) . For all samples , we set a propagation phase contrast microtomography protocol with a sample/detector distance of about 10 m . On ID17 , we had a monochromatic beam ( double-bended Laue crystals ) of 150 keV . On ID19 , we used a filtered pink beam ( Wiggler W150 with a gap of 28 mm , filters: Al , 2 . 8 mm; Cu , 8 mm; W , 1 mm ) with a total integrated energy of 210 keV . Two optic systems were utilized depending on the size of the nodules: a 0 . 5x magnification system with a FreLoN-2K camera resulting in a recorded isotropic pixel size of about 28 µm and a 0 . 3x magnification system with a FreLoN-2K camera resulting in a recorded isotropic pixel size of about 47 µm . The tomographies were computed based on 4998 projections over 360 degrees ( pixel in horizontal x vertical: 1740x300 on ID19; 1800x130 on ID17 ) . The exposure time per projection was 0 . 2 s on ID17 and 0 . 07 s on ID19 . As the vertical field of view could not cover the full height of a nodule , multiple scans were necessary for each specimen , with a minimum overlap of 30% between each scan to correct the vertical profile of the X-ray beam . The reconstructed volumes were stitched together to visualize whole nodules and by optimizing the overall contrast ( i . e . stretching the range of grey values from the 32 bit raw data into a 16 bit full range of values , avoiding too high levels of saturation ) . All three dimensional ( 3D ) images of the reconstructed morphology of R . buccalis fossils were prepared with the AMIRA software , using TIFF images reconstructed from data obtained by propagation phase contrast synchrotron radiation microtomography scans . 3D models were built using the isosurface and segmentation features of AMIRA . During segmentation of R . buccalis , we determined that its conal valves rows are continuous and follow a well-defined helicoid ( clockwise ) trajectory . The identities of each individual valves were assigned whenever the segmented coils reached the same relative position in the spiral ( i . e . concluded a pitch ) . | Modern research has majorly advanced our understanding of how the heart works , and has led to new therapies for cardiac diseases . However , little is known about how the heart has evolved throughout the history of animals with backbones – a group that is collectively referred to as vertebrates . This is partly because the heart is made from soft muscle tissue , which does not fossilize as often as harder tissues such as bones . Even though fossils of soft tissues are rare , paleontologists have already unearthed fossils of other soft organs such as the stomach and umbilical cord . These discoveries suggested that there was hope of finding fossil hearts , and now Maldanis , Carvalho et al . have indeed discovered fossil hearts in two specimens of an extinct species of bony fish called Rhacolepis buccalis . These fish were alive over 113 million years ago during the Cretaceous period , in an area that is now modern-day Brazil . Like all known vertebrates , these R . buccalis fossils have valves between the heart and the major artery that carries blood out of the heart . Such valves are vital because they prevent pumped blood from flowing back into the heart . However , oddly , R . buccalis fossils show five of these valves , which is more than any advanced bony fish that is alive today . Comparing this with the situation in other fish species suggests that vertebrate hearts gradually evolved to become progressively simpler . This discovery shows that it is possible to study heart evolution with fossils . Maldanis , Carvalho et al . hope that their findings will stimulate researchers from all over the world to examine the fossils of well-preserved animals in search of clues to help reconstruct the major steps in the evolution of the vertebrate heart . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"ecology",
"evolutionary",
"biology"
] | 2016 | Heart fossilization is possible and informs the evolution of cardiac outflow tract in vertebrates |
Organisms need to adapt to the ecological constraints in their habitat . How specific processes reflect such adaptations are difficult to model experimentally . We tested whether environmental shifts in oxygen tension lead to events in the adult newt brain that share features with processes occurring during neuronal regeneration under normoxia . By experimental simulation of varying oxygen concentrations , we show that hypoxia followed by re-oxygenation lead to neuronal death and hallmarks of an injury response , including activation of neural stem cells ultimately leading to neurogenesis . Neural stem cells accumulate reactive oxygen species ( ROS ) during re-oxygenation and inhibition of ROS biosynthesis counteracts their proliferation as well as neurogenesis . Importantly , regeneration of dopamine neurons under normoxia also depends on ROS-production . These data demonstrate a role for ROS-production in neurogenesis in newts and suggest that this role may have been recruited to the capacity to replace lost neurons in the brain of an adult vertebrate .
Animals that experience episodes of low oxygen concentration use different strategies to protect their organs , particularly those that are metabolically highly active , such as heart and brain . These species are in general capable of adjusting their metabolic rate and to cope with the accumulation of anaerobic by-products ( Larson et al . , 2014 ) . Re-oxygenation upon return to normoxia may lead to the production of harmful reactive oxygen species ( ROS ) , and it has been proposed that these animals need to repair tissues that might become damaged during re-oxygenation ( Bickler and Buck , 2007 ) . In this article , we set out to test this latter hypothesis in the highly regenerative aquatic salamander , the red-spotted newt . We specifically asked whether hypoxia and re-oxygenation leads to events in the newt brain that share common features with processes taking place after experimental ablation and subsequent regeneration of neurons under normoxia . Adult red-spotted newts are post-metamorphic amphibians with lungs . They remain active all year round and can be found in deep water under ice during winter—an environment known to become hypoxic ( Berner and Puckett , 2010 ) . The red-spotted newt also possesses a wide spectrum of abilities of regenerating complex structures , including the central nervous system ( Parish et al . , 2007; Brockes and Kumar 2008 ) . Of particular importance in the context of the present study is their capacity to replace specific neuronal subtypes in the brain following chemical ablation ( Parish et al . , 2007; Berg et al . , 2010 ) . Neuronal regeneration leads to complete restoration of the original status in terms of functional recovery and in the terms of reaching the correct number of neurons in all brain regions tested so far ( Berg et al . , 2011 ) . Regeneration of neurons is fuelled by the activation and subsequent neurogenesis by neural stem cells ( NSCs ) , the so-called ependymoglia cells ( Berg et al . , 2010; Kirkham et al . , 2014 ) . Ependymoglia cells line the brain ventricles , express the intermediate filament protein GFAP ( glial fibrillary acidic protein ) , and have radial extensions , reaching the pial surface ( Parish et al . , 2007 ) . Regeneration in homeostatically non-germinal niches is independent of the normal constitutive neurogenesis occurring in the forebrain ( Kirkham et al . , 2014 ) , thus the newt brain is an ideal model for studying both constitutive and injury-induced adult neurogenesis , as well as the relationship between the two . In order to test whether environmental shifts in oxygen tension lead to events in the adult newt brain that share features with processes taking place during neuronal regeneration under normoxia , we carried out studies on neurogenesis both during shifting and normal oxygen tension . We find that modulation of oxygen tension leads to loss of neurons , activation of microglia , accumulation of ROS in ependymoglia cells concomitant with their cell cycle reentry , and increased neurogenesis in the forebrain . Inhibition of microglia activation does not abolish ependymoglia activation upon re-oxygenation , and NSCs cultured as neurospheres respond by increased proliferation in vitro , both observations indicating a cell autonomous role for ROS in NSCs . We further show that ROS production is required for cell cycle reentry by ependymoglia cells as well as for neuronal regeneration in the normally quiescent midbrain also during normoxia . Thus , we show that ROS production is an important component of NSC regulation and propose that this role of ROS may have been recruited during evolution to the capacity of regenerating neuronal subpopulations .
First , we tested whether red-spotted newts were able to cope with hypoxia . We placed animals into an aquarium sealed with a plastic lid in which the oxygen level was manipulated by perfusing the water with nitrogen gas . The oxygen concentration was monitored by electrodes , which were feeding back to a unit controlling the gas supply to the water ( Figure 1—figure supplement 1A ) . We tested various regimens of shifting oxygen tension and found that newts were able to cope with hypoxic conditions as low as 10% of the normal oxygen tension provided that the decrease was gradual over a period of 48 hr . Hence , animals were kept in 10% of the normal oxygen tension for five days , and brought subsequently back instantly to normoxic conditions , and analyzed at different time points ( Figure 1—figure supplement 1B ) . In order to determine whether hypoxia and subsequent re-oxygenation causes injury to the brain , we performed TUNEL ( Terminal deoxynucleotidyl transferase dUTP nick end labeling ) staining , which identifies cells in the late phase of apoptosis ( Zhao et al . , 2001; Arama and Steller , 2006 ) . We found a 2 . 0-fold increase in the number of TUNEL+ cells after 5 days of hypoxia in the forebrain . The number of apoptotic cells was further increased to 3 . 3-fold of the normal , after 1 day of re-oxygenation ( Figure 1A , B ) . To determine whether neurons are lost during hypoxia/re-oxygenation , we carried out double immunostaining for the pan-neuronal marker NeuN and TUNEL . We found that the number of neurons with apoptotic phenotype was elevated showing a 3 . 2-fold increase after re-oxygenation compared to control animals ( Figure 1C ) . These results indicated that hypoxia/re-oxygenation leads to neuronal loss in the newt brain . 10 . 7554/eLife . 08422 . 003Figure 1 . Hypoxia/re-oxygenation-induced neuronal cell death and microglia response . ( A ) TUNEL+ cells are shown in the forebrain parenchyma at low magnification . The high-magnification image shows TUNEL+/NeuN+ nucleus ( arrow ) . Note the disappearing NeuN staining in the TUNEL+ cell . ( B ) Quantification of TUNEL+ cells after hypoxia and hypoxia/re-oxygenation . n = 4 , **p < 0 . 01 , ***p < 0 . 001 . ( Unpaired t-test ) . ( C ) Quantification of NeuN+/TUNEL+ cells after hypoxia and hypoxia/re-oxygenation . n = 4 , *p < 0 . 05 . ( Mann–Whitney test ) . ( D ) Quantification of microglia activation after hypoxia and hypoxia/re-oxygenation . n = 4–5 , *p < 0 . 05 . ( Unpaired t-test ) . ( E ) Low-magnification image illustrating microglia proliferation in control and experimental animals . Scale bar = 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 00310 . 7554/eLife . 08422 . 004Figure 1—source data 1 . Hypoxia/re-oxygenation-induced neuronal cell death and microglia response . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 00410 . 7554/eLife . 08422 . 005Figure 1—figure supplement 1 . Experimental system for manipulation of oxygen tension . ( A ) A gas canister ( 1 ) is connected to an air diffuser ( 2 ) that bubbles nitrogen gas in to the water . The water is mixed with magnetic stirrer ( 3 ) and an oxygen electrode measures oxygen tension in aquarium ( 4 ) . The oxygen electrode is connected to an oxygen regulator ( 5 ) and the regulator is in turn connected to a solenoid valve that regulates the outlet of gas into the aquarium ( 6 ) . ( B ) Schematic illustration of the experimental design . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 005 Activation of microglial cells is a hallmark of an injury response in the vertebrate brain ( Gonzalez-Scarano and Baltuch , 1999 ) . We have previously reported that microglial cells expressing IBA1 ( ionized calcium-binding adaptor molecule 1 ) become activated following selective , toxin-mediated ablation of neuronal subpopulations in the newt brain ( Kirkham et al . , 2011 ) . Hence , we next tested whether hypoxia/re-oxygenation leads to a microglia response by assessing the number of proliferating IBA1+ cells in the brain . We found that the number of proliferating IBA1+ cells increased 2 . 8-fold by hypoxia/re-oxygenation ( Figure 1D , E ) . These results collectively show that hypoxia/re-oxygenation leads to loss of neurons in the newt brain and activation of a microglia response . Ependymoglia cells give rise to neurons both during homeostasis and following loss of neurons ( Parish et al . , 2007; Berg et al . , 2010; Kirkham et al . , 2014 ) and we next asked whether hypoxia/re-oxygenation leads to increased ependymoglia cell proliferation in the forebrain . We did not find any statistically significant changes in the number of proliferating ependymoglia cells immediately after hypoxia ( Figure 2B ) . In contrast , after hypoxia followed by re-oxygenation , we observed a 1 . 8-fold increase in the number of proliferating ependymoglia cells as assessed by double immunostaining with antibodies against PCNA ( proliferating cell nuclear antigen ) and GFAP ( Figure 2A , C ) . In order to corroborate these observations , we also carried out pulse labeling with the nucleotide analogue EdU , which incorporates into the DNA during S-phase . Animals were injected with EdU 2 hr before sacrificing them . In accordance with the conclusion based on PCNA staining , we observed a 1 . 8-fold increase in the number of proliferating ependymoglia cells ( Figure 2—figure supplement 1 ) . 10 . 7554/eLife . 08422 . 006Figure 2 . Hypoxia/re-oxygenation-induced ependymoglia activation and neurogenesis . ( A ) Low-magnification images illustrating increased proliferation of ependymoglia cells after re-oxygenation . Arrows point to PCNA+/GFAP+ cells . ( B , C ) Quantification of PCNA+/GFAP+ ependymoglia cells showing increased ependymoglia proliferation after re-oxygenation but not after hypoxia . n = 4–5 , *p < 0 . 05 . ( Unpaired t-test for B and Mann–Whitney test for C ) . ( D ) Low-magnification images illustrating increased number of EdU+/Hu+ cells in the forebrain parenchyma after re-oxygenation . Arrows point to EdU+/Hu+ cells . ( E ) Quantification of EdU+/Hu+ cells indicating increased neurogenesis in the forebrain parenchyma after re-oxygenation n = 5 , *p < 0 . 05 . ( Unpaired t-test ) . Scale bar = 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 00610 . 7554/eLife . 08422 . 007Figure 2—source data 1 . Hypoxia/re-oxygenation-induced ependymoglia activation and neurogenesis . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 00710 . 7554/eLife . 08422 . 008Figure 2—source data 2 . Re-oxygenation leads to increased proliferation assessed by EdU incorporation . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 00810 . 7554/eLife . 08422 . 009Figure 2—figure supplement 1 . Re-oxygenation leads to increased proliferation assessed by EdU incorporation . ( A ) Images illustrating increased EdU incorporation into ependymoglia cells following re-oxygenation . Arrows point to EdU+/GFAP+ cells . ( B ) Quantification of the number of EdU+/GFAP+ cells . n = 4 , *p < 0 . 05 . ( Unpaired t-test ) . Scale bar = 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 009 Next , we wanted to test whether the loss of neurons and concomitant injury response also led to increased neurogenesis . We pulsed animals with EdU for five days following re-oxygenation . After 35 days of chase , we detected that the number of newly formed neurons increased by 2 . 4-fold compared to the control as assessed by the number of cells that had incorporated EdU and were expressing the pan-neuronal marker Hu ( Figure 2D , E ) . Hence , we conclude that in addition to loss of neurons , hypoxia/re-oxygenation also leads to increased neurogenesis . It has been hypothesized that hypoxia followed by re-oxygenation leads to production of ROS , which on one hand may lead to tissue damage and on the other hand act as signaling molecule ( Li and Jackson , 2002; D'Autreaux and Toledano , 2007 ) . To address whether ROS is produced during hypoxia/re-oxygenation , we incubated brains with the super oxide sensitive dye , hydroethidine ( HEt ) , which upon oxidation produces red fluorescence . We observed increased HEt signal already 6 hr of re-oxygenation , which remained persistent over several days . As illustrated in Figure 3A , HEt signal shows ROS accumulation in brain sections , with marked enrichment in ependymoglia cells three days after re-oxygenation . Similar results were obtained when HEt was injected intravenously 60 min prior to sacrificing the animals ( data not shown ) . 10 . 7554/eLife . 08422 . 010Figure 3 . ROS-dependent ependymoglia proliferation and neurogenesis . ( A ) Hydroethidine ( HEt ) shows increased reactive oxygen species ( ROS ) levels after re-oxygenation particularly in ependymoglia cells . Apocynin inhibits ROS accumulation . ( B ) Quantification of ventricular PCNA+/GFAP+ cells showing that apocynin does not inhibit homeostatic ependymoglia proliferation . n = 4 . ( Unpaired t-test ) . ( C ) Low-magnification images illustrating that apocynin decreases the hypoxia/re-oxygenation-induced ependymoglia cell proliferation . Arrows point to PCNA+/GFAP+ cells . ( D ) Quantification of ventricular PCNA+/GFAP+ cells showing that apocynin decreases the hypoxia/re-oxygenation induced ependymoglia cell proliferation . n = 4 , *p < 0 . 05 . ( Unpaired t-test ) . ( E ) Images illustrating accumulation of mitochondrial ROS as indicated by Mitosox signal after re-oxygenation . Administration of the mitochondrially targeted antioxidant , Mitotempo reduces mitochondrial ROS . Note the co-localization of Mitosox signal with the mitochondrial marker mitofusin-1 . Arrows point to mitofusin-1+/mitosox+ cells . ( F ) Administration of Mitotempo does not change ependymoglia proliferation . n = 4 ( Unpaired t-test ) . ( G ) Low-magnification images illustrating that apocynin decreases the hypoxia/re-oxygenation induced neurogenesis in the forebrain parenchyma . Arrows point to EdU+/Hu+ cells . ( H ) Quantification of EdU+/Hu+ cells showing that apocynin decreases the hypoxia/re-oxygenation induced neurogenesis in the forebrain parenchyma . n = 4–5 , *p < 0 . 05 . ( Unpaired t-test ) . Scale bar = 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 01010 . 7554/eLife . 08422 . 011Figure 3—source data 1 . ROS-dependent ependymoglia proliferation and neurogenesis . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 01110 . 7554/eLife . 08422 . 012Figure 3—source data 2 . ROS detection in GFAP+ and DCX+ cells in vitro . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 01210 . 7554/eLife . 08422 . 013Figure 3—figure supplement 1 . ROS detection in differentiating neurons in vitro . ( A ) Images illustrating HEt signal in GFAP+ and DCX+ cells . ( B ) Quantification of HEt signal intensity shows no difference in signal intensity . n = 3 . ( Mann–Whitney test ) . Scale bar: 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 013 Next , we aimed to address the relationship between ROS and ependymoglia cell proliferation . Following re-oxygenation , we injected animals with apocynin , which is an inhibitor of ROS production by interfering with the NADPH oxidase complex ( NOX ) ( Muijsers et al . , 2000 ) . First , we observed that twice daily apocynin injection over a period of 3 days led to reduced ROS levels as detected by HEt staining ( Figure 3A ) . Double immunostaining with PCNA and GFAP revealed that apocynin on its own did not reduce cell proliferation during normoxic conditions ( Figure 3B ) . Importantly , after re-oxygenation , the number of proliferating ependymoglia cells was reduced by 2 . 0-fold ( Figure 3C , D ) in apocynin-injected compared to vehicle-injected animals . In addition to NOX , ROS may also be generated by the mitochondria ( Prozorovski et al . , 2015 ) and we first tested whether mitochondrial ROS increased during re-oxygenation . To detect mitochondrial ROS , we incubated newt brains with mitochondrial superoxide indicator Mitosox ( Robinson et al . , 2008 ) . We observed excellent co-localization of the signal with the mitochondrium marker mitofusin-1 ( Rojo et al . , 2002 ) ( Figure 3E ) . Animals that were undergoing hypoxia/re-oxygenation displayed accumulation of mitochondrial ROS compared to control animals , which could be blocked with administration of the mitochondrially targeted antioxidant , Mitotempo ( Dikalova et al . , 2010 ) ( Figure 3E ) . However , in contrast to the effect of NOX-inhibition with apocynin , Mitotempo administration did not reduce the number of PCNA+ ependymoglia cells ( Figure 3F ) . To address whether inhibition of ROS production ultimately interfered with neurogenesis , we carried out pulse/chase experiments with EdU . Animals were pulsed with EdU during 5 days following re-oxygenation and injected with apocynin for twice per day over a period of 8 days starting directly after re-oxygenation . Following a 35-day chase , we found that the number of EdU+/Hu+ cells was reduced by 2 . 0-fold compared to vehicle-injected animals ( Figure 3G , H ) . These results show that the neurogenic response to hypoxia/re-oxygenation induced neuronal loss depends on ROS production . Previous studies indicated increased ROS production in newborn neurons ( Tsatmali et al . , 2006 ) , and we aimed to quantify ROS in newborn neurons in the newt brain but this was not feasible due to technical reasons . As an alternative , we used an in vitro culture system . Newt ependymoglia cells form neurospheres under appropriate conditions as described earlier ( Kirkham et al . , 2014 ) . Sphere growth occurs over time as the cells within proliferate , and shifting cells to growth factor free media induces differentiation . In such cultures , we compared the intensity of the ROS indicators HEt in GFAP+ cells and in cells expressing the early neuronal marker doublecortin ( DCX ) . These analyses did not show any difference in signal intensity indicating that ependymoglia cells and young neurons do not differ in terms of ROS production . ( Figure 3—figure supplement 1 ) . Previous studies in zebrafish have demonstrated a critical role of inflammatory cells , such as microglia , in NSC activation following traumatic brain injury ( Kyritsis et al . , 2012 ) . We next addressed whether microglia activation plays an important role in ependymoglia cell proliferation following hypoxia/re-oxygenation in the newt . To suppress microglia activation , we administered dexamethasone to animals twice daily for 5 days prior to shifting them to hypoxia and twice daily for three days post re-oxygenation . In accordance with previous reports ( Kirkham et al . , 2011; Kyritsis et al . , 2012 ) , the number of proliferating microglial cells was reduced 3 . 8-fold compared to vehicle-injected animals ( Figure 4A , B ) . In contrast , the proliferative response by ependymoglia cells to hypoxia/re-oxygenation was not altered ( Figure 4C , D ) . Consistently , we did not observe any significant decrease of microglia activation in apocynin-treated animals compared to the controls ( Figure 4—figure supplement 1A ) . These results indicated that activation of ependymoglia cells was independent of microglia activation following hypoxia/re-oxygenation . 10 . 7554/eLife . 08422 . 014Figure 4 . Suppression of microglia activation does not inhibit hypoxia/re-oxygenation–induced ependymoglia proliferation . ( A , B ) Dexamethasone inhibits microglia activation indicated by decreased number of IBA1+/PCNA+ cells . Low-magnification images illustrate that dexamethasone decreases microglia proliferation in A and quantification is shown in B . n = 6 , *p < 0 . 05 . ( Unpaired t-test ) . ( C , D ) Dexamethasone does not inhibit ependymoglia activation , indicated by unchanged number of ventricular GFAP+/PCNA+ cells . Low-magnification images illustrate that dexamethasone does not decrease ependymoglia proliferation in C and quantification is shown in D . n = 6 . ( Unpaired t-test ) . ( E–G ) Hypoxia/re-oxygenation increases proliferation of GFAP+ cells in neurospheres in a ROS dependent manner . Images illustrating proliferation of GFAP+ cells in control and experimental neurosphere cultures are shown in E . Quantifications of PCNA+/GFAP+ cells are shown in F , G . n = 3 , *p < 0 . 05 . ( Unpaired t-test ) . Scale bar = 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 01410 . 7554/eLife . 08422 . 015Figure 4—source data 1 . Suppression of microglia activation does not inhibit hypoxia/re-oxygenation-induced ependymoglia proliferation . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 01510 . 7554/eLife . 08422 . 016Figure 4—source data 2 . Apocynin does not inhibit microglia proliferation in vivo but abrogates neurosphere-formation after hypoxia/re-oxygenation . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 01610 . 7554/eLife . 08422 . 017Figure 4—figure supplement 1 . Apocynin does not inhibit microglia proliferation in vivo but abrogates neurosphere-formation after hypoxia/re-oxygenation . ( A ) The number of proliferating microglia cells assessed by PCNA+/IBA1+ cells is not affected by apocynin treatment . n = 4 . ( Unpaired t-test ) . ( B ) Apocynin does not inhibit neurosphere formation in normoxic conditions . n = 4 . ( Unpaired t-test ) . ( C ) Apocynin abrogates hypoxia/re-oxygenation-induced increase in neurosphere formation . n = 6 , *p < 0 . 05 ( Unpaired t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 017 In order to corroborate these findings , we studied the effect on hypoxia/re-oxygenation in neurosphere cultures . When we placed neurospheres into 1% hypoxic chamber for 24 hr and shifted subsequently back to normoxic conditions , we observed a 1 . 4-fold increase in the number of proliferating GFAP+ cells in the neurospheres compared to control cultures . This increase was abolished in apocynin-treated cultures indicating a ROS-production dependent response ( Figure 4E , F ) . Apocynin on its own did not reduce proliferation ( Figure 4G ) . In accordance to these observations , we observed an increase in the number of spheres being formed following re-oxygenation and that this effect was abolished by apocynin treatment ( Figure 4—figure supplement 1B , C ) . Collectively , the above results suggest that the increase in NSC proliferation following re-oxygenation was cell autonomous . Next , we asked whether regeneration of neurons is dependent on ROS production also during normoxic conditions . To do so , we ablated midbrain dopamine neurons by injecting the neurotoxin , 6-OHDA ( 6-hydroxydopamine ) , without applying the hypoxia/re-oxygenation protocol . We previously showed that , in contrast to the forebrain , the newt midbrain is non-germinal and essentially quiescent ( Parish et al . , 2007 ) . However , after administration of 6-OHDA , which kills dopamine neurons in the midbrain within 3 days after injection , newts regenerate lost midbrain dopamine neurons within four weeks . Regeneration is fuelled by the local activation of normally non-proliferating ependymoglia cells , which subsequently undergo a neurogenic program ( Parish et al . , 2007; Berg et al . , 2010 ) . First , we noticed that injection of 6-OHDA led to accumulation of ROS in ependymoglia cells ( Figure 5A ) . Next , we tested whether inhibition of ROS biosynthesis during the regeneration phase interfered with the ablation-responsive cell cycle reentry by ependymoglia cells , by treating animals with apocynin for five days starting from day 4 post-ablation . We found a 3 . 3-fold reduction in the number of proliferating ependymoglia cells in apocynin- vs control-injected animals ( Figure 5B , C ) . In addition , while apocynin treatment alone did not reduce the number of TH ( Tyrosine Hydroxylase ) -expressing neurons in sham-lesioned control ( Figure 5F ) , apocynin inhibited regeneration of dopamine neurons after 6-OHDA-injection assessed by a reduction in the number of midbrain neurons expressing TH ( Figure 5D , E ) . These data show that injury responsive ependymoglia proliferation and neuronal regeneration depends on ROS production also during normoxia . 10 . 7554/eLife . 08422 . 018Figure 5 . ROS dependent regeneration of midbrain dopamine neurons during normoxia . ( A ) HEt staining illustrating increased ROS levels in ependymoglia cells following ablation of dopamine neurons with 6-OHDA compared to sham-injured brains . Apocynin treatment abrogates lesion-induced increase of ROS levels . ( B ) Images illustrating that cell cycle reentry by quiescent midbrain ependymoglia after 6-OHDA-injection is inhibited by apocynin . ( C ) Quantification of cell cycle reentry by quiescent midbrain ependymoglia in the presence and absence of apocynin after 6-OHDA-injection . n = 4 , *p < 0 . 05 . ( Unpaired t-test ) . ( D ) Images illustrating that apocynin inhibits regeneration of midbrain dopamine neuron 21 days post ablation . ( E ) Quantification of TH+ cells 21 days post ablation of dopamine neurons in the presence or absence of apocynin . n = 4–5 , *p < 0 . 05 . ( Unpaired t-test ) . ( F ) Apocynin on its own does not change the number of TH+ cells in the midbrain in sham-ablated animals . n = 4 . ( Unpaired t-test ) . Scale bar = 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 01810 . 7554/eLife . 08422 . 019Figure 5—source data 1 . ROS-dependent regeneration of midbrain dopamine neurons during normoxia . DOI: http://dx . doi . org/10 . 7554/eLife . 08422 . 019
Previous reports provided evidence that ROS signaling and the redox state influence stem cell fate . However , conflicting data exist as to whether ROS impairs or contributes to normal stem cell function ( Kim and Wong , 2009; Chuikov et al . , 2010; Dickinson et al . , 2011; Le Belle et al . , 2011; Walton et al . , 2012; Wang et al . , 2013 ) . We used systemic manipulation of oxygen tension as a means to manipulate ROS levels in a naturally regenerating organism . In contrast to most previous studies , which were heavily based on mixed cultures of stem and progenitor cells , we here took advantage of the fact that the vast majority of the ependymoglia cells lining the ventricles of the adult newt brain are rarely dividing stem cells ( Kirkham et al . , 2014 ) . Our data support the view that increased ROS signaling is important for the activation of NSCs , ultimately leading to replacement of lost neurons . There are multiple sources of ROS in the cell and the evoked cellular response depends on the subcellular localization of ROS production . Specifically , it has been suggested that mitochondrial- and NOX-derived ROS play opposing roles in NSC and progenitor cell proliferation with mitochondria-derived ROS inhibiting and NOX-derived ROS promoting proliferation of neural progenitor proliferation ( Hou et al . , 2012; Prozorovski et al . , 2015 ) . In the newt brain however , mitochondrially targeted antioxidant administration neither increased nor reduced the proliferation of ependymoglia cells , although it apparently reduced mitochondrial ROS accumulation following re-oxygenation . This is in sharp contrast to the effect we found upon NOX inhibition . Since inhibiting NOX-derived ROS production counteracted re-oxygenation induced proliferation , our data indicate a dominating role for NOX-derived ROS in the control of ependymolgia cell proliferation . Other studies showed ROS accumulation in newly formed neurons , indicating a function in neuronal differentiation ( Tsatmali et al . , 2006; Tsatmali et al . , 2005 ) . While we could not evaluate the effect of ROS production on neuronal differentiation in vivo , our assessment did not show any increase in ROS levels in young neurons compared to NSCs in vitro . Although these observations do not demonstrate that ROS signaling is not important for differentiation , the results indicate that differentiation of neurons does not require increased ROS production . Our experimental strategy allowed us to address a potential link between the apparent neuronal regeneration capacity of these animals and the ecological challenges within their normal habitat known to provide varying oxygen tension ( Berner and Puckett , 2010 ) . Completely faithful recapitulation of such constrains is not possible under laboratory conditions , not the least due to the variability of such events in nature . However , our findings are consistent with the model suggesting that shifts between hypoxic and normoxic conditions cause tissue damage ( Bickler and Buck , 2007 ) . We can conclude that return to physiological oxygen tension from hypoxia leads to injury response in the newt brain , shown by loss of neurons , microglia activation , cell cycle reentry by NSCs , and increased neurogenesis . Importantly , increased ROS levels are also detectable in NSCs not only after re-oxygenation but also during replacement of midbrain dopamine neurons during constantly normoxic condition . Hence , it is plausible that ROS-production in NSC has been co-opted to the capacity of replacing lost neuronal population in the newt brain . It should also be pointed out that many hypoxia tolerant vertebrates are good regenerators . Two examples for this are zebrafish and crucian carp , both of which having marked brain regeneration capacity ( Kirsche and Kirsche , 1961; Wolburg and Bouzehouane , 1986; Kroehne et al . , 2011 ) . Zebrafish , which live in tropical waters known to become hypoxic during nighttime , is able to survive at low levels of oxygen for days and the crucian carp remains active for months in anoxic water ( Nilsson and Renshaw , 2004; Cao et al . , 2008; Chu et al . , 2010 ) . ROS signaling has also been linked to regeneration in several contexts ( Vriz et al . , 2014 ) and the dependence of appendage regeneration on ROS accumulation has been found in Xenopus larvae ( Love et al . , 2013 ) . The potential evolutionary relevance of this finding can also be discussed in the context of why certain animals are highly regenerative , displaying a broad spectrum of regenerative abilities in many of their tissues and body parts , while other animals are not . Hydra , planaria , zebrafish , and salamanders show exceptional regenerative responses as they can regrow several body parts . The seemingly random , phylogenetically uneven distribution of animals capable of regenerating multiple structures in their bodies suggests that a regeneration capacity of an organism on such scale could be a purely ancestral phenomenon , which has been lost in most species for reasons that are unclear at present ( Sanchez Alvarado , 2000; Simon and Tanaka , 2013 ) . Nevertheless , specific micro-evolutionary selection mechanisms , such as loss of neurons during hypoxia/re-oxygenation as we demonstrate in the present work , may contribute to how an inherent regeneration capacity is manifested or whether it is manifested at all . Several recent findings support this view . First , closely related salamander species have non-overlapping range of regeneration capacities . For example , adult newts but not axolotls are able to regenerate the lens of the eye ( Grogg et al . , 2005 ) . Similarly , some planarian species have more extensive regeneration capacities than others ( Liu et al . , 2013; Sikes and Newmark , 2013; Umesono et al . , 2013 ) . Second , we found that axolotls and newts display key cellular and molecular differences during limb regeneration ( Sandoval-Guzman et al . , 2014 ) . Third , it has been shown that a central molecular component of salamander limb regeneration , Prod1 , is only found in the salamander genome ( Garza-Garcia et al . , 2010 ) suggesting local evolution of limb regeneration in salamanders . Fourth , work on pectoral fin regeneration in zebrafish revealed an intriguing example of a sex-specific obstruction of regeneration , likely due to interference with a signaling pathway maintaining key secondary sexual attributes ( Kang et al . , 2013 ) . In a cross-species comparative setting , it is also noteworthy that cell cycle reentry by NSCs appears to be dependent on accumulation of inflammatory cells and microglia activation in the zebrafish brain ( Kyritsis et al . , 2012 ) . The hypoxia/re-oxygenation experimental paradigm that we employed does not provide evidence for such an interaction in the newt brain and our data rather suggest a cell autonomous process . Although the experimental manipulations used in zebrafish and in the newt brain here are different from each other , and we cannot rule out the possibility that microglia could activate NSCs in the newt brain under certain conditions , our observations indicate that the two animal species embark on at least partially non-overlapping signaling mechanism during neuronal regeneration . While the systemic manipulation of oxygen tension and ROS accumulation led to increased neuronal death , we could not detect NSCs with apoptotic phenotype . This difference indicates that NSCs resist to ROS related damages . Future studies should address the identity of the molecular programs underlying NSCs survival and cell cycle re-activation as a response to increased ROS levels in the brain .
All experiments were performed on adult red-spotted newts , Notophthalmus viridescens ( Charles Sullivan , Nashville , TN , USA ) according to European Community and local ethics committee guidelines . Newts were placed in an aquarium sealed with plastic lid . Nitrogen gas was perfused into the water via an air diffuser to make the environment hypoxic . Gas flow was regulated by a valve , which was controlled by an O2-sensor in the aquarium via an electrode . O2 tension was gradually reduced during 48 hr to finally reach 10% of normal level and subsequently brought back to normoxia as indicated in Figure 1—figure supplement 1 . Brains were dissected out and incubated in 100 μM HEt ( Thermo Fisher Scientific , Waltham , MA ) or 10 μM Mitosox ( Thermo Fisher Scientific , Waltham , MA ) solution for 5 to 15 min in a dark chamber at room temperature . Then , they were fixed in 4% formaldehyde and sectioned . Alternatively , HEt 10 mg/kg was injected intravenously and the animals were left sedated for 1 hr . Animals were then perfused and brains were isolated and sectioned . Apocynin ( Sigma , 5 mg/kg ) was injected intraperitoneally immediately after hypoxia twice per day for 3 days . Mitotempo ( Sigma , 5 mg/kg ) was injected intraperitoneally immediately after hypoxia twice per day for 3 days . Dexamethasone ( Sigma , 2 mg/kg ) was injected intraperitoneally twice per day for 5 days before newts were shifted to hypoxia and for 3 days immediately after hypoxia . EdU ( Invitrogen , Carlsbad , CA , 50 mg/kg ) was injected intraperitoneally twice per day during reperfusion , from day 4 till day 8 and animals chased for 35 days for assessing neuronal differentiation . For assessing cell proliferation , EdU ( 50 mg/kg ) was injected intraperitoneally 2 hr before sacrifice . 6-OHDA was injected intracranially as described earlier ( Berg et al . , 2010 ) . During dopamine neuron regeneration experiments , apocynin ( 5 mg/kg ) was administered between day 4 and day 9 after 6-OHDA-injection . Newts were sedated with 0 . 1% Tricane ( Sigma , St . Louis , MO ) solution and perfused with 4% formaldehyde and cryo-protected in sucrose at 4°C overnight . 20-μm serial coronal sections were made alternating on five slides . Sections were post-fixed with 4% formaldehyde solution for five minutes followed by 3 × 5 minutes wash in PBS . Sections were treated with 0 . 1% Triton X-100 in PBS ( Sigma ) for 15 min at RT . For PCNA staining , sections were incubated with 2M HCl in 0 . 5% Triton X-100 in PBS for 20 min at 37°C and washed 3 × 3 minutes with PBS . All sections were blocked in blocking solutions , containing 5% donkey serum , 0 . 5% Triton X-100 in PBS for 30 min at RT . Subsequently , sections were incubated with one of the following primary antibodies in blocking solutions overnight: mouse anti-PCNA ( 1:500; Millipore , Temecula , CA ) , rabbit anti-Mitofusin-1 ( 1:500; Cell signaling , Danvers , MA ) , goat anti-DCX ( 1:500; Santacruz , Paso Robles , CA ) , rabbit anti-IBA1 ( 1:500; Wako , Richmond , VA ) , rabbit anti-TH ( 1:500; Millipore ) , mouse anti-NeuN ( 1:500; Millipore ) , and mouse anti-HuC/HuD ( 1:500; Millipore ) . Next day , sections were washed 3 × 5 minutes in PBS , and incubated with the added appropriate secondary antibody ( 1:500; Molecular probes , Eugene , OR ) in blocking solutions for 2 hr at room temperature . EdU staining was performed by incubating sections with 100 mM Tris , 1 mM CuSO4 , 50-100 μM fluorescent azide , and 100 mM ascorbic acid as prescribed in ( Salic and Mitchison , 2008 ) . TUNEL staining was performed according to the manufacturer's protocol ( Roche ) . Primary cell culture was performed as previously described ( Kirkham et al . , 2014 ) . Isolated cells were plated in 25 cm2 flasks and left at 25°C with 2% CO2 . After 24 hr , the cells were shifted to 1% oxygen for 24 hr . Subsequently , cells were shifted to normoxia and left for 2 weeks . Fresh medium was added every fourth day and the number of neurospheres was assessed after 14 days . For blocking ROS production , apocynin was added to a final concentration of 100 μM immediately after hypoxia for 3 days . For proliferation assay , EdU was added to neurosphere cultures to a final concentration of 20 μM and spheres were transferred to Poly-D-Lysine-coated slides . To measure ROS intensity , live cells were incubated with 30 μM HEt solution for 30 min at RT . Cells were washed with L15 medium and immediately fixed with 4% PFA , proceeded for immunocytochemistry . The number of positive cells was quantified under 20x magnification with the optical fractionator method on systemic random sampling of every fifth sections along the rostro–caudal axis . Images for cell counting were captured with LSM-700 using ZEN software . To analyze the double-labeled cells , 20x confocal images along the entire Z-axis using 1-μm intervals were taken and counted . For each animal , totally 10 sections in forebrain for hypoxia/re-oxygenation studies and 5 sections in midbrain for regeneration studies following 6-OHDA-injection were analyzed . To assess ROS signal in cultured cells , HEt fluorescence intensity was measured in the cell nuclei and averaged from the data of 10–25 cells/biological sample . Images were processed with either Photoshop ( Adobe ) or with Image J using linear adjustments . Animals were randomly chosen for each experiment and allocated into groups in a non-biased manner . Normal distribution of sample data was determined by using the Shapiro–Wilk test . We performed unpaired two-tailed t-tests for samples that were normally distributed , and Mann–Whitney test performed for samples that were not normally distributed . Sample size ( n ) is indicated in each experiment . Error bars represent SEM . Results are considered statistically significant at p < 0 . 05 . | During the winter , red-spotted newts remain active in water that is covered by ice . The oxygen levels under the ice tend to drop and so the newts adjust their metabolism to cope with these conditions . However , when oxygen levels return to normal , this may result in the newts producing larger amounts of chemically reactive molecules called reactive oxygen species ( ROS ) . These molecules form naturally as a by-product of oxygen metabolism , but in high quantities they can damage cells and tissues . It has been proposed that red-spotted newts and other animals that experience periods of low oxygen may have evolved processes to repair such damage . Unlike us , red-spotted newts are able to replace nerve cells in the brain that have died or been injured . This regeneration is fuelled by stem cells called ependymoglia cells , which divide to produce new nerve cells . Here , Hameed et al . investigated whether the return of oxygen to normal levels after a period of low oxygen can damage nerve cells in the newts , and whether this is followed by regeneration . The experiments show that nerve cells in the newt brain do indeed die when oxygen levels return to normal . Also , the brain activates an injury response that triggers the ependymoglia cells to divide . During this process , the ependymoglia cells accumulate ROS and their ability to divide is impaired if the production of ROS is blocked . The replacement of injured brain cells in normal oxygen conditions also depends on increased ROS levels . Together , Hameed et al . 's findings demonstrate a key role for ROS production in controlling the regeneration of damaged nerve cells in the red-spotted newt . A future challenge is to identify the genes that control the survival and activation of ependymoglia cells in response to increased ROS levels in the brain . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"cell",
"biology"
] | 2015 | Environmental changes in oxygen tension reveal ROS-dependent neurogenesis and regeneration in the adult newt brain |
Gaining insights into the regulatory mechanisms that underlie the transcriptional variation observed between individual cells necessitates the development of methods that measure chromatin organization in single cells . Here I adapted Nucleosome Occupancy and Methylome-sequencing ( NOMe-seq ) to measure chromatin accessibility and endogenous DNA methylation in single cells ( scNOMe-seq ) . scNOMe-seq recovered characteristic accessibility and DNA methylation patterns at DNase hypersensitive sites ( DHSs ) . An advantage of scNOMe-seq is that sequencing reads are sampled independently of the accessibility measurement . scNOMe-seq therefore controlled for fragment loss , which enabled direct estimation of the fraction of accessible DHSs within individual cells . In addition , scNOMe-seq provided high resolution of chromatin accessibility within individual loci which was exploited to detect footprints of CTCF binding events and to estimate the average nucleosome phasing distances in single cells . scNOMe-seq is therefore well-suited to characterize the chromatin organization of single cells in heterogeneous cellular mixtures .
Extensive transcriptional variation between individual cells has been observed using single cell RNA-seq . These data facilitate identification of functional subpopulations in seemingly homogeneous cell populations ( Shalek et al . , 2014 ) , or characterization of the cellular composition of complex tissues ( Jaitin et al . , 2014; Treutlein et al . , 2014; Macosko et al . , 2015 ) . To gain mechanistic insights into regulatory features that underlie cellular heterogeneity it is essential to measure chromatin organization in individual cells . A number of methods that map chromatin organization in populations of cells have been adapted for single cells , including ATAC-seq ( Cusanovich et al . , 2015; Buenrostro et al . , 2015b ) , DNase-seq ( Jin et al . , 2015 ) , methylome sequencing ( Smallwood et al . , 2014; Farlik et al . , 2015 ) , and ChIP-seq ( Rotem et al . , 2015 ) . Interpretation of these data in single cells is complicated because of the near binary and extremely sparse signal ( Cusanovich et al . , 2015; Buenrostro et al . , 2015b; Maurano and Stamatoyannopoulos , 2015 ) . Nucleosome Occupancy and Methylome-sequencing ( NOMe-seq ) ( Kelly et al . , 2012 ) employs the GpC methyltransferase ( MTase ) from M . CviPI to probe chromatin accessibility ( Kelly et al . , 2012; Kilgore et al . , 2007 ) . The GpC MTase methylates cytosines in GpC dinucleotides in non-nucleosomal DNA in vitro . Combined with high-throughput bisulfite sequencing this approach has been used to characterize nucleosome positioning and endogenous methylation in human cell lines ( Kelly et al . , 2012; Taberlay et al . , 2014 ) and in selected promoters of single yeast cells ( Small et al . , 2014 ) . NOMe-seq data have several unique features that are advantageous considering the challenges associated with single cell measurements ( Figure 1a ) . First , NOMe-seq simultaneously measures chromatin accessibility ( through GpC methylation ) and endogenous CpG methylation . Chromatin accessibility indicates whether a putative regulatory region might be utilized in a given cell ( ENCODE Project Consortium , 2012 ) , while endogenous DNA methylation in regulatory regions has been connected to a variety of regulatory processes often associated with repression ( Schübeler , 2015 ) . The ability to combine complementary assays within single cells is essential for a comprehensive genomic characterization of individual cells since each cell represents a unique biological sample which is almost inevitably destroyed in the process of the measurement . Second , each sequenced read might contain several GpCs which independently report the accessibility status along the length of that read . NOMe-seq therefore captures additional information compared to purely count-based methods , such as ATAC-seq and DNase-seq , which increases the confidence associated with the measurements and allows detection of footprints of individual transcription factor ( TF ) binding events in single cells . Third , the DNA is recovered and sequenced independently of its methylation status , which is a pre-requisite to distinguish between true negatives ( i . e . closed chromatin ) and false negatives ( i . e . loss of DNA ) when assessing accessibility at specified locations in single cells . This is especially important in single cells where allelic drop-out is pervasive . In single cells , NOMe-seq can therefore measure the fraction of accessible regions among a set of covered , pre-defined genomic locations . In this proof- of-principle study , I showed that NOMe-seq , which previously had only been performed on bulk samples ( Kelly et al . , 2012; Taberlay et al . , 2014 ) , can be performed on single cells . In addition to endogenous methylation at CpG dinucleotides , single cell NOMe-seq ( scNOMe-seq ) measured chromatin accessibility at DHSs and TF binding sites in individual cells , and detected footprints of CTCF binding at individual loci . Finally , the average phasing distance between nucleosomes within individual cells can also be estimated from scNOMe-seq data . 10 . 7554/eLife . 23203 . 003Figure 1 . Overview of scNOMe-seq procedure . ( a ) Schematic of GpC methyltransferase-based mapping of chromatin accessibility and simultaneous detection of endogenous DNA methylation . ( b ) Schematic of scNOMe-seq procedure introduced in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 00310 . 7554/eLife . 23203 . 004Figure 1—figure supplement 1 . FACS profile from Hoechst stained nuclei to assess DNA content . Nuclei were stained with Hoechst 33342 DNA dye and nuclei with DNA content corresponding to the G1-phase of the cell cycle were sorted into individual wells in a 96 well plate . Aggregates and debris were removed using gates on forward and side scatter . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 00410 . 7554/eLife . 23203 . 005Figure 1—figure supplement 2 . Schematic of experimental set up . A total of 19 individual cells from GM12878 were profiled in this study , 12 of these cells were exposed to GpC MTase and seven were subjected to the same process without exposure to MTase . For K562 11 cells were profiled all of which were subjected to GpC MTase treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 00510 . 7554/eLife . 23203 . 006Figure 1—figure supplement 3 . Number of covered GpC and CpG dinucleotides is proportional to the number of total bases covered . Number of covered cytosines in GpC and CpG dinucleotides plotted against the total number of nucleotides covered per sample . This comparison suggests that there is no strong bias towards or against GpC and CpG dinucleotides . This plot also shows that the coverage was about 2-fold higher for K562 cells compared to GM12878 . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 006
To test whether the GpC methylation observed in GpC MTase treated samples ( Figure 2—figure supplement 1 ) captured known chromatin accessibility patterns , I focused on DNaseI hypersensitive sites ( DHSs ) that were previously identified in GM12878 and K562 cell lines ( ENCODE Project Consortium , 2012 ) . DHSs were associated with strong enrichment of GpC methylation , both in data from pooled and individual GM12878 ( Figure 2a , b , Figure 2—figure supplement 2 ) and K562 cells ( Figure 2—figure supplements 3 and 4 ) . Conversely , endogenous CpG methylation decreased around the center of the DHSs in agreement with previous reports ( Stadler et al . , 2011; Ziller et al . , 2015 ) ( Figure 2a and Figure 2—figure supplement 3 ) . These data show that scNOMe-seq detected chromatin accessibility at DHSs . To assess how many of the known DHSs regions were recovered in a single cell , I first filtered DHSs that contained GpC dinucleotides within their primary sequence and thus could be theoretically detected by NOMe-seq . The frequent occurrence of GpC di-nucleotides renders the majority ( >85% ) of DHSs detectable by NOMe-seq ( Figure 2—figure supplements 5 and 6 ) . Of the theoretically detectable DHSs , 10 . 6% ( 20388/191566 ) and 17 . 3% ( 33182/191598 ) had one or more GpCs covered and , using a more stringent criterion , 5 . 2% ( 9083/174896 ) and 9 . 5% ( 16608/174828 ) were covered at four or more GpCs in individual GM12878 cells and K562 cells , respectively ( Figure 2c ) . Chromatin accessibility signal can vary along the length of a given DHSs due to binding of transcription factors ( Neph et al . , 2012 ) and the specific position of a GpC within a DHS will thus affect its chance of being methylated . To account for this variability and to obtain more robust estimates of GpC methylation only DHSs with at least four covered GpC were used for the subsequent analyses and referred to as ‘covered DHSs’ . 10 . 7554/eLife . 23203 . 007Figure 2 . scNOMe-seq data reveal how accessibility in single cells underlies observed DNaseI hypersensitivity in a population of cells . ( a ) Average GpC methylation level ( blue ) and CpG methylation level ( orange ) at DHSs in GM12878 cells . Regions are centered on the middle of DNase-seq peak locations . Shown is the average methylation across a 2 kb window of 12 GM12878 cells . ( b ) Heatmap displaying the average GpC methylation level across the same regions as in a ) . Each row corresponds to an individual GM12878 cell . Cells were grouped by similarity . ( c ) Proportion of DHSs covered by scNOMe-seq sequencing reads in each cell . The proportion displayed corresponds to the fraction of DHSs covered by at least 1 or 4 GpCs in a given cell . Only DHSs with at least 1 GpC ( red ) or 4 GpCs ( cyan ) within their primary sequence were taken in consideration . Error bars represent standard deviation . ( d ) Average GpC methylation at DHSs grouped into quartiles based on associated DNase-seq peak scores from lowest to highest scores . ‘Shuffled’ represents methylation data in genomic regions obtained by random placements of DHS peak intervals . Data shown are from GM12878 cells . ( e ) Fraction of accessible sites in individual GM12878 cells ( red ) and K562 cells ( cyan ) . Shown are the means and standard deviation based on all cells . ( f ) Scatter plot showing relationship between GpC methylation levels and DHS peaks score for each covered DHS . Plot shows data from all individual GM12878 cells . Red trend line is shown to visualize the relationship between GpC methylation and endogenous CpG methylation . ( g ) Scatter plot showing relationship between CpG methylation levels and DHS peaks score for each covered DHS . Plot shows data from all individual GM12878 cells . Red trend line is shown to visualize the relationship between CpG methylation and peak scores . ( h ) Plot illustrates the relationship between endogenous CpG methylation and GpC methylation at DHS loci . Plot shows combined data from all GM12878 cells . Correlation was calculated based on Pearson correlation ( r = −0 . 13 ) i ) Average CpG methylation at DHS loci grouped based on GpC scores within single cells . Each dot represents the average CpG methylation level for a single cell . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 00710 . 7554/eLife . 23203 . 008Figure 2—figure supplement 1 . Average CpG and GpC methylation levels in single cells . Boxplots representing the methylation level at CpG and GpC dinucleotides for groups of cells ( GM12878 w/ and w/o MTase , K562 w/ MTase ) . GM12878 and K562 cells show different levels of CpG methylation . The difference in CpG methylation between GM12878 w/o MTase and GM12878 w/ MTase treatment was largely driven by two cells . These cells were kept as no other criterion suggested their removal . GpC MTase treated cells shows a clear enrichment of GpC methylation while GM12878 cells not exposed to MTase do not show levels above 1% . These might reflect incomplete conversion , minimal cross-contamination during the parallel preparation , or activity of endogenous methyltransferases . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 00810 . 7554/eLife . 23203 . 009Figure 2—figure supplement 2 . Heatmaps of average GpC and CpG methylation across DHS regions in GM12878 cells . Each row represents data from an individual cell , both treated and control samples are plotted together . Cells were grouped using hierarchical clustering based on GpC methylation ( left ) and CpG methylation ( right ) within 2 kb regions around DHSs . As expected GpC methylation clearly separates MTase treated and untreated samples . Endogenous CpG methylation does not differ systematically between MTase treated and untreated samples . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 00910 . 7554/eLife . 23203 . 010Figure 2—figure supplement 3 . Average GpC and CpG methylation across DHS regions in K562 cells . Average GpC methylation level ( blue ) and CpG methylation level ( orange ) at DNase Hypersensitive sites ( DHSs ) in K562 cells . Regions are centered on the middle of DNase-seq peak locations . Shown is the average methylation across a 2 kb window of the pool of 11 K562 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 01010 . 7554/eLife . 23203 . 011Figure 2—figure supplement 4 . Heatmaps of average GpC and CpG methylation across DHS regions in K562 cells . Each row represents data from an individual cell . Cells were grouped using hierarchical clustering based on GpC methylation ( left ) and CpG methylation ( right ) within 2 kb regions around DHSs . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 01110 . 7554/eLife . 23203 . 012Figure 2—figure supplement 5 . Distribution of counts of GpCs within DHSs in GM12878 and K562 cells . Histogram shows the number of GpCs per DHS in GM12878 cells ( left ) and K562 cells ( right ) . While each GpC dinucleotide can be measured on both strands and would therefore yield a count of two cytosines this histogram only displays counts per GpC ( using the cytosines on the forward strand ) . This is to account for the fact that in single cells DHSs will be covered at most by one or two reads that originate from the same fragment . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 01210 . 7554/eLife . 23203 . 013Figure 2—figure supplement 6 . Proportion of DHSs at different cutoffs for GpCs and CpGs . Bar graphs display proportion of DHSs that contain at least the number of GpCs ( left ) and CpGs ( right ) indicated . The proportions are given in relation to the total number of DHSs in each cell line . Numbers in the bars refer to the number of DHSs at that GpC and CpG threshold . As described in Figure 2—figure supplement 5 and methods , only cytosines in GpC and CpG di–nucleotides on the forward strand were counted for this analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 01310 . 7554/eLife . 23203 . 014Figure 2—figure supplement 7 . Relationship between DNase-seq peak score and GpC and CpG methylation in GM12878 and K562 cells . ( a ) Average GpC methylation and ( b ) , ( c ) endogenous CpG methylation at DHSs grouped into quartiles based on associated DNase-seq peak scores from lowest to highest scores . ‘Shuffled’ represents methylation data in genomic regions obtained by random placements of DHS peak intervals . ( a ) and ( c ) show data for 11 K562 cells and ( b ) shows data for 12 GM12878 cells . Each point represents average score from a single cell . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 01410 . 7554/eLife . 23203 . 015Figure 2—figure supplement 8 . Correlation between GpC methylation and DHS peak score . Shown are correlation coefficients for comparisons between single cell and bulk NOMe-seq data with DNase-seq peak score for each covered location for ( a ) GM12878 and ( b ) K562 . Each dot represents value for a single Pearson correlation . The correlation between GpC methylation and DHS peaks scores was significantly lower in single cells compared to bulk NOMe-seq data . No correlation was observed between GpC methylation and DHS peak score using randomized DHS locations . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 01510 . 7554/eLife . 23203 . 016Figure 2—figure supplement 9 . Cumulative distribution of average GpC methylation in DHSs in GM12878 and K562 cells . Plot of cumulative distribution of GpC methylation for individual GM12878 and K562 cells at DHSs with at least four covered GpC . GM12878 and K562 cells exposed to GpC MTase show similar distributions . About 50% of all cells show no or low methylation ( < = 25% ) . GM12878 cells not exposed to GpC MTase do not show any significant number of DHSs with GpC methylation . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 01610 . 7554/eLife . 23203 . 017Figure 2—figure supplement 10 . Proportion of accessible DHSs remains stable across range of thresholds for methylation levels and covered GpCs per site . Different thresholds for GpC methylation and number of covered GpC required per individual DHS were used to test how much the number of resulting ‘accessible’ DHSs depended on these parameters . Threshold for GpC methylations was varied between 25% and 50% while the number of required GpCs for a DHSs to be considered in this analysis was varied between 1 and 8 . The proportion of accessible sites is plotted for each set of parameters in GM12878 cells ( left ) and K562 cells ( right ) . Proportion of accessible sites remained relatively stable across the range of parameters . Note that the categories with low GpCs count thresholds contain all sites above this threshold . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 01710 . 7554/eLife . 23203 . 018Figure 2—figure supplement 11 . Cell-to-cell variability in DHSs accessibility reflects DNaseI hypersensitivity of the region . Pair-wise jaccard distances between GM12878 ( a ) and K562 ( b ) cells , respectively , were calculated based on DHSs accessibility in individual cells . DHSs were grouped by DNase-seq peak scores and DHSs were considered accessible if the average methylation for that locus was above 40% . Only DHSs with at least four covered GpCs were included in this analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 01810 . 7554/eLife . 23203 . 019Figure 2—figure supplement 12 . Comparison of correlations between single cell NOMe-seq and bulk NOMe-seq data sets . Shown are the correlation coefficients for comparisons between GpC methylation in single cell and bulk NOMe-seq data within DNase-seq peak location for a ) GM12878 and b ) K562 . Each dot represents value for a single Pearson correlation . Comparison was performed on original DHS loci ( left ) and on randomized DHS loci ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 01910 . 7554/eLife . 23203 . 020Figure 2—figure supplement 13 . GpC methylation correlates with DHS peaks scores in individual cells . High DHS peak scores are associated with higher GpC methylation in single cells . Scatter plot showing relationship between GpC methylation levels and DHS peaks scores for each covered DHS . Each plot shows data from an individual GM12878 cell . Red trend line to aid visualization of the relationship between GpC methylation and peak scores . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 02010 . 7554/eLife . 23203 . 021Figure 2—figure supplement 14 . Endogenous CpG methylation is inversely correlated with DHS peak scores in individual cells . High DHS peak scores are associated with lower endogenous CpG methylation in single cells . Scatter plot showing relationship between CpG methylation levels and DHS peaks score for each covered DHS . Each plot shows data from an individual GM12878 cell . Red trend line to aid visualization of the relationship between GpC methylation and peak scores . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 02110 . 7554/eLife . 23203 . 022Figure 2—figure supplement 15 . Comparison of CpG and GpC methylation status at individual DHS in single GM12878 cells . Smoothened scatterplot illustrates the relationship between endogenous CpG methylation and GpC methylation at DHS loci . Each plot shows data from a single GM12878 cell . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 022 In single cells , the average GpC methylation at covered DHSs was strongly correlated with the observed DNaseI accessibility at these sites in bulk populations ( Figure 2d , Figure 2—figure supplements 7 and 8 ) . The opposite trend was observed for endogenous CpG methylation which was lowest for DHSs with the highest DNaseI accessibility ( Figure 2—figure supplement 7 ) . The correlation between GpC methylation and DNaseI accessibility was lower for scNOMe-seq data compared to bulk NOMe-seq data in the same cell line ( Figure 2—figure supplement 8 ) . At the level of individual sites the distribution of GpC methylation suggested that around 50% of the covered DHS showed less than 25% GpC methylation in individual cells ( Figure 2—figure supplement 9 ) . To estimate the proportion of covered DHSs that were concurrently accessible in a single cell I applied a fixed threshold of 40% GpC methylation above which sites were considered accessible ( Materials and methods ) . At this GpC methylation threshold 32–44% and 26–37% of all covered DHSs were determined to be accessible in single GM12878 and K562 cells , respectively . As expected these results depended to some degree on the cutoffs used for GpC methylation and the number of required GpCs per DHS . However , even under the most lenient conditions less than 50% of DHSs were accessible in individual cells ( Figure 2—figure supplement 10 ) . Grouping the DHSs based on DNaseI accessibility in bulk samples , confirmed that the degree of DNaseI accessibility related closely to the frequency of DHS accessibility in single cells ( Figure 2e ) . This analysis leveraged the NOMe-seq-specific property that the DNA sequence is recovered independently of its accessibility status . It provided direct evidence for the notion that the degree of DNaseI accessibility observed in DNase-seq of bulk samples reflects the frequency with which a region is accessible in individual cells . Consequently , chromatin accessibility between cells is less variable at regions with high DNaseI accessibility in bulk samples ( Figure 2—figure supplement 11 ) . Correspondingly , correlation of GpC methylation between individual cells is stronger at DHS loci compared to randomized locations ( Figure 2—figure supplement 12 ) . Chromatin accessibility and endogenous methylation show characteristic patterns at gene promoters and within gene bodies ( Schübeler , 2015; ENCODE Project Consortium , 2012 ) . To test whether these features can be observed in scNOMe-seq data , I first plotted the average GpC and CpG methylation around transcription start sites ( TSS ) . The average GpC methylation showed the expected increase of chromatin accessibility directly upstream of the TSS ( Figure 3a , Figure 3—figure supplement 1 ) . In contrast , and as expected , the endogenous CpG methylation decreased towards the TSS ( Figure 3b ) . To visualize the distribution of CpG methylation throughout entire gene loci , I plotted the aggregated CpG methylation across regions containing the entire gene body and 50 kb upstream and 50 kb downstream of each gene ( Figure 3c , Figure 3—figure supplement 1 ) . Endogenous methylation was specifically reduced at the narrow promoter region and gradually increased throughout the gene body . Downstream of the transcription end site ( TES ) the average level CpG methylation level fell back to the non-genic background level . Endogenous CpG methylation is typically increased within highly expressed genes ( Schübeler , 2015 ) . This trend was clearly apparent in the single cell data where gene body methylation was highest in highly expressed genes ( Figure 3d , Figure 3—figure supplement 1 ) . Correspondingly , in promoter regions ( −500 bp to +150 bp ) chromatin accessibility ( GpC methylation ) increased with the transcript level of the adjacent gene ( Figure 3e , Figure 3—figure supplement 2 ) . In contrast to chromatin accessibility , endogenous methylation was lowest in promoters of genes with high transcript levels ( Figure 3f ) . These data show that scNOMe-seq recapitulated known characteristics of chromatin accessibility and endogenous methylation at gene promoters and within gene bodies . 10 . 7554/eLife . 23203 . 023Figure 3 . Single cell NOMe-seq reveals chromatin features closely linked to gene expression . ( a ) Average GpC methylation level at TSS in GM12878 cells . Regions are centered on the TSS locations . Shown is the average methylation across a 2 kb window of 12 GM12878 cells . ( b ) Same as in a ) but displaying the endogenous CpG methylation level . ( C ) Average endogenous CpG methylation at gene loci in individual GM12878 cells . Shown is the average methylation across gene bodies ( represented as meta genes ) and 50 kb regions upstream and downstream of each gene . Each line represents the aggregated CpG methylation data for a single GM12878 cell ( TES: transcription end site ) . ( d ) Boxplot displays average CpG methylation in gene bodies . Genes were grouped into quartiles based on their transcript levels in bulk . Dots represent the average CpG methylation value for individual cells . ( e ) Boxplot displays average GpC methylation in promoter regions ( −500 bp to +150 bp ) . Genes were grouped into quartiles based on their transcript levels in bulk . ( f ) Similar to ( e ) but displayed are the levels of endogenous CpG methylation . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 02310 . 7554/eLife . 23203 . 024Figure 3—figure supplement 1 . Endogenous methylation in gene bodies of single K562 cells . ( a ) Average endogenous CpG methylation at gene loci in individual K562 cells . Shown is the average methylation across gene bodies ( represented as meta genes ) and 50 kb regions upstream and downstream of each gene . Each line represents the aggregated CpG methylation data for a single K562 cell ( TES: transcription end site ) . ( b ) Boxplot displays average CpG methylation in gene bodies . Genes were grouped into quartiles based on their transcript levels in bulk . Dots represent the average CpG methylation value for individual cells . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 02410 . 7554/eLife . 23203 . 025Figure 3—figure supplement 2 . Chromatin accessibility in promoters correlates with transcript levels of adjacent genes . ( a ) Average GpC methylation level at TSS genes in GM12878 cells . Regions are centered on the TSS locations and genes were grouped into quartiles based on their transcript levels in bulk GM12878 cells . ( b ) The same plot as in ( a ) based on scNOMe-seq data from K562 cells and , correspondingly , transcript levels in K562 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 025 A potentially powerful application for single cell genomic approaches is the label-free classification of single cells from heterogeneous mixtures of cells solely based on the measured feature ( Cusanovich et al . , 2015; Buenrostro et al . , 2015a; Jaitin et al . , 2014; Macosko et al . , 2015 ) . Of note , using a union set of DHSs from both cell types was sufficient to classify individual GM12878 and K562 cells into their respective cell types based on GpC methylation ( Figure 4a , Figure 4—figure supplement 1 ) . While this assessment might have been influenced in part by the separate processing of the cell types , both cell types showed preferential enrichment of GpC methylation at their respective DHSs compared to DHSs identified in the other cell type ( Figure 4b ) . Similar to GpC methylation , endogenous CpG methylation at multiple sets of genomic features was sufficient to separate the cells into the respective cell types ( Figure 4c , Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 23203 . 026Figure 4 . single cell GpC and CpG methylation signal is sufficient to group GM12878 and K562 cells according to their origin . ( a ) Heatmap shows similarity scores ( pair-wise Jaccard distances ) for accessibility between all GM12878 and K562 cells measured on the union set of DHSs from GM12878 and K562 cells . Cells were grouped based on unsupervised hierarchical clustering . ( b ) Average GpC methylation at the DHSs from GM12878 cells and K562 cells , respectively , was calculated for all individual GM12878 and K562 cells . The resulting two values for GpC methylation are displayed for each cell . While the average methylation levels at K562 DHSs for both cell types appear similar , GM12878 and K562 are separable based on these data when accounting for different levels of genome-wide GpC methylation in GM12878 and K562 cells . Importantly , for cells from either cell type the methylation levels are higher in the DHSs of the cell type of origin than in the DHSs of the other cell type . ( c ) Heatmap shows correlation coefficients between all GM12878 and K562 cells for pair-wise comparison of CpG methylation levels . Genome was divided into 10 kb bins and only bins with sufficient coverage in both cells were used for a given pair ( > = 20 covered CpGs ) . Cells were grouped based on unsupervised hierarchical clustering . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 02610 . 7554/eLife . 23203 . 027Figure 4—figure supplement 1 . Single GM12878 and K562 cells can be grouped based on GpC methylation and endogenous methylation . ( a ) Heatmap shows correlation coefficients for GpC methylation between all GM12878 and K562 cells measured on the union set of DHSs from GM12878 and K562 cells . Cells were grouped based on unsupervised hierarchical clustering . Only DHS with at least four covered GpCs in both cells were used for pair-wise comparison . ( b ) Same as in ( a ) but based on CpG methylation level in the union set of DHSs , at least two covered CpGs in both cells were required to include a DHS in the pair-wise comparison . ( c ) Heatmap shows correlation coefficients between all GM12878 and K562 cells for pair-wise comparison of CpG methylation levels in gene bodies . Only loci with sufficient coverage in both cells were used for a given pair ( > = 10 covered CpGs ) . Cells were grouped based on unsupervised hierarchical clustering . All correlation coefficients were calculated using Pearson correlation . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 027 To examine in detail whether scNOMe-seq captures features of chromatin accessibility that are specifically associated with transcription factor binding , I analyzed scNOMe-seq data at transcription factor binding sites ( TFBS ) . The average GpC methylation around CTCF ChIP-seq peaks ( ENCODE Project Consortium , 2012 ) in single cells recapitulated the accessibility previously observed in NOMe-seq bulk samples ( Kelly et al . , 2012 ) : Accessibility increased strongly towards the CTCF binding sites while the location of the CTCF motif at the center of the region showed low accessibility suggesting that CTCF binding protected from GpC MTase activity and thus creating a footprint of a CTCF binding event , both when averaged across data from all single cells ( Figure 5a and Figure 5—figure supplement 1 ) and in individual cells ( Figure 5b and Figure 5—figure supplement 2 ) . In contrast , endogenous CpG methylation was generally depleted around the center of CTCF binding sites ( Figure 5a and Figure 5—figure supplement 1 ) . Similar accessibility profiles , albeit less pronounced compared to CTCF , were observed for additional transcription factors , for example EBF1 and PU . 1 ( Figure 5—figure supplement 3 ) . These analyses provided evidence that , in aggregate , scNOMe-seq detected chromatin accessibility characteristic of CTCF binding in single cells . To test whether scNOME-seq data detected CTCF footprints at individual motifs loci , GpC methylation at motifs within CTCF ChIP-seq peaks was compared to the GpC methylation level in the regions flanking each motif ( Figure 5c ) . On average , two-thirds of CTCF motif instances within these accessible regions showed no GpC methylation , suggesting that CTCF binding prevented the GpC MTase from methylating the cytosines within the binding motif and thus creating a footprint ( Figure 5d and f ) . Of note , motifs associated with a footprint had significantly higher scores than motifs without a footprint suggesting that the motif score is a strong determinant of CTCF binding within these accessible regions ( p= 5 . 429e-12 , paired t-test ) ( Figure 5e , g and Figure 5—figure supplement 4 ) . Of note , the CTCF footprints could be observed at individual loci within individual cells and were shared across cells ( Figure 5h and Figure 5—figure supplement 5 ) . 10 . 7554/eLife . 23203 . 028Figure 5 . scNOMe-seq detected characteristic accessibility patterns at CTCF transcription factor binding sites and measured CTCF footprints at individual loci . ( a ) Average GpC methylation level ( blue ) and CpG methylation level ( orange ) at CTCF binding sites in GM12878 cells . Regions are centered on motif locations . Shown is the average methylation across a 2 kb window of the pool of 12 GM12878 cells . ( b ) Heatmap displaying the average GpC methylation across CTCF binding sites . Each row corresponds to an individual GM12878 cell and rows are grouped by similarity . ( c ) Schematic outline the measurement of CTCF footprints in accessible regions . M denotes CTCF binding motifs within CTCF ChIP-seq regions and U and D indicate 50 bp upstream and downstream flanking regions . footprint score was determined by subtracting the average GpC methylation in the flanking regions from the GpC methylation at the motif . ( d ) Heatmap displays GpC methylation in accessible regions found in a representative GM12878 cell ( GM_1 ) . Each row represents a single CTCF motif instance within a CTCF ChIP-seq region . Average methylation values for the motif and the 50 bp upstream and downstream regions are shown separately . Regions are sorted based on the footprint score . Displayed are only regions that had sufficient GpC coverage and that were considered accessible based on the methylation status of the flanking regions . ( e ) Heatmap reporting the CTCF motif scores for the motif regions in ( d ) . Regions are sorted in the same order as in ( d ) . ( f ) Average number of accessible regions at CTCF motifs and the average number of those with a detectable footprint per individual GM12878 cell . Error bars reflect standard deviation . ( g ) Average CTCF motif scores in regions with and without CTCF footprint for all 12 GM12878 cells . Each line connects the two data points from an individual cell . Regions with footprint are associated with higher motif scores ( p= 5 . 429e-12 , paired t-test ) . ( h ) Combined display of scNOMe-seq data from this study and DNase hypersensitivity data , nucleosome occupancy , and CTCF ChIP-seq data from ENCODE . Upper panel shows a ~10 kb region containing a CTCF binding site . DNaseI hypersensitivity data and nucleosome density show characteristic distribution around CTCF binding sites in GM12878 cells . Lower panel shows the GpC methylation data of 5 individual cells that had sequencing coverage in this region , 4 of the cells provide GpC data covering the CTCF motif located in the region . scNOMe-seq data tracks show methylation status of individual GpCs . Each row corresponds to data from a single cell . These data indicate that binding of CTCF is detected in all 4 cells . Data are displayed as tracks in the UCSC genome browser ( http://genome . ucsc . edu ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 02810 . 7554/eLife . 23203 . 029Figure 5—figure supplement 1 . Average GpC methylation and endogenous CpG methylation at CTCF sites in pooled K562 cells . Average GpC methylation level ( blue ) and CpG methylation level ( orange ) at CTCF binding sites in K562 cells . Regions are centered on motif locations . Shown is the average methylation across a 2 kb window of the pool of 11 K562 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 02910 . 7554/eLife . 23203 . 030Figure 5—figure supplement 2 . Average GpC methylation level at CTCF binding sites in individual K562 cells . Heatmap shows the average GpC methylation across a 2 kb window centered on the CTCF motif location . Each row corresponds to an individual K562 cell and rows are grouped by hierarchical clustering . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 03010 . 7554/eLife . 23203 . 031Figure 5—figure supplement 3 . Average GpC methylation and endogenous CpG methylation at additional transcription factor binding sites in pools of GM12878 and K562 cells . Average GpC methylation level ( blue ) and CpG methylation level ( orange ) at ( a ) PU . 1 binding sites in GM12878 cells and b ) EBF1 binding sites . Regions are centered on motif locations . Shown is the average methylation across a 2 kb window of the pool of 12 GM12878 cells . ( c ) Same plot as in ( a ) but based on PU . 1 binding sites in K562 cells . Regions are centered on motif locations . Shown is the average methylation across a 2 kb window of the pool of 11 K562 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 03110 . 7554/eLife . 23203 . 032Figure 5—figure supplement 4 . Scores at CTCF motifs with footprints are significantly higher than those without . Boxplot representing the CTCF motif scores in regions with and without CTCF footprint of an individual GM12878 cell ( GM_1 , the same cell as shown in Figure 3d and e ) . Within individual cells regions with footprint are associated with higher motif scores ( p < 2 . 2 e-16 , t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 03210 . 7554/eLife . 23203 . 033Figure 5—figure supplement 5 . Loci with CTCF footprint in single cells . At each locus scNOMe-seq data from this study and DNase hypersensitivity data from ENCODE are shown . scNOMe-seq data tracks show methylation status of individual GpCs . Each row corresponds to data from a single cell , Colors indicate the methylation status of each GpC ( yellow: methylated; blue: unmethylated ) . Data are displayed as tracks in the UCSC genome browser . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 033 The pattern of GpC methylation adjacent to CTCF sites suggested that scNOMe-seq also detected the well-positioned nucleosomes flanking these regions ( Figure 5a ) ( Kelly et al . , 2012 ) . This observation was confirmed by the oscillatory distribution of the average GpC and CpG methylation around locations of well-positioned nucleosomes identified from MNase-seq data ( ENCODE Project Consortium , 2012 ) ( Figure 6a ) . While nucleosome core particles are invariably associated with DNA fragments of 147 bp , nucleosomes are separated by linker DNA of varying lengths , resulting in different packaging densities between cell types and between genomic regions within a cell ( Valouev et al . , 2011; Schones et al . , 2008 ) . To determine whether scNOMe-seq data can be used to measure the average linker length , average distances between nucleosome midpoints in single cells ( phasing distances ) were estimated by correlating the methylation status between pairs of cytosines in GpC di-nucleotides at offset distances from 3 bp to 400 bp ( Figure 6c , d and Figure 6—figure supplements 1 and 2 ) . The estimated phases fell between 187 bp and 196 bp ( mean = 196 . 7 bp ) in GM12878 cells , and between 188 bp and 200 bp ( mean = 194 . 2 bp ) in K562 cells ( Figure 6e ) . These estimates are in general agreement with phase estimates derived from MNase-seq data in human cells ( Valouev et al . , 2011 ) . In addition , estimated phasing distances varied within individual cells depending on the chromatin context , similar to observation from bulk MNase-seq data ( Valouev et al . , 2011 ) ( Figure 6f ) . 10 . 7554/eLife . 23203 . 034Figure 6 . Nucleosome phasing in single cells . ( a ) Average GpC methylation level and ( b ) CpG methylation level at well-positioned nucleosomes in GM12878 cells . Regions are centered on midpoints of top 5% of positioned nucleosomes . Shown is the average methylation across a 2 kb window of the pool of 12 GM12878 cells . ( c ) , ( d ) Correlation coefficients for the comparison in methylation status between GpCs separated by different offset distances for GM12878 ( c ) and K562 ( d ) cells . Each line represents a single cell . Data are smoothened for better visualization . ( e ) Distribution of estimated phase lengths for GM12878 and K562 cells . ( f ) Nucleosome phasing in GM12878 in genomic regions associated with different chromatin states defined by chromHMM ( ENCODE ) . Boxplot represents the distribution of estimated phase lengths from all 12 GM12878 cells and overlaid points indicate values of each individual cells . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 03410 . 7554/eLife . 23203 . 035Figure 6—figure supplement 1 . Number of nucleotide pairs used for correlation at each offset distance . Plotted is the number of nucleotide pairs that are found at each offset distance and used to calculate the correlation coefficient at that distance . The number of comparison declines precipitously . While each read has a maximum of 100 bp all samples were sequenced in paired-end mode and even though the alignment was performed in single end mode additional comparisons can therefore be made in most cases based on GpCs covered in the read from the opposite side of a fragment . In many cases , the data become too sparse beyond 400 bp . Each line represents data from an individual cell . Data from GM12878 and K562 cells are plotted on the left and right , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 03510 . 7554/eLife . 23203 . 036Figure 6—figure supplement 2 . Offset distances with a high proportion of GpC pairs that share methylation status indicate average phasing distance . Shown is the proportion of nucleotide pairs at each offset distance in which both cytosines are methylated . These measurements yield curves very similar to the distribution of Pearson correlation coefficients ( Figure 6c and d ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23203 . 036
In this study , I demonstrated that scNOMe-seq simultaneously measures chromatin accessibility by GpC methylation as well as endogenous CpG and DNA methylation in single cells . scNOMe-seq detected chromatin accessibility at DHSs and TFBS and , in aggregate , these data recapitulated NOMe-seq data obtained from bulk cells ( Kelly et al . , 2012 ) . scNOMe-seq data also detected footprints of CTCF binding , and was used to estimate nucleosome phasing distances . Similar to other single cell genomic methods , scNOMe-seq relies on annotations obtained from bulk measurements ( Cusanovich et al . , 2015; Buenrostro et al . , 2015b; Smallwood et al . , 2014; Farlik et al . , 2015 ) . A limitation of single cell genomic methods is their sparse coverage which leads to high allelic drop-out . For methods in which the signal is based on counting the sequenced fragments , such as ATAC-seq and DNase-seq , this poses a challenge since true negatives at a specific location cannot be distinguished from false negatives that are a consequence of read loss . Compared to these methods , scNOMe-seq has the unique advantage , that reads are recovered independently of the signal and allelic drop-out events therefore can be distinguished from closed or inaccessible chromatin configurations . The frequency of accessible sites in the population of DHSs can be estimated . Using this approach only about 30–50% of DHSs detected in the population were found accessible in a single cell , depending on the thresholds chosen to call a site accessible . While this assessment would have been possible using bulk NOMe-seq data , scNOMe-seq offers important possibilities for future applications . For example , to compare accessibility across multiple loci within a single cell and the use of heterogeneous cellular mixtures as input material . As expected , the chance of a covered DHS to being open or closed is not equally distributed across all DHSs from the population . Instead , DHSs with strong DNaseI accessibility showed a higher frequency of accessibility in single cells compared to those sites with low DNaseI accessibility in the population ( Figure 2e ) suggesting that the peak height is indeed directly related to the frequency with which a site is accessible in individual cells . In agreement with this observation a large proportion of variability observed between cells was attributable to DHSs with low DNaseI accessibility in bulk samples ( Figure 2—figure supplement 11 ) . In principle , variation between cells could be due to differential GpC MTase enzyme activity . However , the genome-wide levels of GpC methylation reached comparable levels in all cells and the variability between cells was not equally distributed across all DHS ( Figure 2d , Figure 2—figure supplement 1 ) Measuring similarity of chromatin accessibility between cells was sufficient to group GM12878 and K562 cells based on their cell type of origin ( Figure 3a ) . In this particular case , the separation is confounded with experimental batches . However , higher average GpC methylation in DHSs for the respective cell type compared to the DHSs of the other cell type indicated that scNOMe-seq can differentiate the two cell types ( Figure 2—figure supplement 14 ) . Similarly , endogenous CpG methylation at different genomic features ( DHS , 10 kb windows , gene bodies ) was sufficient to distinguish between the two cell types . This approach should be extendable to scNOMe-seq data from samples containing mixtures of cell types . scNOMe-seq measures chromatin accessibility at GpC di-nucleotides along the entire length of a sequencing read . Since most features that bind DNA are smaller than the length of 100 bp ( 200 bp within 200–500 bp regions in the case of paired end reads ) , the regions covered by sequence-specific transcription factors and nucleosomes can be captured within a single fragment . This allows one to directly detect binding of TFs provided that their sequencing motif contains at least one GpC di-nucleotide . I demonstrated the feasibility of this approach using CTCF binding sites . Of note , most motifs within regions of CTCF ChIP-seq peaks were protected from GpC methylation ( ‘footprint’ ) ( Figure 5 ) . In agreement with an inferred binding event as the cause for this protection , scores for CTCF motifs that were associated with a footprint were significantly higher than for motifs without a footprint . Depending on the motif specificity of a given TF and provided that their motifs contain a GpC dinucleotide , similar measurements should be feasible for many TFs and could be used to infer the activity of a range of transcription factors in single cells or to measure combinatorial binding of two or more TFs . Estimation of the average nucleosome phasing distances allows one to study chromatin compaction and complements the measurements of chromatin accessibility at regulatory regions and DNA methylation . The estimates from individual cells fit very well with measurements made from MNase-seq data in bulk samples ( Valouev et al . , 2011 ) . It remains to be established whether the variation in phasing distances between individual cells is of biological or technical nature ( Figure 6e ) . These proof-of-principle experiments have been performed using commercial kits for bisulfite conversion and library amplification , additional optimization or alternative amplification approaches ( Smallwood et al . , 2014 ) are likely to increase the yield substantially . Compared to other single cell methods , for example ATAC-seq , scNOMe-seq does not enrich for accessible chromatin regions and thus requires significantly more sequencing coverage . Ultimately , it should be possible to integrate the GpC MTase treatment into microfluidic workflows and combine this method with scRNA-seq , similar to recently published methods that combine scRNA-seq and methylome- sequencing ( Angermueller et al . , 2016 ) . This study was primarily designed to test the feasibility of NOMe-seq in single cells and only a small number of nuclei where sequenced for each cell line . As a consequence , this set up could not be used to study cell-to-cell variation in detail . scNOMe-seq will be particularly useful for studies that aim to simultaneously measure chromatin accessibility and DNA methylation . This approach will be especially powerful for the characterization of chromatin organization in single cells from heterogeneous mixtures or complex tissues , for example to samples of brain tissues or primary cancer cells .
GM12878 ( RRID:CVCL_7526 ) and K562 ( RRID:CVCL_0004 ) cells were obtained directly from Coriell and ATCC , respectively . No further confirmation of the authenticity of these cell lines or mycoplasma testing has been performed . GM12878 were grown in RPMI medium 1640 ( Gibco ) , supplemented with 2 mM L-Glutamine ( Gibco ) , and Penicilin and Streptavidin ( Pen Strep , Gibco ) , and 15% fetal bovine serum ( FBS , Gibco ) . K562 were grown in RPMI medium 1640 of the same composition but with 10% FBS . Cells were grown at 37 C and in 5% CO2 . NOMe-Seq procedure was performed based on protocols for CpG methyltransferase M . SSsI described in Miranda et al . ( 2010 ) and the GpC methyltransferase from M . CviPI ( Kelly et al . , 2012 ) , with some modification . Between 2 × 10^6 and 5 × 10∧6 cells were harvested by centrifuging the cell suspension for 5 min at 500x g . Cells were washed once with 1x PBS , re-suspended in 1 ml lysis buffer ( 10 mM Tris-HCl pH 7 . 4 , 10 mM NaCl , 3 mM MgCl2 ) and incubated for 10 min on ice . IGEPAL CA-630 ( Sigma ) was added to a final concentration of 0 . 025% and the cell suspension was transferred to a 2 ml Dounce homogenizer . Nuclei were released by 15 strokes with the pestle . Success of lysis was confirmed by inspection under a light microscope . Nuclei were collected by centrifuging the cell suspension for 5 min at 800x g at 4 C and washed twice with cold lysis buffer without detergent . One million nuclei were resupended in reaction buffer to yield a suspension with a final concentration of 1x GpC MTase buffer ( NEB ) , 0 . 32 mM S-Adenosylmethionine ( SAM ) ( NEB ) , and 50 ul of GpC methyltransferase ( 4 U/ul ) ) from M . CviPI ( NEB ) . The final reaction volume was 150 ul . The suspension was carefully mixed before incubating for 8 min at 37 C after which another 25 ul of enzyme and 0 . 7 ul of 32 mM SAM were added for an additional 8 min incubation at 37C . To avoid disruption of nuclei incubation was stopped by adding 750 ul of 1x PBS and collecting the nuclei at 800 xg . Supernatant was removed and nuclei were re-suspended in 500 ul 1x PBS containing Hoechst 33342 DNA dye ( NucBlue Live reagent , Hoechst ) . Nuclei were kept on ice until sorting . For preparation of bulk libraries in GM21878 cell , nuclei preparation and GpC MTase treatment was performed as described above . Nuclei were lysed immediately after incubation and DNA was isolated using Phenol/Chloroform purification . Nuclei were sorted at the Flow Cytometry core at the University of Chicago on a BD FACSAria or BD FACSAria Fusio equipped with a 96-well-plate holder . To obtain individual and intact nuclei gates were set on forward and side scatter to exclude aggregates and debris . DAPI/PacBlue channel or Violet 450/500 channel were usedto excited the Hoechst 33342 DNA dye and to gate on cells with DNA content corresponding to cells in G1 phase of the cell cycle in order to maintain similar DNA content per cell and to remove potential heterogeneity attributable to cell cycle . Cells were sorted into individual wells pre-filled with 19 ul of 1x M-Digestion buffer ( EZ DNA Methylation Direct Kit , Zymo Research ) containing 1 mg/ml Proteinase K . Following collection , the plates were briefly spun to collect droplets that might formed during handling . Nuclei were lysed by incubating the samples at 50 C for 20 min in a PCR cycler . DNA was subjected to bisulfite conversion by adding 130 ul of freshly prepared CT Conversion reagent ( EZ DNA Methylation Direct Kit , Zymo ) to the lysed nuclei . Conversion was performed by denaturing the DNA at 98 C for 8 min followed by 3 . 5 hr incubation at 65 C . DNA isolation was performed using the EZ DNA Methylation Direct Kit ( Zymo Research ) following the manufacturer’s instruction with the modification that the DNA was eluted in only 8 ul of elution buffer . Libraries were prepared using the Pico Methyl-seq Library prep Kit ( Zymo Research ) following the manufacturer’s instruction for low input samples . Specifically , the random primers were diluted 1:2 before the initial pre-amplification step and the first amplification was extended to a total of 10 amplification cycles . Libraries were amplified with barcoded primers allowing for multiplexing . The sequences can be found in Supplementary file 1 , primers were ordered from IDT . The purification of amplified libraries was performed using Agencourt AMPureXP beads ( BeckmannCoulter ) , using a 1:1 ratio of beads and libraries . Concentration and size distribution of the final libraries was assessed on an Bioanalyzer ( Agilent ) . Libraries with average fragment size above 150 bp were pooled and sequenced . Libraries were sequenced on Illumina HiSeq 2500 in rapid mode ( K562 cells ) and HiSeq4000 ( GM12878 cells ) . Sequences were obtained using 100 bp paired-end mode . For processing and alignment each read from a read pair was treated independently as this slightly improved the mapping efficiency . Before alignment , read sequences in fastq format were assessed for quality using fastqc ( http://www . bioinformatics . babraham . ac . uk/projects/fastqc/ ) . Reads were trimmed to remove low quality bases and 6 bp were clipped from the five prime end of each read to avoid mismatches introduced by amplification . In the case of GM12878 cells 6 bp were clipped from either end of the read . Only reads that remained longer than 20 bp were kept for further analyses . These processing steps were performed using trim_galore version 0 . 4 . 0 ( http://www . bioinformatics . babraham . ac . uk/projects/trim_galore/ ) with the following settings: trim_galore --quality 30 --phred33 --illumina --stringency 1 -e 0 . 1 --clip_R1 6 --gzip --length 20 -- output_dir outdir Sample . fastq . gz . The trimmed fastq files were aligned using the bisulfite aligner bismarck version 0 . 15 . 0 ( Krueger et al . , 2012 ) which calls bowtie2 ( Langmead and Salzberg , 2012 ) internally . Reads were aligned to the human genome ( genome assembly hg38 ) . Reads were aligned in single read mode using default settings . The amplification protocol used to generate the scNOMe-seq libraries yielded non-directional libraries and alignment was performed with the option —non_directional ( bismark --fastq --prefix SamplePrefix --output_dir output_dir -- non_directional --phred33-quals --score_min L , 0 , –0 . 2 --bowtie2 genome_file trimmed . fastq . gz ) . Some libraries contained small amounts of DNA from C . elegans as spike-ins , however these were not used during the analysis . Duplicates were removed using samtools version 0 . 1 . 19 ( Li et al . , 2009 ) on sorted output files from bismark ( samtools rmdup Sample Prefix . sorted . bam Sample Aligned_rmdup . bam ) . Coverage and methylation status of all cytosines was extracted using bismark_methylation_extractor ( Krueger et al . , 2012 ) ( bismark_methylation_extractor -s --ignore 6 --output outdir --cytosine_report --CX --genome_folder path_to_genome_data SampleAligned_rmdup . bam ) . The resulting coverage files were used to extract the methylation status of cytosines specifically in GpC and CpG di-nucleotides using the coverage2cytosine script which is part of Bismark ( Krueger et al . , 2012 ) . The resulting coverage files contained cytosines in GCG context which are ambiguous given that they represent a cytosine both in GpC and CpG di- nucleotides . Coordinates of these ambiguous positions were identified using oligoMatch ( Kent et al . , 2002 ) and these positions were removed from the coverage files . The number of unconverted cytosines ( estimated based on apparent methylation rates in non-GpC and non-CpG context ) was low in all libraries ( <1% ) . However , it was noted that unconverted cytosines were not randomly distributed but associated with entirely unconverted reads . Regions covered by a read with more than three unconverted cytosines in non-CpG and non-GpC context were removed from further analysis as well . The genotype was not taken into account as its effect on calling the methylation status incorrectly was deemed negligible for the analyses performed here . ScNOMe-seq data were compared to a number of genomic features in GM12878 and K562 cells collected by Encode ( ENCODE Project Consortium , 2012 ) which were downloaded through the UCSC data repository ( Karolchik et al . , 2014 ) . These datasets are listed in Supplementary file 1 . While the scNOMe-seq data were aligned against human genome assembly hg38 , some of the datasets were only available on genome assembly hg19 and the coordinates of these datasets were lifted from hg19 to hg38 using liftOver ( Kent et al . , 2002 ) ( default re-mapping ratio 0 . 95 ) . Nucleosome positions based on MNase-seq data in GM12878 were determined with DANPOS version 2 . 2 . 2 ( Chen et al . , 2013 ) using default settings . Resulting intervals were lifted to hg38 . After removing summit locations with occupancy values above 300 , the top 5% ( 713361 ) of nucleosome positions based on their summit occupancy value were used . GpC and CpG methylation density across intervals encompassing DNase hypersensitivity sites ( DHSs ) , transcription factor binding sites ( TFBS ) , and well positioned nucleosomes was calculated across the 2 kb regions centered on the middle of these regions using the scoreMatrixBin function in the genomation package ( Akalin et al . , 2015 ) in R ( R Core Team , 2015 ) . Data were aggregated in 5 bp bins for each region and across all regions covered in a single cell . The average methylation level in pre-defined intervals ( DHSs , TFBS ) was determined by computing the average GpC or CpG methylation for each interval together with the number of GpC/CpGs covered in this interval using the map function in bedtools ( Quinlan and Hall , 2010 ) . If no other cut-offs were given , DHSs were considered ‘covered’ and used in analyses when at least 4 GpCs occurring within the predefined interval were covered by sequencing data in an individual cell . Because the frequency of CpG di-nucleotides is significantly lower , only 2 CpGs were required in order for a DHSs to be considered covered for analyses that focused on endogenous DNA methylation . To count the number of cytosines within the primary sequence of a given DHSs only cytosines on the forward strand were counted . While each GpC dinucleotide can be measured on both strands and would therefore yield a count of two cytosines the data are sparse and each location will get at most a single read . This approach should therefore give a more conservative estimate of the possible GpC coverage . For analyses that used the scores of the peak regions , the peak scores reported the datasets from bulk samples were used ( ENCODE Project Consortium , 2012 ) . For analyses that were centered on transcription factor binding motifs the PWMs were obtained from the JASPAR database ( Tan , 2014 ) ( Tan ) for the TFs CTCF ( MA0139 ) , EBF1 ( MA0154 ) , and PU . 1 ( MA0080 ) . Genome-wide scanning for locations of sequence matches to the PWMs was performed using matchPWM in the Biotstring package ( Pages et al . , 2016 ) in R with a threshold of 75% based on the human genome assembly hg38 . All plots were prepared using ggplot2 ( Wickham , 2009 ) , with the exception of heatmaps displaying the average methylation density around genomic features in individual cells which were prepared using heatmap . 2 in gplots ( Warnes et al . , 2016 ) . Similarity in accessible chromatin between cells was calculated based on Jaccard similarity . Jaccard similarity index ( Equation 1 ) was calculated between pairs of samples by first obtaining the intersection of DHSs covered in both samples of a pair with more than 4 GpCs . Each feature was annotated as open or closed , depending on the methylation status ( > = 40% methylation ) and only pairs in which at least one of the members was open were considered for this comparison . ( 1 ) jac ( A , B ) = ( A∩B ) ( A∪B ) The similarity between samples from GM12878 and K562 cells was calculated based on the union set of DHSs from both cell lines . The similarity indexes of all pairwise comparisons were used to compute the distances between each cell . The resulting clustered data were displayed as a heat map . CTCF footprints were measured by comparing the GpC methylation level in each motif to the methylation level in the 50 bp flanking regions immediately upstream and downstream of the motif . Overlapping motifs were merged into a single interval before determining the coordinates for flanking regions . To ensure sufficient GpC coverage for each interval the resulting three adjacent intervals for each locus were required to contain at least one covered GpC each , and four covered GpCs in total . This analysis only included regions that were accessible based on the methylation status of the flanking regions ( at least 50% ) . A CTCF 'footprint score' was determined by simply subtracting the average GpC methylation of the flanking regions from the GpC methylation of the motif . scNOMe-seq data were displayed in the UCSC genome browser ( Kent et al . , 2002 ) by converting the GpC methylation coverage file into a bed file and using the methylation value as score . To facilitated the visualization of the data in the context of previous Encode data the methylation files were lifted to hg19 . The tracks shown together with scNOMe-seq data are Open Chromatin by DNaseI HS from ENCODE/OpenChrom ( Duke University ) for DNaseI hypersensitivity , Nucleosome Signal from ENCODE/Stanford/BYU , and CTCF ChIP-seq signal from Broad Histone Modification by ChIP-seq from ENCODE/Broad Institute . All data are from GM12878 cells . Nucleosome phasing estimates were obtained by first calculating the correlation coefficients for the methylation status of pairs of GpCs ad different offset distances . These values were computed using a custom python script ( Source code 1 – NucPhasing . py ) . Essentially , pairs of sequenced cytosines in GpC di-nucleotides were collected for each offset distance from 3 bp to 400 bp cytosine . At each offset distance the correlation of the methylation status was calculated across all pairs . Correlation coefficients were plotted against the offset distances revealing periodic changes in the correlation coefficient . The smoothened data were used to estimate the phasing distances by obtaining the offset distance corresponding to the local maximum found between 100 bp and 300 bp . To determine phase lengths of nucleosomes in different chromatin contexts the GpC coverage files were filtered for positions falling into categories defined by chromHMM ( ENCODE Project Consortium , 2012; Ernst et al . , 2011 ) before obtaining the correlation coefficients . Raw data and methylation coverage files are available at GEO ( https://www . ncbi . nlm . nih . gov/geo/ ) under the accession number GSE83882 . | DNA contains all the information and instructions needed to build an organism and enables its cells to fulfil their role . There are many different cell types in animals , and each type only needs a small portion of the information found in the DNA to do its job . Hence , only the sections – also known as genes – specific to a particular role need to be active or ‘expressed’ in any given cell type . The sets of genes that are active in a cell determine what that cell can do and make it different from other cell types . Short regions that surround the genes act as ‘switches’ to control their activity . To study how these switches work , researchers have developed techniques that can measure when specific ones are on or off in different cell types . A technique called NOMe-seq can simultaneously identify active and inactive regions in the genome by measuring markers that are specific for active and inactive regions . However , this approach was created for large samples containing thousands of cells , so-called bulk samples . Since the pattern of gene expression can vary between individual cells even if they are of the same cell type , it is important to analyse each individual cell . Sebastian Pott has now adapted the NOMe-seq technique for use in single cells . The modified protocol was used to study human cells that have been well studied and can be grown in the laboratory . The results of this proof-of-principle study showed that the adapted version of the technique obtained similar results as when used on bulk samples . This demonstrates that the method can reliably measure active and inactive regions in single cells across the genome . The next step will be to use this tool to study how genes are affected in diseases like cancer , where gene expression can be variable between individual cells . For example , understanding the differences between individual cells within a cancer sample might help to understand why some cancers react to treatments better than others . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"tools",
"and",
"resources",
"genetics",
"and",
"genomics"
] | 2017 | Simultaneous measurement of chromatin accessibility, DNA methylation, and nucleosome phasing in single cells |
Although shape perception is considered a function of the ventral visual pathway , evidence suggests that the dorsal pathway also derives shape-based representations . In two psychophysics and neuroimaging experiments , we characterized the response properties , topographical organization and perceptual relevance of these representations . In both pathways , shape sensitivity increased from early visual cortex to extrastriate cortex but then decreased in anterior regions . Moreover , the lateral aspect of the ventral pathway and posterior regions of the dorsal pathway were sensitive to the availability of fundamental shape properties , even for unrecognizable images . This apparent representational similarity between the posterior-dorsal and lateral-ventral regions was corroborated by a multivariate analysis . Finally , as with ventral pathway , the activation profile of posterior dorsal regions was correlated with recognition performance , suggesting a possible contribution to perception . These findings challenge a strict functional dichotomy between the pathways and suggest a more distributed model of shape processing .
Shape is the most fundamental perceived property of objects , and , accordingly , shape processing is crucial for successful visual object recognition ( Palmer , 1999 ) . Deriving information about the shapes of objects in the input has long been considered to be under the purview of the ventral visual pathway ( Goodale and Milner , 1992; Ungerleider and Mishkin , 1982 ) and many functional imaging studies have provided evidence to support this claim ( e . g . , Freud et al . , 2013; Grill-Spector et al . , 1998; Kourtzi and Kanwisher , 2000; Lerner et al . , 2001; Malach et al . , 1995 ) . Specifically , these investigations have uncovered a gradient of shape sensitivity in the ventral pathway , with posterior regions in the early to mid-visual cortex ( i . e . , V1-hV4 ) being less responsive to the shape of the object than more anterior regions , such as the Lateral Occipital Cortex ( LOC ) and the Fusiform Gyrus ( FG ) . Additionally , shape sensitivity in these latter , anterior regions is correlated with perceptual abilities , and a lesion to these areas results in an impairment in object perception ( e . g . , Freud et al . , 2017a; Goodale et al . , 1991; Konen et al . , 2011 ) . It appears , however , that shape perception is not solely a product of the computations mediated by the ventral pathway . Human neuroimaging studies ( Freud et al . , 2015; Jeong and Xu , 2016; Konen and Kastner , 2008; Zachariou et al . , 2014; Zachariou et al . , 2017; Bracci and Op de Beeck , 2016; Bracci et al . , 2017 ) , primate electrophysiological studies ( Durand et al . , 2007; Janssen et al . , 2008; Janssen et al . , 2000; Van Dromme et al . , 2016 ) ; for review see Theys et al . , 2015 ) , comparative studies between human and primates ( Denys et al . , 2004; Sawamura et al . , 2005 ) and human neuropsychology ( Freud et al . , 2017a ) have all uncovered object representations mediated by the dorsal pathway , even under conditions in which no visuomotor response is required . However , many questions remain concerning the topographical organization of the dorsal object representations , the nature of the visual properties encoded in these representations and their contribution to perceptual behavior . In the current study , we addressed these questions in two studies , each combining functional magnetic resonance imaging ( fMRI ) and psychophysical measures . We hypothesized that the large-scale topographical organization of dorsal object representations obeys a spatial gradient in which posterior regions of the pathway derive shape representations and contribute to visual perception , whereas more anterior regions derive representations that are better tuned to subserve visuomotor behaviors ( Freud et al . , 2016 ) . Our motivation stems from two key issues concerning dorsal pathway function: the first is the need to reconcile recent findings of dorsal pathway activation under conditions of perception ( as noted above ) with the well-established findings that the dorsal pathway is engaged in object-directed actions such as grasping and manipulation ( e . g . , Culham et al . , 2003; Fabbri et al . , 2016; for a review , see Gallivan and Culham , 2015 ) . The second concerns the connectivity that anchors the parietal representational continuum at one end with more caudal regions connected to visual cortex ( Greenberg et al . , 2012 ) and , at the other end , with more rostral regions anatomically and functionally connected to motor cortex ( Davare et al . , 2010 ) . We adopt an approach that has been used previously to characterize the neural basis of shape processing in ventral cortex in response to stimuli in which shape information is parametrically eliminated by increasingly distorting images of objects ( Lerner et al . , 2001 ) ( see Figure 1A for example ) . The decrease in BOLD activation with increased scrambling serves as an index of shape sensitivity ( Denys et al . , 2004; Grill-Spector et al . , 1998; Lerner et al . , 2001; Malach et al . , 1995; Murray et al . , 2002 ) . Following these previous studies , in the first experiment , we adopted a parametric box scrambling procedure to map the large-scale organization of shape processing along the ventral and dorsal visual pathways and the contribution of these representations to shape perception . In addition , we employed more advanced multivariate analyses to further elucidate the nature of representations and the similarities or differences in these representations in the two pathways . However , the box-scrambling method has some inherent limitations . First , it increases the number of edges , and early visual cortex is especially sensitive to this type of information ( Grill-Spector et al . , 1998; Lerner et al . , 2001 ) . Second , box scrambling disrupts crucial shape attributes such as good continuation and figure-ground segregation ( Koffka , 1935; Qiu and von der Heydt , 2005; Read et al . , 1997 ) , and , unsurprisingly , eliminates the identity information of the stimulus . Furthermore , in the box scrambling experiment , participants were shown each object at every level of scrambling and so there may have been priming or adaptation for the same object across levels of scrambling . To circumvent these limitations and verify the results , in the second experiment we used a diffeomorphic scrambling manipulation that increasingly precludes the identification of the object but preserves some fundamental shape properties by repeatedly applying a flow field generated from a set of two-dimensional cosine components with random phase and amplitude ( Stojanoski and Cusack , 2014 ) . Consequently , some shape information ( i . e . , the presence of an object or figure that is clearly differentiated from the background and has well-defined boundaries , relatively uniform texture and center of mass; see methods and Figure 1B for quantitative analyses of object distortion ) is still largely available even in the most distorted version , but the object itself is parametrically distorted and unrecognizable . This manipulation enabled us to re-examine the large-scale organization of the two pathways , and to disentangle sensitivity to shape versus identity information . Last , each object was presented at only one level of distortion ( counterbalanced across participants ) to remove possible priming effects .
A key question that emerges is whether the neural patterns we have uncovered bear any relation to perceptual performance . Previous studies have observed a robust correlation between fMRI activation in object-selective regions of the ventral pathway and object recognition abilities ( Avidan et al . , 2002; Grill-Spector et al . , 2000 ) . Because this correlation has not been examined for the dorsal pathway in the context of non-spatial perceptual tasks ( rather than , for example , working memory , Jeong and Xu , 2016 ) , we correlated the fMRI activation for the different levels of scrambling with recognition abilities measured outside the scanner . The dorsal pathway is known to be highly responsive to images of tools , which convey visuomotor information ( Chao and Martin , 2000; Lewis , 2006; Mruczek et al . , 2013 ) . Our stimulus set included pictures of both tools and non-tool objects , and it could be argued that the shape sensitivity along the dorsal pathway may be a product of visuomotor associations related specifically to tools rather than of more general shape processing . Closer scrutiny of the image statistics suggests that a direct comparison between tools and objects is ill-advised in the present study , because the image statistics from the two categories differed dramatically ( e . g . , number of pixels: objects [mean = 50 , 768 , SD = 15 , 665] and tools [mean = 17 , 475 , SD = 10 , 158] ) and shape attributes ( e . g . , tools are usually elongated , while objects have more diverse shapes ) ( Bracci and Op de Beeck , 2016; Chen et al . , 2017 ) . Nevertheless , to ensure that the activation in the dorsal pathway was not solely a result of visuomotor associations , we reanalyzed the data of the two experiments using only the fMRI blocks in which objects ( and not tools ) were presented . Notwithstanding the loss of statistical power , the reanalysis fully reproduced the analyses described above at the voxel-wise and ROI level ( see Figure 2—figure supplement 1 upper panel , Figure 3—figure supplement 1-upper panel ) . The results of the two experiments , therefore , cannot be ascribed to the presence of tool stimuli that convey visuomotor associations , and , instead , bolster to the conclusion that posterior regions of dorsal cortex are responsive to the shape of the visual input .
Consistent with early investigations ( e . g . , Grill-Spector et al . , 1998; Grill-Spector and Weiner , 2014; Lerner et al . , 2001; Murray et al . , 2002 ) , shape sensitivity , was not evident in the caudal parts of the ventral pathway ( i . e . , early visual cortex ) and emerged in the rostral and lateral parts of the ventral pathway , reflecting the hierarchical nature of object processing . However , in the transition from high-level visual regions ( i . e . , parahippocampal gyrus , fusiform gyrus ) to even more anterior temporal regions , shape sensitivity was still present , albeit reduced ( i . e . , a flatter slope ) . Interestingly , in recent years , accumulating evidence has shown that the anterior and medial regions of the temporal lobe ( i . e . , perirhinal and entorhinal cortices ) may also play a role in object recognition ( Behrmann et al . , 2016 ) ; for a review see Murray et al . , 2007 ) . The results of the present investigation are compatible with the view that these more anterior representations may be confined to high-level object properties ( such as conjunction of multiple features see , Barense et al . , 2007; or familiarity , see Martin et al . , 2013 ) and to memory-based representations , and less to shape and geometric information per se . The novel findings pertain more to the nature of shape processing along the dorsal pathway . In particular , shape sensitivity increased from early visual cortex to extrastriate cortex , reached its peak sensitivity in the posterior parietal cortex and then gradually decreased in more anterior regions , closer to the central sulcus . These findings extend previous studies in both humans and non-human primates that have demonstrated that , similar to the ventral pathway , the dorsal pathway derives object representations . ( e . g . , Denys et al . , 2004; Freud et al . , 2015; Konen and Kastner , 2008; Theys et al . , 2015; Van Dromme et al . , 2016 ) . The similarity between the two pathways suggest that the topographical organization is constrained by similar factors , namely cortical distance from the visual cortex and connectivity to other cortical systems ( i . e . , the motor system in the dorsal pathway and semantic system in the ventral pathway ) . Finally , the similarity in the large-scale organization of the two pathways was further confirmed by an RSA approach that revealed that the representational structure of regions in the posterior dorsal pathway was highly correlated with the representational structure of lateral-ventral ROI . Notably , despite these similarities , differences were also observed between the two pathways: shape sensitivity was greater in the ventral than dorsal pathway , reflecting the centrality of this pathway in object perception , and shape sensitivity reached its maximum values in more anterior regions of the ventral than dorsal pathway . Some fMRI investigations have provided evidence that pictures of tools activate regions in the dorsal pathway ( Chao and Martin , 2000; Valyear et al . , 2007 ) , even when no overt visuomotor task was required . Nevertheless , one might speculate that the observed activation might still reflect visuomotor plans that are associated with the tool being displayed . More recent studies , however , have documented dorsal activation for both 2D and 3D images , even when the stimuli do not have any visuomotor association and the task is not action-based ( Freud et al . , 2015; Konen and Kastner , 2008; Zachariou et al . , 2014 ) . Moreover , when BOLD responses to tools were compared to non-tools ( but graspable ) objects , no difference was observed between the two categories in the posterior parts of the dorsal pathway , but greater activation was observed for tools than for objects in more anterior regions ( Mruczek et al . , 2013 ) . In the current work , shape sensitivity within the dorsal pathway was found both for tools and non-tool objects . Together , these results suggest that the representations subserved by the posterior regions of the dorsal pathway are not limited to visuomotor associations evoked by the object . Instead , the neural representation may reflect the processing of different shape cues such as the 3D status of the object ( Berryhill et al . , 2009; Freud et al . , 2017a; Konen and Kastner , 2008; Van Dromme et al . , 2016 ) and/or object elongation ( Chen et al . , 2017; Fabbri et al . , 2016 ) . Recent studies have used multivariate analytic approaches to elucidate the visual and cognitive features represented by the dorsal ( and ventral ) regions . For example , Bracci and Op de Beeck , 2016 mapped the cortical sensitivity for a stimulus set in which shape and category ( e . g . , animals , musical instruments , tools ) were dissociated . Similar to the present findings , posterior LOC was more correlated with a shape model than a category model . In the dorsal pathway , the TOS , which corresponds to posterior ROIs in the probabilistic atlas used here , was also highly correlated with the shape model , and not with the category model . Nevertheless , in this previous investigation , some anterior dorsal ROIs were more correlated with the category model , than the shape model . As discussed below , dorsal pathway representations were found to be highly sensitive to task properties ( Bracci et al . , 2017 ) and therefore , the discrepancies between the present investigation and Bracci and Op de Beeck’s findings might be related to the nature of the task: while Bracci and Op de Beeck asked participants to compare the real-size of preceding images ( encouraging object-based processing ) , in the present investigation , we deliberately avoided explicit processing of the stimuli and had participants complete an orthogonal fixation-based task . The comparison between the box scrambling experiment and the diffeomorphic scrambling provided additional information on the nature of shape representations derived by the dorsal pathway . The diffeomorphic transformation distorted the object’s identity , while preserving the presence of some shape information ( to a greater degree than was true of the box scrambling manipulation ) . In contrast to the inferior surface of the ventral pathway , along the dorsal pathway a decrease in shape sensitivity was found for this manipulation , compared with the results obtained from the box scrambling experiment . Hence , our results suggest that dorsal pathway representations are tied to the presence of a single coherent shape , even if it is not identifiable as a familiar object and lacks important visual properties . The residual perceptual abilities of patients with visual agnosia are consistent with such interpretation . In particular , the dissociable dorsal representations can support sensitivity to particular attributes ( such as the 3D structure of an object ) but cannot support intact recognition abilities ( Freud et al . , 2017a ) . Note that this interpretation ought to be considered speculative as the shape attribute is confounded with other visual properties that constitute a shape in the diffeomorphic scrambling experiment: our image analysis procedures revealed that a host of factors , related to low- , mid- and high-level shape cues , such as defined perimeter , entropy and homogeneous texture , were also better preserved for the diffeomorphic than box-scrambled stimuli . Moreover , and in contrast to previous investigations , the RSA in the present study was based on a block design experiments , and therefore , could not uncover the representational content of a specific exemplar . Future studies should , therefore , explore the importance of different visual properties to dorsal pathway representations , and , by doing so , evaluate whether dorsal and ventral representations differ quantitatively ( or also qualitatively ) . According to Goodale and Milner ( 1992 ) , the visual pathways should be described in terms of their functions , rather than in terms of the visual information that is represented . The question then is , what behavioral functions are subserved by dorsal pathway representations ? Here , we provide novel evidence for a correlation between recognition abilities and fMRI shape sensitivity in the posterior part of the dorsal pathway , even when the correlation of these variables with scrambling levels was partialled out . Despite the inability to infer causality from those correlations , they extend seminal findings that found correlations between perceptual performance and fMRI activation in different regions of the ventral pathway ( Grill-Spector et al . , 2000 ) and point to a plausible functional contribution of the dorsal pathway to object perception . Consistent with this suggestion , a patient with an occipitoparietal lesion was impaired in the perception of objects defined by monocular and binocular cues ( Berryhill et al . , 2009 ) . Moreover , recent studies , using TMS in humans and reversible deactivation in non-human primates , have successfully established a causal relationship between the dorsal activation and perceptual classification ( Van Dromme et al . , 2016; Zachariou et al . , 2017 ) . Notably , as evident from visual agnosia patients and from the results of the diffeomorphic scrambling experiment , the dorsal pathway appears not to be sufficient to support intact perception and object recognition , and the ventral pathway remains more central to this function . Even so , future studies should assess the contribution of dorsal pathway representations to different perceptual tasks in humans . This research would help elucidate the functional role of the dorsal pathway representations in object perception . In both experiments reported here , there was a decrease in shape sensitivity in the anterior parts of the dorsal pathway , in both univariate and multivariate analyses . This was particularly evident in the aIPS , a region associated with computations related to the representations of objects in the context of visuomotor tasks ( Culham et al . , 2003; for a recent review see , Gallivan and Culham , 2015 ) . However , several other neuroimaging investigations in humans and electrophysiological studies non-human primates have revealed shape sensitivity in the anterior parts of the IPS ( Durand et al . , 2007; Freud et al . , 2017b; Janssen et al . , 2000; Orban , 2011; Theys et al . , 2013; Theys et al . , 2015 ) . This apparent contradiction raises the possibility that the nature of the task modulates shape sensitivity in the dorsal pathway , and , indeed , a recent study found that the aIPS coded object shape and elongation during grasping but not during passive viewing ( Fabbri et al . , 2016 ) . Moreover , task demands were found to modulate the nature of representations in other regions of the dorsal pathway , under perceptual ( rather than visuomotor ) tasks , further suggesting that dorsal pathway representations might be sensitive to tasks demands ( Bracci et al . , 2017 ) . Notably , in contrast to these findings , a recent study revealed that the ventral pathway is less affected by task demands ( Bugatus et al . , 2017 ) , revealing potential differences in the way task modulates the representations in the two pathways . The present study uncovered novel evidence for the nature of shape representations along the dorsal pathway and its similarity to representations along the ventral pathway . In two fMRI experiments , using a variety of analytical approaches , we found that posterior extrastriate regions of the dorsal pathway are highly sensitive to shape information ( but less to object identity ) , that the magnitude of their activation is correlated with perceptual behaviors and that in some regions , the representations derived are highly similar to those in some regions of ventral cortex . This sensitivity decreases in anterior regions of the dorsal pathway reflecting a gradual shift from representation-for-perception to representation-for action , leading to the conclusion that there is a representational continuum from more posterior areas tuned to visual properties of the objects in the input to more anterior areas tuned to motor aspects of the observed objects .
Twenty-two right-handed participants ( box scrambling experiment: Eleven participants , nine males; mean age: 31 , range: 19–46 years , diffeomorphic scrambling experiment: Eleven participants , five males; mean age: 25 , range: 19–46 years ) . The data obtained from two additional participants were not analyzed due to excessive head movements ( >3 mm ) during multiple scans . One additional participant ( diffeomorphic scrambling experiment ) did not complete the behavioral session , and therefore , for this experiment , the correlation between behavioral performance and fMRI signal was calculated based on the results of ten participants . All participants had normal or corrected-to-normal vision and were financially compensated for their participation . Informed consent was obtained prior to the study . All experimental procedures were approved by the Institutional Review Board of Carnegie Mellon University . fMRI raw data are available at https://doi . org/10 . 1184/R1/c . 3889873 . v1 and were processed using BrainVoyager 20 . 2 software ( Brain Innovation , Maastricht , Netherlands; RRID:SCR_013057 ) , MRIcron ( RRID:SCR_002403 ) , complementary in-house software written in Matlab ( The MathWorks , Inc , Natick , MA , USA; RRID:SCR_001622; see source code ) and R Development Core Team ( 2009 ) . Preprocessing included 3D-motion correction and filtering of low temporal frequencies ( cutoff frequency of 2 cycles per run ) . No spatial smoothing was applied to allow the voxel-wise analysis . All scans were transformed to Montreal Neurological Institute ( MNI ) space ( Fonov et al . , 2011 ) . Three main analytical approaches were employed: a novel voxel-wise approach that allows a continuous mapping of shape processing along the two pathways , a more traditional ROI analysis and a multivariate representational similarity analysis ( RSA ) . | We rely on our sense of vision to perceive the world around us and the objects within it . We also use vision to guide our interactions with objects . One of the most influential theories in cognitive neuroscience is the idea that separate pathways within the brain support these two processes . The ventral pathway is in charge of vision-for-perception . It analyses the features that help us recognize objects , such as their color , size or shape , enabling us to identify the hammer in a toolbox , for example . The dorsal pathway is responsible for vision-for-action . It processes features that help us interact with objects , such as their movement and location , enabling us to use the hammer to strike a nail . However , recent studies have suggested that the ventral and dorsal pathways may not be as independent as originally thought . Freud et al . now test this idea by examining if the dorsal vision-for-action pathway can also perceive and process objects . Healthy volunteers viewed pictures of objects while lying inside a brain scanner . Some of the objects in the pictures were intact , whereas others had been distorted . If a brain region shows greater activation when viewing intact objects than distorted ones , it implies that that region is sensitive to the normal shapes of objects . Freud et al . found that both the ventral and dorsal pathways were sensitive to shape , with some areas in the two pathways showing highly similar responses . Furthermore , the shape sensitivity of certain regions within the dorsal pathway correlated with the volunteers’ ability to recognize the objects . This suggests that regions distributed across both pathways – and not just the ventral one – may contribute to object recognition . The two-pathways hypothesis has governed our understanding of vision and of other sensory systems including hearing for several decades . By challenging the binary distinction between the two pathways , the results of Freud et al . suggest that models of sensory processing may require updating . This improved understanding may ultimately improve diagnosis and treatment of perceptual disorders such as agnosia , in which patients struggle to recognize objects . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | The large-scale organization of shape processing in the ventral and dorsal pathways |
Particulate matter ( PM ) air pollution causes cardiopulmonary mortality via macrophage-driven lung inflammation; however , the mechanisms are incompletely understood . RNA-sequencing demonstrated Acod1 ( Aconitate decarboxylase 1 ) as one of the top genes induced by PM in macrophages . Acod1 encodes a mitochondrial enzyme that produces itaconate , which has been shown to exert anti-inflammatory effects via NRF2 after LPS . Here , we demonstrate that PM induces Acod1 and itaconate , which reduced mitochondrial respiration via complex II inhibition . Using Acod1-/- mice , we found that Acod1/endogenous itaconate does not affect PM-induced inflammation or NRF2 activation in macrophages in vitro or in vivo . In contrast , exogenous cell permeable itaconate , 4-octyl itaconate ( OI ) attenuated PM-induced inflammation in macrophages . OI was sufficient to activate NRF2 in macrophages; however , NRF2 was not required for the anti-inflammatory effects of OI . We conclude that the effects of itaconate production on inflammation are stimulus-dependent , and that there are important differences between endogenous and exogenously-applied itaconate .
Exposure to particulate matter ( PM ) air pollution is associated with significant morbidity and mortality and is one of the top preventable causes of death in the world ( McGlade and Landrigan , 2019 ) . The World Health Organization estimates that exposure to air pollution is responsible for 4 . 2 million premature deaths worldwide every year ( WHO , 2018 ) . Air pollution has been identified as a leading cause of global disease burden , and as the fifth mortality risk factor particularly in low- and middle-income countries ( Cohen et al . , 2017 ) . The majority of PM-associated morbidity and mortality is due to cardiopulmonary diseases including asthma , chronic obstructive lung disease , lung cancer , congestive heart failure and ischemic/thrombotic cardiovascular disease ( myocardial infarction , ischemic stroke ) ( Cohen et al . , 2017; Hamanaka and Mutlu , 2018 ) . We have previously shown that lung macrophages are required for PM-induced lung inflammation and consequently acute thrombotic events ( Mutlu et al . , 2007; Chiarella et al . , 2014; Soberanes et al . , 2019 ) . PM induces the release of pro-inflammatory cytokines including interleukin-6 ( IL-6 ) , which is required for the PM-induced prothrombotic state and resultant acceleration of vascular thrombosis ( Mutlu et al . , 2007; Chiarella et al . , 2014 ) . Furthermore , we found that PM affects mitochondrial function in lung macrophages characterized by increased oxygen consumption rate and generation of reactive oxygen species ( ROS ) , which are both required for PM-induced IL-6 production ( Soberanes et al . , 2019 ) . Despite significant improvement in our understanding about the mechanisms by which PM induces pro-inflammatory cytokines such as IL-6 , the mechanisms that regulate PM-induced lung inflammation are not completely understood . To better understand these mechanisms , we performed RNA-sequencing in macrophages following PM exposure . Here , we demonstrate that PM regulates the expression of genes that have not been previously reported . One of the top genes induced by PM was Irg1 ( Immune-responsive Gene 1 ) or Acod1 ( Aconitate decarboxylase 1 ) , which encodes a mitochondrial enzyme that produces itaconate from the Tricarboxylic Acid ( TCA ) cycle metabolite cis-aconitate ( Strelko et al . , 2011; Michelucci et al . , 2013 ) . Itaconate has been reported to be produced in macrophages following LPS stimulation and to have anti-inflammatory effects ( Lampropoulou et al . , 2016; Mills et al . , 2018 ) . We confirmed that PM induces both mRNA and protein expression of Acod1 in macrophages , and increases intracellular and media levels of itaconate . Our metabolic analysis showed that itaconate inhibits mitochondrial complex II ( succinate dehydrogenase ) and that Acod1 is an important regulator of cellular respiration following PM exposure . Treatment of macrophages with a cell-permeable itaconate derivative , 4-octyl itaconate ( OI ) attenuated PM- and LPS-induced cytokine production . Acod1-/- cells , which lack endogenous itaconate production , exhibited exaggerated IL-1β production following LPS exposure; however , no effect on PM-induced inflammation was observed . As recently described , OI induced NRF2 protein and expression of its target genes such as Nqo1 and Hmox1 ( Mills et al . , 2018 ) ; however , we found that NRF2 was not required for the anti-inflammatory effects of OI . Furthermore , using Acod1-/- cells , we found that endogenous itaconate production is not required for NRF2 protein or its target gene expression following treatment with either PM or LPS . These results suggest that recently reported anti-inflammatory and NRF2-inducing effects of exogenous itaconate ( OI ) may not represent the effects of endogenously-produced itaconate and therefore OI should not be used as a surrogate for Acod1 and endogenous itaconate . Collectively , our results suggest that stimulus and location of itaconate production play major roles in governing the effect of itaconate on inflammation in macrophages . Furthermore , in contrast to recent studies , we suggest that NRF2 is not a major regulator of the anti-inflammatory effects of itaconate .
Previous studies that evaluated the effect of PM on macrophages have evaluated a limited number of regulated genes , primarily focusing on pro-inflammatory cytokines . To gain a non-biased understanding of the mechanisms by which PM induces lung inflammation , we treated BMDMs from C57BL/6 mice with PM or vehicle control for 24 hours and performed RNA-Seq to analyze PM-induced changes in the transcriptome . Differentially expressed gene ( DEG ) analysis revealed 370 unique genes significantly regulated by PM . Gene ontology analysis of RNA-seq data showed upregulation of pathways relating to inflammatory responses , metabolism , and cytokine stimulation ( Figure 1A ) . The most highly upregulated gene induced by PM exposure was Acod1 ( Figure 1B and Figure 1—source data 1 ) . Other highly upregulated genes included proinflammatory cytokine genes such as Tnfa , Il1b as well as Nqo1 , a transcriptional target of Nuclear factor ( erythroid-derived 2 ) -like 2 ( NRF2 ) ( Figure 1B ) . Acod1 encodes a mitochondrial enzyme , which catalyzes the conversion of cis-aconitate to itaconate , a mitochondrial metabolite that has recently been shown to regulate inflammatory responses ( Michelucci et al . , 2013 ) . To confirm that PM upregulates mRNA and protein expression of Acod1 , we first treated BMDMs with PM for 4 , 8 or 24 hours , and then assessed the expression of Acod1 through qPCR and western blot over time . PM induced Acod1 gene expression as early as 4 hours after treatment , peaking at 8 hours ( Figure 1C ) . Western blot analysis showed expression of ACOD1 protein was delayed compared to mRNA expression and was detectable beginning 24 hours after PM treatment ( Figure 1D ) . To determine whether Acod1 protein induction was associated with increased cellular levels of its metabolic product , we measured intracellular levels of itaconate in BMDMs 24 hours after treatment with PM using capillary electrophoresis–mass spectrometry ( CE-MS ) ( Human Metabolome Technologies , Boston , MA ) . We found that PM caused a significant increase in intracellular itaconate levels ( Figure 1E ) , correlating with the upregulation of ACOD1 . Furthermore , we detected increased concentrations of itaconate via mass spectrometry in the medium of macrophages treated with PM , indicating that itaconate is released from cells ( Figure 1F ) . Collectively , these findings demonstrate that PM causes a time-dependent expression of ACOD1 and production of itaconate , prompting us to further study whether itaconate plays a role in the regulation of PM-induced inflammation . Since itaconate has been shown to be a weak inhibitor of complex II , succinate dehydrogenase ( SDH ) ( Cordes et al . , 2016; Lampropoulou et al . , 2016 ) , we hypothesized that PM exposure may regulate mitochondrial respiration in macrophages through induction of Acod1 and production of itaconate . We first confirmed that itaconate inhibits SDH in BMDMs using the Seahorse XF Plasma Membrane Permeabilizer to assess the effect of itaconate on individual mitochondrial respiratory complexes . To measure the effect of itaconate on complex II/SDH activity , we permeabilized cells in the presence of rotenone ( to eliminate the contribution of complex I ) , ADP , and succinate as substrate . Measurement of oxygen consumption rate ( OCR ) showed an immediate decrease in OCR following injection of either itaconate or malonate , a known SDH inhibitor used as a positive control . Addition of oligomycin did not cause a further reduction in OCR ( Figure 2A ) . These data suggest that itaconate , like malonate , is indeed an SDH inhibitor . When cells were permeabilized in media containing only the complex I substrates pyruvate/malate , injection of itaconate caused a smaller decrease in OCR relative to that seen during complex II-dependent respiration . This reduction was again similar to what was observed with malonate ( complex II inhibitor ) . Injection of oligomycin resulted in a significant drop in OCR . These results are consistent with itaconate mediating its inhibitory effects on respiration via complex II ( Figure 2B ) . Because there are no known mechanisms of transporting itaconate into the cell , studies have used different forms of itaconate that are cell membrane-permeable to examine the role of itaconate in vitro . 4-octyl itaconate ( OI ) has been shown to cross the plasma membrane and increase intracellular concentrations of itaconate ( Mills et al . , 2018 ) . We thus measured OCR in intact cells in the presence or absence of OI ( 0 . 25 mM , a dose shown by Mills et al to reduce LPS-induced inflammation [Mills et al . , 2018] ) . OI injection caused a small reduction in basal OCR , while basal ECAR increased slightly , which is likely as a compensatory response to reduction in OCR ( Figure 2C and D ) . Oligomycin then decreased the OCR of both groups to the same level consistent with OI-induced reduction in coupled respiration . Maximal OCR determined following FCCP , an uncoupler , was lower in OI-treated BMDMs compared to control BMDMs ( Figure 2E ) . Together , these results suggest that itaconate is sufficient to reduce mitochondrial OCR via inhibition of complex II/SDH . Since itaconate inhibits mitochondrial respiration and ACOD1 is induced late following PM exposure , we sought to determine whether PM exposure exerts time-dependent changes in mitochondrial oxygen consumption rate in macrophages . We thus measured OCR in BMDMs following PM treatment at 1 hour and 24 hours , when ACOD1 expression is undetectable and present , respectively . As we have previously shown ( Soberanes et al . , 2019 ) , after 1 hour of exposure , cells treated with PM exhibited increased basal oxygen consumption rate compared with control BMDMs ( Figure 3A ) . Consistent with a role of glycolysis in promoting inflammation , ECAR was also induced by PM exposure after 1 hour . Interestingly , after 24 hours of PM treatment , ECAR levels remained elevated; however , both basal and maximal OCR were decreased relative to control-treated cells . ( Figure 3B ) . This decrease in OCR following 24 hours of PM treatment was similar to the decrease in OCR after itaconate treatment ( Figure 2C ) suggesting that ACOD1 induction and itaconate production are important regulators of macrophage metabolism following PM treatment . We next sought to determine whether PM-induced ACOD1 expression and endogenous itaconate production is required for PM-induced changes in mitochondrial metabolism at 24 hours . In order to answer this question , we used BMDMs from Acod1-/- mice , which lack the ability to produce itaconate . We first measured TCA cycle intermediates in WT and Acod1-/- BMDMs at 24 hours following treatment with PM or vehicle . PM-induced production of itaconate was detectable only in WT but not in Acod1-/- BMDMs , confirming that ACOD1 is required for itaconate production ( Figure 4A , B ) . Consistent with the role of itaconate as an inhibitor of SDH , succinate accumulated in WT BMDMs , but not Acod1-/- cells following PM treatment ( Figure 4A , C ) . This finding provides further support for endogenously-produced itaconate inhibiting SDH and leading to accumulation of succinate , as SDH catalyzes the oxidation of succinate to fumarate . In contrast , Acod1-/- BMDMs had higher levels of TCA metabolites downstream of succinate ( fumarate , malate ) ( Figure 4A ) , consistent with the idea that due to the absence of itaconate in these cells , SDH and consequently the oxidation of succinate to fumarate are not inhibited . These results are consistent with previous reports using LPS as a stimulus ( Lampropoulou et al . , 2016 ) . Overall , our metabolomics data suggest that PM causes a break in the TCA cycle that is Acod1 dependent . To determine whether PM-induced ACOD1 expression and production of itaconate are required for the reduction of OCR at 24 hours following PM , we performed a mitochondrial stress test in WT and Acod1-/- BMDMs . While WT BMDMs exhibited significantly reduced OCR following PM , Acod1-/- cells treated with PM for 24 hours did not show any reduction in OCR compared to control treatment ( Figure 4D ) . Collectively , these results suggest that endogenously-produced itaconate is required for PM-induced reduction in OCR . While a recent study showed that nitric oxide production as a result of LPS exposure can modulate metabolic reprogramming and affect oxidative phosphorylation in macrophages ( Bailey et al . , 2019 ) , we found that inducible nitric oxide synthase ( iNOS ) protein was produced only in BMDMs stimulated with LPS . PM did not induce iNOS expression in either wild-type or Acod1-/- BMDMs ( Figure 4—figure supplement 1 ) . Thus , nitric oxide production is likely to be minimal with PM as a stimulus , and would not play a large role in the regulation of oxidative phosphorylation following PM . We have previously shown that PM induces an inflammatory response in macrophages , including the release of proinflammatory cytokines such as IL-6 and TNFα ( Mutlu et al . , 2007; Chiarella et al . , 2014 ) . Since the expression of ACOD1 was time-dependent , we first sought to determine whether the effect of PM on cytokine expression was also time-dependent . We thus treated BMDMs with PM for 4 , 8 or 24 hours and then analyzed the pro-inflammatory cytokine mRNA expression . We found that the mRNA expression of Il6 , Tnfa and Il1b increased at 4 and 8 hours , and then decreased at 24 hours suggesting that PM-induced expression of cytokines is time-dependent similar to the expression of ACOD1 ( Figure 5A ) . Since the reduction in inflammatory cytokine mRNA coincided with the induction of ACOD1 and itaconate production , we hypothesized that itaconate may play a role in the reduction in cytokine expression observed at 24 hours . To test this hypothesis , we first evaluated the effect of exogenous itaconate on the early PM-induced inflammatory response . We pre-treated BMDMs with OI for 2 hours before treating with PM for 4 hours and measured mRNA expression of pro-inflammatory cytokines by qPCR . Pretreatment with OI attenuated the PM-induced mRNA expression of Tnfa , Il6 and Il1b , suggesting that accumulation of itaconate with the induction of ACOD1 may be responsible for reduced cytokine mRNA expression at 24 hours following PM treatment ( Figure 5B ) . To determine whether absence of itaconate accumulation would exaggerate PM-induced inflammation at later timepoints , we measured cytokine mRNA and protein level in both wild-type and Acod1-/- BMDMs 24 hours after exposure to PM . Interestingly , loss of itaconate accumulation did not result in increased mRNA expression of Il6 , Tnfa or Il1b or protein level in media in Acod1-/- BMDMs ( Figure 5C , D ) . While loss of Acod1/itaconate did not affect PM-induced IL-6 protein levels , there was a reduction in TNFα protein level in media ( Figure 5D ) . Overall , these results suggest that endogenous itaconate does not attenuate the PM-induced inflammatory response . This was a surprising finding as Acod1/itaconate deficiency has been shown to augment inflammatory cytokine expression in response to LPS ( Lampropoulou et al . , 2016 ) . To ensure that our findings were not the result of experimental differences other than stimulus , we treated wild-type and Acod1-/- BMDMs with LPS for 24 hours and measured Il6 , Tnfa or Il1b mRNA expression . In contrast with PM treatment , Acod1-/- cells exhibited augmented IL6 and Il1b mRNA expression after LPS treatment ( Figure 5E ) , consistent with previously published results ( Lampropoulou et al . , 2016 ) . There was no effect of Acod1 deletion on LPS-induced Tnfa expression ( Figure 5E ) , also consistent with previous findings ( Lampropoulou et al . , 2016 ) . The loss of Acod1/itaconate also resulted in increased IL-6 and IL-1β protein level in media ( Figure 5F ) . These findings suggest that the effect of itaconate on inflammation is stimulus-dependent . Importantly , these results also suggest that the effect of endogenously-produced itaconate on inflammation is different than that of exogenously-applied itaconate . To better understand the role of endogenous and exogenously-applied itaconate on PM-induced response in macrophages , we performed RNA-sequencing in Acod1-/- and WT BMDMs exposed to PM ( Figure 6A and Figure 6—source data 1 ) . We found only 51 DEGs between Acod1-/- and WT BMDMs following PM ( Figure 6B and Figure 6—source data 2 ) . Both inflammatory genes , including Il1b and Tnfa , and NRF2 target genes , including Nqo1 and Gclm , were not significantly different between Acod1-/- and WT BMDMs . Next , to determine the effect of exogenous itaconate on PM-induced gene expression , we performed RNA-sequencing in WT BMDMs exposed to PM following 2 hours of pretreatment with vehicle or OI . Transcriptomic analysis showed 1 , 030 DEGs between groups with and without OI ( Figure 6C and Figure 6—source data 3 ) . OI pretreatment significantly downregulated inflammatory gene expression in PM-treated BMDMs and upregulated NRF2 target genes . The effect of OI on these genes was not significantly different between WT and Acod1-/-cells . These results suggest that the effect of exogenous itaconate on the PM-induced transcriptomic response is markedly different from the effect of endogenous itaconate ( Figure 6A ) . Recent studies suggest that itaconate exerts its inhibitory effects on LPS-induced inflammation via activation of NRF2 , a transcriptional factor that plays a key role in antioxidant defense ( Mills et al . , 2018 ) . Both PM and LPS induce mitochondrial production of reactive oxygen species which promote inflammatory gene expression ( Hsu and Wen , 2002; Soberanes et al . , 2012 ) ; however , PM contains a mixture of metals and other compounds which have their own redox-modulating properties ( Jeng , 2010 ) . We thus hypothesized that differential effects of PM and LPS on NRF2 activation may explain the lack of effect of endogenous itaconate on inflammation in PM-treated Acod1-/- BMDMs . We found that PM upregulated NRF2 and its target genes ( Nqo1 , Gclm , Hmox1 ) equally in WT and Acod1-/- BMDM ( Figure 7A , B ) . These results suggested that itaconate-independent activation of NRF2 might explain the fact that no augmented inflammatory gene expression was observed in Acod1-/- cells after PM exposure . As our experiments in Acod1-/- BMDMs showed that Acod1 is not required for PM-induced NRF2 activation , we sought to determine whether endogenous itaconate production is required for LPS-induced NRF2 activation . Surprisingly , we found that similar to PM-induced NRF2 activation , Acod1 expression was not required for LPS-mediated induction of NRF2 protein or target gene expression ( Figure 7C , D ) . NRF2 protein levels were similarly upregulated in both LPS-treated WT and Acod1-/- cells ( Figure 7C ) . While Gclm expression was slightly reduced in Acod1-/- cells , Hmox1 induction was not affected by Acod1 deficiency and Nqo1 was more highly induced by LPS in Acod1-/- cells ( Figure 7D ) . Furthermore , we found that NRF2 induction after LPS treatment occurs prior to Acod1 induction ( Figure 7E ) . Treatment with LPS induced a time-dependent expression of ACOD1 and NRF2 . Interestingly , expression of NRF2 occurred at an earlier time point compared to ACOD1 ( 4 hours vs . 8 hours ) . Maximal expression of NRF2 also preceded the maximal expression of ACOD1 ( 8 hours vs 24 hours ) . Taken together , our results suggest that although NRF2 activation has been proposed to be the mechanism by which itaconate exerts its inflammatory effect , NRF2 induction occurs prior to ACOD1 , and furthermore , ACOD1 is not required for NRF2 activation downstream of either PM or LPS . To date , the majority of studies investigating the role of Acod1 and itaconate on inflammation , including those linking NRF2 to the mechanisms by which itaconate may regulate inflammation , have largely focused on the effects of exogenously-applied cell membrane-permeable forms of itaconate ( e . g . , dimethyl itaconate and OI ) ( Lampropoulou et al . , 2016; Bambouskova et al . , 2018; Mills et al . , 2018; Zhao et al . , 2019 ) . NRF2 has been suggested to regulate the effect of OI on IL-1β expression; however , as we saw that OI reduced PM and LPS-induced expression of other cytokines , including TNFα , the expression of which is not regulated by endogenous itaconate , we examined whether the effect of OI on these cytokines was regulated by NRF2 . We found that , consistent with previous findings ( Mills et al . , 2018 ) , OI treatment was sufficient to induce NRF2 protein and expression of NRF2 target genes ( Nqo1 , Gclm , and Hmox1 ) ( Figure 8A , B ) . OI induced NRF2 protein expression to a similar extent as PM , and the combination of OI and PM caused a further increase in NRF2 protein levels and expression of its target gene , Nqo1 ( Figure 8A , B ) . To determine whether NRF2 is required for the OI-mediated attenuation of inflammatory cytokine expression , we transfected BMDMs with two independent siRNAs targeting Nfe2l2 , or a non-targeting control . NRF2 protein was confirmed to be eliminated by Western blot analysis , as Nfe2l2 siRNA-transfected cells did not upregulate NRF2 protein following 4 hours of PM treatment ( Figure 8C , Figure 8—figure supplement 1 ) . Despite the loss of NRF2 , we found that OI still exhibited its anti-inflammatory effects on PM-induced inflammation . OI attenuated PM-induced proinflammatory cytokine ( Tnfa , Il6 , Il1b ) expression in both control and NRF2 knockdown cells ( Figure 8D ) . These data suggest that NRF2 is not required for exogenous itaconate to attenuate PM-induced inflammation and are in contrast to recent results suggesting that OI exerts its anti-inflammatory effect on LPS-induced inflammation by upregulating NRF2 ( Mills et al . , 2018 ) . Because of the differences between PM and LPS , we then looked at the effect of OI on LPS-induced inflammation . While OI similarly decreased LPS-induced inflammation ( Figure 9A ) , in contrast to PM treatment , LPS alone did not induce significant NRF2 target gene activation after 4 hours . No additional effect was seen on NRF2 target gene expression with the combination of LPS and OI ( Figure 9B ) . Furthermore , consistent with our findings with PM , OI reduced inflammatory gene expression in response to LPS independent of NRF2 induction ( Figure 9C , Figure 9—figure supplement 1 ) . Taken together , our results suggest that although NRF2 is induced by OI , this activation is not required for the anti-inflammatory effect of exogenous itaconate . Lungs are the primary site of entry for PM , and we have previously demonstrated that alveolar macrophages ( AMs ) are metabolically distinct from BMDMs ( Woods et al . , 2020 ) . We thus sought to determine whether the role of itaconate in AMs is similar to what we observed in BMDMs . We first isolated alveolar macrophages from WT and Acod1-/- mice and then treated them with PM as we did in BMDMs ( Figure 10—figure supplement 1 ) . As expected , there was no Acod1 expression in Acod1-/-alveolar macrophages and treatment with PM induced a significant increase in the expression of Acod1 mRNA in WT alveolar macrophages ( Figure 10—figure supplement 1A ) . PM increased the expression of inflammatory cytokine genes ( Il6 , Tnfa , and Il1b ) as well as IL-6 and TNFα protein levels in media both in WT and Acod1-/-alveolar macrophages ( Figure 10—figure supplement 1B , C ) . There was no difference in cytokine mRNA or protein expression between WT and Acod1-/-alveolar macrophages . As we observed in BMDMs , PM caused similar increases in the expression of NRF2 target genes in WT and Acod1-/-alveolar macrophages ( Figure 10—figure supplement 1D ) . We next investigated whether itaconate plays a role in inflammation in an in vivo model . We intratracheally instilled either PBS or PM ( in PBS ) into the lungs of WT and Acod1-/- mice as we previously described ( Mutlu et al . , 2006; Mutlu et al . , 2007 ) . Twenty-four hours after instillation , we obtained bronchoalveolar lavage ( BAL ) fluid to measure cytokine levels as well as to isolate AMs for gene expression analysis . AM expression of Acod1 was significantly induced by PM in WT mice only ( Figure 10A ) . Inflammatory cytokine genes ( Il6 , Tnfa , and Il1b ) were significantly induced by PM in AMs from both WT and Acod1-/- mice . There was no difference in levels of inflammatory cytokine genes between the WT and Acod1-/- groups ( Figure 10B ) . Analysis of cytokine levels in the BAL fluid by ELISA showed similar increases in IL-6 and TNFα following PM treatment , with no differences between WT and Acod1-/- mice ( Figure 10C ) . Collectively , these data suggest that itaconate does not play a role in PM-induced inflammation in vitro or in vivo .
Exposure to PM air pollution is associated with significant morbidity and mortality ( Hamanaka and Mutlu , 2018 ) . Air pollution exposure is one of the top preventable causes of death in the world ( McGlade and Landrigan , 2019 ) . It is estimated that exposure to air pollution causes 4 . 2 million premature deaths worldwide every year according to the World Health Organization ( Hamanaka and Mutlu , 2018; WHO , 2018 ) . Since limiting exposure to PM cannot be achieved immediately , a better understanding of the mechanisms by which PM causes morbidity and mortality is required in order to combat the effects of air pollution on public health . As lungs are the entry site for PM , it is not surprising that lung inflammation plays an important role in PM-induced adverse health effects , which are predominantly cardiopulmonary diseases . We have previously reported that macrophages , and in particular , their production of IL-6 , are required for the PM-induced lung inflammation and resultant acute thrombotic cardiovascular events ( Mutlu et al . , 2007; Chiarella et al . , 2014; Soberanes et al . , 2019 ) . However , the mechanisms regulating PM-induced inflammation in the lung are not completely understood . In this study , using an unbiased approach with RNA-seq , we first discovered Acod1 as the gene most induced by PM in macrophages . In a time-dependent fashion , PM induced the mRNA and protein expression of ACOD1 , which was not evident at earlier time points ( 4 , 8 hours ) but was detected at 24 hours . Confirming a functional role , PM-induced expression of ACOD1 led to production of itaconate in our metabolomics measurements . Assessment of the effect of PM on mitochondrial function showed that PM has a time-dependent effect on mitochondrial respiration . While PM increased mitochondrial OCR at 1 hour , it reduced OCR at 24 hours . Similarities in the time dependency between the effects of PM on ACOD1/itaconate and mitochondrial respiration led us to explore whether itaconate is responsible for the late reduction of OCR following PM exposure . Using Acod1-/- macrophages , we found that itaconate is required for the PM-induced reduction in mitochondrial OCR at 24 hours . Consistent with a role of itaconate as an inhibitor of succinate dehydrogenase , we found that Acod1-/- BMDMs lack succinate accumulation following PM treatment . This finding is in agreement with findings by Cordes et al . , which showed a loss of succinate accumulation in response to LPS stimulation in Acod1-/- macrophages ( Cordes et al . , 2016 ) . Collectively , our data support a key role for Acod1 and itaconate in the metabolic reprogramming of macrophages in response to both LPS and PM stimulation . Interestingly , although we found that endogenous itaconate production was an important regulator of mitochondrial metabolism and respiration in PM-treated macrophages , we were unable to find an effect of endogenous itaconate on PM-induced inflammation . This was in contrast to the effect of Acod1 deletion on LPS-induced inflammation , in which IL-6 and IL-1β induction was augmented in the absence of endogenous itaconate production . Our results show a stimulus dependency for the effect of itaconate that has not been previously demonstrated . Thus , although PM and LPS both promote inflammatory cytokine expression and itaconate-dependent metabolic reprogramming , the effects of itaconate on inflammation differ between the two stimuli . To enhance our understanding of the role of itaconate in the lungs with PM exposure , we used an in vivo model of PM exposure in WT and Acod1-/- mice . Similar to our in vitro data , in vivo exposure of mice to PM shows that while Acod1 is highly upregulated in the lung macrophages following PM exposure , the presence or lack of Acod1 did not affect inflammation in the lung . Expression of IL-6 and TNFα were equally increased with PM exposure , regardless of ACOD1 expression status . To date , the majority of studies investigating the role of Acod1 and itaconate on metabolism have largely focused on the effects of exogenously-applied cell membrane-permeable forms of itaconate ( e . g . , dimethyl itaconate and OI ) ( Lampropoulou et al . , 2016; Bambouskova et al . , 2018; Zhao et al . , 2019 ) . Following the initial studies with dimethyl itaconate ( Lampropoulou et al . , 2016; Bambouskova et al . , 2018; Zhao et al . , 2019 ) , a more recent study questioned whether treatment with dimethyl itaconate actually increases intracellular levels of itaconate ( ElAzzouny et al . , 2017 ) . In response to this limitation of dimethyl itaconate , Mills and colleagues proposed the use of OI as a cell permeable form of itaconate and demonstrated that OI does increase intracellular levels of itaconate ( Mills et al . , 2018 ) . Our results showing that OI reduces PM-induced cytokine production are in agreement with Mills and colleagues suggesting that the exogenous cell membrane-permeable version of itaconate has anti-inflammatory effects . However , given the differing effects of endogenous itaconate production versus OI treatment on PM-treated macrophages , our results suggest additional cautions should be taken when extrapolating the effects of cell-permeable itaconate analogs to the effects of endogenously-produced itaconate . This is highlighted by the fact that while OI is sufficient to induce NRF2 activation in macrophages , Acod1 is not required for NRF2 activation in response to PM or LPS . It would be expected that if a major effect of endogenous itaconate production were NRF2 activation , this should have been absent in Acod1-/- BMDMs . This loss of function experiment was notably lacking in the previous report linking itaconate with NRF2 . Our results suggest that location of itaconate plays a major role in regulating its downstream effects . While our results agree with Mills et al that OI activates the NRF2 antioxidant pathway by upregulation of NRF2 protein and its target genes , we found that NRF2 was dispensable for the anti-inflammatory effects of OI . Mills et al assessed the anti-inflammatory effects of OI by Western blot for intracellular IL-1β . Our findings examining the expression of multiple cytokines at the mRNA level suggest that NRF2 does not play a major role in regulating the anti-inflammatory effect of OI . In conclusion , we found that exposure of macrophages to PM induces a mitochondrial enzyme , ACOD1 and production of a mitochondrial metabolite , itaconate , which is required for PM-induced metabolic reprogramming . Although endogenous itaconate reprograms mitochondrial function in PM-treated macrophages , it does not have a major effect on PM-induced gene expression or inflammatory response in vitro or in vivo . Future work will be required to determine whether endogenous itaconate regulates other macrophage functions . Our finding that itaconate is secreted from macrophages suggests that important , unexplored effects of itaconate production may occur in non-macrophage cells . Our results also caution about the interpretation of results achieved using exogenous cell membrane-permeable forms of itaconate as they may not represent the effects of endogenously-produced itaconate . Further , our results suggest that our understanding of the role of NRF2 in the cellular response to itaconate is far from complete . While OI is sufficient to induce NRF2 activation in macrophages , endogenous itaconate production is not required for PM or LPS-mediated NRF2 activation .
All animal experiments and procedures were performed according to the protocols approved by the Institutional Animal Care and Use Committee at the University of Chicago . We used primary murine cells ( bone marrow-derived macrophages ( BMDMs ) ) , which we isolated as we have recently reported ( Woods et al . , 2020 ) . Hematopoietic cells were isolated from bone marrow of C57BL/6NJ ( Stock No: 005304 ) and Acod1-/- ( Stock No: 029340 ) ( both from Jackson Laboratory ) mice and cultured with M-CSF ( 20 µg/L , BioLegend , catalog number 576408 ) in vitro for 8–10 days to generate BMDMs . Tissue-resident alveolar macrophages were isolated by standard BAL fluid collection , as we have previously described ( Chiarella et al . , 2014; Soberanes et al . , 2019 ) , and cultured in media for 2 hr before treatment . For all experiments , cells were cultured in complete medium containing RPMI ( Gibco , catalog number A10491 ) , supplemented with 10% heat-inactivated FBS ( Gemini , catalog number 100–106 ) and 1% penicillin-streptomycin ( Gemini , 400–109 ) . Reagents were purchased from Sigma-Aldrich , including: 4-Octyl itaconate ( catalog number SML2338 ) , Itaconic Acid ( catalog number I29204 ) , Malonic Acid ( M1296 ) . Particulate matter ( SRM 1649a , Urban Dust ) was from National Institute of Standards Technology ( NIST ) . Lipopolysaccharide was purchased from Santa Cruz ( sc-3535 ) . Full details of reagents used are included in the Key Resources Table . Cells were scraped into RIPA buffer ( Thermo-Scientific , catalog number 89900 ) with protease and phosphatase inhibitors ( Thermo-Scientific , 1861284 ) , sonicated for 10 s on a Fisher Scientific 100 model at speed setting 2 . Samples were resolved by SDS-PAGE on 10% polyacrylamide gels and transferred to nitrocellulose ( Bio-Rad , catalog number 1620167 ) . Primary antibodies used were mouse anti-β-actin monoclonal antibody ( Sigma , catalog number A5441; lot number 037K488; 1:10 , 000 ) , ACOD1 polyclonal antibody ( Invitrogen , catalog number PA5-49094 , 1:500 ) , NRF2 monoclonal antibody ( Abcam , catalog number ab62352 , 1:1000 ) , iNOS Antibody ( Cell Signaling Technology , catalog number 2982S , 1:1000 ) . Secondary antibodies used were anti-rabbit IgG HRP-linked antibody ( Cell Signaling Technology , catalog number 7074S ) and anti-mouse IgG HRP-linked antibody ( Cell Signaling Technology , catalog number 7076S ) . Total RNA was extracted with TRI Reagent ( Zymo Research , R2050-1-200 ) . RNA was then isolated using the Zymo Direct-zol RNA Miniprep Kit ( Zymo Research , catalog number R2053 ) and reverse-transcribed using Bio-Rad iScript Reverse Transcription Supermix ( Bio-Rad , catalog number 1708841 ) in a Bio-Rad C1000 Touch Thermal Cycler . Quantitative mRNA expression was determined by real-time qPCR using iTaq Universal SYBR Green Supermix ( Bio-Rad , catalog number 172–5121 ) . qPCR primer sequences used are as follows: Cells were treated in complete medium for 24 hours , and the media was collected . IL-6 and TNFα cytokine levels in the media were then measured with DuoSet ELISA kits ( R and D systems , catalog number DY406 and DY410 ) according to manufacturer’s protocol . RNA was isolated and submitted for sequencing ( 50 bp SE ) . Quality assessment of sequencing files was assessed with FastQC . Reads were mapped to the GRCm38 reference genome using STAR aligner ( Dobin et al . , 2013 ) , and quality was assessed using Picard tools and RSeQC ( Wang et al . , 2012 ) . Genes were quantified using featureCounts , and low expression genes were removed at cpm = 1 . 5 . Differentially expressed genes ( DEGs ) were identified using DESeq2 analysis at FC > 2 and FDR adjusted p-value p<0 . 05 . BMDMs were plated at 3 × 106 cells on 60 mm tissue culture plates for 2 hours , then treated with PM ( 20 μg/cm2 ) for 24 hours prior to metabolite extraction . Cells were washed with 5% mannitol solution and metabolites were extracted with 400 μl methanol . 275 μl internal standard was added , then the extracts were centrifuged at 2 , 300 × g for 5 min . The supernatant was transferred to pre-washed centrifugal filter units ( HMT , Human Metabolome Technologies , Boston , MA ) and centrifuged at 9 , 100 × g at 4°C for 2 hours . Centrifuged samples were sent to HMT for processing and analysis . This measurement was performed by Mass spectrometry , Metabolomics and Proteomics Facility in Research Resources Center of University of Illinois at Chicago . All samples are analyzed by Waters ACQUITY UPLC BEH C18 Column , 130Å , 1 . 7 μm , 2 . 1 mm X 100 mm coupled to an Agilent 1260 UPLC system , which was operated at a flow rate of 500 uL/min . A linear gradient of 1–60% buffer B ( 100% MeOH ) was applied . MS data were acquired by MRM scan ( Negative ESI spray voltage 4 . 5kV , temperature 550 degrees , m/z range 50–300 ) monitoring signature product ions 129 > 69 ( Quantifier ) and 129 > 59 ( Qualifier ) transitions . The quantification was achieved using peak area of monitored transitions . All the samples were analyzed by triplicate . BMDMs ( 1 × 106 cells ) were transfected with siRNA ( 250 pmol ) and Amaxa Mouse Macrophage Nucleofector Kit ( Lonza , catalog number VPA-1009 ) using a Lonza Nucleofector 2b device ( Lonza , #AAB-1001 ) on the mouse macrophage ( Y-001 ) setting . Cells were cultured for 2 days post transfection before treatment . Success of siRNA transfections were confirmed with western blots and qPCR . siRNAs were purchased from Dharmacon: D-001810-01-05 ( non-targeting siRNA ) ; J-040766-08-0002 ( Nfe2l2 #1 ) , J-040766-06-0002 ( Nfe2l2 #2 ) . The Seahorse XFe24 Extracellular Flux Analyzer ( Seahorse Bioscience , North Billerica , MA ) , was used to measure oxygen consumption rates ( OCR ) and extracellular acidification rates ( ECAR ) as we have previously described ( Nigdelioglu et al . , 2016; Hamanaka et al . , 2019; Soberanes et al . , 2019; Woods et al . , 2020 ) . Macrophages were seeded at a density of 4 × 104 per well on a Seahorse XF24 Cell Culture Microplate ( Woods et al . , 2020 ) . Mitochondrial stress tests were performed according to manufacturer’s protocol in XF DMEM base medium ( Agilent 103334–100 ) containing Glutamine ( 2 mM ) , Sodium Pyruvate ( 1 mM ) , and Glucose ( 25 mM ) . Compounds of interest were injected , followed by sequential injections of Oligomycin ( 1 . 5 μM ) , FCCP ( 1 . 5 μM ) and Antimycin A/Rotenone ( 1 . 25 μM ) . Membrane permeabilization assays were performed in 1x MAS buffer containing mannitol ( 220 mM ) , sucrose ( 70 mM ) , monopotassium phosphate ( 10 mM ) , magnesium chloride ( 5 mM ) , HEPES ( 2 mM ) , EGTA ( 1 mM ) , with the addition of Seahorse XF Plasma Membrane Permeabilizer ( Agilent , catalog number 102504–100 ) and Fatty Acid Free BSA ( 0 . 2% ) . Substrates and reagents used in permeabilization assays include succinate ( 10 mM ) , rotenone ( 2 mM ) , pyruvate ( 10 mM ) , malate ( 0 . 5 mM ) , ADP ( 4 mM ) , itaconate ( 10 mM ) , malonate ( 10 mM ) , oligomycin ( 2 mM ) , antimycin A ( 2 mM ) . All substrates and reagents were pH-adjusted to 7 . 2 prior to assay with 5M potassium hydroxide ( KOH ) solution . Permeabilization assays were performed with cycles of 0 . 5 min mix , 0 . 5 min wait , and 2 min measure intervals , as per manufacturer’s protocol . C57BL/6NJ and Acod1-/- mice were treated via intratracheal instillation with only PBS or PM in PBS as we have previously described ( Mutlu et al . , 2006; Mutlu et al . , 2007 ) . PM was freshly diluted to a 2 . 5 mg/ml stock , and 2 doses of 20 μl PM stock was given intratracheally to each mouse , for a total of 100 μg in 40 μl PBS per mouse . For the PBS controls , 2 doses of 20 μl PBS were given intratracheally to each mouse . Following 24 hours of exposure , BAL fluid was obtained by flushing the lungs with 0 . 5 mM EDTA in ice cold PBS as we have previously described ( Chiarella et al . , 2014; Soberanes et al . , 2019 ) . The first wash of 500 μl was collected and centrifuged , where the supernatant was then used for ELISA studies . The cell pellet was combined with the cells obtained from the subsequent eight washes , and used for analysis of mRNA expression through qPCR analysis . Data were analyzed using Prism 8 ( GraphPad Software , Inc ) . All data are shown as mean ± standard error of the mean ( SEM ) . Significance was determined by unpaired two-tailed Student’s t test ( for comparisons between two samples ) , or by one-way ANOVA using Bonferroni correction for multiple comparisons . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . | Air pollution is a major global health problem that causes around 4 . 2 million deaths each year . Once inhaled , pollution particles can remain in the lungs , where they cause inflammation , tissue damage , and ultimately chronic disease . Macrophages , a population of immune cells in the lungs , are involved in this inflammatory process . Itaconate is a molecule with potential anti-inflammatory effects , produced by mammalian cells including macrophages . Recent studies have shown that a modified form of the molecule , 4-octyl itaconate , reduces inflammation when applied to cells exposed to lipopolysaccharide , a component of infectious bacteria that is , usually , a strong trigger of inflammation . These experiments used the 4-octyl modification to ensure that itaconate could get into the cells . Itaconate’s anti-inflammatory action is thought to work by activating a signaling process in cells called the NRF2 pathway . NRF2 is a protein made by ‘active’ macrophages , that is , macrophages already primed to respond to foreign particles . NRF2 in turn increases production of factors that ‘damp down’ inflammation , all of which are collectively termed the NRF2 anti-inflammatory pathway . Although macrophages in the lungs are linked with inflammation caused by air pollution , their role – and that of itaconate – is still not well-understood . Sun et al . therefore wanted to determine if itaconate helps macrophages control pollution-induced inflammation . Initial experiments treated mouse macrophage cells with pollution particles . Analyzing gene activity in these cells showed that exposure to pollution did indeed switch on the Acod1 gene , which encodes the enzyme that makes itaconate . It also turned on genes for other molecules involved in inflammation . Pre-treating macrophages with 4-octyl itaconate before pollution exposure reduced inflammation and also , as expected , turned on the NRF2 pathway . To determine whether cells’ own production of itaconate affected lung inflammation , macrophages were isolated from mutant mice lacking Acod1 . Comparing these cells , which could not make itaconate , with normal cells revealed that removing itaconate did not change the inflammatory response to pollution . Activity of the NRF2 pathway also remained similar in both types of cells . This showed that itaconate produced by macrophages likely has different effects on lung inflammation from other forms of the compound . These findings represent a step forward in understanding how pollution interacts with immune cells in the lungs . They reveal that the source of anti-inflammatory factors can be just as important in shaping immune responses as the type of factor . These results highlight the need for further , detailed work on the mechanisms underlying pollution-induced disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"immunology",
"and",
"inflammation"
] | 2020 | Endogenous itaconate is not required for particulate matter-induced NRF2 expression or inflammatory response |
Developmental programs sculpt plant morphology to meet environmental challenges , and these same programs have been manipulated to increase agricultural productivity ( Doebley et al . , 1997; Khush , 2001 ) . Hormones coordinate these programs , creating chemical circuitry ( Vanstraelen and Benková , 2012 ) that has been represented in mathematical models ( Refahi et al . , 2016; Prusinkiewicz et al . , 2009 ) ; however , model-guided engineering of plant morphology has been limited by a lack of tools ( Parry et al . , 2009; Voytas and Gao , 2014 ) . Here , we introduce a novel set of synthetic and modular hormone activated Cas9-based repressors ( HACRs ) in Arabidopsis thaliana that respond to three hormones: auxin , gibberellins and jasmonates . We demonstrate that HACRs are sensitive to both exogenous hormone treatments and local differences in endogenous hormone levels associated with development . We further show that this capability can be leveraged to reprogram development in an agriculturally relevant manner by changing how the hormonal circuitry regulates target genes . By deploying a HACR to re-parameterize the auxin-induced expression of the auxin transporter PIN-FORMED1 ( PIN1 ) , we decreased shoot branching and phyllotactic noise , as predicted by existing models ( Refahi et al . , 2016; Prusinkiewicz et al . , 2009 ) .
The body plans of plants are inherently plastic , making them amenable to optimization for a wide range of natural or artificial environments . Extrinsic and intrinsic cues are integrated by developmental programs to maximize the fitness of wild plants ( Vanstraelen and Benková , 2012 ) . Domestication of crops frequently relies on altering such programs to create more productive morphologies for agriculture , such as the dramatic reduction in bushiness of maize ( Doebley et al . , 1997 ) or the dwarfing of cereals that drove the green revolution ( Khush , 2001 ) . Developmental programs are coordinated in large part by a set of hormones ( Vanstraelen and Benková , 2012 ) . Accumulation of a given hormone by de novo synthesis or transport influences the expression or activity of developmental master controller genes , analogous to wires in a circuit . Auxin , perhaps the best-studied hormone , controls many developmental programs that drive agriculturally relevant traits ( Weijers and Wagner , 20152016 ) . Many mathematical models connecting auxin signaling and transport at the molecular level to specific developmental phenotypes at the whole plant level have been developed ( Refahi et al . , 2016; Prusinkiewicz et al . , 2009; Smith et al . , 2006 ) . These models highlight the importance of subtle parameters , like the strength of specific feedback loops in hormone signaling networks , in determining plant morphology . While the ability of hormones to trigger and tune developmental programs makes altering hormonal signaling an attractive target for re-engineering the plant form , there are significant hurdles to overcome in such approaches . Native hormone signaling pathways are comprised of co-expressed and redundant components , embedded in highly reticulate cross-regulatory relationships with other signaling pathways , and have several layers of feedback ( Weijers and Wagner , 20152016 ) . For example , the auxin signaling pathway is comprised of three families of proteins , ARFs , AUX/IAAs , and TIR1/AFBs , all of which have multiple members with redundant regulatory roles and are cross regulated by a plethora of other signals ( Koltai , 2015; Naseem et al . , 2015 ) . Thus , there is a need for tools that can predictably alter how a specific hormone regulates a gene of interest to facilitate re-wiring plant development ( Brophy et al . , 2017 ) . To date , such efforts have been largely limited to reducing or increasing expression of components of the native hormone signaling machinery ( Voytas and Gao , 2014 ) , an approach ill-suited for tuning the strength of connections within a network and easily confounded by redundancy and buffering within a network . In trying to circumvent redundancy , researchers are often forced to construct high order mutants of the multiple genes underlying the function of a single network hub . This approach reduces the precision of experimental or engineering interventions , as these genes are frequently only partially redundant with one another , and , thus this approach introduces more off-target effects . Chimeric promoters with altered hormonal regulation of a gene of interest have been used with some success ( Ulmasov , 1997; Rushton et al . , 2002 ) . However , the paucity of detailed mechanistic maps connecting promoter architecture and chromatin state , and the high heterogeneity in these factors between genes , means that promoter design remains a bespoke approach with an associated high design and development cost for each network of interest . Additionally , these methods often require adding an extra copy of the gene of interest in a novel chromatin context , making it difficult to make definitive mechanistic conclusions . These challenges have made it difficult to study the significance of hormone regulation on specific genes , particularly in regard to the impact of transcriptional feedback loops on differentiation and morphogenesis . For all of these reasons , the potential predictive power of mathematical models has not been fully leveraged in the engineering of morphologies of agronomic interest . To facilitate more sophisticated interventions in plant developmental programs , we designed a set of synthetic and modular hormone-activated Cas9-based repressors ( HACRs , pronounced ‘hackers’ ) .
We previously validated the design of similar synthetic auxin-sensitive transcription factors in Saccharomyces cerevisiae ( Khakhar et al . , 2016 ) . Guided by this work , we fused the deactivated Cas9 ( dCas9 ) protein from Streptococcus pyogenes ( Gilbert et al . , 2013 ) to a highly sensitive auxin-induced degron ( Moss et al . , 2015 ) and the first 300 amino acids of the TOPLESS repressor ( TPL ) ( Pierre-Jerome et al . , 2014 ) ( Figure 1A ) . The dCas9 associates with a guide RNA ( gRNA ) that targets the HACR to a promoter with sequence complementarity where it can repress transcription . Upon auxin accumulation , the degron sequence targets the HACR for ubiquitination and subsequent proteasomal degradation . Thus , in parallel to the natural auxin response , auxin triggers relief of repression on HACR target genes . Transgenic plants were generated with HACRs and a gRNA targeting a constitutively expressed Venus-Luciferase reporter , and , as expected , auxin treatment increased overall fluorescence ( Figure 1B , C ) . A time-course using luciferase to quantify de-repression of the reporter supported these results with a significant spike in reporter signal ( p<0 . 001 , n = 10 ) peaking approximately 80 min post auxin exposure ( Figure 1D , E ) . A HACR with a stabilized degron ( Moss et al . , 2015 ) showed significantly lower reporter signal upon auxin treatment ( p=0 . 01 , n = 10 ) ( Figure 1F ) . The modular nature of HACRs should allow substitution of the degron with any sequence that has a specific degradation cue . We tested this hypothesis by building HACR variants with degrons sensitive to two other plant hormones: jasmonates ( JAs ) ( Katsir et al . , 2008 ) and gibberellins ( GAs ) ( Murase et al . , 2008 ) . Treatment of transgenic plants with exogenous hormones matched to the expressed variants significantly increased reporter signal as compared to control treatments ( Figure 1H , I , J , Figure 1—figure supplement 1 ) . To rewire the connections between the hormone circuitry and developmental master controllers , HACRs must be able to respond to local differences in endogenous hormone levels . To visualize subtle differences in HACR sensitivity at the cellular level , we built a ratiometric auxin HACR by combining our previous design with a second reporter ( tdTomato ) driven by the same UBQ1 promoter driving the Venus reporter , with the only difference being that its gRNA target site was mutated ( Figure 2A ) . An estimation of relative auxin levels was then calculated by normalizing the Venus reporter signal in each cell to that of the tdTomato signal in the same cell , minimizing any effect of differential expression of the UBQ1 promoter in different cell types . Using these lines , we visualized tissues at different developmental stages where auxin distributions had been previously described using auxin reporters like DII-VENUS or R2D2 ( Liao et al . , 2015 ) . Auxin accumulation assayed by the HACR largely matched previous reports , such as the reverse fountain pattern of reporter signal in the root tip ( Band et al . , 2014 ) ( Figure 2B ) and higher signal in the vasculature as compared to the epidermis of the elongation zone ( Band et al . , 2014 ) ( Figure 2C ) . We also observed high reporter signal in emerging lateral root primordia consistent with the auxin accumulation that triggers this developmental event ( Dubrovsky et al . , 2008 ) ( Figure 2D , E ) . To further explore the capacity of HACRs to respond to differences in endogenous hormone levels , we visualized the activity of auxin , GA and JA HACRs targeting a Venus reporter . Auxin accumulates in the apical domain of the early embryo and eventually resolves in later stages to the tips of the developing cotyledons , vasculature , and future root apical meristem ( Liao et al . , 2015 ) – the same patterns that were observed in plants expressing an auxin HACR ( Figure 2F–J ) . In plants expressing a GA HACR , we observed a strong reporter signal in the early endosperm , consistent with the expression of GA biosynthesis enzymes ( Hu et al . , 2008 ) ( Figure 2K–M , Figure 2—figure supplement 1 ) . There are few reports of developmental regulation of JA distribution; however , we did detect accumulation of reporter signal in the developing ovule of plants expressing a JA HACR ( Figure 2—figure supplement 1 ) . Specifically , reporter signal appeared to be localized to the inner- and outermost layers of the integuments that surround the developing seed . We also observed that the JA HACR reporter was strongly induced in leaves subjected to mechanical damage ( Figure 2N–Q ) , a condition known to induce high levels of JA ( Katsir et al . , 2008 ) . Beyond their application as sensors of endogenous hormone distributions , HACRs should also be capable of reprogramming how such signals are translated into plant morphology . To test this , we turned to shoot architecture , an agronomically important trait with a well-established connection to auxin . Fewer side-branches allow for higher density planting ( Khush , 2001 ) and more regular arrangement of lateral organs ( phyllotaxy ) facilitates efficient mechanized harvest ( Burks et al . , 2005 ) . The molecular mechanisms that control branching and phyllotaxy are well studied and have been mathematically modeled ( Refahi et al . , 2016; Prusinkiewicz et al . , 2009 ) . These models predict that a key parameter controlling both these processes is the strength with which auxin promotes its own polar transport ( Bennett et al . , 2014 ) , which we will refer to as feedback strength . One molecular mechanism that contributes to this feedback is the auxin-induced increase in expression of the auxin transporter PIN-FORMED1 ( PIN1 ) ( Vieten et al . , 2005 ) . Thus far , it has been impossible to tune the strength of auxin-mediated transcriptional feedback on PIN1 , and thus impossible to fully test its role in regulating shoot architecture or its potential for engineering this trait . To test whether we could rationally alter shoot architecture by changing feedback strength , we generated transgenic plants with a HACR targeting PIN1 ( Figure 3A ) , as well as a model that produced a qualitative hypothesis of the impact of this intervention ( Supplementary note 1 ) . Our model predicts that this perturbation will decrease the activation of expression of PIN1 by auxin and dampen the dose response relationship between auxin and PIN1 expression ( Figure 3—figure supplement 1B , C ) . Quantitative PCR results on transgenic plants support these predictions , as the modest but significant reduction in PIN1 expression observed in plants expressing a PIN1 gRNA can be erased with exogenous auxin treatment ( Figure 3—figure supplement 1D ) . Our model and these results highlight the substantial difference between regulation by a hormone-responsive transcription factor and a static repressor . Static repressors would consistently suppress target gene expression at all hormone levels . In contrast , HACRs dampen both the dynamic and steady state dose response relationship between hormone concentration and gene expression akin to modulating the gain in a circuit ( Figure 3—figure supplement 1B , C ) . In relation to shoot architecture models , the effect of an auxin-regulated HACR targeting PIN1 should be a reduction in feedback strength . In Prusinkiewicz et al . ( Prusinkiewicz et al . , 2009 ) , auxin-regulated feedback is modeled as a post-translational mechanism dependent on the flux of auxin through the cell membrane . The magnitude of this flux is proportional to the recruitment of PIN1 to the membrane . According to their simulations , feedback strength is directly proportional to the number of branches the plant will develop . This effect is hypothesized to result from the reduced ability of lateral buds to establish auxin efflux into the main stem , an essential step in bud outgrowth ( Figure 3D ) . While the transcriptional mode of feedback we are altering with our HACR is not directly encoded in the Prusinkiewicz et al . model , we hypothesized that decreasing transcriptional feedback strength would have qualitatively similar results to decreasing post-translational feedback strength . Thus , we expected a decrease in the number of branches in lines where auxin HACRs were targeted to PIN1 . This is exactly what we observed ( Figure 3—figure supplements 2 and 5 ) . In lines with the strongest phenotypes , we observed roughly half the total number of branches per plant ( Figure 3E ) . No difference in the number of branches was observed for lines that had a HACR with a stabilized auxin degron regulating PIN1 expression , suggesting this phenotype was not simply due to repression of PIN1 ( Figure 3—figure supplement 3 ) . Feedback strength is also an important control parameter for the process of phyllotactic patterning . In the inhibition zone model , each primordium ( Figure 3F , green circles ) creates an inhibition zone around itself by depleting auxin ( Figure 3F , shown in orange ) from its surroundings , thereby preventing enough auxin to accumulate to form a new primordium . This zone is created by a feedback driven flow of auxin towards the primordium . The cells that are capable of forming new primordia are present in a region called the central zone periphery ( Figure 3F , black ring ) surrounding the shoot apical meristem ( Figure 3F , green circle in the back ring ) . The overlapping inhibition zones from all the existing nearby primordia leave only certain regions of the central zone periphery capable of forming new primordia ( Figure 3F , dashed green circles on yellow arcs ) . A mathematical model by Refahi et al ( Refahi et al . , 2016 ) . divides the central zone periphery into discrete units or cells and calculates a probability for each cell to form a new primordium at every timepoint . This probability is used to simulate the growth of the plant and estimate the expected frequency of phyllotactic patterning errors , such as co-initiation of primordia ( Figure 3F , as shown in the grey meristem ) . This occurs when there is more than one region on the central zone periphery that is competent to form a primordia , leading to two primordia being initiated at the same time . According to the model , the radius of the inhibition zones is inversely proportional to the number of co-initiatiating primordia . In auxin HACR plants with a PIN1 gRNA , we hypothesized that lower feedback strength would lead to a less sharp auxin gradient around each primordium and thus a larger inhibition zone ( Bennett et al . , 2014 ) ( Figure 3F , as shown in the blue meristem ) . Consistent with this prediction , plants with a HACR targeting PIN1 showed a significant reduction in co-initiations ( Figure 3G , Figure 3—figure supplement 4 ) . By making it possible to alter transcriptional feedback strength rather than simply gene expression , the HACR platform enabled exploration of previously inaccessible parameter regimes . This proof-of-concept establishes a new method for modifying a large number of desired traits . Additionally , the modular nature of HACRs allows for independent tuning of hormone sensitivity and repression strength ( Khakhar et al . , 2016 ) , as well as allowing for tissue-specific modulation of target genes . These modifications could substantially extend the range of possible phenotypes and mitigate trade-offs , for example having few branches to fit more plants on a field versus the total number of fruits per plant . The use of HACRs here is among the first examples of utilizing synthetic signaling systems to re-engineer the morphology of a multicellular organism in a model-driven manner , a long standing goal across the fields of pattern formation and tissue engineering , and this strategy should be extensible to a wide variety of organisms , particularly given the success of implementing the auxin-induced degradation module ( AID ) in diverse eukaryotes ( Nishimura et al . , 2009 ) . In agricultural settings , farmers already manipulate development or defense pathways by applying hormones or their synthetic mimics . HACRs could be used to connect these treatments with the expression of genes , such as those involved in defense , to create inducible traits . Additionally , HACRs could be extended to any other hormone that utilizes degradation-based signaling , such as salicyclic acid , strigalactones and karrikins . The wide range of degradation cues , the ease of targeting any gene , and the likely conserved function across angiosperms should mean that HACRs have the capacity to reprogram a plethora of developmental traits in a broad range of crop species .
Expression cassettes for the gRNAs , HACRs and the reporters were built using Gibson assembly ( Gibson et al . , 2009 ) . These were then linearized by restriction enzyme digestion and assembled into a yeast artificial chromosome based plant transformation vector with kanamycin resistance using homologous recombination based assembly in yeast ( Shih et al . , 2016 ) . The PIN1 gRNA expression vector and the additional tdTomato expression vector for the ratiometric lines were built using Golden-Gate assembly ( Engler et al . , 2008 ) into the pGRN backbone ( Hellens et al . , 2000 ) with hygromycin resistance . The gRNA expression cassettes contain a sgRNA driven by the U6 promoter and have a U6 terminator . The HACR expression cassettes are driven by the constitutive UBQ10 ( AT4G05320 ) promoter and have a NOS terminator . All HACR variants contain the same deactivated SpCas9 ( dCas9 ) domain ( Gilbert et al . , 2013 ) translationally fused at the N-terminus to an SV40 nuclear localization signal . The hormone degron domain and the repressor domain were fused to the C terminus of dCas9 , with the respective degron domain in the middle and flexible 6xGS linkers separating the sub-domains . The rapidly degrading NdC truncation of the IAA17 degron ( Moss et al . , 2015 ) was used for all the auxin HACRs described in the paper . The JA HACR contained the degron from the Arabidopsis JAZ9 protein ( AT1G70700 ) ( Katsir et al . , 2008 ) . The GA HACRs contained either GAI ( At1g14920 ) ( Murase et al . , 2008 ) or RGA1 ( At2g01570 ) ( Murase et al . , 2008 ) cloned from Arabidopsis cDNA . The HACR repression domain was the nucleic acid sequence corresponding to the first 300 amino acids of the TOPLESS repressor ( TPL , At1g15750 ) ( Pierre-Jerome et al . , 2014 ) . We chose this repression domain as TPL is the co-repressor used in native auxin and JA signal transduction pathways . The reporter cassette that was regulated by the HACRs contained a yellow fluorescent protein ( Venus ) translationally fused to a nuclear localization sequence on its N-terminus and firefly luciferase translationally fused on its C-terminus with flexible linkers . The reporter was driven by a constitutive UBQ1 ( AT3G52590 ) promoter and had a UBQ1 terminator . The additional reporter in the ratiometric lines was identical to these constructs except Venus-Luciferase was replaced with tdTomato and the gRNA target site in the UBQ1 promoter was mutated . The PIN1 gRNA expression vector contained a U6 promoter and terminator . All HACR reporter lines were built by transforming the yeast artificial chromosome plasmids described above into Agrobacterium tumefaciens ( GV3101 ) and using the resulting strains to transform a Columbia-0 background by floral dip ( Clough and Bent , 1998 ) . Transformants were then selected using a light pulse selection ( Harrison et al . , 2006 ) . Briefly , this involves exposing the seeds to light for 6 hr after stratification ( 4°C for 2 days in the dark ) followed by a three day dark treatment . Resistant seedlings demonstrate hypocotyl elongation in the case of Hygromycin and leaf greening after 5 days in the case of Kanamycin . After selection seedlings were transplanted to soil and grown in long day conditions at 22°C . For all the HACR reporter genotypes ( Figures 1 and 2 ) at least three lines were grown to the T2 and tested for their response to the appropriate hormone treatment with n = 10 for seedlings . To generate the ratiometric auxin HACR lines the additional tdTomato reporter was transformed into Col0 and then lines that were screened for uniform tdTomato expression were crossed into a line that had the HACR targeted to a Venus reporter . Three different auxin HACR backgrounds were transformed with a gRNA targeting PIN1 . The branching of three independent lines , representing three independent PIN1 gRNA insertion events , in each HACR background was characterized in the T2 at n = 5 . Several lines were characterized in the T3 at n > 20 both with and without selection . The number of co-initiations of three independent lines in one HACR background was characterized in the T2 at n = 5 . The number of co-initiating siliques of one of these lines was characterized in the T3 at n = 25 . For imaging the effects of auxin treatment on root tips we selected plants on 0 . 5xLS +0 . 8% bactoagar containing Kanamycin using the light pulse protocol described above . Four days after the seedlings were removed from the dark we transplanted to fresh 0 . 5xLS +0 . 8% bactoagar without Kanamycin and then imaged on a Leica TCS SP5 II laser scanning confocal microscope on an inverted stand . For auxin induction of root tips , the seedlings were sprayed with a 1:1000 dilution in water of either control ( DMSO ) or auxin dissolved in DMSO ( 5 µM final concentration ) and then mounted on slides in water and imaged after 24 hr . For the imaging of ratiometric lines seedlings were germinated without selection and then visually screened using a fluorescence microscope for expression of both reporters . These seedlings were then imaged on a confocal microscope at several positions along the primary root to visualize auxin distributions in the root tip , the elongation zone and in developing lateral roots . The images were taken using a Leica TCS SP5 II laser scanning confocal microscope on an inverted stand . The ratiometric images were generated using the calcium imaging calculator in the Leica software , by background subtracting both the tdTomato and Venus signals and then normalizing the Venus signal by the tdTomato signal . The images of ovules 48 hr after pollination were obtained by emasculating flowers prior to anther dehiscence followed by hand pollination 12 hr after . After 48 hr , the ovules from the pistils of these flowers were dissected using hypodermic needles under a dissection microscope and then mounted on slides in 80 mM sorbitol and imaged with confocal microscopy as in Beale et al . ( Beale et al . , 2012 ) . To image the developing embryos , ovules were dissected from siliques at the appropriate developmental stages , individually dissected and mounted onto slides in MS0 media before being analyzed by confocal microscopy . All confocal microscopy images presented in this work are maximum projections of sub-stacks from regions of interest . Luciferase based time course assays were used to characterize the dynamics of HACR response to exogenous or endogenous hormone stimulus . All imaging was done using the NightOWL LB 983 in vivo Imaging System , which uses a CCD camera to visualize bioluminescence . For the data collected for Figure 1 and Figure 3—figure supplement 1 , assays were performed on seedlings . Here , T2 plants were selected by Kanamycin selection using the previously described light pulse protocol . These were then transplanted to fresh plates without antibiotic four days after selection and sprayed with luciferin ( 5 µM in water ) in the evening . The next morning , after approximately 16 hr , they were sprayed again with luciferin . After 5 hr they were imaged for one hour ( 10 min exposure with continuous time points ) , then sprayed with a control treatment ( a 1:1000 dilution of DMSO in water ) and then imaged for five hours . These same plates were then re-sprayed with luciferin ( 5 µM in water ) and left overnight . The next day these same plates were again imaged with an identical protocol as the previous day , except they were sprayed with a 1:1000 dilution of hormone in water ( 5 µM Indole-3-acetic acid ( auxin ) , 30 µM coronatine ( JA ) or 100 µM GA3 post dilution ) rather than control . Luminescence of each seedling was recorded over time and reported as values normalized to the time-point prior to treatment . For the mechanical damage assay of the jasmonate HACR in Figure 2 , plants were treated identically as described above except that instead of being sprayed with hormones , leaves on the plant were mechanically crushed using forceps . All the data collected was analyzed and plotted using python ( Khakhar , 2018; https://github . com/arjunkhakhar/HACR_Data_Analysis; *copy archived at https://github . com/elifesciences-publications/HACR_Data_Analysis ) . For the luciferase assays , all the time courses were normalized the reading before induction to make them comparable . All p-values reported were calculated in python using the one-way ANOVA function from the SciPy package ( Oliphant , 2007 ) . ( https://docs . scipy . org/doc/scipy/reference/generated/scipy . stats . f_oneway . html ) To characterize branching in plant lines with and without an auxin HACR regulating PIN1 , we selected T2 transformants for lines that had a gRNA targeting PIN1 and the parental HACR background that had no gRNA . The plants that passed the selection were transplanted onto soil and then characterized as adults at the point that there were on average four stems on the no gRNA control lines . In all cases the parental controls that lack a gRNA and the lines derived from them , by transforming with a gRNA targeting PIN1 , were all grown in parallel and phenotyped on the same day to ensure the data collected was comparable . Additionally , while we do not believe that the selection would have a significant effect on the phenotyping data as we collected it more than a month after the plants had been transplanted off selection plates onto soil , both the lines with a PIN1 targeting gRNA and the parental controls they were compared to were selected in parallel to control for any confounding effect . Phenotyping involved counting the number of branches on the plant . We quantified the number of branches on five T2 plants for three different lines with a HACR targeted to regulate PIN1 in two different HACR backgrounds , in parallel with the parental HACR background . The line with the strongest phenotype was propagated to the T3 generation with its parental HACR background and the same experiment was repeated with an n = 25 . To quantify the number of co-initiating siliques we measured the internode length between the first 20 siliques on a single axillary stem and every instance of two siliques emerging from the same point on the stem ( an internode length less than 1 mm which we found to be the threshold for visual discrimination ) was considered a co-initiation . The line that showed the strongest phenotype was propagated to the T3 generation with its parental HACR background and the same experiment was repeated with an n = 25 . To prove the phenotypes we were observing were independent of selection conditions we also characterized branching of T2 and T3 plant lines that were not selected on antibiotic selections . These plant lines were transplanted off 0 . 5x LS plates ten days after germination . They were then grown till adulthood and then phenotyped and genotyped for the presence of the HACR and PIN1 gRNA . All plants that were phenotyped were grown in long day conditions on Sunshine #4 mix soil in rose pots and watered every other day on a watering table . All qPCR assays were performed on seedlings seven days after they been selected using the light pulse procedure ( fifteen days post germination ) . For each biological replicate five seedlings that passed selection were transplanted off the selection plate and into 4 ml of 0 . 5xLS with either mock of 50 nM 2-4D . They were then incubated in well lit , humidity-controlled conditions for 3 hr and then the seedlings were blotted and flash frozen in liquid nitrogen . The RNA was extracted from these seedlings using the Illustra RNAspin Mini Kit from GE . cDNA was then prepared from 1 ug of RNA using the iScript cDNA synthesis kit from Biorad and then used to run a qPCR with the iQ SYBR Green Supermix also from Biorad on a Biorad qPCR machine . Each sample was analyzed for expression of PIN1 and PP2A which was used to normalize PIN1 levels . A standard curve was generated using the pooled samples for each primer set to determine amplification efficiency . The primers used are listed below: PIN1_q_R: AACATAGCCATGCCTAGACC PIN1_q_F: CGTGGAGAGGGAAGAGTTTA PP2A_q_R: AACCGCTTGGTCGACTATCG PP2A_q_F: AACGTGGCCAAAATGATGC ABS44 - https://benchling . com/s/yXKJkba5 ABS50 - https://benchling . com/s/897tnlX2 PHD5 - https://benchling . com/s/HnODIKMV PHD3 - https://benchling . com/s/HOEPc5FA PHD6 - https://benchling . com/s/Ge8pztYw pGRN_H-pU6:pPIN1_gRNA_Target1-tU6 - https://benchling . com/s/3RBYAIkF pGRN_H-pUBQ1_AlteredGrnaTargetSite:NLS-tdTomato-tUBQ1 - https://benchling . com/s/Pd0Ms4Qs | The genetic information of plants contains sets of instructions that shape a growing seedling . These ‘developmental programs’ are under the control of a range of hormones , such as auxin . Typically , the information from the hormones is relayed to the genetic material through proteins called transcription factors , which can act on DNA to turn specific genes on or off . Scientists have a good understanding of the roles of hormones , and they have created mathematical models that predict how changes in hormone levels affect the shape of a plant . However , it is still difficult to manipulate hormones inside a plant and test these models . Here , Khakhar et al . created artificial transcription factors , referred to as HACRs , and put them into a plant called Arabidopsis thaliana . An HACR is made of different molecular modules stitched together . Each module has a precise role; for example , one turns off a specific gene , while another targets the HACR for destruction if a given hormone is present . First , Khakhar et al . showed that HACRs could help track the levels of auxin in a developing plant . Arabidopsis plants were genetically engineered so that they would always produce a fluorescent protein . Then , an HACR was created that would switch off the gene for that fluorescent protein , so that no fluorescence would be present in the cell . If auxin was present , the HACR would get degraded , meaning fluorescence would appear . This helped to finely assess the amount of the hormone in various parts of the plant . By changing the modules in the HACRs , this approach could be applied to at least three other types of hormones . Second , HACRs were used to reprogram Arabidopsis and change its appearance . For example , it is well known that auxin controls the number and location of branches on a plant . This complex process depends on how strongly auxin promotes the expression of a gene called PIN1 . Khakhar et al . engineered an HACR that represses PIN1 , and created a mathematical model that described the impact of this intervention . As predicted by the simulation , the HACR changed the strength of the relationship between PIN1 and auxin , which resulted in plants with fewer branches – a trait that is of interest in farming . HACRs are a new type of technology that is likely to work in a wide range of species . Ultimately , these artificial transcription factors could help to engineer plants that can face the disruptions brought by climate change , which would ensure better food security for people around the world . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"tools",
"and",
"resources"
] | 2018 | Synthetic hormone-responsive transcription factors can monitor and re-program plant development |
The beating of motile cilia generates fluid flow over epithelia in brain ventricles , airways , and Fallopian tubes . Here , we patch clamp single motile cilia of mammalian ependymal cells and examine their potential function as a calcium signaling compartment . Resting motile cilia calcium concentration ( [Ca2+] ~170 nM ) is only slightly elevated over cytoplasmic [Ca2+] ( ~100 nM ) at steady state . Ca2+ changes that arise in the cytoplasm rapidly equilibrate in motile cilia . We measured CaV1 voltage-gated calcium channels in ependymal cells , but these channels are not specifically enriched in motile cilia . Membrane depolarization increases ciliary [Ca2+] , but only marginally alters cilia beating and cilia-driven fluid velocity within short ( ~1 min ) time frames . We conclude that beating of ependymal motile cilia is not tightly regulated by voltage-gated calcium channels , unlike that of well-studied motile cilia and flagella in protists , such as Paramecia and Chlamydomonas .
Cilia are ancient microtubule-based cellular appendages found in eukaryotic organisms ( Satir et al . , 2008 ) . Motile cilia , like flagella , drive fluid flow via dynein-ATPase action on their 9 + 2 microtubular structure , while most primary cilia are solitary , nonmotile , and lack central microtubular pairs ( 9 + 0 ) ( Lindemann and Lesich , 2010; Satir and Christensen , 2007 ) . In mammals , almost all cells possess a single primary cilium that houses the Sonic Hedgehog pathway ( Drummond , 2012 ) and mediates aspects of cell-cell signaling . Other nonmotile cilia are found in specialized sensory cells , such as rods and cones of the eye ( Satir and Christensen , 2007 ) . In contrast , multiple copies of motile cilia sprout from ependymal cells lining the brain ventricles , and epithelial cells in the airways and Fallopian tubes ( Brooks and Wallingford , 2014; Satir and Christensen , 2007 ) . A distinct type of motile cilium ( 9 + 0 ) additionally protrudes from cells in the embryonic node during development ( Babu and Roy , 2013 ) . Motile cilia are much like eukaryotic flagella that drive locomotion in spermatozoa and protists ( Brooks and Wallingford , 2014; Satir and Christensen , 2007 ) . Ca2+ influx modulates motility in the flagella of spermatozoa through specialized Ca2+-selective , pH-sensitive CatSper channels to produce hyperactivated motility ( Kirichok et al . , 2006; Miki and Clapham , 2013; Qi et al . , 2007; Ren et al . , 2001 ) . Ca2+ influx also alters the flagellar waveform or ciliary beating direction in Chlamydomonas and Paramecium , respectively , and arrests beating in Mussel gill epithelia ( Bessen et al . , 1980; Inaba , 2015; Naito and Kaneko , 1972; Tsuchiya , 1977; Walter and Satir , 1978 ) . Analysis of Paramecium and Chlamydomonas mutants and electrophysiological recordings identified voltage-gated calcium channels ( CaV ) in cilia/flagellar membranes as required regulators of ciliary beating ( Beck and Uhl , 1994; Dunlap , 1977; Fujiu et al . , 2009; Kung and Naito , 1973; Matsuda et al . , 1998 ) . These observations suggest a conserved Ca2+ channel-dependent mechanism regulating flagellar/ciliary beating . Whether ion channels in motile cilia of mammalian cells changes their beat frequency is not clear , but intraciliary [Ca2+]-dependent changes in motile cilia beating has been reported by several groups ( Di Benedetto et al . , 1991; Girard and Kennedy , 1986; Lansley et al . , 1992; Nguyen et al . , 2001; Schmid and Salathe , 2011; Verdugo , 1980 ) . The question that we seek to answer in this study is whether Ca2+-permeant ion channels are present in motile cilia , and if so , do they change motile cilia behavior . Successful whole-cilia patch clamping of fluorescently-labeled immotile primary cilia revealed nonselective cation currents ( PKD2-L1 + PKD1-L1 heteromeric complexes ) in primary cilia membranes ( DeCaen et al . , 2013; Delling et al . , 2013 ) . Here , we examined ion currents in fluorescently-labeled , voltage-clamped motile cilia of brain ependymal cells and demonstrate that motile cilia are well coupled electrically and by diffusion to the cellular compartment . We show that few CaV channels are present in the cilia membrane , that resting [Ca2+] is only slightly elevated in motile cilia , and that motile cilia [Ca2+] is driven primarily by changes in cytoplasmic [Ca2+] . Excitation of the ependymal cell by membrane depolarization increases ciliary [Ca2+] with only minor changes in motility and fluid movement , suggesting that beating of ependymal motile cilia is not significantly regulated by the activity of ciliary or cytoplasmic CaV channels .
We initially examined ependymal cell GFP-labeled motile cilia from immunolabeled brain sections of transgenic Arl13b-EGFPtg mice ( Delling et al . , 2013 ) . Ciliary localization of Arl13b-EGFP was confirmed by co-staining with the ciliary marker , acetylated tubulin ( Figure 1A ) . We also observed GFP-labeled motile cilia in primary cultures . At day 10 in vitro ( DIV10 ) , ~88% of acetylated tubulin stained multiciliated ependymal cells ( n = 400 cells ) had GFP-labeled cilia ( n = 352 cells ) . Transgene expression varied , with ~one-third of the cells exhibiting weak GFP fluorescence in cilia ( Figure 1B , arrow ) . Some cells had only a few motile cilia ( either mono- , bi- or sparsely ciliated ) that often displayed moderate motility ( ~two–three fold slower beat frequency ) as compared to neighboring multiciliated cells ( Figure 1B , arrow head ) . Sparsely ciliated cells may be the result of ependymal cell maturation , in vitro culturing , or represent a population of previously reported biciliated ependymal cells ( Mirzadeh et al . , 2008 ) . Sparse cilia likely experience different hydrodynamic forces than large groups of synchronized cilia in multiciliated cells ( metachronism ) , which could explain differences in motility ( Guirao and Joanny , 2007; Guirao et al . , 2010 ) . 10 . 7554/eLife . 11066 . 003Figure 1 . Ependymal motile cilia identification and patch clamp . ( A ) Immunolabeling of motile cilia in a section of the lateral ventricle from an Arl13B-EGFPtg mouse . Ciliary localization of Arl13B-EGFP was confirmed using anti-GFP ( green ) and anti-acetylated tubulin antibodies ( red ) . ( B ) Anti-GFP ( green ) and anti-acetylated tubulin staining ( red ) of cultured ependymal cells at DIV10 . GFP labeling of motile cilia varied , with some cells displaying only weak or barely detectable GFP fluorescence in cilia ( arrows ) . Some cells were only sparsely ciliated ( arrowhead ) . ( C ) Staining of cultured multiciliated cells with anti-GFP ( green ) and anti-Spag6 ( red ) . ( D ) Representative staining of a previously recorded ependymal cell grown on a gridded glass bottom dish ( grid size , 50 μm , arrow marks motile cilium ) . Motile cilia of sparsely ciliated cells were GFP ( green ) and Spag6 ( red ) positive ( n = 40/45 ) . Nuclei were labeled with Hoechst dye ( blue , A–D ) . Panels B-D display average intensity z-projections of image stacks . Scale bars , 10 μm ( A–C ) and 5 μm ( D ) . ( E ) Image showing dye diffusion into a motile cilium after successful break-in ( 50 μM Alexa 594 hydrazide , n = 9 ) . Scale bar , 3 μm . ( F ) Example current ( bottom ) recorded in the whole-motile-cilium configuration in response to increasing voltage steps ( top ) . Holding potential , -80 mV . ( G ) Mean steady state current after break-in plotted as a function of command voltage ( n = 4 ) . External solution ( aCSF ) with 3 mM KCl ( black filled squares ) , 70 mM KCl ( blue filled circles ) , and 140 mM KCl ( red filled diamonds ) . Arrows in the graph indicate calculated EK values . Error bars; ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 11066 . 003 Motility and multiciliation hinder patch clamp recordings from individual cilia and thus recordings were from sparsely ciliated cells . We verified the 9 + 2 structural arrangement of the cilia by imaging the cells on coded , gridded glass dishes and subsequent staining ( Video 1 ) . Spag6 , a protein associated with the central pair of microtubules ( Sapiro et al . , 2002 ) , was detected in motile cilia of multi- and sparsely-ciliated cells ( Figure 1C , D ) . Eighty-nine percent of recorded sparsely ciliated cells exhibited Spag6 immunoreactivity in their motile cilia ( Figure 1D; n = 40/45 cells ) , consistent with the presence of a central pair of microtubules . In contrast , only a small number of primary cilia from mIMCD cells were ( weakly ) immunofluorescent in control experiments ( 8% , n = 9/114 , data not shown ) . 10 . 7554/eLife . 11066 . 004Video 1 . Time lapse of a ciliated cell . Ependymal cells were grown on gridded culture dishes ( grid size , 50 μm ) . Sparsely ciliated cells were recorded , fixed and stained with a central pair marker ( Spag6 ) . The example movie corresponds to the staining shown in Figure 1D ( same grid ) . Frame rate 0 . 065 s , playback 1x . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 11066 . 004 Typically , we patched the bulging tip of a motile cilium . High resistance seals were obtained by positioning the patch electrode in the focal plane at a position near the end of the cilia’s stroke , applying suction at the moment the cilium approached the pipette tip ( Video 2 ) . Access to a motile cilium was typically achieved by applying a series of brief voltage pulses . Upon successful ‘break-in’ , we were able to monitor dye diffusion from the electrode into the cilium ( Figure 1E ) . Recording via the whole-motile-cilium yielded a linear current that reversed at hyperpolarized potentials ( Figure 1F , G ) . These currents were primarily carried by K+ ( Figure 1G ) , reminiscent of the ohmic currents reported for ependymal cell bodies ( Genzen et al . , 2009a; Liu et al . , 2006; Nguyen et al . , 2001 ) . 10 . 7554/eLife . 11066 . 005Video 2 . Patch clamping of a motile cilium . Frame rate 0 . 58 s , playback 1x . Scale bar , 3 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 11066 . 005 To determine whether ependymal motile cilia are electrically coupled to the cellular compartment , we measured their electrical properties and compared them to whole-cell recordings ( Figure 2A , B ) . Input resistances immediately upon break-in were on average ~five-fold higher in whole-cilium recordings ( ~180 MΩ , n = 8 ) than in whole-cell recordings ( ~36 MΩ , from the cell body and connected cells , n = 12 ) , reflecting the resistance introduced by the cilium ( see below ) . Uncoupling the cells with flufenamic acid ( FFA ) , an anthranilic acid derivative known to inhibit gap junctions ( Harks et al . , 2001 ) , decreased the measured conductance in both whole-cell and whole-cilium recordings . Bath substitution with TEA-Cl/BaCl2 blocked the remaining K+ conductances ( Figure 2A , B ) . 10 . 7554/eLife . 11066 . 006Figure 2 . Electrical coupling of motile cilia to the cellular compartment . ( A , B ) Mean current-voltage relation recorded from the cell body ( A , n = 6 ) or a motile cilium ( B , n = 5 ) after break-in in aCSF ( filled squares ) , subsequent cell uncoupling with flufenamic acid ( FFA , 100 μM , 2 min , filled circles ) , or block of K+ conductances in TEA-Cl/BaCl2 ( TEA/Ba2+ , filled diamonds; only cell/cilium recordings with input resistance >1 GΩ after TEA-Cl/BaCl2 treatment are plotted ) . Holding potential between 200 ms steps , -80 mV . Note: The voltage in A and B refers to the command voltage . The voltage error , that is , the difference between the command voltage and membrane voltage produces a large error due to the high resistance through the cilium in series with the low , multicellular membrane resistance ( resting K+ conductance , cell-cell connections via gap junctions ) . Thus , large currents are inaccurate: the top traces only serve to show that flufenamic acid uncouples cells . ( C ) Example capacitive currents recorded in response to a 20 mV hyperpolarizing voltage step ( 50 ms ) for the cell body ( black ) and motile cilium ( green ) after uncoupling with flufenamic acid ( FFA , 100 μM ) and block of K+ conductances ( TEA-Cl/BaCl2 ) . The steady state ( time-independent ) current in the motile cilium trace is leak current . ( D–F ) Time constant ( D ) , membrane capacitance ( E ) , and series resistance ( F ) determined from an average of 100–200 sweeps of capacitive current for cell body ( black squares ) and motile cilium ( green squares ) recordings after cell uncoupling with flufenamic acid ( FFA , 100 μM , ~5 min ) and perfusion with TEA-Cl/BaCl2 ( TEA/Ba2+ , n = 7–8 ) . ( G ) Resting membrane potential assessed under current clamp with a gramicidin-perforated patch for cell bodies ( black squares ) and motile cilia ( green squares ) before ( n = 7 cell body , n = 8 motile cilia ) and after ( n = 7 cell body , n = 5 motile cilia ) addition of flufenamic acid ( FFA , 100 μM ) to the bath . Open squares represent the range of individual cells/cilia; filled squares are the mean . Error bars; ± SEM . ( H ) Cartoon illustrating simplified equivalent circuit of access to the cellular compartment via a motile cilium . The cable-like properties of a motile cilium significantly increase the access ( series ) resistance . Block of gap junctions by flufenamic acid removes the contributions from neighboring cells . DOI: http://dx . doi . org/10 . 7554/eLife . 11066 . 006 To directly assess coupling between motile cilia and cell compartments , we analyzed the capacitive current recorded under voltage clamp in the whole-cell and whole-motile-cilium configuration . In response to a hyperpolarizing voltage step , capacitive current relaxation was 8–10 times faster in whole-cell recordings ( 0 . 3 ± 0 . 1 ms in aCSF/FFA and 0 . 4 ± 0 . 1 ms in TEA-Cl/BaCl2 ) than in whole-motile-cilium recordings ( 3 . 1 ± 0 . 3 ms in aCSF/FFA and 3 . 3 ± 0 . 4 ms in TEA-Cl/BaCl2; Figure 2C , D ) . By integrating the current transient to yield net capacitance , we determined that a similar surface area was charged in whole-cell ( 20 . 4 ± 3 . 2 pF in aCSF/FFA , 18 . 5 ± 2 . 5 pF in TEA-Cl/BaCl2 ) and whole-motile-cilium recordings ( 18 . 3 ± 2 . 4 pF in aCSF/FFA , 19 . 0 ± 2 . 8 pF in TEA-Cl/BaCl2; Figure 2E ) . Calculating the series resistance from the time constant and membrane capacitance , we determined a nine–ten fold higher resistance in whole-motile-cilia recordings ( 180 . 3 ± 20 . 6 MΩ in aCSF/FFA and 188 . 2 ± 25 . 0 MΩ in TEA-Cl/BaCl2 ) as compared to whole-cell recordings ( 18 . 0 ± 3 . 4 MΩ in aCSF/FFA and 20 . 2 ± 3 . 7 MΩ in TEA-Cl/BaCl2; Figure 2F ) . Together these findings suggest that motile cilia are indeed electrically coupled to the cellular compartment via a high resistance cable determined by the length and cross-sectional area of a motile cilium ( estimated as ~350 MΩ for an ideally insulated cable; Figure 2H , see Materials and methods ) . In other words , currents recorded in the whole-motile-cilium configuration can be attributed to channel openings in the cell and/or cilia membrane . Finally , we measured an only slightly depolarized resting membrane potential for motile cilia as compared to the ependymal cell body ( -77 ± 3 mV , motile cilia; -88 ± 2 mV , cell body; Figure 2G ) . When K+ currents were blocked in the TEA-Cl/BaCl2 bath , we frequently observed relatively small voltage-dependent inward currents ( Figure 2B ) . Consistent with the finding that cilia and cellular compartments are electrically coupled , we recorded CaV currents in both whole-cell and whole-cilium configurations ( Figure 3A , B; high series resistance in whole-motile-cilium recordings affected the time course and peak of the calcium current; see Materials and methods ) . These currents were potentiated by BayK8644 and reduced by nimodipine and CdCl2 ( Figure 3A , B and Figure 3—figure supplement 1A ) , consistent with the pharmacology of L-type calcium channels ( CaV1 subfamily ) ( Catterall et al . , 2005 ) . Indeed , CaV1 . 2 and CaV1 . 3 α subunit transcripts were detected in the cDNA of cultured ependymal cells ( DIV10; Figure 3F ) . 10 . 7554/eLife . 11066 . 007Figure 3 . CaV-mediated currents and single channels in ependymal cells . ( A , B ) Average peak current in response to voltage steps ( 500 ms ) before ( filled squares , in TEA-Cl/BaCl2 ) and after addition of the CaV potentiator , BayK 8644 ( BayK , 5 μM , open squares ) , recorded from the cell body ( A , n = 7 ) or motile cilia ( B , n = 8 ) . Peak current amplitudes varied substantially ( range pre BayK treatment: cells , -58 pA to -726 pA; cilia , -25 pA to -413 pA ) . The high series resistance in whole-motile-cilium recordings shifts the peak to more hyperpolarized potentials . Holding potential , -80 mV . Cells were uncoupled by flufenamic acid ( FFA , 100 μM ) . Error bars; ± SEM . ( C , D ) Example of 5 consecutive traces recorded in cell-attached ( C ) or cilium-attached ( D , pipette filled with BaCl2 ) . BayK-induced long lasting CaV channel openings were observed in 6 of 8 cell-attached recordings ( BayK , 5 μM , bath ) . CaV channel openings were rare in motile cilia-attached recordings ( n = 1/29 , see Figure 3—figure supplement 1; note: smaller pipettes in motile cilia recordings results in smaller membrane area sampling ) . ( E ) All point amplitude histogram of all traces from the recording shown in C . ( F ) RT-PCR showing amplification of CaV1 . 2 and CaV1 . 3 transcripts from cDNA derived from cultured ependymal cells ( EC , DIV10 ) . cDNAs from skeletal muscle ( CaV1 . 1 ) , heart ( CaV1 . 2 ) , brain ( CaV1 . 3 ) , and eyes ( CaV1 . 4 ) served as positive control tissues ( CT ) . Minus reverse transcriptase negative control ( NC ) . Molecular ladder ( M ) . GAPDH was amplified from all cDNAs . Images cropped for illustration . DOI: http://dx . doi . org/10 . 7554/eLife . 11066 . 00710 . 7554/eLife . 11066 . 008Figure 3—figure supplement 1 . CaV-mediated currents and single channels in ependymal cells . ( A ) Reduction of CaV currents by nimodipine ( 10 μM , 2 min pre-incubation ) and cadmium chloride ( CdCl2 , 100 μM ) ( n = 4 , recorded from the cell body ) . Holding potential , -60 mV . Error bars; ± SEM . ( B ) Example of slightly distorted current observed in some whole-motile-cilium recordings due to the high series resistance . The red trace depicts the delayed opening of CaV channels as a result of the increased series resistance and delayed voltage clamp . Large oscillations indicate transient loss of voltage clamp . ( C , D ) Example of 5 consecutive traces recorded from a motile cilium in the cilium-attached configuration in presence of BayK8644 ( 5 μM , bath and pipette , C ) and open point amplitude histogram of all recorded traces ( D , fraction of full amplitude = 0 . 9 ) . Single channel openings were observed in 1 of 29 recordings . Mean single channel amplitude was 1 . 2 pA . DOI: http://dx . doi . org/10 . 7554/eLife . 11066 . 008 To address the question of whether CaV currents are in motile cilia membranes , cell body membranes , or both , it would be ideal to detach the cilium from the cell . However , we were unable to detach and record from isolated motile cilia , as we were able to do in primary cilia ( DeCaen et al . , 2013 ) . Thus , we recorded from cell and motile cilia membranes in the membrane-attached configuration . Prolonged single channel openings in the presence of BayK8644 averaged 1 . 4 pA at 0 mV ( from the record shown in Figure 3C ) and were frequently observed in cell-attached recordings , but were typically not detected in cilium-attached recordings ( Figure 3C–E ) . Despite many attempts , brief single channel openings reminiscent of L-type channel openings in the absence of BayK8644 ( Hess et al . , 1984 ) were recorded from only a single motile cilium patch ( Figure 3—figure supplement 1C , D ) . Full-length ciliary recordings ( Kleene and Kleene , 2012 ) , were also not feasible given the presence of CaV channels in the cell membrane . Nevertheless , the low percentage of motile cilium patches in which we observed CaV channel openings suggests that CaV channels are not enriched in ependymal motile cilia . If channel densities are identical in cell and cilia membranes , we would only expect ~3 channels in a motile cilium ( see Materials and methods ) . These Ca2+-permeant ion channels and currents in motile cilia are clearly distinct from the nonselective currents recorded from primary cilia ( DeCaen et al . , 2013 ) . To further examine the potential functional consequences of CaV channel activity , we evaluated motile cilia [Ca2+] by targeting a genetically encoded ratiometric Ca2+ sensor to these cilia ( Delling et al . , 2013 ) . Previous work established Somatostatin Receptor 3 ( SSTR3 ) transgene expression in ependymal motile cilia ( O'Connor et al . , 2013 ) . We fused GCaMP6s ( Chen et al . , 2013 ) with mCherry to the C-terminus of SSTR3 and transduced cultured ependymal cells with a recombinant adenoviral vector ( pAd-mSSTR3-mCherry-GCaMP6s ) , resulting in abundant expression of the sensor in motile cilia ( Figure 4A ) . Addition of ionomycin evoked a rapid increase in GCaMP6s fluorescence , confirming the ability of the sensor to report changes in ciliary [Ca2+] ( Figure 4B , C ) . We calibrated the ratiometric sensor ( see Materials and methods ) and determined the average ratio of F_GCaMP6s and F_mCherry in motile cilia under basal conditions ( aCSF , 1 . 4 mM extracellular Ca2+; Figure 4D , E ) . To avoid offsets in the position of beating cilia , we scanned non-sequentially and corrected ratios for bleed-through ( Figure 4—figure supplement 1A , B , and Materials and methods ) . Consistent with a relatively low number of CaV channels and a hyperpolarized resting membrane potential ( at which CaV channels are inactive ) , we determined the resting motile cilia [Ca2+] as 165 nM ( average ratio: 0 . 2 ± 0 . 01; Figure 4D , E ) . In control experiments in which immobilized cilia were scanned sequentially , a similar resting [Ca2+] was determined ( 167 nM , Figure 4—figure supplement 1C , and see below ) . We conclude that motile cilia resting [Ca2+] is only slightly elevated in comparison to cytoplasmic [Ca2+] ( ~100 nM ) ( Clapham , 2007 ) at steady state , in contrast to the ~500 nM elevation observed in primary cilia ( Delling et al . , 2013 ) . 10 . 7554/eLife . 11066 . 009Figure 4 . Motile cilia [Ca2+] can be modified by cytoplasmic [Ca2+] . ( A ) Cluster of recombinant adenovirus-transduced ependymal cells expressing the cilia-targeted fusion construct mSSTR3-mCherry-GCaMP6s . In fixed cells , cilia were recognized by staining with anti-GFP ( green ) and anti-mCherry ( red ) antibodies . ( B , C ) mSSTR3-mCherry-GCaMP6s reported changes in motile cilia [Ca2+] in response to ionomycin ( 2 μM ) . Example images ( B ) showing GCaMP6s ( pseudocolor ) and mCherry fluorescence before ( in aCSF ) and after addition of ionomycin to the bath , and quantified ratio changes ( C , n = 7 cells ) . The ratio of GCaMP6s and mCherry ( R ) was normalized to the initial ratio ( R0 ) . ( D ) Example pseudocolor images of F_GCAMP6s/F_mCherry ratios of ependymal motile cilia in aCSF ( basal ) and under defined free [Ca2+] in the bath . Ratio images were background-subtracted and thresholded . ( E ) Calibration curve showing F_GCAMP6s/F_mCherry ratio plotted as a function of free [Ca2+] ( n = 5–8 for each [Ca2+] ) . Resting motile cilia [Ca2+] was 165 nM at steady state ( n = 30 , red star ) . ( F ) Quantified changes in F_GCAMP6s/F_mCherry ratio in response to Ca2+ uncaging in the cytoplasm ( n = 13 ) . ( G ) Example pseudocolor images from a time lapse recording of an ependymal cell , recorded from the side ( see Materials and methods ) . Ca2+ was uncaged in the cytoplasm at the cilia base ( approx . time point marked by arrowhead , 405 nm illumination for 200 ms ) . ( H ) Cartoon illustrating line scanning . The red circle and red line indicate the typical position of the uncaging stimulus and line scan . ( I ) Example record of a line scan through the cytoplasm and a motile cilium displayed in pseudocolor . The arrowhead marks the position of uncaging at the ciliary base . Ca2+ rapidly diffused from the cytoplasm into the motile cilium ( 593 ± 86 ms to the tip , n = 15 ) . Error bars; ± SEM . Scale bars , 10 μm ( A ) and 5 μm ( B , D , G ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11066 . 00910 . 7554/eLife . 11066 . 010Figure 4—figure supplement 1 . Motile cilia [Ca2+] can be modified by cytoplasmic [Ca2+] . ( A ) Example images of HEK293 cells transfected with GCaMP6s . Images were acquired using the same settings used to calibrate the sensor . Non-sequential scanning resulted in bleed-through of the GCaMP6s signal . Fluorescent intensities are indicated . Scale bar , 10 μm . ( B ) Plotted linear correlation of the GCaMP6s fluorescent intensity and F_bleed-through ( n = 24 ) . ( C ) Sensor calibration using sequential scanning mode after treatment with sodium metavanadate ( SMVD , 100 μM; n = 4–5 for each [Ca2+] ) . Resting motile cilia [Ca2+] , 167 nM ( n = 10 , red star ) . ( D ) Example DIC images taken before and after incubation of ependymal cells with sodium metavanadate ( SMVD , 100 μM ) . Black line indicates the position of the line used to derive the kymographs shown below . Scale bars , 10 μm . ( E ) SMVD treatment greatly reduced the cilia beat frequency ( n = 40 , aCSF and n = 48 , SMVD ) . Beating frequencies of analyzed cilia ranged from 5 . 75–18 . 5 Hz in aCSF and 0–10 Hz after incubation with SMVD . Error bars; ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 11066 . 010 To assess Ca2+ diffusion between the cell body and motile cilia , we loaded transduced ependymal cells with caged Ca2+ ( NP-EGTA-AM ) . A brief uncaging stimulus in the cytoplasm rapidly increased ciliary GCaMP6s fluorescence ( Figure 4F ) . However , imaging motile cilia through the z-axis of the cilia ( that is , top or bottom views ) prevented further analysis of Ca2+ diffusion . In order to track Ca2+ waves , we halted ciliary beating and imaged ependymal cilia from the side ( Figure 4G , H , see Materials and methods ) . As reported previously , sodium metavanadate ( SMVD ) treatment greatly reduced motility , presumably through inhibition of the dynein ATPase ( Gibbons et al . , 1978; Nakamura and Sato , 1993 ) ( Figure 4—figure supplement 1D , E ) . Upon uncaging at the ciliary base ( visualized by additional loading of the cells with the Ca2+ indicator Oregon Green 488 BAPTA-1 AM , OGB-1 ) , Ca2+ entered the cilium and moved from the base to the tip with an apparent velocity of 22 ± 3 μm/s ( Figure 4G–I , Video 3 ) . Thus , fluctuations in cytoplasmic [Ca2+] can readily diffuse across the cell-cilia boundary to modify motile cilia [Ca2+] , as we found for primary cilia ( Delling et al . , 2013 ) . 10 . 7554/eLife . 11066 . 011Video 3 . Time lapse of a cell imaged from the side . Ca2+ was uncaged in the cytoplasm at the cilia base ( approximate location indicated by circle ) with a brief 405 nm laser pulse ( time point 1 . 016 s for 200 ms ) and diffused into the motile cilia . Cilia movement was inhibited by sodium metavanadate ( 100 μM ) . Frame rate 0 . 254 s , playback 1x . Scale bar , 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 11066 . 011 We next asked whether activation of CaV channels increases ciliary [Ca2+] and regulates ciliary motility in ependymal cells . Differences in the rise times of the two Ca2+ indicators ( OGB-1 and GCaMP6s ) precluded a meaningful analysis of signal onsets in motile cilia versus the cytoplasm . The homogeneity of voltage in cell body and cilium , as well as the observed rapid diffusion of Ca2+ from the cytoplasm into motile cilia , suggests that ciliary [Ca2+] will increase quickly , irrespective of the localization of CaV channels ( Figure 4 ) . By recording ependymal cells in current clamp , we measured the response to increasing external [K+] and observed a depolarization of membrane potential from -87 ± 3 mV at rest ( aCSF 3 K+ ) to -48 ± 5 mV , -31 ± 1 mV , -18 ± 0 . 7 mV , and -1 . 7 ± 2 mV in aCSF 20 K+ , 40 K+ , 70 K+ , or 140 K+ , respectively ( Figure 5—figure supplement 1A , B ) . Bath perfusion of depolarizing external [K+] ( >40 mM [K+] ) robustly increased cytoplasmic and ciliary [Ca2+] ( Figure 5A , B and Figure 5—figure supplement 1C , D , F ) . The response was greatly reduced by pre-incubation with nimodipine and was absent in Ca2+-chelated bath saline ( calculated as ~2 . 5 nM free [Ca2+] assuming ~10 μM total Ca2+; Figure 5C , D and Figure 5—figure supplement 1E ) , suggesting that CaV channel activation accounts for the increase in ciliary [Ca2+] . 10 . 7554/eLife . 11066 . 012Figure 5 . Depolarization increases ciliary [Ca2+] , but not ciliary beat frequency or fluid velocity . ( A , B ) Example pseudocolor images of a cell expressing the ratiometric sensor mSSTR3-mCherry-GCaMP6s in motile cilia before and during perfusion of 140 K+ ( A ) and quantification of ciliary F_GCaMP6s/ F_mCherry ratio changes in response to 140 K+ ( B , n = 15 cells ) . Ionomycin ( 1 μM ) was applied as control stimulus . ( C , D ) Example pseudocolor images of a cell pre-treated with nimodipine ( 10 μM , 2 min , C ) and quantification of ratio changes in response to a depolarizing stimulus ( 140 K+ ) after CaV channel block by 10 µM nimodipine ( D , n = 11 cells ) . ( E ) Ciliary beat frequency of cultured ependymal cells was not substantially altered during perfusion of depolarizing [K+] solutions ( 70 or 140 K+ , n = 5 coverslips ) . The black line in the DIC image ( left ) indicates the position of the line used to derive the kymographs . Kymographs were analyzed at the indicated time points ( duration , 1 s ) . ( F ) Example image of a brain slice showing frame by frame position of tracked beads along the lateral ventricle . Dotted lines indicate the area in which beads were tracked ( <30 μm from surface ) . The arrow marks the direction of fluid flow . ( G ) Bead velocities measured under the conditions indicated . The mean is indicated by the black line and the open circles ( gray ) represent the velocities of all tracked beads for each condition ( 3 K+: n = 6 slices , 21 beads; ATP ( 100 μM ) : n = 4 slices , 13 beads; 40 K+: n = 2 slices , 10 beads; 70 K+: n = 5 slices , 16 beads; 70 K+ + 10 μM nimodipine ( 70 N ) : n = 5 slices , 14 beads ) . Error bars; ± SEM . Scale bars , 5 μm ( A , C , E ) and 10 μm ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11066 . 01210 . 7554/eLife . 11066 . 013Figure 5—figure supplement 1 . Depolarization increases ciliary [Ca2+] , but not ciliary beat frequency or fluid velocity . ( A ) Average cell ( gray ) and motile cilia ( green ) membrane potential in varying bath [K+] . ( B ) Example current clamp recording from an ependymal cell body . ( C , D ) Motile cilia F_GCaMP6s/F_mCherry ratio changes in response to 40 K+ ( C , n = 6 cells ) and 70 K+ ( D , n = 7 cells ) . Ionomycin ( 1 μM ) was applied as control stimulus . ( E ) Motile cilia F_GCaMP6s/F_mCherry ratio changes in response to 140 K+ in externally Ca2+-chelated solution ( no added Ca2+ , 0 . 5 mM EGTA , n = 6 ) . ( F ) Quantified changes in cytoplasmic [Ca2+] in response to 140 K+ ( n = 22/23 , 3 coverslips ) . Ependymal cells were loaded with the Ca2+ dye , OGB-1 ( 2 μM ) . Error bars; ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 11066 . 013 Previous studies suggest that Ca2+ increases the beating frequency of motile cilia in airways and Fallopian tubes ( Di Benedetto et al . , 1991; Girard and Kennedy , 1986; Lansley et al . , 1992; Schmid and Salathe , 2011; Verdugo , 1980 ) . Nguyen et al , reported a serotonin-mediated Ca2+-dependent increase in ciliary beating frequency of ependymal motile cilia ( Nguyen et al . , 2001 ) . We thus examined ciliary beating in response to depolarizing external [K+] by plotting kymographs of individual cilia ( Lechtreck et al . , 2008 ) . We observed no significant difference in the beating frequency during perfusion of depolarizing external [K+] solution ( aCSF 70 K+ or aCSF 140 K+ ) . For example , cilia beat frequency was initially 9 . 7 ± 1 Hz ( 0–1 s ) and was unchanged by perfusion with depolarizing high [K+] solution ( 9 . 7 ± 1 Hz at 15–16 s and 9 . 3 ± 0 . 8 Hz at 29–30 s; Figure 5E , see Materials and methods ) . For comparison cilia beat frequency during continuous perfusion of standard external solution ( aCSF 3 K+ ) was 9 . 8 ± 1 Hz ( 0–1 s ) , 10 ± 1 Hz ( 15–16 s ) , and 10 ± 1 Hz ( 29–30 s ) . Kymographs depicting the beat cycles of individual cilia additionally revealed that cilia smoothly transitioned from a forward to a backward stroke independent of external [K+] ( Figure 5E ) . While ciliary beating was unaffected in vitro , we next asked whether activation of CaV channels by a depolarizing stimulus would disturb fluid movement in acute slices of the lateral ventricles . We assessed the velocity of polystyrene beads ( 500 nm ) in proximity to the apical surface ( <30 μm; Figure 5F ) . In agreement with previous studies , we determined an average velocity of 86 ± 5 μm/s in standard bath solution ( aCSF 3 K+ ) ( Lechtreck et al . , 2008 ) . ATP application ( 100 µM ) slightly increased average bead velocity ( 96 ± 8 . 4 μm/s ) , but was not statistically significantly different . In increasing external [K+] , we observed a reduction in average velocity , which was reduced by preincubation with 10 µM nimodopine ( 79 ± 3 μm/s , aCSF 40 K+; 67 ± 5 . 5 μm/s , aCSF 70 K+; 80 ± 4 . 3 μm/s , aCSF 70 K+ + nimodipine; Figure 5G ) . Note , however , that the velocity of individual beads varied considerably ( Figure 5G ) . We cannot rule out an effect of CaV channels in governing overall motile cilia function , but if it exists , is small and variable .
The 9 + 2 structural arrangement of microtubules and inherent dynein ATPase-activity which drives ciliary motion are conserved among motile cilia from protist to humans . However , Ca2+ modification of cilia function in mammalian ependymal cells is quite distinct from that in protists ( Inaba , 2015; Satir and Christensen , 2007 ) . While beat reversal , and changes in waveform critically depend on the activity of ciliary CaV channels in Chlamydomonas and Paramecium ( Fujiu et al . , 2009; Kung and Naito , 1973; Matsuda et al . , 1998 ) , we found little evidence linking Ca2+ in ependymal cell bodies or motile cilia to their function . Moreover , unlike primary cilia in which specialized ciliary channels ( polycystins ) predominate ( DeCaen et al . , 2013; Delling et al . , 2013 ) , we found that CaV channels in the cell body primarily determine changes in motile cilia [Ca2+] . Although we could not accurately quantify the relative CaV channel densities in motile cilia compared to the cell body or test the possibility of non-uniform channel distribution along the cilium shaft , these channels were not enriched at the ciliary tip . Whether sparsely ciliated cells are less well differentiated or represent a subclass of ciliated ependymal cells is unclear , but the low abundance of CaV channels revealed by patch clamping of these cilia is consistent with a cytoplasmic control of ciliary [Ca2+] in imaging experiments from multiciliated cells . We thus conclude that ependymal cilia have distinct electrical properties as compared to flagella ( cilia ) that are used for locomotion in mammalian sperm ( Kirichok et al . , 2006; Miki and Clapham , 2013; Qi et al . , 2007; Ren et al . , 2001 ) and protists ( Beck and Uhl , 1994; Dunlap , 1977; Fujiu et al . , 2009; Kung and Naito , 1973; Matsuda et al . , 1998 ) . Based on direct measurements of cell body and motile cilia membrane potentials , and voltage-activated CaV channels in both compartments , we hypothesize that the motile cilium compartment is passively coupled and primarily controlled by channel activity and [Ca2+] in the cell body , rather than being independently regulated . Thus , a cellular response would promote simultaneous signaling to all cilia , perhaps slowly entraining the cilia and enabling signal propagation to neighboring cells ( Sanderson et al . , 1988; Schmid and Salathe , 2011 ) . Unlike primary cilia , ependymal cilia had only slightly elevated resting [Ca2+] , thus enabling smaller changes in cytoplasmic Ca2+ to equilibrate within cilia . Indeed , membrane depolarization triggered activation of CaV channels , which resulted in a robust change in ciliary [Ca2+] . While we could not differentiate the onset of the Ca2+ response in the cilium from that in the cell body , the results suggest that ciliary [Ca2+] follows cytoplasmic [Ca2+] . Previous studies suggest that different sources of Ca2+ ( internal stores , cell membrane flux ) can modify the beat frequency of mammalian motile cilia ( Di Benedetto et al . , 1991; Lansley et al . , 1992; Nguyen et al . , 2001; Salathe and Bookman , 1995; Schmid and Salathe , 2011 ) . ATP-activated P2X7 receptor-mediated Ca2+ influx made a minor contribution to the ATP-induced increase in beating frequency of ependymal cilia , while activation of the adenosine A2B receptor made a more significant change , perhaps via slower G protein-dependent pathways ( Genzen et al . , 2009b ) . In contrast , we observed that membrane depolarization , or application of ATP , did not significantly alter ependymal motile cilia function . We observed only a slight decrease in bead velocity in the lateral ventricles of acute brain slices that correlated with depolarization and CaV channel activation . Since we did not detect a change in ciliary beating frequency in in vitro cultures of ependymal cells , we suspect that coupling to other cell types or factors mediates the observed reduction . The variability in velocities of individual beads , however , leads us to question the physiological relevance of the observed changes . Finally , although previous studies suggest that ependymal cell differentiation in vitro reflects in vivo conditions ( i . e . postnatal maturation into multiciliated cells [El Zein et al . , 2009; Guirao et al . , 2010; Spassky et al . , 2005] ) , we cannot exclude the possibility of altered channel expression . Since we did not observe significant changes in ciliary beating in response to membrane depolarization in either acute brain slices or in vitro cultures , we suspect that control mechanisms are similar . As molecular oars , 9 + 2 motile cilia power fluid flow across the surface of epithelia . Defects in cilia motility are linked to primary ciliary dyskinesia ( PCD ) and manifest in disease conditions such as chronic sinusitis , male infertility , and hydrocephalus ( Ibanez-Tallon et al . , 2003; Roy , 2009 ) . Here we have examined the electrophysiological and Ca2+ signaling properties of motile cilia in ependymal cells . We conclude that they are quite distinct from those of primary cilia ( DeCaen et al . , 2013; Delling et al . , 2013 ) and the flagella that power motile cells . First , membrane potential and [Ca2+] closely follow those in the cytoplasm . This stems naturally from a relative paucity of ion channels in the cilia compared to the nonselective channels in primary cilia ( DeCaen et al . , 2013 ) . Second , voltage-dependent calcium channels change [Ca2+] in the cytoplasm that propagates readily into the motile cilia , but only marginally alters beat frequency or fluid flow . These findings , although they cannot necessarily be generalized to all motile cilia , suggest that CaV channels in ependymal cells primarily serve other secretory or differentiation roles in the cell body .
Transgenic Arl13B-EGFPtg ( Delling et al . , 2013 ) and C57BL/6 ( wild type ) mice were used in this study . Animal research protocols were approved by the IACUC of Boston Children’s Hospital . Paraformaldehyde ( PFA , 4% ) fixed frozen sections of brain ventricles ( from a juvenile male ) and cultured ependymal cells were permeabilized with 0 . 1–0 . 5% Triton X-100 , blocked in 5% NGS , 1% BSA , 0 . 1% Triton X-100 , 0 . 05% Tween20 , and incubated with unconjugated Fab fragment goat anti-mouse ( Jackson Immuno Research ) to block endogenous mouse IgGs . Primary antibodies ( goat anti-GFP FITC conjugate , Novus Biologicals or rabbit anti-GFP Alexa Fluor 488 conjugate , Life Technologies; rat anti-mCherry , Life Technologies; mouse anti-acetylated tubulin , Sigma-Aldrich; rabbit anti-Spag6 , Sigma-Aldrich ) were applied overnight at 4°C . Sections and cells were washed in PBS-T and incubated with secondary antibody ( Alexa Fluor 568 or 647 conjugates , Life Technologies and Jackson Immuno Research ) . Nuclei were stained with Hoechst 33342 ( Life Technologies ) and sections/cells mounted in Prolong Diamond Antifade ( Molecular Probes ) . Cells were cultured on glass coverslips ( VWR ) or gridded glass-bottom dishes ( Ibidi ) for staining . mIMCD3 cells were fixed and stained accordingly and images acquired with settings in the range of the acquisition settings used for Spag6-stained ependymal cells . All images were acquired with a FluoView1000 confocal microscope ( Olympus ) and processed using ImageJ ( NIH ) . Expression clones were created by cloning mSSTR3-mCherry-GCaMP6s into pENTR 3C and subsequent LR recombination with the destination vector pAD/CMV/V5-DEST vector ( Life Technologies ) . Glycine-serine linkers were introduced between mSSTR3-mCherry and mCherry-GCaMP6s as described previously ( Delling et al . , 2013 ) . The expression vector was digested with PacI and transfected into HEK293 AD cells . The crude adenoviral stock was further amplified by infection of HEK293 cells . Cultured ependymal cells ( derived from wild type pups ) were typically transduced with the recombinant adenoviral vector ( pAD-mSSTR3-mCherry-GCaMP6s ) at DIV6 or 8 and functional expression of the sensor in motile cilia assessed at DIV10 to 12 . Cilia motility and fluid flow were assessed at 60x magnification on an inverted Olympus IX70 microscope . Images were acquired with an OrcaFlash 4 . 0 high speed camera and HCImageLive imaging software ( Hamamatsu ) . RNA was isolated from ependymal cell cultures and dissected tissue ( brain , heart , eyes and skeletal muscle ) using TRIzol reagent ( Life Technologies ) . Ependymal cell cultures were derived from a transgenic Arl13B-EGFPtg mouse pup and tissue collected from a 2 week-old heterozygous Arl13B-EGFPtg male . cDNA was synthesized from total RNA using Superscript III ( Life Technologies ) . The reverse transcriptase was omitted from the cDNA synthesis reaction in controls for cultured ependymal cells . PCR products were amplified with Phusion high-fidelity DNA polymerase ( NEB ) using a touchdown PCR protocol with decreasing annealing temperatures ( 3 cycles at 68 , 65 , and 62 degrees; and 25 cycles at 58 degrees ) . The following intron-spanning primers were used: CaV1 . 1 ( Cacna1s ) , fw 5’ ATGACAACAACACTCTGAACCTC 3’ and rv 5’ GGAAGCCGTAGGCTATGATCT 3’ ( PrimerBank ID 189409135c2 ) ; CaV1 . 2 ( Cacna1c ) , fw 5’ CCTGCTGGTGGTTAGCGTG 3’ and rv 5’ TCTGCCTCCGTCTGTTTAGAA 3’ ( PrimerBank ID 192322a1 ) ; CaV1 . 3 ( Cacna1d ) , fw 5’ GCTTACGTTAGGAATGGATGGAA 3’ and rv 5’ GAAGTGGTCTTAACACTCGGAAG 3’ ( PrimerBank ID 134288874c2 ) ; CaV1 . 4 ( Cacna1f ) , fw 5’ TACTAATCCCATTCGTCGGTCC 3’ and rv 5’ CATAGGCTACGATCTTGAGCAC 3’ ( PrimerBank ID 115648150c2 ) ; GAPDH , fw 5’ TGGCCTTCCGTGTTCCTAC 3’ and rv 5’ GAGTTGCTGTTGAAGTCGCA 3’ ( PrimerBank ID 126012538c3 ) . A 2-Log DNA ladder ( NEB ) and PCR samples were loaded on 1% NuSieve 3:1 agarose gels ( Lonza ) and images acquired with the Syngene GBox system and processed using ImageJ . The PCR reactions for CaV1 . 1 and its controls were run separately and a larger volume ( 1 . 5x ) was loaded on the gel ( weak amplification of CaV1 . 1 transcripts from control tissue ) . | Certain specialized cells in the brain , airways and Fallopian tubes have large numbers of hair-like structures called motile cilia on their surface . By beating in a synchronized manner , these cilia help to move fluids across the surface of the cells: for example , cilia on lung cells beat to clear mucus away , while those in the brain help the cerebrospinal fluid to circulate . Motile cilia in mammals are structurally similar to the flagella that propel sperm cells and certain single-celled organisms around their environments . These flagella have specialized pore-forming proteins called ion channels in their membrane through which calcium ions can move . This flow of calcium ions controls the beating of the flagella . However , it is unclear whether a similar movement of calcium ions across the cilia membrane regulates motile cilia beating in mammals . Doerner et al . have now used a method called patch clamping to study the movement of calcium ions across the membrane of the motile cilia found on a particular type of mouse brain cell . This revealed that unlike flagella , these motile cilia have very few voltage-gated calcium channels; instead , the vast majority of these ion channels reside in the main body of the cell . Furthermore , the level of calcium ions in the motile cilia follows changes in calcium ion levels that originate in the cell body . Overall , Doerner et al . demonstrate that the activity of voltage-gated calcium channels does not control the beating rhythm of the motile cilia in the mouse brain or how quickly the fluid above the cell surface moves . Future work should investigate whether this is also the case for the cells that line the trachea and Fallopian tubes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2015 | Ion channels and calcium signaling in motile cilia |
Mammalian cardiomyocytes become post-mitotic shortly after birth . Understanding how this occurs is highly relevant to cardiac regenerative therapy . Yet , how cardiomyocytes achieve and maintain a post-mitotic state is unknown . Here , we show that cardiomyocyte centrosome integrity is lost shortly after birth . This is coupled with relocalization of various centrosome proteins to the nuclear envelope . Consequently , postnatal cardiomyocytes are unable to undergo ciliogenesis and the nuclear envelope adopts the function as cellular microtubule organizing center . Loss of centrosome integrity is associated with , and can promote , cardiomyocyte G0/G1 cell cycle arrest suggesting that centrosome disassembly is developmentally utilized to achieve the post-mitotic state in mammalian cardiomyocytes . Adult cardiomyocytes of zebrafish and newt , which are able to proliferate , maintain centrosome integrity . Collectively , our data provide a novel mechanism underlying the post-mitotic state of mammalian cardiomyocytes as well as a potential explanation for why zebrafish and newts , but not mammals , can regenerate their heart .
The adult mammalian heart is considered to be a post-mitotic organ , as mammalian cardiomyocytes lose their ability to proliferate shortly after birth ( Li et al . , 1996; Soonpaa et al . , 1996; Zebrowski and Engel , 2013 ) . This is supported by the limited regenerative capacity of the adult mammalian heart ( Senyo et al . , 2014 ) and the fact that primary adult cardiomyocyte-born tumors are extremely rare , if they exist at all ( Dell'Amore et al . , 2011 ) . Understanding the underlying mechanisms governing the post-mitotic state of adult mammalian cardiomyocytes may clarify whether it is possible to induce cardiac regeneration based on cardiomyocyte proliferation as seen in zebrafish and newts ( Poss et al . , 2002; Bettencourt-Dias et al . , 2003; Gamba et al . , 2014 ) . Tremendous efforts have been invested to induce postnatal mammalian cardiomyocyte proliferation . However , besides the fact that cell cycle promoting factors ( e . g . , Cyclins ) are downregulated around birth , while cell cycle inhibitors ( e . g . , Cyclin-dependent kinase inhibitors ) are induced ( Ikenishi et al . , 2012 ) , little is known about the mechanisms that induce cell cycle exit or establish the post-mitotic state in mammalian cardiomyocytes ( van Amerongen and Engel , 2008 ) . The centrosome is a solitary , juxtanuclear organelle in metazoan cell-types . It consists of a pair of tubulin-based structures , called centrioles , encased in a dense , non-membranous , multi-protein cloud called the pericentriolar matrix ( PCM ) , which is further surrounded by a dispersed array of proteins termed centriolar satellites ( Bettencourt-Dias , 2013 ) . Traditionally , the centrosome is known as the microtubule organizing center ( MTOC ) of the cell , and is required for primary cilium formation ( Kim and Dynlacht , 2013 ) . Recently , the centrosome has emerged as a critical signaling hub for the cell cycle regulatory machinery ( Doxsey et al . , 2005 ) . For instance , centrosome localization of Cyclin E and Cyclin A is required for G1/S cell cycle progression ( Pascreau et al . , 2011 ) . Consistent with this role , increasing evidence supports the requirement of a functional centrosome for proliferative potential in mammalian cell-types ( Doxsey et al . , 2005 ) . For instance , cells undergo G0/G1 arrest when ( i ) centrosomes are removed via laser ablation , ( ii ) centrosome integrity is disrupted via knockdown of centrosome proteins , or ( iii ) centriole biogenesis is blocked via a chemical inhibitor ( Hinchcliffe et al . , 2001; Khodjakov and Rieder , 2001; Srsen et al . , 2006; Mikule et al . , 2007; Wong et al . , 2015 ) . To date , the role of centrosome integrity in cell proliferation has always been studied in the context of centrosome component mutants . Here , we show that centrosome integrity is developmentally regulated in mammalian cardiomyocytes , revealing a novel mechanism that renders cells post-mitotic . Our findings might have important implications for efforts to induce therapeutic cardiomyocyte proliferation in adult mammalian hearts .
A normal diploid cell in G0/G1-phase contains one centrosome with two paired centrioles . During S-phase the centrosome duplicates whereby the two parental centrioles form daughter centrioles . Around the transition from G2-phase to mitosis , parental centrioles ‘split’ ( loss of cohesion ) resulting in two separated centrosomes that become part of the spindle poles during mitosis ( Doxsey et al . , 2005 ) . When centrosome integrity is compromised during G0/G1-phase , centrioles can lose cohesion ( i . e . , become unpaired ) adopting a premature ‘split’ configuration , defined as a distance > 2 µm between centrioles ( Graser et al . , 2007 ) . To determine if centrosome integrity changes in cardiomyocytes during development , we analyzed centriole configuration using antibodies against γ-tubulin , a marker of both the centriole and the PCM ( Sonnen et al . , 2012 ) . Centrioles were observed to be in a typical paired configuration in the majority of cultured cardiomyocytes isolated from embryonic day ( E ) 15 , E18 , and postnatal day ( P ) 0 rat hearts ( Figure 1A , B ) as well as in cardiomyocytes from E15 and P0 rat heart sections ( Figure 1—figure supplement 1A , B ) . In contrast , shortly after birth , centrioles were split in the vast majority of P3 and P5 cardiomyocytes in vitro ( Figure 1A , B ) and in vivo ( Figure 1—figure supplement 1A , B ) . Analysis of mother and daughter centriole markers ( Odf2 and Centrobin , respectively ) in isolated rat cardiomyocytes verified that each γ-tubulin signal represented a single centriole ( Figure 1—figure supplement 1C ) . As an internal control , cardiac non-myocytes from the same cultures and tissue sections were examined . The majority of non-myocytes in the heart are fibroblasts , endothelial cells , and smooth muscle cells that all have the ability to proliferate . At all examined developmental stages , non-myocytes showed a typical paired-centriole configuration indicating that the split-centriole phenotype was cardiomyocyte-specific ( Figure 1B and Figure 1—figure supplement 1A , B ) . Collectively , these data demonstrate that centriole cohesion , and thus centrosome integrity , is lost in mammalian cardiomyocytes shortly after birth . 10 . 7554/eLife . 05563 . 003Figure 1 . Loss of centrosome integrity during heart development . ( A ) Analysis of centriole ( γ-tubulin ) configuration in E15- or P5-isolated ventricular rat cardiomyocytes ( Troponin I ) . Nuclei: DAPI . PC: paired-centrioles . SC: split-centrioles . Scale bar: 5 µm . ( B ) Frequency of cells with paired-centrioles during development . Bi: Binucleated . ( C ) Analysis of the localization of the centrosome proteins PCM1 , PCNT ( Pericentrin ) , and CEP135 in isolated cardiomyocytes . ( D ) PCNT localization frequency in cardiomyocytes isolated from different developmental stages . ( E ) Centrosomal PCNT signal intensity in P3-isolated cardiomyocytes with paired- and split-centrioles relative to E15-isolated cardiomyocytes . ( F ) Frequency of paired-centrioles in P0-isolated cardiomyocytes after siRNA-mediated Pcnt knockdown . scr: scrambled . ( G ) Representative images of the analysis in ( F ) . ( H ) PCM1 localization frequency in cardiomyocytes isolated from different developmental stages . ( I ) Analysis of PCM1 and PCNT localization in E15-isolated cardiomyocytes cultured for either 1 or 8 days . ( J ) Frequency of P0-isolated cardiomyocytes with paired-centrioles after 1 day , 3 days , or 6 days in culture . ( K ) RT-PCR analysis of Pcnt B and S isoform expression during rat heart development in vivo . ( L ) Localization of PCNT isoforms . P3-isolated cardiomyocytes immunostained with antibodies against either both PCNT B and S isoforms or only the PCNT B isoform . Yellow arrows: cardiomyocyte nuclei . Red arrows: non-myocyte nuclei . Unless otherwise noted , scale bars: 10 µm; red arrowheads: centrioles; data are mean ± SD , n = 3 , *: p < 0 . 05 . For the experiments ≥ 10 cells ( E ) , ≥ 50 cells ( B , F , J ) , ≥ 100 ( D , H ) cells were analyzed per experimental condition . DOI: http://dx . doi . org/10 . 7554/eLife . 05563 . 00310 . 7554/eLife . 05563 . 004Figure 1—figure supplement 1 . Loss of centrosome integrity during heart development . ( A ) Representative images of centrosomes ( γ-tubulin ) in heart cryosections of P0 rat heart ventricles . Nuclei: DAPI . Cardiac nuclei: Nkx2 . 5 . Arrowheads indicate centrioles . γ-tubulin signals separated by a distance greater than 2 µm were considered as singlets . Scale bars: 10 µm . ( B ) Quantitative analysis of centriole signals and configurations in cardiomyocytes and non-myocytes from cyrosections of E15- , P0- , P3- , P5 , or adult ( 2 months ) rat heart ventricles . CM: cardiomyocyte . NCM: non-myocyte . Results are from three independent animals , data are mean ± SD , ≥ 100 cells were analyzed per experimental condition , *: p < 0 . 05 ( refers to doublets ) . ( C ) Representative images of P3-isolated cardiomyocytes immunostained for mother centriole ( Odf2 ) and daughter centrioles ( Centrobin ) ( green arrowheads ) . SC: split-centrioles . Colored ratios equal daughter or mother centriole: all centrioles . ( D ) Analysis of the localization of the centrosome proteins PCM1 , PCNT ( Pericentrin ) , and CEP135 in isolated cardiomyocytes . Red arrowheads indicate centrioles . ( E ) Representative images of Cdk5Rap2 localization in E15- and P3-isolated rat cardiomyocytes . Yellow arrowhead: Cdk5Rap2 at the nuclear envelope . ( F ) Quantitative analysis of centriolar Cdk5Rap2 signal intensity in E15- and P3-isolated cardiomyocytes . Data are mean ± SD , n = 3 , ≥ 10 cells were analyzed per experimental condition , *: p < 0 . 05 . ( G ) Representative images of PCM1 localization in E15-isolated cardiomyocytes ( Troponin I ) . Centrioles: γ-tubulin . DOI: http://dx . doi . org/10 . 7554/eLife . 05563 . 00410 . 7554/eLife . 05563 . 005Figure 1—figure supplement 2 . Loss of centrosome integrity during heart development . ( A ) Localization of Pericentrin ( PCNT ) B-GFP in non-myocytes and cardiomyocytes from different developmental stages . Cardiomyocyte ( Troponin I ) , centriole ( γ-tubulin ) . Arrowheads indicate centrioles . Yellow asterisk: cardiomyocytes . Blue asterisk: non-myocyte . ( B ) Frequency of PCNT B-GFP-positive centrioles in non-myocytes ( NCM ) and cardiomyocytes ( CM ) from different developmental stages . ≥ 50 cells pooled from several experiments were analyzed per time point . ( C ) Representative images of centriole configuration and PCNT and PCM1 localization in mouse iPSC-derived cardiomyocytes . Yellow scale bars: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05563 . 005 To identify an underlying cause of the split-centriole phenotype , the cellular localization of various centrosome proteins was assessed in isolated cardiomyocytes from different developmental stages . The PCM proteins Pericentrin and Cdk5Rap2 have previously been shown to be required for centriole-cohesion ( Graser et al . , 2007; Matsuo et al . , 2010 ) . Consistent with this function , both PCM proteins localized to the centrosome in E15-isolated cardiomyocytes ( Figure 1C , D and Figure 1—figure supplement 1D , E ) . In contrast , both proteins were localized to the nuclear envelope in P3-isolated cardiomyocytes ( Figure 1C , D and Figure 1—figure supplement 1D , E ) . Although remnants of Pericentrin and Cdk5Rap2 could be observed at the centriole in P3-isolated cardiomyocytes ( Figure 1C , D and Figure 1—figure supplement 1D , E ) , their presence was significantly reduced when centrioles were split ( Figure 1E and Figure 1—figure supplement 1F ) . Pcnt siRNA-mediated knockdown in P0-isolated cardiomyocytes resulted in an increase of split-centrioles ( Figure 1F , G ) , confirming that Pericentrin is required for centriole-cohesion in cardiomyocytes . In contrast to PCM proteins , the centriole-associated proteins CEP135 , Odf2 , and Centrobin were not observed at the nuclear envelope in P3-isolated cardiomyocytes ( Figure 1C and Figure 1—figure supplement 1C , D ) . Collectively , these results indicate that loss of centriole cohesion is accompanied by redistribution of centrosome proteins to the nuclear envelope . Centrosome localization of Pericentrin is dependent on the centriole satellite protein PCM1 ( Dammermann and Merdes , 2002; Barenz et al . , 2011 ) . Consistent with this , PCM1 localized at the centrosome and the nuclear envelope in E15- and P3-isolated cardiomyocytes , respectively ( Figure 1C , H and Figure 1—figure supplement 1D ) . Further , nuclear envelope localization of PCM1 occurred by P0 ( Figure 1H ) , prior to that of Pericentrin ( Figure 1D ) . Occasionally , PCM1 was observed in a semi-belt pattern at the nuclear envelope proximal to the centrosome in E15-isolated cardiomyocytes ( Figure 1—figure supplement 1G ) , indicating a transitional state . These data suggest that in cardiomyocytes , PCM1 is also required for the localization of Pericentrin to the centrosome . To test this , we overexpressed the major isoform of Pericentrin associated with the centrosome , Pericentrin B , in E15- , P0- , and P3-isolated cardiomyocytes and non-myocytes . Pericentrin B-GFP localized to centrioles in non-myocytes as well as E15-isolated cardiomyocytes ( Figure 1—figure supplement 2A , B ) . In contrast , Pericentrin B-GFP did not localize to the centrioles in the vast majority of P0- and P3-isolated cardiomyocytes but rather created cellular aggregates which were not observed in E15 cardiomyocytes or non-myocytes . These data indicate that the machinery required for centrosome integrity is lost in postnatal cardiomyocytes . To determine if relocalization of centrosome proteins to the nuclear envelope and loss of centriole-cohesion is caused by cell autonomous mechanisms , long-term culturing experiments were performed . Long-term culturing of E15-isolated and P0-isolated cardiomyocytes resulted in PCM1 and Pericentrin relocalization to the nuclear envelope ( Figure 1I ) and loss of paired-centrioles ( Figure 1J ) , respectively . Further , mouse iPSC-derived cardiomyocytes had split-centrioles and the centrosome proteins PCM1 and Pericentrin localized to the nuclear envelope ( Figure 1—figure supplement 2C ) . These results indicate that loss of centrosome integrity begins during fetal development and progresses in a cell autonomous manner . Previously , it has been demonstrated that Pcnt is alternatively spliced ( Miyoshi et al . , 2006 ) , resulting in two isoforms; Pericentrin B which resembles the human Pericentrin Kendrin , and Pericentrin S , which lacks an N-terminus region . Pericentrin B is ubiquitously expressed at all developmental stages . In contrast , expression of Pericentrin S starts late in fetal development and was found to be specific for adult heart and skeletal muscle ( Miyoshi et al . , 2006 ) . Therefore , we speculated that expression of the Pericentrin S isoform may coincide with changes in centrosome integrity during the development of cardiomyocytes . RT-PCR analysis revealed the appearance of Pcnt S expression shortly before birth in the heart ( Figure 1K ) . Antibodies specific for the Pericentrin B isoform indicated that Pericentrin S , and not Pericentrin B , was the predominant Pericentrin isoform at the nuclear envelope in P3-isolated cardiomyocytes ( Figure 1L ) . These results indicate that Pcnt is alternatively spliced during perinatal heart development and this change is related to its relocalization to the nuclear envelope . Loss of centrosome integrity in postnatal cardiomyocytes suggested that centrosome function might be compromised as well . Nearly all cell-types studied to date are capable of forming a primary cilium when arrested in G0/G1-phase ( Bowser and Wheatley , 2000 , Nigg and Stearns , 2011 ) . This suggests that cardiomyocytes , which are arrested shortly after birth in G0/G1 phase ( Takeuchi , 2014 ) , should be capable of ciliogenesis . Yet , given that PCM1 , which is required for ciliogenesis ( Kim et al . , 2008 ) , is lost from the centrosome during development , we speculated that ciliogenesis is suppressed in cardiomyocytes . Induction of ciliogenesis via serum starvation resulted in the formation of a primary cilium in E15-isolated cardiomyocytes and non-myocytes ( Figure 2A ) . However , the frequency of cardiomyocytes capable of ciliogenesis decreased with heart development ( Figure 2B ) with less than 1% of isolated postnatal cardiomyocytes , which were arrested in G1/G0 as confirmed by lack of Ki67 expression and FACS analysis ( Figure 2—figure supplement 1A–D ) , forming a primary cilium . Further , no postnatal binucleated cardiomyocytes were observed to be capable of ciliogenesis ( Figure 2A ) . In contrast to cardiomyocytes , the frequency of cardiac non-myocytes capable of ciliogenesis was high at all developmental stages investigated ( Figure 2B ) . 10 . 7554/eLife . 05563 . 006Figure 2 . Loss of centrosome function during heart development . ( A ) Identification of primary cilium ( Arl13b ) in E15- and P5-isolated cardiomyocytes ( Troponin I ) and non-myocytes . Centrioles: γ-tubulin . White circle: nuclei . ( B ) Frequency of ciliated cardiomyocytes and non-myocytes isolated from hearts at different developmental stages . Data are mean ± SD , n = 3 , *: p < 0 . 05 . ≥ 200 cells were analyzed for each condition . ( C ) Localization of the cellular MTOC ( Pericentrin [PCNT] ) in E15- and P3-isolated cardiomyocytes . Microtubules: acetylated-α-tubulin; yellow arrowheads: PCNT-positive MTOC . ( D ) Localization of microtubule regrowth ( yellow arrowhead ) . E15- or P3-isolated cardiomyocytes treated with nocodazole or nocodazole followed by wash-out . ( E ) Analysis of microtubule regrowth at centrioles . P3-isolated cardiomyocytes were treated as in ( D ) . Yellow arrowheads: Cdk5Rap2-positive centrioles in non-myocytes; White arrowheads: Cdk5Rap2-positive centrioles in cardiomyocytes . ( F ) Pericentrin is required for a functional MTOC . P3-isolated cardiomyocytes transfected with scrambled or PCNT siRNAs and analyzed for microtubule formation . Scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05563 . 00610 . 7554/eLife . 05563 . 007Figure 2—figure supplement 1 . Loss of centrosome function during heart development . ( A ) Representative images of postnatal cardiomyocytes stained for cell cycle activity ( Ki67 ) and cardiomyocytes ( α-actinin ) including a quantitative analysis for Ki67-positive cardiomyocytes . Data are mean ± SD , n = 3 . ≥ 200 cells were analyzed per experimental condition . Nuclei: DAPI . Scale bars: 20 µm . ( B ) Quantitative analysis of the purity of cardiomyocyte cultures based on α-actinin . Data are mean ± SD , n = 3 . ≥ 150 cells were analyzed per experimental condition . ( C ) Cell cycle analysis of cardiomyocytes based on DNA content by FACS analysis . Peaks: G0/G1 ( green ) , S-phase ( yellow ) , G2/M ( blue ) . ( D ) Quantitative analysis of ( C ) . Data are mean ± SD , n = 3 . Per experimental condition 10 , 000 cells were analyzed . ( E ) Localization of microtubule regrowth in non-myocytes of P3-isolated cardiomyocyte cultures treated with nocodazole or nocodazole followed by wash-out . Scale bars: 10 µm . ( F ) Frequency of MTOC localization in cardiomyocytes ( CM ) and non-myocytes ( NCM ) from different developmental stages after nocodazole wash out . n = 3 , ≥ 50 cells were analyzed per experimental condition . ( G ) Frequency and extent of centriole-based microtubule formation in P3-isolated CM and NCM after nocodazole wash out . > 50 cells were analyzed per condition . ( H ) Representative images of centriole-based microtubule formation as quantitated in ( G ) . Microtubules ( acetylated-α-tubulin ) , Centrioles ( Cdk5Rap2 ) . Arrowheads indicate centrioles . ( I ) Relative frequency of cardiomyocytes with nuclear envelope-based MTOC after Pcnt siRNA-mediated knockdown . Scr: Scramble siRNA . ***: p < 0 . 005 . n = 3 , ≥ 100 cells were analyzed per experimental condition . DOI: http://dx . doi . org/10 . 7554/eLife . 05563 . 007 Pericentrin and Cdk5Rap2 are required for the centrosome to function as the cellular MTOC ( Takahashi et al . , 2002; Choi et al . , 2010 ) . As these proteins localize to the nuclear envelope during neonatal development , we hypothesized that the cellular MTOC is transferred from the centrosome to the nuclear envelope . In accordance with this hypothesis , microtubules were found to predominantly emanate from the centrosome in E15-isolated cardiomyocytes as also observed in non-myocytes . In contrast , in P3-isolated cardiomyocytes microtubules were found to predominantly emanate from the nuclear envelope ( Figure 2C ) . The local shift of the MTOC during development was confirmed by a MTOC-regrowth assay ( Figure 2D and Figure 2—figure supplement 1E , F ) , which further demonstrated that microtubules do not emanate from centrioles in P3-isolated cardiomyocytes ( Figure 2E and Figure 2—figure supplement 1G , H ) . Finally , siRNA-mediated knockdown of Pcnt in P3-isolated cardiomyocytes demonstrated that Pericentrin is required for a functional MTOC at the nuclear envelope ( Figure 2F and Figure 2—figure supplement 1I ) . Collectively , these results indicate that , in addition to centrosome integrity , centrosome function is progressively compromised in cardiomyocytes during development . Given that traditional centrosome functions of ciliogenesis and microtubule organization are lost in postnatal cardiomyocytes , we hypothesized that there would be a relationship between centrosome integrity and proliferative potential . To test this , the cell cycle marker Ki67 was assessed in cardiomyocytes from different developmental stages . Cardiomyocyte proliferative potential decreased with neonatal development ( Figure 3A ) . Moreover , P3-isolated cardiomyocytes with paired-centrioles exhibited greater proliferative potential than those with split centrioles ( Figure 3B , C ) at a frequency similar to that of E15- and P0-isolated cardiomyocytes ( Figure 3A , C ) —of which the vast majority have paired-centrioles ( Figure 1B ) . In contrast , cardiac non-myocyte proliferative potential , as well as the percentage of non-myocytes with paired centrioles , did not decrease during neonatal development ( Figure 3—figure supplement 1A ) . Subsequently , we tested whether centrosome integrity is required for cardiomyocyte proliferative potential . Centrosome integrity was disrupted by either siRNA-mediated knockdown of Pcnt , or overexpression of a RFP-tagged dominant negative C-terminal Pericentrin ( RFP-PeriCT ) , which displaces endogenous Pericentrin localization ( Gillingham and Munro , 2000; Mikule et al . , 2007 ) . Immunofluorescence analysis confirmed that RFP-PeriCT localizes to the centrioles in P0-isolated cardiomyocytes ( Figure 3—figure supplement 1B ) . Both methods used to disrupt centrosome integrity suppressed P0-isolated cardiomyocyte proliferative potential ( Figure 3D–F ) . Taken together , these data indicate that loss of centrosome integrity promotes G0/G1 cell cycle arrest in mammalian cardiomyocytes . 10 . 7554/eLife . 05563 . 008Figure 3 . Absence of centrosome integrity results in cell cycle aberrations . ( A ) Correlation between proliferative potential ( Ki67 ) and centrosome integrity ( split-centrioles ) . Overlay of the frequency of Ki67-positive E15- , P0- , P3- , or P5-isolated cardiomyocytes in response to 20% fetal bovine serum ( FBS ) ( bars ) and the frequency of cardiomyocytes with split-centrioles at different developmental stages according to Figure 1B ( red line ) . ( B ) Representative images indicating that 20% FBS-stimulated P3-isolated cardiomyocytes ( Troponin I ) with paired-centrioles ( γ-tubulin ) exhibited greater proliferative potential ( Ki67 ) than those with split centrioles . Scale bars: 10 µm . ( C ) Frequency of Ki67-positive P3-isolated cardiomyocytes in response to 20% FBS in the presence or absence of p38 MAP kinase inhibitor ( p38i ) . ( D–F ) Pericentrin ( PCNT ) is required for cardiomyocyte proliferative potential . ( D ) Relative frequency of Ki67-positive P0-isolated cardiomyocytes treated with scrambled or PCNT siRNA in response to 20% FBS . ( E ) Representative images of 20% FBS-stimulated P0-isolated cardiomyocytes transfected with a construct driving RFP-tagged dominant negative C-terminal Pericentrin ( RFP–PeriCT ) expression and stained for Ki67 . Scale bars: 10 µm . ( F ) Quantitative analysis of ( E ) . ( G , H ) Analysis of centrioles ( γ-tubulin ) during metaphase ( H3P ) . ( G ) Representative images of P3-isolated cardiomyocytes in metaphase in the presence or absence of p38i upon stimulation with 10% FBS . Scale bars: yellow: 50 µm; red: 10 µm . ( H ) Quantitative analysis of ( G ) . ( I ) Representative images of centrioles in adult cardiomyocytes in metaphase stimulated with FGF1 plus p38i . Chromosomes: DAPI . Scale bars: 20 µm . Data are mean ± SD , n = 3 , *: p < 0 . 05 . For the experiments ≥ 20 cells ( F ) , ≥ 25 cells ( D ) , ≥ 40 cells ( H ) , or ≥ 50 cells ( A , C ) were analyzed per experimental condition . DOI: http://dx . doi . org/10 . 7554/eLife . 05563 . 00810 . 7554/eLife . 05563 . 009Figure 3—figure supplement 1 . Absence of centrosome integrity results in cell cycle aberrations . ( A ) Correlation between proliferative potential ( Ki67 ) and centrosome integrity ( split centrioles ) . Overlay of the frequency of Ki67-positive P0- , P3- , P5-isolated non-myocytes in response to 20% fetal bovine serum ( FBS ) ( bars ) and the frequency of non-myocytes with split-centrioles at different developmental stages according to Figure 1B ( red line ) . Data are mean ± SD , n = 3 . > 50 cells were analyzed per experimental condition . n . s . : not statistically significant ( p > 0 . 05 ) . ( B ) Representative images of P0-isolated cardiomyocytes transfected with a construct driving RFP-tagged dominant negative C-terminal Pericentrin ( RFP–PeriCT ) expression and stained for cardiomyocytes ( Troponin I ) and centrioles ( γ-tubulin ) . Nuclei were stained with DAPI . Scale bars: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05563 . 009 Disruption of centrosome integrity promotes p38MAP kinase ( p38 ) -mediated G1/S cell cycle arrest ( Mikule et al . , 2007; Pascreau et al . , 2011 ) . Previously , it has been demonstrated that p38 inhibition ( p38i ) allows postnatal cardiomyocyte proliferation ( Engel et al . , 2005 ) . This raised the question whether p38i can over-ride post-mitotic arrest in cardiomyocytes with split-centrioles . p38i enhanced the proliferative potential of P3-isolated cardiomyocytes with split-centrioles ( Figure 3C ) . These results indicate that suppression of the p38-mediated stress-activated pathway can promote cell cycle progression in cardiomyocytes that lack centriole-cohesion . A bipolar mitotic spindle is required for high fidelity chromosome segregation ( Pihan , 2013 ) . Thus , we determined if the observed changes in centrosome integrity affects spindle pole number during mitosis in postnatal cardiomyocytes when cell cycle progression is induced . In response to serum stimulation , P0- and P3-isolated cardiomyocytes exhibited a typical metaphase consisting of two γ-tubulin foci ( Figure 3G , H ) . p38i had no effect on the number of γ-tubulin foci in P0-isolated cardiomyocytes during metaphase ( Figure 3H ) . In contrast , p38i resulted in a significant increase in metaphases containing multiple γ-tubulin foci in P3-isolated cardiomyocytes ( Figure 3G , H ) . Similarly , adult-isolated cardiomyocytes induced to re-enter the cell cycle by p38i + FGF1 also exhibited multiple γ-tubulin foci during metaphase ( Figure 3I ) . These data indicate that over-riding cell cycle arrest in cardiomyocytes results in multiple spindle poles . In contrast to mammals , adult newts and zebrafish can regenerate their heart through cardiomyocyte proliferation ( Poss et al . , 2002; Bettencourt-Dias et al . , 2003 ) . This suggests that they either do not establish a post-mitotic state by disassembling their centrosomes or they are able to reverse this mechanism upon injury . To test which scenario occurs in these species , we analyzed their centrosomes . Paired centrioles could readily be identified in adult newt and zebrafish cardiomyocyte nuclei ( Figure 4A ) . Further , the frequency of newt and zebrafish cardiomyocyte nuclei with intact centrosomes was similar to that of non-myocyte nuclei ( Figure 4B ) . In addition , MTOC-regrowth assays demonstrated that centrosomes in zebrafish and newt cardiomyocytes are functional ( Figure 4C and Figure 4—figure supplement 1A , B ) . Collectively , these results suggest that , unlike mammalian cardiomyocytes , newt and zebrafish cardiomyocytes maintain centrosome integrity throughout adulthood . 10 . 7554/eLife . 05563 . 010Figure 4 . Centrosome integrity is maintained in adult newt and zebrafish cardiomyocytes . ( A ) Representative images of centrosomes ( γ-tubulin ) in heart cryosections of adult transgenic cmlc2:dsRedExp-nuchsc4 zebrafish , of adult newt hearts , and adult mouse hearts . Nuclei: DAPI , cardiac nuclei: DsRed or Nkx2 . 5 . Green-framed expansions: newt and zebrafish cardiomyocyte nuclei with paired-centrioles . Yellow-framed expansion: mouse non-myocyte nucleus with paired-centrioles . White asterisk: cardiomyocyte nucleus . Yellow scale bar: 10 µm . Red scale bar: 2 µm . ( B ) Quantitative analysis of nuclei associated with intact centrosomes in cryosections as shown in ( A ) . Data are mean ± SD , n = 3 , ≥ 300 cells were analyzed per experimental condition , *: p < 0 . 05 . ( C ) Representative images documenting localization of microtubule regrowth ( β-tubulin ) in cultured adult zebrafish ( Troponin I ) and newt ( phalloidin-TRITC ) cardiomyocytes . Adult zebrafish were treated with serum ( control ) , nocodazole , or nocodazole followed by wash out . Adult newt cardiomyocytes were treated with serum ( control ) , ice , or ice followed by return to normal ( norm ) temperature . Nuclei: DAPI . Yellow arrowheads: localization of centrosome . Yellow scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05563 . 01010 . 7554/eLife . 05563 . 011Figure 4—figure supplement 1 . Centrosome integrity is maintained in adult newt and zebrafish cardiomyocytes . ( A ) Frequency of MTOC localization in adult zebrafish cardiomyocytes ( CM ) and non-myocytes ( NCM ) after nocodazole wash out . Cardiomyocytes from 12 hearts were pooled and 50 CMs and 50 NCMs were analyzed . ( B ) Frequency of MTOC localization in adult newt CM and NCM after microtubule depolymerization with ice . Cardiomyocytes from 10 hearts were pooled and > 50 CMs and > 50 NCMs were analyzed . ( C ) Expression of dominant negative Pericentrin ( RFP-PeriCT ) via RNA injection caused developmental delay in zebrafish embryos at 24 hr post fertilization ( hpf ) compared to RFP RNA-injected control embryos . ( D ) Percentage of embryos showing a normal or developmental delayed phenotype in each group . ( E ) Representative images of RFP and RFP-PeriCT RNA-injected zebrafish embryos stained for mitotic cells ( H3P ) . ( F ) Quantification of H3P events in the area of interest as defined in ( E ) . ( G ) Representative images of clone classes in RFP DNA-injected embryos at 20 hpf . ( H ) Percentages of embryos in different classes in RFP or RFP-PeriCT DNA-injected embryos . Data are mean ± SEM; n = total number of embryos analyzed . ***: p < 0 . 001 . Scale bars: yellow/black: 250 µm; white: 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05563 . 011 We then sought to determine if zebrafish cells require centrosome integrity for proliferation in vivo . To determine this , we injected RNA encoding RFP-tagged dominant negative C-terminal Pericentrin ( RFP-PeriCT ) into one-cell stage zebrafish embryos . Compared to control embryos ( RFP RNA-injected ) , a significantly larger number of RFP-PeriCT RNA-injected embryos showed developmental delay at 24 hr post fertilization ( hpf ) and a significantly reduced number of H3P-positive cells ( Figure 4—figure supplement 1C–F ) . Since RFP-PeriCT RNA did not cause necrosis or lethality , it is plausible that the observed developmental delay was caused , at least partly , by reduced proliferation of RFP-PeriCT-expressing cells . To further test an effect on proliferation , we injected DNA constructs driving RFP or dominant negative Pericentrin ( RFP-PeriCT ) expression from the CMV promoter into one-cell stage embryos and analyzed the size of RFP-positive clones of cells at 20 hpf . Embryos were sorted into four classes according to the size of the clones , namely no visible clones ( class I ) , few scattered cells ( class II ) , medium-sized clones ( class III ) , and large clones ( class IV ) . RFP DNA-injected controls exhibited comparable numbers of embryos displaying no clones ( class 1 ) , scattered ( class II ) , medium-sized ( class III ) , and larger clones ( class IV ) . In contrast , the majority of RFP-PeriCT DNA-injected embryos contained either no clones ( class I ) or few scattered clones ( class II ) , while larger clones ( class III and IV ) were rarely seen ( Figure 4—figure supplement 1G , H ) . This suggests that clonal expansion of RFP-PeriCT expressing cells is reduced , which substantiates the hypothesis that expression of RFP-PeriCT reduces cell proliferative potential . These results suggest that zebrafish cells require centrosome integrity for proliferation in vivo .
Increasing evidence supports the requirement of a functional centrosome for cellular proliferative potential ( Doxsey et al . , 2005 ) . For instance , somatic cell-types arrest in G0/G1 when centrosomes are removed or disrupted ( Hinchcliffe et al . , 2001; Khodjakov and Rieder , 2001; Srsen et al . , 2006; Mikule et al . , 2007; Wong et al . , 2015 ) . Further , centrosome degradation occurs during female meiosis—a process believed to inhibit parthenogenesis—with mitotic cycles restored upon fertilization when the sperm donates a centrosome during zygote formation ( Klotz et al . , 1990; Clift and Schuh , 2013 ) . Moreover , relocalization of the MTOC to non-centrosomal loci has been described for several post-mitotic , highly differentiated , cell-types ( Bartolini and Gundersen , 2006 ) . This body of literature coupled with our results suggests that centrosome integrity can be developmentally regulated to achieve a post-mitotic state . Centrosome disassembly appears to be a very effective way to achieve a post-mitotic state . But why do cardiomyocytes disassemble their centrosomes ? Upon birth , the neonatal heart , and the cardiomyocytes therein , undergo increased hemodynamic stress . Effective cardiomyocyte function in response to increased hemodynamic stress may require a cytoskeletal architecture more conducive to handling postnatal physical stresses ( e . g . , a nuclear envelope-based MTOC ) . Thus , centrosome disassembly may be a result of cytoskeletal reorganization . In this scenario , proliferative potential might be sacrificed for postnatal function . It has long been considered that skeletal myoblasts and cardiomyocytes have distinct molecular mechanisms to control their proliferative growth . This theory has been primarily based on the observation that proliferation and differentiation ( contraction ) are mutually exclusive in skeletal muscle but occur in parallel in cardiomyocytes ( Ueno et al . , 1988 ) . However , our results together with others ( Tassin et al . , 1985; Bugnard et al . , 2005; Srsen et al . , 2006; Fant et al . , 2009; Zaal et al . , 2011 ) challenges this dogma , as establishment of a post-mitotic state in both muscle cell-types coincides with MTOC relocalization to the nuclear envelope and centrosome disassembly . The observation that p38i promotes cardiomyocyte proliferative potential in the absence of a functional centrosome provides some optimism for heart regeneration via proliferation of endogenous cardiomyocytes . However , we observe that an absence of centrosome integrity correlates with multiple spindle poles . Although multiple spindle poles are generally resolved into a semi-clustered/pseudo-bipolar conformation , this process is nevertheless highly prone to the formation of merotelic spindle-kinetochore attachments , which can promote chromosome missegregation ( Ganem et al . , 2009 ) . Indeed , postnatal cardiomyocytes induced to proliferate exhibit chromosome segregation abnormalities ( Engel et al . , 2006 ) . While cell viability can exist when aneuploidy is limited ( e . g . , as seen in Down Syndrome ) , aneuploidy is generally not well tolerated ( Torres et al . , 2008 ) . Therefore , proliferation of adult cardiomyocytes may not necessarily result in viable daughter cells . Thus , in the absence of a functional centrosome , whether a particular manipulation induces aneuploidy , and to what degree , may be a critical factor in determining its regenerative therapeutic potential . The origin of multiple spindle poles ( i . e . , multiple γ-tubulin foci ) during cardiomyocyte mitosis is not entirely clear . Interestingly , in addition to centriole-cohesion , Pericentrin and Cdk5Rap2 are also required for centriole-engagement ( Barrera et al . , 2010; Lee and Rhee , 2012 ) . During S-phase , centrioles are duplicated , with each mother centriole forming a daughter centriole , which remains closely attached ( i . e . , engaged ) until mid-anaphase ( Kuriyama and Borisy , 1981; Sluder , 2013 ) . When mother-daughter centrioles are engaged , only the mother centriole exhibits a strong γ-tubulin signal ( Wang et al . , 2011 ) . Thus , one speculation is that when cells that lack centriole-cohesion enter mitosis , they have an increased likelihood of premature loss of centriole-engagement , thus accounting for the four γ-tubulin signals observed at metaphase . However , it has to be considered that induction of mitosis in bi-nucleated cardiomyocytes , which increasingly appear after birth , results in four γ-tubulin signals in metaphase corresponding to four duplicated centrioles . The ability of zebrafish and newts to regenerate their heart has gained extensive interest in recent years . One major question is what distinguishes mammalian cardiomyocytes from those of zebrafish and newts with regards to their proliferative potential . Our data demonstrate that the state of cellular differentiation of cardiomyocytes from various species is not evolutionary conserved . The fact that adult zebrafish and newt cardiomyocytes maintain their centrosome integrity indicates that factors promoting adult zebrafish cardiomyocyte proliferation might not necessarily induce adult mammalian cardiomyocyte proliferation . Recently , it has been shown that planarians can develop and regenerate in the absence of centrosomes ( Azimzadeh et al . , 2012 ) . This has questioned the requirement of a centrosome for cell proliferation during development and regeneration . However , there are significant differences between planarians and vertebrates in cell cycle control . For example , planarians express only one single repressive E2F , whereas mammals and zebrafish express several repressive and activating E2Fs . In addition , planarians do not have Cyclin E or Cyclin A homologs . Further , cyclin-dependent kinase CDK2 ( to which Cyclin E/A usually binds ) is expressed at such low levels in planarians that it is considered functionally dead ( Zhu and Pearson , 2013 ) . This is important as in mammals centrosome localization of Cyclin E and Cyclin A is required for G1/S cell cycle progression ( Pascreau et al . , 2011 ) . Thus , at least in the case of Cyclin E and Cyclin A , planarians lack centrosome-regulated cell cycle factors . In addition , our data indicate that centrosome integrity is required for proliferation during zebrafish development . This is in agreement with the recent observation that plk4 depletion in zebrafish impairs centriolar biogenesis during development and increases premature cell cycle exit ( independent of ciliogenesis defects ) resulting in reduced zebrafish size ( Martin et al . , 2014 ) . Other examples of centriole-associated genes whose depletion causes cell cycle defects resulting in impaired development are stil ( Pfaff et al . , 2007; Vulprecht et al . , 2012; Sun et al . , 2014 ) and cetn2 ( Delaval et al . , 2011 ) . While there is no evidence for the requirement of functional centrosomes in cardiac regeneration in zebrafish , there are data that indirectly suggest that centrosome function might be required for cardiac regeneration . For example it has been shown that inhibition of mps1 and plk1 , factors implicated in centrosome assembly and maturation ( Pike and Fisk , 2011; Joukov et al . , 2014; Kong et al . , 2014 ) , impairs cardiac regeneration in zebrafish after apex resection ( Poss et al . , 2002; Jopling et al . , 2010 ) . Thus , it will be interesting in the future to test if destruction of centrosome integrity will indeed abolish the ability of zebrafish to regenerate their heart . Taken together , this study suggests that relocalization of the MTOC disrupts centrosome integrity which , in turn , promotes a post-mitotic state in mammalian cardiomyocytes . Given the increasing role of the centrosome in cell cycle control , understanding how centrosome integrity is regulated during development may reveal new mechanisms to regulate cell proliferation with implications for regeneration and cancer .
The investigation conforms with the Guide for the Care and Use of Laboratory Animals published by the Directive 2010/63/EU of the European Parliament and according to the regulations issued by the Committee for Animal Rights Protection of the State of Hessen ( Regierungspraesidium Darmstadt ) as well as Baden-Württemberg ( Regierungspraesidium Tübingen ) . Extraction of organs and preparation of primary cell cultures were approved by the local Animal Ethics Committee in accordance to governmental and international guidelines on animal experimentation ( protocol TS—5/13 Nephropatho; Zebrafish protocol number o . 183 ) . Adult ( 1–2 year ) zebrafish ( Danio rerio ) hearts were isolated from the transgenic line Tg ( cmlc2:dsRedExp-nuchsc4 ) ( Takeuchi et al . , 2011 ) . Adult newt cardiomyocytes were isolated from red-spotted newts ( Notophthalmus viridescens , Charles Sullivan , Nashville , TN , USA ) . Ventricular cardiomyocytes and whole-hearts were obtained from embryonic day 15 ( E15 ) , E18 , postnatal day 0 ( P0 ) , P3 , P5 , and adult Sprague–Dawley rats ( from Charles River Laboratories , Cologne , Germany or own bred ) . Mammalian and zebrafish ventricular cardiomyocytes were isolated as described previously ( Engel et al . , 2005; Sander et al . , 2013 ) . Newt ventricular cardiomyocytes were isolated using the following procedure . To prevent contaminations , animals were kept 24 hr in advance in a Sulfamerazine bath ( 5 g/l , Sigma , St . Louis , MO , USA ) for disinfection . Organ removal was performed under deep anesthesia , by incubating the animals in a Tricaine solution ( 1 g/l , Sigma ) with pH 7 . 4 for 15–20 min . After decapitation , ventricles were removed , washed several times with 65% L15 Leibovitz media ( Gibco , Grand Island , New York , USA ) with antibiotics ( 2% penicillin/streptomycin and ciprofloxacin ( 10 µg/ml ) ) and incubated over night at 25°C . Thereafter , enzymatic digestion followed with a sterile mixture of collagenase ( 1 mg/ml , Sigma ) , elastase ( 0 . 1 mg/ml , Sigma ) and DNase ( 0 . 1 mg/ml , Sigma ) with glucose ( 3 mg/ml , Sigma ) and BSA ( 1 . 5 mg/ml , Sigma ) in aPBS ( 75% PBS ) for 6 hr at 27°C . After mechanical dissociation , and several washing step , cells were plated on laminin ( 15 µg/ml , Sigma ) coated 8-well chamber slides ( Nunc ) and cultured for 5 days with 65% MEM with Glutamaxx ( Gibco ) containing 10% FCS and antibiotics ( 2% penicillin/streptomycin and ciprofloxacin [10 µg/ml] ) at 25°C with 5% CO2 , media change took place once after 3 days . Mouse iPSC-derived cardiomyocytes ( Axiogenesis AG , Cologne , Germany ) were thawed and plated according to manufacturer's instructions . Mammalian cardiomyocytes were cultured on 1 mg/ml fibronectin ( Sigma ) -coated glass coverslips . Isolated cardiomyocytes were seeded and cultured in DMEM/F-12 , GlutamaxTM-I ( Life Technologies , Darmstadt , Germany ) + Penicillin ( 100 U/ml ) /Streptomycin ( 100 µg/ml ) ( Pen/Step ) ( Life Technologies ) for 2 days prior to experimentation . To analyze proliferative potential of whole cardiomyocyte populations , cardiomyocytes were cultured with 20% fetal bovine serum ( FBS ) ( Sigma ) for 2 days . To analyze proliferative potential between cardiomyocytes with paired and split centrioles , cardiomyocytes were cultured with 20% FBS + 2 mM hydroxyurea ( HU ) ( Sigma ) , for 2 days . As centriole-cohesion is normally lost at G2/M , G1/S arrest with hydroxyurea prevents misinterpreting normal loss of centriole-cohesion ( which occurs in G2 ) for precocious loss of centriole-cohesion ( which occurs during G1 ) . To analyze mitotic aberrations , neonatal cardiomyocytes were cultured with 10% FBS for 2 days , and adult cardiomyocytes were seeded in 10% horse serum and 20 µM cytosine β-D-arabinofuranoside ( Sigma ) for 2 days and then stimulated with 50 ng/ml FGF1 ( R&D Systems , Abingdon , UK ) + p38i ( 5 µM SB203580 , Tocris Biosystems , Bristol , UK ) for 2 days as previously described ( Engel et al . , 2005 ) . To analyze microtubule regrowth in mammalian and zebrafish cardiomyocytes , cells were cultured for 2 days and then treated with 5 µg/ml nocodazole ( Sigma ) for 2 . 5 hr . Subsequently , cells were washed with nocodazole-free media for 5–10 min to allow microtubule regrowth . To analyze microtubule regrowth in newt cardiomyocytes , cells were cultured for 5 days and then placed on ice for 3 hr . Subsequently , cells were returned to 25°C for 10 min . For siRNA studies cells were cultured for 2 days prior to transfection of siRNAs ( Qiagen , Venlo , Limburg , Netherlands ) using Lipofectamine RNAiMAX ( Life Technologies ) . Cycling and MTOC regrowth assays utilizing siRNAs were conducted 2 days after siRNA transfection . Plasmids were transfected with Lipofectamine LTX ( Life Technologies ) on the day of seeding . Cycling assays were conducted 2 days after plasmid transfection . To induce ciliogenesis , cardiomyocytes were seeded and cultured in DMEM GlutaMAXTM-I ( Gibco ) + Pen/Strep for 3 days . Hearts from zebrafish , newts , Sprague Dawley rats ( Charles River Laboratories ) , or C57BL/6J mice ( Charles River Laboratories ) were oriented perpendicularly in relation to their long axis , embedded in an O . C . T . compound tissue-freezing medium , and frozen in liquid nitrogen . Hearts were sectioned with a Leica CM 3000 cryostat ( 10 µm ) . Cryosections were fixed in 3 . 7% formalin ( Sigma ) for 10 min at room temperature ( RT ) . Isolated cells were fixed in either pre-chilled methanol for 5 min at −20°C or 3 . 7% formalin ( Sigma ) for 10 min at RT . Immunostaining was performed as described previously ( Engel et al . , 2005 ) utilizing 3% BSA ( Sigma ) /PBS instead of goat-serum as blocking buffer . Formalin-fixed cells were permeabilized prior to antibody staining with 0 . 2% Triton X-100 ( Sigma ) /PBS ( 10 min , RT ) . Primary antibodies: goat anti-Troponin I ( 1:250 , Abcam , Cambridge , UK ) , rabbit anti-Troponin I ( 1:250 , Santa Cruz Biotechnology , Heidelberg , Germany ) , rabbit anti-Cdk5Rap2 ( 1:500 , Millipore , Hessen , Germany ) , rabbit anti-Pericentrin ( 1:700 , Covance , Princeton , NJ , USA ) , mouse anti-γ-tubulin ( 1:500 , Santa Cruz Biotechnology ) , rabbit anti-PCM1 ( 1:500 , Santa Cruz Biotechnology ) , rabbit anti-Odf2 ( 1;500 , ProteinTech Group , Manchester , UK ) , rabbit anti-Centrobin ( 1:500 , Sigma ) , goat anti-Nkx2 . 5 ( 1:100 , Santa Cruz Biotechnology ) , rabbit anti-CEP135 ( 1:500 , Abcam ) , rabbit anti-phospho-histone H3-Serine 10 ( 1:1000 , Santa Cruz Biotechnology ) , rabbit anti-Mef2 ( 1:500 , Santa Cruz Biotechnology ) , mouse anti-β-tubulin ( KMX ) ( 1:500 , Millipore ) , Phalloidin-TRITC ( 1:300 , Sigma ) . Rabbit anti-Pericentrin ( 1:500; MmPeriC1 ) against both B and S isoforms was produced as previously described ( Mühlhans et al . , 2011 ) . Mouse anti-Pericentrin against the Pericentrin B isoform ( 1:500; MmPeri N-term clone 7H11 or 8D12 ) was made against the first 233 amino acids of mouse Percentrin B ( AN: NP_032813 or BAF36559 ) . Primary immune complexes were detected with ALEXA 350- , ALEXA 488- , ALEXA 594- , or ALEXA 647-conjugated antibodies ( 1:500 , Life Technologies , Carlsbad , CA , USA ) . DNA was stained with 0 . 5 µg/ml DAPI ( 4′ , 6′-diamidino-2-phenylindole ) ( Sigma ) . Images were captured on a Keyence BZ9000 Fluorescence Microscope ( Keyence , Osaka , Japan ) , using 63× or 100× objectives . Images were arranged with ImageJ ( Public Domain ) and Adobe Illustrator ( Adobe , San Jose , CA , USA ) . For the quantitative analysis of protein intensity in cardiomyocytes and non-myocytes at centriolar loci , a line-plot was generated using software ( Keyence ) which traversed the centrioles ( identified by γ-tubulin staining ) and the average Cdk5Rap2 or Pericentrin background intensity was subtracted from signals corresponding to γ-tubulin loci . Cardiomyocyte Cdk5Rap2 or Pericentrin signal intensity was normalized to that of non-myocytes from the same cultures . RNA was isolated from different developmental stages of rat ( E11 to E20 , n ≥ 10; P5 , P10 , and adult , n ≥ 3 ) using TRIzol ( Life Technologies ) . RT-PCR was performed following standard protocols . A set of three primers was used to detect Pcnt B and Pcnt S in the same reaction . Primers used were Pcnt B-F , 5′-CATGGCTCTGCACAATGAAG-3′ , Pcnt S-F 5′-CAGGGCTGTTCCATATGTTC-3′ , Pcnt-R 5′-GAAGTCTCCTCAGGGCATCTC-3′ . Neonatal cardiomyocytes were cultured for 3 days after isolation . On the third day the cells were washed in PBS , trypsinized , fixed in ice-cold 70% EtOH/15% PBS , and centrifuged ( 10 min , 700×g , 4°C ) . The cell pellet was resuspended in PBS and centrifuged again . Cells were then resuspended in extraction buffer ( 50 mM Na2HPO4: 25 mM citric acid ( 9:1 ) , 0 . 1% Triton X-100 , 0 . 01% NaN3 , pH 7 . 8 ) and incubated for 15 min at RT . Cells were centrifuged and the cell pellet was incubated in 250 µl of complete DNA staining buffer ( 10 mM PIPES , 0 . 1 M NaCl , 2 mM MgCl2 , 0 . 1% Triton X-100 , 0 . 02% NaN3 , pH 6 . 8 ) , 15 µl of RNase A ( 10 mg/ml ) and 12 µl of propidium iodide ( 1 mg/ml ) for 30 min at RT . Afterwards the cell suspension was transferred to a FACS tube and 150 µl of PBS was added . Per sample 10 , 000 events were analyzed with a BD FACSCanto II ( BD Transduction , Heidelberg , Baden-Württenberg , Germany ) and analysed with the FlowJo software . RFP-tagged dominant negative C-terminal Pcnt was PCR-amplified from the RFP-PeriCT construct , using specific primers ( forward: 5′gggcccgaattcGCAAACATGGTGACGTCACCGGTCGCCACCATG3′ , reverse 5′ gggcccctcgagTCATCGGGTGGCAGGATTTCTTTGAAG 3′ ) to introduce an EcoRI site at the 5′ end and an XhoI site at the 3′ end and cloned into the EcoRI and XhoI sites of the pCS2+ vector . Subsequently , mCherry ( Clontech , Saint-Germain-en-Laye , France ) was introduced into the EcoRI and BglII sites replacing mDsRed . Capped sense RNA was synthesized in vitro using mMessage mMachine kits ( Life Technologies ) . RNA ( 90 pg ) or DNA ( 25 pg ) was injected into the cytoplasm of one-cell stage zebrafish embryos using standard procedures . Embryos were raised at 28 . 5°C until indicated time and were checked for RFP expression . Embryos were raised at 28 . 5°C until 24 hpf and were fixed in 4% PFA in PBS for 2 hr at RT . After 3 washes in PBST ( PBS + 0 . 1% Tween ) , embryos were washed in water and were permeabilized in prechilled acetone at −20°C before immunostaining . Mitotic cells were identified using anti-H3P antibodies ( 1:200 , Cell-Signaling ) which were detected with secondary antibodies conjugated to Alexa 555 ( 1:1000 , Invitrogen ) . Nuclei were visualized with DAPI . Images of single optical plane at the notochord were acquired with Leica SP5 confocal ( Leica , Wetzlar , Germany ) . For quantification , the number of H3P-positive cells was counted manually in a region dorsal to the notochord , 500 µm from the tip of the tail . Data of at least three independent experiments are expressed as mean ± SD . Statistical analysis was determined using Students t-test using Excel ( Microsoft , Redmond , WA , USA ) or ANOVA followed by Post-hoc t-test and Bonferroni correction . For DNA injection experiment into zebrafish embryos , statistical significance was tested using chi-squared test . | Muscle cells in the heart contract in regular rhythms to pump blood around the body . In humans , rats and other mammals , the vast majority of heart muscle cells lose the ability to divide shortly after birth . Therefore , the heart is unable to replace cells that are lost over the life of the individual , for example , during a heart attack . If too many of these cells are lost , the heart will be unable to pump effectively , which can lead to heart failure . Currently , the only treatment option in humans with heart failure is to perform a heart transplant . Some animals , such as newts and zebrafish , are able to replace lost heart muscle cells throughout their lifetimes . Thus , these species are able to fully regenerate their hearts even after 20% has been removed . This suggests that it might be possible to manipulate human heart muscle cells to make them divide and regenerate the heart . Recent research has suggested that structures called centrosomes , known to be required to separate copies of the DNA during cell division , are used as a hub to integrate the initial signals that determine whether a cell should divide or not . Here , Zebrowski et al . studied the centrosomes of heart muscle cells in rats , newts and zebrafish . The experiments show that the centrosomes in rat heart muscle cells are dissembled shortly after birth . Centrosomes are made of several proteins and , in the rat cells , these proteins moved to the membrane that surrounded the nucleus . On the other hand , the centrosomes in the heart muscle cells of the adult newts and zebrafish remained intact . Further experiments found that that breaking apart the centrosomes of heart muscle cells taken from newborn rats stops these cells from dividing . Zebrowski et al . 's findings suggest that the loss of centrosomes after birth is a possible reason why the hearts of adult humans and other mammals are unable to regenerate after injury . In the future , these findings may aid the development of methods to regenerate human heart muscle and new treatments that may limit division of cancer cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"short",
"report",
"cell",
"biology"
] | 2015 | Developmental alterations in centrosome integrity contribute to the post-mitotic state of mammalian cardiomyocytes |
Antimicrobial peptides ( AMPs ) are broad spectrum antibiotics that selectively target bacteria . Here we investigate the activity of human AMP LL37 against Escherichia coli by integrating quantitative , population and single-cell level experiments with theoretical modeling . We observe an unexpected , rapid absorption and retention of a large number of LL37 peptides by E . coli cells upon the inhibition of their growth , which increases population survivability . This transition occurs more likely in the late stage of cell division cycles . Cultures with high cell density exhibit two distinct subpopulations: a non-growing population that absorb peptides and a growing population that survive owing to the sequestration of the AMPs by others . A mathematical model based on this binary picture reproduces the rather surprising observations , including the increase of the minimum inhibitory concentration with cell density ( even in dilute cultures ) and the extensive lag in growth introduced by sub-lethal dosages of LL37 peptides .
Antimicrobial peptides ( AMPs ) are natural amino-acid based antibiotics that are part of the first line of defense against invading microbes in multicellular systems ( Zasloff , 2002; Brogden , 2005 ) . In humans , AMPs are found in many organs that are in contact with the outside world , including airways , skin , and the urinary tract ( Hancock and Lehrer , 1998; Zasloff , 2002; Brogden , 2005; Jenssen et al . , 2006; Ganz , 2003; Epand and Vogel , 1999 ) . The short sequence of the AMPs ( typically <50 amino acids ) along with the flexibility in the design and synthesis of new peptides has spurred attention towards understanding the detailed mechanism of AMPs action which can lead to the rational design of novel antibiotic agents ( Zasloff , 2002; Brogden , 2005; Hancock and Sahl , 2006 ) . A hallmark of the AMPs antibacterial mechanism is the role of physical interactions . Structures of AMPs exhibit two common motifs: cationic charge and amphiphilic form ( Zasloff , 2002; Brogden , 2005 ) . The cationic charge enables them to attack bacteria , enclosed in negatively charged membranes , rather than mammalian cells , which possess electrically neutral membranes . The amphiphilic structure allows AMPs to penetrate into the lipid membrane structures ( Matsuzaki et al . , 1995; Shai , 1999; Ludtke et al . , 1996; Heller et al . , 2000; Taheri-Araghi and Ha , 2007; Huang , 2000; Yang et al . , 2001 ) . Despite our detailed knowledge about interactions of AMPs with membranes , we lack a comprehensive picture of the dynamics of AMPs in a population of cells . We are yet to determine the extent to which the physical interactions of AMPs disrupt biological processes in bacteria and the degree to which electrostatic forces govern the diffusion and partitioning of AMPs among various cells . Specifically , it was suggested by Matsuzaki and Castanho et al . that the density of cells in a culture can alter the activity of AMPs through distributions among different cells ( Matsuzaki , 1999; Melo et al . , 2009 ) . We have recently examined the role of adsorption on various cell membranes theoretically ( Bagheri et al . , 2015 ) . Experimental investigations using bacteria and red blood cells by Stella and Wimley groups ( Savini et al . , 2017; Starr et al . , 2016 ) directly demonstrated the decisive role of cell density on the effectivity of antimicrobial peptides . In this work , we utilize complementary experimental and modeling approaches to understand the population dynamics of activity of AMPs from a single-cell perspective . Like all antibiotic agents , AMPs need a minimum concentration ( MIC ) to inhibit growth of a bacterial culture . For some antibiotics , including AMPs , the MIC is dependent on the cell density . Often referred to as the ‘inoculum effect’ , these phenomena are a trivial consequence of overpopulated cultures . However , in dilute cultures , MICs have been reported to reach a plateau independent of cell density ( Savini et al . , 2017; Starr et al . , 2016; Udekwu et al . , 2009; Artemova et al . , 2015 ) , unless the cell population becomes so small that stochastic single-cell effects become important ( Coates et al . , 2018 ) . For a precise measurement of the inoculum effect , we extended microplate assays by Wiegand et al . ( 2008 ) to obtain a functional form of the MIC in terms of the initial cell density ( the ‘inoculum size’ ) . Contrary to our expectations , we observed that the MIC for the LL37 peptide ( AnaSpec , California ) remains dependent on Escherichia coli density , even in dilute cultures where the average cell-to-cell distance is above 50 μm , much greater than the average cell dimensions ( ∼1×1×5 μm ) ( Taheri-Araghi et al . , 2015 ) . With no direct interactions among the cells and nutrients in excess for all , this dependence suggests that the effective peptide concentration is somehow compromised in a cell density dependent manner . By tracking a dye-tagged version of LL37 peptide ( 5-FAM-LC-LL37 , AnaSpec ) , we found that the inhibition of growth of E . coli cells was followed by the translocation of a large number of AMPs into the cytoplasmic compartment of cells , thus reducing the peptide concentration in the culture , which works in favor of other cells . In the sense of such dynamics , MIC refers to a sufficient concentration of AMPs for absorption into all the cells . Below the MIC , peptide is absorbed by only a fraction of cells , leaving an inadequate amount of AMPs to inhibit the growth of remaining cells . We have directly observed that cultures with sub-MIC concentrations of dye-tagged LL37 peptides ( 5-FAM-LC-LL37 ) exhibit a heterogenous population combining non-growing cells containing many LL37 peptides and growing cells without LL37 peptides .
The MIC of an antibiotic may depend on the cell density for various reasons: the distribution of antibiotic molecules among bacteria ( Udekwu et al . , 2009; Clark et al . , 2009; Melo et al . , 2009; Jepson et al . , 2016 ) ; or as a result of cellular enzyme secretion ( e . g . , β-lactamase in case of lactame resistant antibiotics ( Clark et al . , 2009; Artemova et al . , 2015 ) ; or due to the chemical composition of the culture and regulation of gene expression ( for instance in late exponential or stationary phases ) ( Karslake et al . , 2016; Artemova et al . , 2015 ) . In this work we focus on dilute cultures where the dependence of MIC on inoculum size reflects the number of antimicrobial molecules either absorbed or degraded by each individual cell . The MIC for AMPs is in the micromolar range , ∼1014 AMPs/ml . Early exponential cultures contain ∼106 cells/ml , which amounts to the ratio of ∼108 AMPs/cell . At such a high ratio , only binding or degradation of AMPs of the same order of magnitude per individual cells can lead to the inoculum effect ( Starr et al . , 2016; Savini et al . , 2018 ) . To map out the functional form of the inoculum effect , we implemented a two-dimensional dilution scheme on a 96-well plate with a linear dilution of LL37 peptides in columns 7 and 12 followed by a 2/3 dilution series of cells and LL37 peptides over two distinct regions , columns 12 to 8 and 7 to 1 ( Figure 1A ) . ( See Materials and methods and Appendix 1 for the details of the cell counting and plate preparation ) . An early exponential E . coli culture in rich defined media ( RDM , Teknova ) was diluted to specific cell densities to cover a relatively even distribution of inoculum sizes . Each well on the microplate corresponds to a unique combination of densities of LL37 peptides and E . coli cells ( Figure 1B ) . The cultures in the wells were monitored for 24 hr by an automated plate reader ( EPOCH 2 , BioTek ) in terms of optical density at 600 nm wavelength ( OD600 ) , while the plate was incubated with orbital shaking at 37°C . Growth or inhibition of growth in each well is evidently distinguishable: growing cultures reach a yield comparable to each other but the non-growing cultures do not show any consistent increase in OD600 ( Figure 1C ) . The results , averaged over four similar trials , demonstrate a distinct increase of the MIC as a function of inoculum size ( Figure 1B ) . ( Detailed data presented in Figure 1—figure supplement 1 ) The solid data points in Figure 1B refer to the wells with a growing culture and different marker symbols refer to the number of repeated trial outcomes resulting in growing cultures . A theoretical model developed later in this work nicely fits the average MIC . A separate set of experiments with dense cultures showed that the MIC increases to 3 . 69±0 . 43 μM and 7 . 09±1 . 88 μM for the inoculum sizes of 12 . 2×106 and 24 . 4×106 cells/ml ( 8 replicates with three biological repeats were used for each reported value ) . This is one order of magnitude increase in the MIC value , which can be critical in medical applications . An interesting feature observed in the results was the extended lag phase , up to several hours , introduced by the sub-MIC concentrations of LL37 peptides ( see Figure 1C ) . Despite such growth delay , the average doubling time of the cells ( TD ) did not change significantly , remaining under 30 min in most cases ( see Figure 1D ) . We tested and confirmed the stability of peptides over the duration of the experiment ( Appendix 2 ) . Hence , we hypothesized that this behavior was attributed to heterogenous cell death , where the growth of a fraction of the cells is inhibited , while the rest of the cells recover the normal population growth after a time delay that is correlated to the number of dead cells . This hypothesis is investigated further at the single-cell level . In the microplate experiments , direct cell-to-cell interactions are minimal as the cells are on average over 50 μm apart from each other ( corresponding to inoculum size in Figure 1A , B ) . All the electrostatic interactions are completely shielded . We asked whether the inoculum effect is due to the absorption of peptides into the cells ( Clark et al . , 2009 ) . AMP absorption into bacteria has been previously discussed and quantified using various techniques . Different prokaryotes were reported to absorb 1−20×107 AMPs/cell ( Steiner et al . , 1988; Savini et al . , 2018; Starr et al . , 2016; Tran et al . , 2002; Melo et al . , 2011; Roversi et al . , 2014 ) , which is high enough to initiate the inoculum effect . Here we also quantified the absorption of a dye-tagged analogue of LL37 ( 5-FAM-LC-LL37 , AnaSpec ) by colorimetric measurement of 14 μM of AMPs before and after incubation with E . coli and separation by centrifugation ( see Figure 1E , F and Appendix 3 for details ) . We observed a reduction in the supernatant concentration of AMPs that was proportional to the inoculum size with an average rate of 7 . 6±2 . 1×108 AMPs/cell ( calculated based on a linear fit to the data ) . Since the colorimetric measurements rely on clear solutions , we had to infer the amount of absorbed AMPs by subtracting the supernatant concentration from the initial concentration ( Figure 1F left axis ) . Note that the amount of absorbed AMPs in cultures with AMP concentrations higher than the MIC can well exceed the required amount of AMPs to kill a cell . We further investigated peptide absorption by tracking dye-tagged peptide action on live cells . To this end , we brought E . coli cells from an exponential culture to an imaging platform where they were treated with an above-MIC concentration ( 10 μM ) of 5-FAM-LC-LL37 under agarose gel containing RDM growth media . We closely monitored cell growth and distribution/localization of peptides by phase contrast and fluorescent time-lapse microscopy ( Figure 2A ) . By analyzing 383 cells , we observed that the inhibition of growth is followed by a rapid translocation of peptides into target cells , as quantified by a jump in the cell’s fluorescent signal ( Figure 2B ) . As a result , fluorescent signals showed a bimodal distribution over the course of the experiment . A large degree of temporal , cell-to-cell heterogeneity was also observed as the peptide translocation time varied for about 30 min for different cells ( Figure 2B ) . The fluorescent signal remained high after translocation confirming retention of peptides in the cells . The density of bound AMPs , however , cannot be estimated from the fluorescent signal due to the possibility of self-quenching of the dyes . As such , the signal strength is reduced in the regions with densely packed fluorophores ( Swiecicki et al . , 2016 ) . The instantaneous growth rate of individual cells was non-monotonic and data collapsed onto each other once plotted with reference to peptide translocation time ( Figure 2C ) . There is a drop to negative values , indicating the shrinking of cells , which was found to be synchronized with the uptake of peptides . The growth rate reaches a steady value of zero in ∼10 min . As a whole , E . coli cells exhibited a binary physiological state over the course of the peptide action in terms of growth rate and peptide uptake . That is , the cells were found to be in either of these distinct states: ( 1 ) growing , with no significant peptide uptake; and ( 2 ) non-growing , followed by an abrupt peptide uptake . This is quantitatively evident in the scatter plot of the instantaneous growth rate as a function of fluorescence intensity , where cells segregate into two separate clusters ( Figure 2D ) . At the peptide concentration of 10 μM ( above the MIC ) all cells were initially in state ( 1 ) and then transitioned to state ( 2 ) within one generation . Population-level data from microplates were suggestive , but not conclusive , of a heterogeneous growth inhibition in cultures with a sub-MIC concentration of AMPs . Hence , we proceeded with single-cell experiments as noted above with a sole modification of using a lower , sub-MIC concentration of 5-FAM-LC-LL37 ( 4 . 0 μM for the dye-tagged peptide ) . As such , most individual cells grew to form micro-colonies . The striking observation was the phenotypic heterogeneity in the isogenic population of cells in each micro-colony ( Figure 3A ) . As a colony expanded , growth of some cells was inhibited and peptides translocated into them . We observed a similar transition from state ( 1 ) to state ( 2 ) as previously seen in above-MIC cultures ( Figure 2 ) . The difference is that at sub-MIC the transition occurred for only a fraction of the cells . Analysis of 13 separate micro-colonies , consisting of a total of 280 cells ( over the course of the experiment ) , showed that the relative size of the colonies initially increased exponentially until the appearance of non-growing cells ( Figure 3B ) . The fluorescence intensity of non-growing cells showed an abrupt transition , as in the case of above MIC cultures , with comparable relative changes in the signal , which suggests that a similar number of peptides are taken by each cell ( Figure 3C ) . In order to test whether the absorption of peptides can explain the inoculum effect , we developed a mathematical model with minimal single-cell assumptions ( Figure 4A ) , which describes the time evolution of the mean concentration [B] of growing bacteria and the mean concentration [P] of free AMPs in the solution . The model describes two processes: ( 1 ) bacteria are assumed to divide with a constant rate kD=ln2/TD , where TD≈23min is the average doubling time; ( 2 ) AMPs kill growing bacteria with a rate kk , and afterwards each dead cell quickly takes up N AMPs ( see Figures 2B and 3C ) . These AMPs are bound to the membrane as well as to the cytoplasm of the cell and are not recycled to attack other cells . The time evolution of concentrations of growing bacteria [B] and available AMPs [P] is described by the following equations: ( 1 ) d[B]dt=kD[B]−kk[B][P] , ( 2 ) d[P]dt=−Nkk[B][P] . This model predicts two different outcomes ( see Figure 4B ) depending on the initial concentrations of bacteria and AMPs: ( 1 ) The population of bacteria goes extinct for a sufficiently large concentration of AMPs , that is above MIC; ( 2 ) The population of bacteria can recover in a low concentration of AMPs , that is below MIC . The two unknown parameters of the model were fitted to best approximate the MIC dependence on the inoculum size ( Figure 4C ) . The fit resulted in a killing rate kk=0 . 040μM−1min−1 and N=3 . 8×107 AMPs absorbed per dead cell . Note that the estimated number of absorbed AMPs from the fitting is much smaller than the estimated number of 7 . 6±2 . 1×108 AMPs/cell from the colorimetric measurements . This discrepancy might arise from the fact that cells can divide several times before they stop growing and absorb peptides , which was neglected in the estimate from the colorimetric measurements . Note that in the limit , where the initial concentration of bacteria goes to zero , the MIC value approaches the finite value kD/kk=0 . 75μM ( see Equation 1 ) . This is consistent with a previous model by Stella group ( Savini et al . , 2017 ) , which considered that bacteria get killed once the number of peptides bound to cell membrane reaches a certain threshold . This threshold is expected to be smaller than the number N of absorbed AMPs by dead cells , which includes peptides bound to cell membrane as well as peptides bound to intracellular content . ( Note that the number of surface bound peptides correlates with the concentration of free peptides in solution ) . We further examined whether the model can reproduce other experimental data without additional fitting . In particular we tested whether the model could predict the growth delay in surviving bacterial population when the concentration of AMPs is increased ( see Figure 1C ) . The predictions of our model agree reasonably well with experimental results for the growth delay of population ( see Figure 4D ) , given the simplicity of the model . Deviations likely originate from the fact that cell doubling times are scattered ( Figure 1D ) , while our model assumes a fixed cell doubling time of 23 min . Due to the exponential growth , the growth delay is highly sensitive to the cell doubling time . The temporal heterogeneity we reported in growth inhibition and peptide retention is a key factor for the emergence of the surviving subpopulation . The wide distribution of ∼30 min ( above MIC cultures , Figure 2B ) is puzzling , as all the cells experienced the same environmental conditions . We looked at the correlations of the translocation time with two cell size measures to investigate any dependence on the physiological conditions of the target cells . A negative correlation was observed between peptide translocation time and initial cell length , indicating that small cells can resist the action of AMPs and they continue growing until a later time point ( Figure 5A left panel ) . In contrast , cell length at translocation time did not show any correlation with the translocation time ( Figure 5A right panel ) . This clearly demonstrated that the action of 5-FAM-LC-LL37 is cell-cycle and cell-age dependent . This seems in agreement with the findings made by the Weisshaar lab , where LL37 peptides were observed to first bind to the septum of dividing cells . Thus , the higher chance to act on larger , dividing cells , as opposed to small growing cells ( Sochacki et al . , 2011 ) . The stronger binding of LL37 peptids to the septum area is not well understood but may have multiple physical origins . Among various possibilities , the Wong lab has shown that LL37 peptides preferentially bind to membranes with negative Gaussian curvature , a geometry that can be found in the septum of rod shaped microorganisms ( Yang et al . , 2008; Schmidt et al . , 2011 ) . To investigate the cell-size dependence of action of AMPs we added a sublethal dosage of cephalexin to the growth media ( RDM ) that was shown to slightly increase cell sizes without the loss of viability by delaying cell divisions ( Si et al . , 2017 ) . Our measurements using the microfluidic ‘mother machine’ ( Wang et al . , 2010 ) also showed that 1 μg/mL of cephalexin results in the 17% increase in the length of the newborn cells ( Figure 5B ) , from 3 . 37 μm to 3 . 96 μm , without any appreciable change of the growth rate . Based on the microplate experiments , the MIC of cells grown in the presence of cephalexin increased from 1 . 87±0 . 23 μM to 2 . 09±0 . 19 μM for those grown in regular RDM at the inoculum size of 6 . 1×106 .
Heterogeneities in bacterial response to antibiotics can be critical if leading to the survival of a subpopulation that can recover population growth ( Coates et al . , 2018 ) . In this work , we discovered an unexpected absorption and retention of an antimicrobial peptide ( LL37 and the dye-tagged analogue , 5-FAM-LC-LL37 ) in E . coli cells , which under sub-MIC concentrations led to the emergence of two distinct subpopulations in an isogenic bacterial culture: a group of cells that retain peptides after their growth is inhibited and a group of surviving cells that grow owing to the reduction of the ‘free’ peptide concentration by the other group . This ‘passive cooperation’ is an interesting feature of an E . coli culture where cells do not have any form of active communication , unlike ion-channel based cooperation in Bacillus subtilis biofilms ( Prindle et al . , 2015; Liu et al . , 2017; Liu et al . , 2015 ) . At the cellular and molecular scales , a distinct feature of the action of AMPs is its collective nature , where a large number of AMPs are required to first bind to and then disrupt the cell membranes to kill the cell . Absorption of AMPs has been discussed and quantified previously for different microorganisms and peptides using various techniques ( Steiner et al . , 1988; Savini et al . , 2018; Starr et al . , 2016; Tran et al . , 2002; Melo et al . , 2011; Roversi et al . , 2014 ) . Utilizing live , single-cell microscopy , we observed the temporal and cell-to-cell heterogeneity in the peptide absorption into E . coli cells , which goes beyond the membrane binding . The integration of the population and single-cell data , combined with the theoretical modeling , presented in this work , provides strong evidence that the inoculum effect in the case of LL37 pepides arises from the retention of these peptides in target cells . Despite the seeming complexity of the partitioning of LL37 peptides in a population of bacterial cells , the picture at the single-cell level is simple and binary , consistent in cultures with above-MIC and sub-MIC concentrations of AMPs: occurrence of growth inhibition depends on the free peptide concentration and is followed by an abrupt , permanent translocation of peptides into the target cells . Our quantification of peptide retention ( ∼3 . 8×107 peptides/cell ) is within the range reported using various methods ( Steiner et al . , 1988; Savini et al . , 2018; Starr et al . , 2016; Tran et al . , 2002; Melo et al . , 2011; Roversi et al . , 2014 ) . Yet , this large number raises the question of which molecules inside of the cells are interacting with LL37 peptides . The negative charge of DNA as well as some proteins can provide binding sites for LL37 peptides . To categorically distinguish between these two , we utilized an E . coli strain lacking the septum positioning minCDE system , which produces enucleated mini-cells ( Figure 6A ) . The mini-cells do not contain DNA as confirmed by the localization of a fluorescently tagged histone-like protein hupA ( Figure 6A , see Materials and methods for the details of the strains genotype ) . Translocation of 5-FAM-LC-LL37 peptides into mini-cells showed qualitatively similar absorption and retention as seen in regular cells ( Figure 6B ) , which suggests the presence of significant interactions of AMPs with the intracellular content other than DNA . Yet , the rate at which AMPs are absorbed into the mini-cells is slower than that in the neighboring mother cells with DNA content ( Figure 6B ) , perhaps indicating the role of the negative charge of DNA . While our results provide a quantitative picture of LL37 partitioning and acting in an E . coli population , they open new questions on the molecular and evolutionary basis of their activity . First , the strong absorption of peptides into the cytoplasmic area raises questions on the nature and impact of this intercellular binding: what specific proteins and domains are peptides binding to ? How does the peptide binding perturb the protein functions ? Second , the population survivability as a result of peptide absorption raises questions on the evolutionary dynamics of this phenomena: how do E . coli cells evolve to achieve this cooperative fit ? How does this phenomena affect multi-species cultures with prokaryotic and eukaryotic organisms ? Finally , our findings imply an important dynamics for the activity of LL37 peptides ( and possibly other AMPs ) in the multicellular host organisms . Considering the peptide absorption into the target cells , AMP concentration should not be assumed the only key factor , but the rate of the expression of the AMPs by the host is also decisive in determining the effectivity of AMPs . The expression rate competes with the rate of absorption of the AMPs in the bacterial cells . In the results presented in this work , we focused on closed systems where the total number of AMPs remained constant . As a future direction , one could examine growth inhibition of bacterial cultures experiencing an influx of the AMPs .
In this work , we used derivatives of a prototrophic Escherichia coli K12 strain , NCM3722 , that was constructed , sequenced , and extensively tested by Kustu and Jun labs ( Soupene et al . , 2003; Brown and Jun , 2015 ) . In all microplate and single-cell experiments , we used ST08 , a nonmotile derivative of NCM3722 ( ΔmotA ) ( a gift from Suckjoon Jun’s Lab at the University of California , San Diego ) , except for experiments with mini-cell producing strains where we used ST20 . This strain possesses a deficiency in septum positioning system ( ΔminCDE ) and a DNA marker ( hupA-mRuby2 ) . ST20 was constructed using standard P1 transduction to transfer the gene deletion minCDE::aph from PAL40 ( from Petra Levin’s Lab at the Washington University , St . Louise ) to ST12 , a construct from ST08 with the infusion of mRuby2 fluorescent protein with the histone like protein hupA ( from Suckjoon Jun’s Lab ) . In all experiments , a MOPS based rich defined media ( RDM ) was used , developed by Fred Neidhardt ( Neidhardt et al . , 1974 ) , which is commercially available from Teknova Inc . The average generation time of the E . coli strain used in this study was 23 min in RDM at 37°C . All cells in the experimental samples were grown to early or mid log phase prior to the start of the experiment in a 37°C water bath shaker , set to 240 rpm . Antimicrobial peptides used in this study included LL37 and a dye-tagged version 5-FAM-LC-LL37 . All peptides were purchased from AnaSpec , California . The net peptide content of the purchased material was quantified at 75% by elemental analysis of the carbon , hydrogen , and nitrogen content ( CHN analysis ) through the manufacturing company . The product was shipped in vials as a dry powder . 400 μM stocks suspended in autoclaved double distilled H2O were prepared in-house . The stock was stored in a −20°C cold storage until ready for use . Special handling precautions for the antimicrobial peptides , the stock solution , or solutions containing them were used . Low protein binding supplies including pipette tips ( Low Retention Aerosol Barrier , FisherBrand ) , Eppendorf tubes ( Protein LoBind , Eppendurf ) , and microplates ( Ultra-Low attachment , Corning ) were used . To analyze the microplate data we developed a custom MATLAB ( MathWorks Inc . ) code to measure the growth rate and T0 . 1 for each individual well . The latter quantifies the time at which the OD600 reaches 0 . 1 , as a parameter that shows a ‘delay’ in the growth of the culture . The growth rate is calculated in the standard manner by fitting an exponential line against the growth curve that resembles stable , exponential growth . We have conducted four experimental repeats to construct the data reported in Figure 1 of the manuscript . The full experimental data for each of these four repeats are depicted Figure 1—figure supplement 1 , which shows the growth curves for all 96 wells and the calculated values for the growth curve and T0 . 1 . Each box in the plot refers to one well , where the vertical axis corresponding to each well denotes the log scale of the OD600 ( after subtracting the blank value ) and horizontal axis resembles the time during the 24 hr the experiment took place . The red curves correspond to the growth curve while the black overlap shows the detected exponential growth region and the blue line represents the exponential fit with respect to the data . The two numbers in each box represent the calculated values for the doubling time ( in the top ) and the T0 . 1 ( in the bottom ) , both in terms of minutes . The T0 . 1 is calculated from the initial readings generated by the plate reader . The boxes with no number and gray curve refer to wells that did not show any growth . In order to determine the relationship between the MIC and cell density , we first had to find the correlation between cell density and optical density measured at a wavelength of 600 nm ( OD600 ) . A high precision cell counting experiment was performed using a hemocytometer ( Hausser Scientific 3900 ) and high magnification microscopy . Using different cell cultures with different values of OD600 , we were able to correlate values obtained for OD600 measured on a spectrophotometer ( Genesys 20 , Fisher Scientific ) with the density of E . coli ST08 cells . For the cell counting protocol , cells from the experimental culture at mid exponential phase were harvested , fixed with 0 . 025% formaldehyde , and diluted to four varying concentrations , corresponding to OD600=0 . 115 , 0 . 066 , 0 . 033 , and 0 . 022 . Each of these cultures were transferred to a hemocytometer , allowing for the enumeration of cells . The number of cells where counted in the specified grids of the hemocytometer with dimensions of 50×50×20μm3 . The imaging was performed on a Nikon-Ti microscope with a 40× objective lens using transmission light phase contrast microscopy . For each specific OD600 value , we have used 10 fields of view , each containing 42 grids . The histogram of the cell population in each grid is shown in Figure 1—figure supplement 2A . The average number of cells per grid compared with the average density of cells for each culture produced a strong linear correlation with the measured OD600 values obtained . The linear regression of the cell density as a function of OD600 results in the conversion factor of 6 . 09×108cells/ml ( Figure 1—figure supplement 2B ) . To monitor the inhibition of growth for E . coli cells by antimicrobial peptides in a live microscopy setting we chose to use agarose pads , using 3% low melt agarose gel , to immobilize them . Inspired by previous works ( Moffitt et al . , 2012; Priest et al . , 2017 ) , we developed a system for patterning and housing specific amounts of agarose gel suitable for long term microscopy needs . The patterns are parallel channels that allow cells to spread and move away from one another under the agarose pad , while aligned in certain directions to help with the image analysis and cell segmentation . The housing also reduces the chance of evaporation of the liquid culture , thus allowing the gel to be used for hours at 37°C during microscopy . An inverted microscope ( Nikon Ti-E ) equipped with the Perfect Focus System ( PFS 3 ) , a 100x oil immersion objective lens ( NA 1 . 45 ) , and an Andor Zyla sCMOS camera were used for imaging . The light source that was used for the phase contrast microscopy was made possible with the help of LED transmission light ( TLED , Sutter Instruments 400–700 nm ) and Spectra X light engine ( Lumencor ) , which was used for fluorescent imaging . The illumination condition for phase contrast was 50 ms exposure with an illumination intensity set to 10% of the max TLED intensity . The fluorescent images for the 5-FAM dye were taken with the excitation wavelength of 485 while using a quad band filter ( DAPI/FITC/TRITC/Cy5 , 84000 , Chroma Technologies ) . For temperature control , the microscope was incubated in a custom made plexiglass housing with a forced air heater and temperature controller ( Air-Therm SMT , World Precision Instruments ) set to 37±0 . 1∘°C . The temperature was constantly monitored during the course of the experiment . We developed custom high-throughput image analysis software optimized for segmenting and analysis of individual cells whose growth are inhibited by LL37 or 5-FAM-LC-LL37 under patterned gel agarose gel ( Taheri-Araghi , 2018; copy archived at https://github . com/elifesciences-publications/CellSegmentation ) . For the heterogeneous colonies of growing and non-growing cells , the boundary of the cells were first highlighted manually to increase contrast for segmentation . The procedure includes standard , predefined image enhancement and processing steps available as functions in MATLAB ( MathWorks Inc . ) image processing toolbox . The overall steps are as follows: The instantaneous growth rate is calculated assuming exponential growth for the cells and taking the derivative of the cell length as a function of time . To decrease the noise in the calculations , an exponential curve was fitted to three consecutive datapoint of cell length as a function of time . Differential equations ( Equations 1–2 ) describing the population dynamics model were analyzed in Mathematica with the function NDSolve . For each initial concentration of bacteria and peptides , the evolution of both populations were analyzed during the subsequent Tmax=1000 min . If the concentration of peptides [P ( Tmax ) ] at the end of simulation was smaller than kD/kk , then the bacterial population survived ( see Equations 1–2 ) . On the other hand , if the concentration of peptides [P ( Tmax ) ]>kD/kk , then the bacterial population most likely went extinct . The MIC concentration for a given initial concentration of bacteria was obtained by requiring [P ( Tmax ) ]=kD/kk , which was found using the bisection method . Finally , the unknown model parameters kk and N were obtained by minimizing the error between the values of MIC from the model and the experimental data ( see Figure 4C ) by using the function FindMinimum . | Many organisms use molecules called antimicrobial peptides as a first line of defense against harmful bacteria . For example , in humans , these natural antibiotics are found in the airways , on the skin and in the urinary tract . Because antimicrobial peptides are positively charged , they function in a different way compared to conventional antibiotics . They are attracted to the negatively charged surface of a bacterium , where they latch onto and penetrate through the membrane that encapsulates the microbe . While this mechanism is well studied at the molecular level , very little is known about how antimicrobial peptides spread and interact in a population of bacteria . Snoussi et al . combined several approaches to investigate the dynamics of antimicrobial peptides in Escherichia coli populations of varying densities . Experiments on single cells showed that peptides stopped the growth of bacteria , which were found to be more susceptible during the late stages of their life cycle . The dying cells then absorbed and retained a large number of antimicrobial peptides . This left fewer ‘free’ peptides that could target the other E . coli cells . In fact , when there were not enough peptides to kill all the bacteria , two sub-populations quickly emerged: one group that had stopped dividing – ‘soaking up’ the peptides – and another group that could grow unharmed . This new type of cooperation between threatened E . coli bacteria is passive , as it does not rely on any direct interactions between cells . The results by Snoussi et al . are relevant to medicine , because they highlight the relative importance for the body to produce enough new antimicrobial peptides to replenish the molecules trapped in bacteria . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2018 | Heterogeneous absorption of antimicrobial peptide LL37 in Escherichia coli cells enhances population survivability |
Truncating mutations in the giant sarcomeric protein Titin result in dilated cardiomyopathy and skeletal myopathy . The most severely affected dilated cardiomyopathy patients harbor Titin truncations in the C-terminal two-thirds of the protein , suggesting that mutation position might influence disease mechanism . Using CRISPR/Cas9 technology , we generated six zebrafish lines with Titin truncations in the N-terminal and C-terminal regions . Although all exons were constitutive , C-terminal mutations caused severe myopathy whereas N-terminal mutations demonstrated mild phenotypes . Surprisingly , neither mutation type acted as a dominant negative . Instead , we found a conserved internal promoter at the precise position where divergence in disease severity occurs , with the resulting protein product partially rescuing N-terminal truncations . In addition to its clinical implications , our work may shed light on a long-standing mystery regarding the architecture of the sarcomere .
The use of genetics in clinical medicine depends on knowledge that an identified mutation confers risk of disease . Recent guidance issued by the American College of Medical Genetics emphasizes that the strongest form of evidence supporting causality of a mutation is whether it results in a truncation of the protein ( nonsense , frameshift or canonical splice-site mutations ) , specifically for genes where loss-of-function mutations are known to cause disease ( Richards et al . , 2015 ) . Truncating mutations within alternatively spliced exons and at the extreme C-terminus of the protein must be interpreted with caution , as they may have little to no impact on protein function . However , the guidelines do not address whether truncating mutations at different positions along the length of the protein might differ in phenotype severity or mechanism of action , a finding that could complicate interpretation of truncation mutations for a broad range of inherited diseases . Truncations in the giant sarcomeric protein Titin ( TTN ) reveal patterns that are at odds with this uniform assignment of causality . TTN lies along the length of the sarcomere and is thought to mediate passive tension of the muscle fiber , while also serving as a binding scaffold for a large number of proteins , including actin and myosin ( Lewinter and Granzier , 2013 ) . Individual regions of TTN are annotated according to their localization to the Z-disc , I-band , A-band , and M-line regions of the sarcomere visualized on immuno-electron micrographs Fürst et al . , 1988 . Truncating and/or missense mutations in TTN result in cardiac ( dilated cardiomyopathy [DCM] ) and skeletal muscle ( myofibrillar myopathy ) disease ( Chauveau et al . , 2014 ) . Recent estimates suggest that truncation mutations in TTN may explain as much as 25% of DCM ( Herman et al . , 2012 ) , a disease which affects upwards of 1 in 2500 individuals ( Codd et al . , 1989 ) . Although truncations are found along the length of TTN in DCM patients , in patients with the most severe forms of DCM , truncations reside within the C-terminal two-thirds of the protein ( from amino acid 14 , 760 onwards , Figure 1A ) , specifically in the distal I-band and A-band ( we will refer to these as C-terminal truncations ) ( Herman et al . , 2012; Roberts et al . , 2015 ) . Multiple TTN exons are alternatively spliced ( Bang et al . , 2001 ) , with many having little to no inclusion within cardiac isoforms , and splicing has been proposed to at least partially explain the exclusion of N-terminal mutations in patients with end-stage DCM ( Herman et al . , 2012 ) . Nonetheless , over 5500 amino acids in the N-terminal region of the protein are constitutive ( percent spliced in or PSI >0 . 9 , Figure 1B ) , implying the importance of other mechanisms . 10 . 7554/eLife . 09406 . 003Figure 1 . Generation of six stable zebrafish lines with truncating mutations in ttna , the zebrafish orthologue of TTN , using CRISPR/Cas9 technology . ( A ) Schematic of the TTN meta-transcript ( Ensembl ID ENST00000589042 ) , including all known exons with the exception of the Novex-3 exon , which is exclusive to the Novex-3 isoform ( Ensembl ID ENST00000360870 ) . Sarcomeric regions are denoted by different colors . A heatmap of PSI values computed from RNA-Seq data of two DCM hearts is shown below the cartoon representation of the transcript . Above the schematic , amino acid positions of truncating mutations of TTN in end stage dilated cardiomyopathy identified in a prior study ( Roberts et al . , 2015 ) are highlighted ( gold squares ) . We restricted our representation to nonsense , frameshift or canonical splice-site mutations and did not include predicted disruptions of putative splice enhancers . A truncating mutation mapping to the Novex-3 exon is not shown . Below the schematic , the corresponding human amino acid positions of zebrafish exons disrupted through CRISPR/Cas9 technology in our work are depicted , including six stable mutant lines ( n1 , n2 , n3 , c1 , c2 , c3 , gold diamonds ) and individual exons targeted by high dose guide RNA/Cas9 injection ( gold triangles ) . ( B ) Many of the exons not mutated in end-stage DCM are constitutive ( i . e . show minimal alternative splicing ) . Histogram representing the PSI frequency distribution for TTN amino acids N-terminal to amino acid 14 , 760 , which is the most N-terminal amino acid in which a definitive truncation mutation is found in an end-stage DCM patient . The median PSI value was computed for each TTN exon using RNA-Seq datasets from hearts of two patients with DCM ( GEO Accession GSM1380722 , GSM1380723 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 00310 . 7554/eLife . 09406 . 004Figure 1—figure supplement 1 . Domain organization of human TTN ( top ) , zebrafish ttna ( middle ) and zebrafish ttnb ( bottom ) . In each case a meta-transcript is shown , which includes all possible exons . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 004
We sought to address the basis of this variation of severity with mutation position . For a disease such as DCM , in which inheritance is most commonly observed as adult-onset autosomal dominant with variable penetrance , deciphering such mechanisms in mouse models can be challenging , as heterozygotes tend to show minimal phenotypes and homozygotes tend to be embryonic lethal ( Gramlich et al . , 2009 ) . We selected the vertebrate zebrafish model for several reasons . Firstly , zebrafish has high sequence conservation of sarcomeric proteins with mammals and recapitulate disease phenotypes with loss-of-function of multiple established Mendelian cardiomyopathy genes ( Dahme et al . , 2009 ) . In fact , forward genetic screens have generated zebrafish Titin mutants , although in only one case is the position of the mutation known ( Xu et al . , 2002; Steffen et al . , 2007; Myhre et al . , 2014 ) . Secondly , zebrafish can live for up to two weeks with a non-contractile heart , thus allowing mechanistic studies of homozygote mutants . Thirdly , the relatively low cost and large clutch size of zebrafish allow the ability to generate multiple mutant lines and provide statistical power for robust conclusions . As a result of an ancestral genome duplication event , zebrafish include two titin genes: ttna and ttnb ( Xu et al . , 2002; Steffen et al . , 2007; Seeley et al . , 2007 ) . Ttna shares considerable similarity with the human TTN gene ( Figure 1—figure supplement 1 , showing ‘meta-transcripts’ with all possible exons included ) . The zebrafish and human proteins are similar in length ( ∼32 , 000–35 , 000 amino acids depending on the splice isoform ) and share 56% identity and 82% homology . They have a nearly identical domain organization , with three immunoglobulin-like of domains ( proximal , mid , and distal ) , an elastic PEVK region , N2A and N2B domains which distinguish cardiac and skeletal isoforms , a fibronectin-immunoglobulin repeat region that mediates binding to myosin , and a C-terminal kinase domain . In both species , alternative splicing targets the middle Ig-like and PEVK domains , which modulate passive tensile properties of the sarcomere ( Granzier and Labeit , 2004 ) . Cardiac isoforms typically include much of this elastic region as well as the N2B ± N2A domains , while skeletal isoforms include the N2A domain but exclude N2B as well as much of the elastic region . Multiple mutants isolated from forward genetic screens coupled with morpholino knockdown analysis have demonstrated that ttna is essential for cardiac sarcomere formation in zebrafish ( Xu et al . , 2002; Myhre et al . , 2014; Seeley et al . , 2007 ) . Although the paralagous zebrafish ttnb gene is dispensable for cardiac sarcomere development ( Steffen et al . , 2007; Seeley et al . , 2007 ) , it plays an essential role in skeletal muscle sarcomerogenesis . It is expressed at comparable levels to ttna in skeletal muscle ( as opposed to ∼2/3 of ttna abundance in heart ) . In contrast to human TTN or zebrafish ttna , the ttnb protein has abbreviated versions of the middle Ig-like and PEVK elastic domains ( Figure 1—figure supplement 1 ) and thus more closely resembles skeletal rather than cardiac muscle isoforms of ttna . Morpholino-based skipping of the N2A exon of either ttna or ttnb results in severe disruption of skeletal muscle sarcomeric architecture , highlighting the mutual contribution of the two genes ( Seeley et al . , 2007 ) . Using CRISPR/Cas9 technology ( Jinek et al . , 2012 ) , we engineered six truncating mutations within the major cardiac TTN orthologue , ttna , ( Figure 1A , Table 1 ) , so that the resulting protein product would be truncated at either the Z-disk ( n1 ) , proximal or mid I-band ( n2 and n3 ) or proximal , mid , or distal A-band ( c1 , c2 , and c3 ) . From wild-type ( WT ) zebrafish embryo transcriptome data , we anticipated that all exons would be constitutive in both cardiac and skeletal muscle ( PSI >0 . 99 , Table 1 ) . Mutant lines were bred for two generations to minimize any off-target effects from the genome editing , and we studied the F3 generation for cardiac and skeletal muscle consequences . 10 . 7554/eLife . 09406 . 005Table 1 . Characteristics of six stable zebrafish truncation mutant lines . Size of insertion/deletion , PSI of corresponding exon in embryonic zebrafish heart and skeletal muscle , protein change using standard nomenclature ( 23 ) , and corresponding human amino acid protein for the TTN meta-transcript are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 005Mutant IDMutation typePSI of mutated exon ( heart/skeletal muscle ) Zebrafish protein change#Orthologous human amino acid position*n17bp insertion1 . 000/1 . 000p . E1463fsX31697n21bp insertion1 . 000/1 . 000p . A2815GfsX33048n32bp deletion1 . 000/1 . 000p . Q7597GfsX89693c12bp deletion0 . 996/1 . 000p . E15193GfsX817996c24bp deletion1 . 000/0 . 995p . T18311GfsX821215c38bp insertion1 . 000/0 . 996p . A29052RfsX832003#Amino acid numbering is relative to ENSDART00000109099 . *Amino acid numbering is relative to TTN meta-transcript . As expected ( Xu et al . , 2002; Gramlich et al . , 2015 ) , ttna truncations showed no gross phenotype in heterozygotes , with all heterozygotes attaining sexual maturity . We thus focused on homozygote mutants , with the expectation that partial rescue from ttnb could potentially elicit phenotypic differences among mutants , especially in skeletal muscle where it is expressed at comparable levels to ttna and has a documented essential role . N-terminal ( ttnan1/n1 , ttnan2/n2 , ttnan3/n3 , collectively referred to as ttnan/n ) and C-terminal ( ttnac/c ) truncations all showed severe deficits in cardiac contractility ( Video 1 ) and typically died within two weeks of birth . Immunostaining revealed severe cardiac sarcomeric disarray ( Figure 2A ) . 10 . 7554/eLife . 09406 . 006Figure 2 . C-terminal ttna truncations result in a severe skeletal muscle phenotype while N-terminal truncations are indistinguishable from wild-type . Fixed heart ( A ) and skeletal muscle ( B ) samples of 72 hpf ttnawt/wt , ttnan/n and ttnac/c mutant embryos were analyzed by immunostaining for α–actinin , which highlights Z-disc architecture . The cardiac sarcomere was disarrayed in all mutants . However , in skeletal muscle ttnac/c mutants demonstrated severe sarcomeric disarray while ttnan/n mutants retained sarcomeric architecture . Scale bar: 10 uμm . ( C ) All targeted TTN exons are constitutive ( i . e . not alternatively spliced ) in both cardiac and skeletal muscle . PSI values computed for each mutated exon using RNA-Seq data for dissected hearts and trunk skeletal muscle for various mutant genotypes at 72 hpf . Wild-type fish were analyzed at both 72 hpf and adulthood . Analysis was limited to samples with a sufficient number of exon-exon junction reads to accurately estimate PSI ( Pervouchine et al . , 2013 ) . ( D ) Nonsense-mediated decay reduces mutant transcript levels to ∼20–25% of wild-type , but does not vary substantially across mutations . Targeted RNA-Seq was used to determine the ratio of reads derived from the mutant allele vs . the wild-type allele in ttnan/wtand ttnac/wt heterozygote mutants , which serves as an estimate of nonsense-mediated decay efficiency . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 00610 . 7554/eLife . 09406 . 007Video 1 . Cardiac contraction for wild-type , ttnan/n and ttnac/c mutants . All embryos were imaged at 72 hpf . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 007 In contrast to their cardiac phenotypes , ttna mutants demonstrated a sharp division in development of skeletal muscle disease . All three ttnan/n mutants were capable of grossly normal motion while the ttnac/c truncations were entirely unable to move ( Video 2 ) . Immunostaining supported the distinction , with ttnac/c mutants having highly distorted sarcomeric architecture while ttnan/n architecture closely resembled wild-type ( Figure 2B ) . To confirm transcription of all mutant isoforms , assess nonsense-mediated decay , and rule out selective alternative splicing of ttnan/n mutant exons , we performed RNA-Seq of heart ( cardiac muscle ) and trunk ( skeletal muscle ) of 3 day old embryos , as well as adult fish . All mutated exons appeared constitutive in both heart and skeletal muscle , across all genotypes ( Figure 2C ) . We next examined the extent of nonsense-mediated decay ( NMD ) for each mutant allele by performing targeted RNA sequencing of heterozygote embryos and comparing the number of reads for wild-type and mutant exons . In keeping with typical estimates for NMD efficiency ( Zetoune et al . , 2008 ) , we saw ∼75–80% of mutant transcript being degraded ( Figure 2D ) . However , no substantial variation was seen across the mutants . 10 . 7554/eLife . 09406 . 008Video 2 . Motility assessment for wild-type , ttnan/n and ttnac/c mutants . All embryos were imaged at 72 hpf . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 008 Given that variability in alternative splicing or NMD could not explain the selective skeletal muscle disarray in ttnac/c mutants , we hypothesized that this variation in severity could be due to ttnac/c mutants interfering with the action of ttnb , although the ratio of mutant to wild-type transcript of ∼20% made this an unlikely scenario . Knockdown of ttnb by morpholino disrupted ttnan/n skeletal muscle morphology indicating that the intact skeletal muscle sarcomeric architecture in ttnan/n mutants arose at least in part from rescue by ttnb protein ( Figure 3A ) , although the phenotype was not nearly as severe as that of ttnac/c mutants . To investigate a dominant negative mechanism for ttnac/c mutants , we performed a morpholino knockdown of ttna , targeting a splice site of an N-terminal constitutive exon ( exon 4 ) , which would be expected to introduce an N-terminal frameshift on top of the same C-terminal mutant allele . If the differences between ttnan/n and ttnac/c mutants arose simply from inhibitory effects of the ttnac/c truncated product , then introducing an upstream frameshift should disrupt production of the dominant negative protein product even further and restore sarcomeric architecture . Knockdown by morpholino was effective ( ∼78% knockdown efficiency , data not shown ) and the combined action of NMD and morpholino knockdown would be expected to reduce mutant ttna protein to only ∼5% of the level of ttnb protein . Nonetheless we were unable to rescue skeletal muscle architecture in ttnac/c mutants ( Figure 3B ) . We thus concluded that ttnac/c mutants do not act as dominant negatives and that an alternative explanation was needed . 10 . 7554/eLife . 09406 . 009Figure 3 . C-terminal ttna truncations do not act as dominant negatives . ( A ) Knockdown of ttnb by morpholino injection worsens skeletal muscle sarcomeric architecture in ttnan/n mutants . Fixed skeletal muscle samples of 72 hpf ttnan1/n1 mutant embryos were analyzed by immunostaining for α–actinin . ( right ) Cartoon representation of ttna and ttnb proteins , with a premature N-terminal stop codon in ttna ( gold star ) . The purple lines indicate morpholino disruption of the ttnb transcript . Scale bar: 10 uμm . ( B ) Knockdown of ttna by morpholino injection does not rescue skeletal muscle architecture in ttnac1/c1 mutants . ttnac1/c1 embyros were injected with a ttna splice-site morpholino to the exon 4-intron 4 junction at the 1–2 cell stage and embryos were examined at 72 hpf . At this morpholino dose , knockdown efficacy was estimated at close to 80% and nearly complete cessation of cardiac contraction was achieved in >90% of wild-type embryos ( data not shown ) . Immunostaining for α–actinin revealed no improvement in skeletal muscle architecture . ( right ) Cartoon representation of mutant ttna proteins , with a premature C-terminal stop codon ( gold star ) . The purple lines indicate disruption of the ttna transcript using a morpholino that is expected to introduce an N-terminal truncation upstream of the C-terminal truncation . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 009 Another possible explanation for the discrepancy in disease severity between ttnan/n and ttnac/c mutants could be an internal promoter that produces a C-terminal isoform , which could partially rescue N-terminal truncations , especially in conjunction with expression of ttnb . Even though N- and C-terminal truncating mutations would both disrupt full-length ttna , the latter would also disrupt a C-terminal isoform while the former would not . Interestingly invertebrates , including C . elegans and D . melanogaster , have distinct genes that encode for protein products corresponding to the N-terminal ( Z-disc and proximal I-band ) or C-terminal ( distal I-band and A-band and M-line ) portions of Titin ( Bullard et al . , 2002 ) . The Novex-3 isoform in humans ( discussed below ) does in fact include only the Z-disc and proximal I-band portions of TTN , but no isoform corresponding only to the C-terminal portion of TTN has ever been reported in vertebrates . Using the exon boundaries defined by ttnan3/n3 and ttnac1/c1 , we examined zebrafish and mouse cardiac and skeletal muscle transcriptomic data for patterns of reads indicating a promoter upstream of one of the ttna exons . We noticed an accumulation of reads 5’ to exon 116 ( e116 , ENSDART00000109099 ) in zebrafish skeletal muscle ( Figure 4A ) and in the corresponding exon of mouse heart ( e223 , ENSMUST00000099981 , Figure 4A ) . We hypothesized that these reads arose from transcription from an alternative promoter within the e115-e116 intron . 5’-RACE confirmed an alternative transcription start site ( TSS ) , 110 bp upstream of e116 ( within the e115-e116 intron ) in zebrafish skeletal muscle , with an in-frame initiator methionine located 7 amino acids upstream of e116 . To establish conservation in mammals , we repeated 5’-RACE on adult mouse and fetal human cardiac tissue and found a TSS in the upstream intron ( 257bp upstream of e223 in mouse; 256bp upstream of the orthologous exon , e240 , in humans ) , with a putative initiator methionine 12 amino acids upstream of the relevant exon . Using PCR , we confirmed the presence of a multi-exonic transcript that includes this alternative promoter site in zebrafish , mouse , and human cardiac tissue ( Figure 4B ) . 10 . 7554/eLife . 09406 . 010Figure 4 . An internal promoter at the distal I-band explains phenotypic differences between N- an C-terminal ttna mutants . ( A ) RNA-Seq data from 72 hpf zebrafish trunk ( left ) and E12 . 5 mouse hearts ( right ) depicting accumulation of reads within the intron , upstream of the orthologous exons 116 ( zebrafish ) and 223 ( mouse ) . The Ttn gene is on the negative strand and so reads within the upstream intron are shown to the right of the black bar , which indicates the position of the relevant exon . ( B ) Transcription from an alternative Titin promoter occurs in zebrafish , mouse , and human heart . ( left ) PCR amplification using a primer in the upstream exon or an internal primer at or near the internal TSS ( as determined by 5’-RACE ) as forward primer and a shared reverse primer was performed using cDNA from zebrafish , mouse and human heart . For all three species , products of the expected size were found , supporting transcription from an internal promoter . ( right ) Cartoon representation of PCR amplification scheme , with exon numbering according to zebrafish transcript . The zebrafish 5’ UTR is shorter than that of mouse and human . ( C ) ( left ) H3K4me3 ChIP-Seq data from human fetal leg muscle ( GEO Accession GSM1058781 ) , indicating an active promoter overlies exon 240 , which is orthologous to exon 116 in zebrafish . The peak at the far right of the figure represents the conventional human TTN promoter . ( right ) DNAse-Seq data from human fetal heart ( GEO Accession GSM665830 ) indicates highly accessible chromatin overlying exon 240 , which typically demarcate enhancers or promoters ( Neph et al . , 2013 ) . The accessible chromatin peak at the far right is the conventional TTN promoter . The gold triangle in both panels indicates the genomic position of the human internal promoter TSS identified through 5’-RACE . ( D ) Phenotypic divergence in skeletal myopathy in ttna truncations occurs at the exons flanking the alternative promoter . Cas9 protein and high dose gRNAs corresponding to exon 115 or exon 116 were injected into 1 cell embryos . At 72 hpf , embryos with complete or near complete cessation of cardiac contraction were collected and immunostained for α–actinin to assess skeletal muscle architecture . ( E ) An ATG morpholino targeting the intronic region upstream of the novel initiator methionine is sufficient to create skeletal muscle disarray on a ttnan3/n3 background . 1–2 cell embryos from a ttnan3/wt cross were injected with an intronic morpholino and ttnan3/n3 homozygotes immunostained for α–actinin to assess sarcomere architecture in the skeletal muscle . No skeletal muscle disarray was seen in wild-type fish injected with the internal ATG morpholino ( data not shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 01010 . 7554/eLife . 09406 . 011Figure 4—figure supplement 1 . H3K4me3 Chip-Seq analysis of human fetal muscle ( top , GEO Accession GSM1160200 ) and DNAse-Seq analysis of fetal human heart ( bottom , GEO Accession GSM665817 ) identifies an internal chromatin accessible peak overlying the alternative TTN promoter . A similar peak profile was seen in multiple H3K4me3 and DNAse-Seq data sets ( data not shown ) : GEO Accessions GSM530661 ( fetal heart ) , GSM665824 ( fetal heart ) , GSM701533 ( fetal trunk ) , GSM772735 ( fetal heart ) , GSM1027322 ( pediatric heart ) , and GSM530654 ( fetal heart ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 01110 . 7554/eLife . 09406 . 012Figure 4—figure supplement 2 . The transition between mild and severe skeletal phenotypes occurs at intron 115 , the site of the alternative promoter . One-cell zebrafish embyros were injected with recombinant Cas9 protein and high-dose gRNA , targeting exons 113 , 114 or 122 ( e115 and e116 are shown in Figure 4 ) . Immunostaining for alpha–actinin reveals disrupted skeletal muscle architecture only for e122 . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 012 We next consulted publicly available chromatin accessibility and activation data to find supporting evidence for an internal TTN promoter at this site . DNAse-Seq is a powerful tool to identify regions of accessible chromatin , which typically correspond to promoter or enhancer regions ( Neph et al . , 2013 ) . Examination of multiple fetal human cardiac and skeletal muscle DNA-Seq datasets revealed a prominent peak overlying e240 , the orthologous exon to e116 in humans ( Figure 4C , Figure 4—figure supplement 1 ) . A similar peak was found in H3K4me3 Chip-Seq data , a mark for active promoters ( Zhou et al . , 2010 ) , from skeletal muscle and cardiac tissue ( Figure 4C , Figure 4—figure supplement 1 ) . Unbiased epigenetic data thus provides additional strong support for an internal promoter . For convenience , and as a counterpart to the N-terminal Novex-3 isoform , we hereafter refer to this C-terminal Titin isoform as the Cronos isoform . If a protein product arising from the ttna internal promoter ameliorates disease severity in ttnan/n but not ttnac/c mutants , we should be able to pinpoint the phenotypic transition at e116 – i . e . truncation mutants N-terminal to e116 should have mild phenotypes while those at or C-terminal to e116 should have severe phenotypes . Interestingly , the corresponding human genomic position ( the e239-e240 intron in ENST00000589042 ) represents the exact transition point for where disease severity markedly increases in DCM TTN truncation patients ( Figure 1A ) . We used high dose guide RNA ( gRNA ) and Cas9 protein injections to exons 113 , 114 , 115 , 116 , and 122 and examined cardiac and skeletal muscle architecture and function in the resulting fish , focusing on embryos with nearly complete cessation of cardiac beating , as these would be expected to have a high likelihood of homozygous genomic disruption ( given that only homozygote and not heterozygote ttnan/n or ttnac/c mutants have a cardiac phenotype ) . As expected , e113 , e114 and e115 F0 mutants showed intact skeletal muscle architecture while e116 and e122 mutants had severe skeletal muscle disruption ( Figure 4D , Figure 4—figure supplement 2 ) . To confirm that the Cronos isoform protein product was actually translated , we used agarose protein electrophoresis and found that a higher mobility ( smaller size ) protein product was indeed found in wild-type and ttnan/n mutants but not in ttnac/c mutants ( Figure 5A ) . Finally , a translation start site morpholino targeted upstream of the actual alternative initiator methionine ( and 38 nucleotides upstream of the intron-exon junction ) was sufficient to completely impair motility and disrupt skeletal muscle architecture in the ttnan3/n3 mutant ( Figure 4E , Video 3 ) , further supporting the hypothesis that the Cronos product can rescue skeletal muscle architecture in ttnan/n mutants . No obvious impact of the morpholino on alternative splicing of neighboring exons was seen ( data not shown ) . 10 . 7554/eLife . 09406 . 013Figure 5 . Premature termination codons in ttna differentially affect the levels of Cronos and full-length ttna depending on mutation location . ( A ) 1 . 0% agarose gel reveals higher molecular weight bands corresponding to different splice isoforms of ttnb ( all fish ) and ttna ( wild-type only ) . The arrow represents Cronos isoform of ttna , which is absent in C-terminal truncation mutants and in zebrafish treated with an ATG morpholino against this isoform . Ttna and ttnb expression results in a complex distribution of splice isoforms , whose identity and composition vary over early development ( Steffen et al . , 2007 ) . ( B ) NMD results in a high ratio of Cronos to full-length ttna transcript levels in N-terminal but not C-terminal truncation mutants . Quantitative real-time PCR was used to estimate relative levels of the Cronos and full-length ttna trancripts for wild-type and mutant embryos . No efficiency difference was seen for the corresponding primer pairs . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 01310 . 7554/eLife . 09406 . 014Video 3 . Motility for wild-type and ttnan/n mutants injected with internal promoter ATG morpholinoDOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 014 Given that NMD substantially reduces the levels of ttna transcripts with premature termination codons ( Figure 2D ) , we reasoned that we should be able to distinguish N- and C-terminal truncation mutants by the ratio of Cronos to full-length transcript . In ttnan/n mutants , which would degrade full-length ttna but not Cronos , we would anticipate a higher Cronos:full-length ratio than in ttnac/c mutants mutants , which should degrade both transcripts approximately equally . As expected , we see a much higher ratio of Cronos:full-length transcript levels in ttnan/n mutants than in ttnac/c mutants ( Figure 5B ) . We next surveyed the expression levels of Cronos in cardiac and skeletal muscle in zebrafish at different developmental stages using both in situ hybridization ( Figure 6A ) and real-time PCR ( Figure 6B ) . Cronos is expressed at high levels in zebrafish skeletal muscle during development ( ∼2:1 ratio of Cronos to full-length ttna ) , but drops off sharply during adulthood ( ∼1:70 ratio ) . Although the Cronos expression in zebrafish heart also appears to be developmentally regulated , it is much lower than in skeletal muscle , with a ∼1:5 ratio at 72 hpf and 1:30 ratio in adulthood . This low cardiac Cronos expression ( coupled with a minor role of ttnb in zebrafish hearts ) is a likely reason for the severe and indistinguishable cardiac sarcomere disruption in ttnan/n and ttnac/c mutants . 10 . 7554/eLife . 09406 . 015Figure 6 . Tissue and developmental profile of Cronos and ttna expression in zebrafish and mouse heart and skeletal muscle . ( A ) In situ hybridization at 72 hrpf ( left ) and 24 and 48 hpf ( right ) reveal prominent expression of both Cronos and full length ttna isoforms in zebrafish somites . In contrast , Cronos expression in developing zebrafish heart ( arrow ) is at low levels . ( B ) Quantitative PCR estimates of the ratio of Cronos to full-length ttna reveals prominent expression in developing zebrafish skeletal muscle , but markedly reduced expression in developing heart and in adult heart and skeletal muscle . ( C ) Quantitative PCR estimates of the ratio of Cronos to full-length Ttn reveals prominent expression in developing mouse hearts , which diminishes through development . Skeletal muscle Cronos expression is comparable in early postnatal mouse , but diminished in adulthood and varies by skeletal muscle bed . EDL = extensor digitorum longus . LV/RV = left/right ventricle . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 015 Given that the primary observation that motivated this work was variation in phenotypic severity in cardiomyopathy patients , we next looked to see if there was appreciable Cronos expression in mammalian hearts . We found that Cronos levels were much higher in developing mouse hearts than in zebrafish , with nearly equal transcript levels of Cronos and full-length Ttn ( Figure 6C ) at embryonic day 12 . 5 . Thereafter mouse cardiac expression of Cronos decreases to ∼20–30% of full-length Titin . Skeletal muscle expression varies across development and in different muscle beds with a 1:5 ratio in mouse hindlimb at P2 , and between 1:16 and ∼1:1000 ratios in adult extensor digitorum longus ( fast-twitch ) and soleus ( slow-twitch ) muscle , respectively . Although our zebrafish work explains why a milder phenotype would be expected in individuals with truncation mutants upstream of the alternate promoter , it does not answer whether such N-terminal truncations would still result in deleterious manifestations in heterozygote patients . Specifically , an open question remains whether all TTN truncations lead to cardiac or skeletal muscle phenotypes , or conversely , do any TTN truncations lack cardiac or skeletal muscle consequences . In a disease such as dilated cardiomyopathy , with a variable age of onset , it can be challenging to derive general conclusions from observational data of control subjects , many of whom are young or middle-aged . Furthermore , self-reports of ‘health’ may not reveal limitations in an individual’s ability to attain full exercise potential , especially with aging . To assess whether any TTN truncations are compatible with a high degree of cardiovascular fitness throughout life , we turned to a population genetics approach , and sequenced the complete TTN exome ( as well as that of 104 other Mendelian cardiac disease genes , Supplementary file 1A ) in 199 competitive senior athletes who entered as contestants in the Huntsman World Senior Games ( median age 73 , Table 2 ) . We compared the frequency and distribution of TTN truncation mutations to those found in unselected controls and healthy volunteers from a recent sequencing study ( CTLlit ) ( Roberts et al . , 2015 ) . The overall rate of TTN truncations did not differ between senior athletes and CTLlit ( 1 . 5 vs 1 . 2% , p = 0 . 91 ) . However , the distribution of TTN truncations differed significantly between groups ( Figure 7 , Figure 7—figure supplement 1 ) : TTN truncations in senior athletes mapped exclusively to the extreme C-terminal exon of the rare Novex-3 isoform ( Bang et al . , 2001 ) , at a nearly 8-fold higher rate than CTLlit ( p = 0 . 003 ) . This exon is found only in the Novex-3 isoform , which is expressed at low levels in the human heart ( <10% of the levels of major TTN isoforms [Roberts et al . , 2015] ) , and we would thus expect truncations in this exon would have little effect on TTN function . We next analyzed the ratio of mutants up- and downstream of Cronos in affected DCM patients , literature controls and senior athletes ( Figure 7 ) . As expected , in end-stage DCM we observed a much higher rate of mutants C-terminal to Cronos than N-terminal ( 30:1 ) , followed by unselected DCM ( 37:11 ) , literature controls ( 6:11 , 2:2 , 10:7 for FHS , healthy volunteers and JHS ) and senior athletes ( 0:3 ) . 10 . 7554/eLife . 09406 . 016Table 2 . Demographic characteristics of 199 senior athletes genotyped by capture-based targeted sequencing for this study . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 016Number199Age73 ( IQR 68-77 ) Sex60% MaleRace97% European AmericanSportsTrack and field , volleyball , cycling , softball , marathon , …10 . 7554/eLife . 09406 . 017Figure 7 . TTN truncations are found preferentially upstream of the Cronos promoter in controls but downstream of Cronos in DCM cases . Ratio of the number of truncating mutations found C-terminal to the position of the Cronos promoter to those found N-terminal , in DCM cases and controls derived from the literature ( Roberts et al . , 2015 ) and in senior athletes . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 01710 . 7554/eLife . 09406 . 018Figure 7—figure supplement 1 . Mapping of TTN truncation mutations seen in literature cases and controls ( Roberts et al . , 2015 ) and senior athletes onto a schematic of the TTN meta-transcript . Compared to literature controls , TTN truncations in senior athletes map at a significantly higher fraction to the Novex-3 exon ( 100% vs . 13% , p = 0 . 003 ) . The vertical dotted line delineates the position of the first exon downstream of the Cronos promoter . FHS = Framingham Health Study . JHS = Jackson Heart Study . DOI: http://dx . doi . org/10 . 7554/eLife . 09406 . 018
Loss of function experiments across a number of systems have demonstrated Titin’s importance in diverse aspects of sarcomere development and function , including the ordered assembly of sarcomeric constituents ( Xu et al . , 2002; van der Ven et al . , 2000 ) , cardiomyocyte systolic ( Hinson et al . , 2015 ) and diastolic function ( Radke et al . , 2007 ) , and resistance to neurohormonal and vasoconstrictor stresses ( Gramlich et al . , 2009 ) . Our work further contributes the discovery of a conserved internal promoter in TTN , whose protein product is essential for skeletal muscle sarcomere development in zebrafish . The Cronos promoter is prominently expressed in developing mouse heart , and its location coincides strikingly with the position of TTN truncations in patients with the most severe forms of dilated cardiomyopathy . Our working model is that disruption of both the full-length TTN and Cronos protein products results in more severe disease than disruption of the full-length product alone . Throughout this work , we have focused on skeletal muscle architecture and function in zebrafish ttna homozygote mutants , given that this tissue demonstrates a sharp and rapidly developing phenotypic distinction between N- and C-terminal Titin truncations . The early onset of this unambiguous phenotypic difference combined with the ease of genetic and transcriptomic manipulation using CRISPR/Cas9 and morpholino technology allowed efficient mapping of the novel Cronos internal promoter as well as testing of alternative mechanistic hypotheses for this phenomenon . Nonetheless , implicit in our work is the assumption that findings from zebrafish skeletal muscle sarcomere development in ttna homozygote mutants ( albeit with the paralogous ttnb present for rescue ) are relevant towards understanding disease development in humans with TTN mutations . Prominent differences between these systems must be noted . With the exception of certain childhood-onset forms of skeletal and cardiac myopathy arising from homozygous TTN truncations ( Carmignac et al . , 2007; Ceyhan-Birsoy et al . , 2013 ) , the mode of inheritance in TTN truncation mutation patients is heterozygous with marked variable expressivity and incomplete penetrance ( Herman et al . , 2012; Roberts et al . , 2015; Itoh-Satoh et al . , 2002; Norton et al . , 2013; van Spaendonck-Zwarts et al . , 2014 ) . Moreover , disease onset usually occurs in adulthood . Such an inheritance pattern is typical of many of the familial forms of cardiomyopathy , and is believed to imply additional modifying effects , potentially from diverse sources as autoimmunity and viral infection as well as contributions from additional genetic variants ( Mrh et al . , 2015 ) . Nonetheless , taking all of our results together , we find it a highly plausible explanation that a superimposed deficiency in expression of the Cronos protein product would result in the more severe forms of human cardiac disease seen in DCM patients with C-terminal TTN truncations . With these limitations noted , our results have two clinical implications for titinopathy patients . Firstly , by precisely mapping the location of the alternate promoter to the e239-e240 intron in humans , we are in a better position to give mutation-specific guidance to TTN truncation patients regarding the potential severity of their disease , with the caveat that additional factors ( genetic and environmental ) also modulate phenotypic severity . Along these lines , given the overlapping distribution of mutations in unselected cases and controls , it is unlikely that highly accurate patient prognoses would be realized from genotype information alone ( Figure 7—figure supplement 1 ) , although there does appear to be nearly complete separation of mutation distribution for the more extreme phenotypes ( end-stage DCM and senior athletes ) . Secondly , our work implies that strategies that augment activity of the internal TTN promoter and levels of the Cronos isoform might provide therapeutic benefit in TTN truncation patients , although it is unclear whether there is a restricted developmental window where this might be possible . In addition to clinical ramifications , our work may shed light on a long-standing mystery regarding sarcomeric architecture . Electron microscopy reconstructions of serial sections of skeletal muscle dating back over 50 years presented a puzzle: the thin/thick filament arrangement in the sarcomere alters its geometry , progressing from a tetragonal lattice arrangement at the Z-band to a trigonal arrangement in the A-band ( Pringle , 1968 ) , with the transition taking place near the I-A junction ( Traeger et al . , 1983 ) . The implication for TTN would be that there would be a resulting increase in the copies of TTN protein , with four at the Z-band and six at the A-band and M-line ( Liversage et al . , 2001 ) , which is difficult to reconcile with the fact that a single TTN molecule spans the entire length , from Z-band to M-line . Our data provides a simple explanation for this observation , with the Cronos protein product selectively increasing copies of TTN at the distal I-band , A-band and M-line . Our work also proposes an alternative explanation for the consistently observed ‘T2’ band on protein gels , which migrates at the expected molecular weight of Cronos . This has invariably been labeled a C-terminal degradation product , as it reacts with A-band antibodies ( Opitz et al . , 2004 ) . However , in keeping with Cronos expression patterns in development , the T2 band was prominently seen in developing mouse ( Lahmers et al . , 2004 ) and rat hearts ( Opitz et al . , 2004 ) , but was markedly reduced in adulthood . It is possible that , at least in some situations , T2 and Cronos represent the same isoform . These results also raise a number of questions related to the function of the Cronos isoform . Given its more marked expression in early development , is it important only for the initial sarcomerogenesis or in subsequent aspects of muscle function ? Is it likely to be reactivated under stress ? And how fast does it turn over compared to full-length TTN ? Models with selective disruption and tagging of this isoform should begin to answer these questions . Finally , it is possible that unrecognized internal promoters influence disease severity in truncating mutations of other genes . It is likely a more nuanced diagnostic and prognostic interpretation of truncating mutations will be needed .
sgRNA’s were designed using an in-house Python script , which scans genomic sequences for 23 bp sequences with a terminal NGG ( PAM sequence ) , and prioritizes a GG at the 5’ end because of the use of T7 RNA polymerase for gRNA synthesis . All sgRNA’s had only a single genome match . Oligonucleotides corresponding to the sgRNA sequence were annealed and then cloned into vector DR274 ( Addgene plasmid 42250 ) as previously described ( Hwang et al . , 2013 ) . For sgRNA synthesis , 100 ng of plasmid template was used in a T7 in vitro transcription reaction ( MAXIscript T7 , Life Technologies , Carlsbad , CA , AM1312M ) . RNA products were digested with DNAse I and purified by column ( RNA Clean and Concentrator , Zymo Research , Irvine , CA , R1016 ) . Recombinant Cas9 was prepared as previously described ( Jinek et al . , 2012 ) . A mixture of recombinant Cas9 ( 0 . 5 ug/uL ) and sgRNA 5- ( 40 ng/uL ) was prepared and injected ( 1 nL total volume ) into one-cell stage embryos ( Ekwill strain ) . Cutting efficiency was assessed by pooling injected embryos at 24 hr post fertilization ( hpf ) , extracting genomic DNA , amplifying the genomic targeted region by PCR ( see Supplementary file 1B for primers ) , and using the Cel1 assay ( Oleykowski et al . , 1998 ) to detect evidence of targeted mutations . To pinpoint the exon at which skeletal muscle disarray occurs , recombinant Cas9 and high dose gRNAs ( 100 ng/uL ) corresponding to exons 113 , 114 , 115 , 116 , 122 and the predicted start codon for the alternative promoter were each injected into ∼120 one-cell stage embryos . Given that homozygotes of either N- or C-terminal truncation mutants of ttna cause cessation of cardiac contraction , we focused our attention on those fish with obvious contractile defects and further phenotyped these through motility assays and immunohistochemistry ( see below ) . For each of the targeted exons , 10 embryos with cessation of contraction were analyzed by immunostaining . Zebrafish founders were outcrossed onto the Ekwill background , and subsequent progeny outcrossed again . F2 founders were then in-crossed to obtain F3 generation homozygote mutants , which were used for phenotypic and genomic analysis . Genotyping was performed using the Cel1 assay ( Oleykowski et al . , 1998 ) and/or allele-specific PCR analysis ( Gaudet et al . , 2007 ) . Embryos were fixed and stained using standard procedures ( Deo et al . , 2014 ) . Briefly , anesthetized embryos at 3 dpf were fixed in 4% paraformaldehyde ( PFA ) for 12 hr . Fixed tissue was washed in PBS + 0 . 1% Triton X-100 ( PBST ) , permeabilized with 10 ug/mL proteinase K in PBST , and then refixed with 4% PFA for 20 min after which it was subjected to standard blocking and antibody incubations . Primary antibodies used were mouse anti-human alpha-actinin ( Life Technologies , MA1-22863 , 1:500 dilution ) and titin T11 ( Sigma-Aldrich , St . Louis , MO , T9030 , 1:100 ) and T12 ( 1:10 , a generous gift from Elisabeth Ehler and Dieter Fuerst ) . The secondary antibody was donkey-anti-mouse conjugated to an Alexa-488 fluorophore ( Life Technologies , R37114 , 1:500 ) . Videos at 100–250 frame per second of beating hearts were taken from live 72 hr post-fertilization ( hpf ) embryos using an Axiozoom V16 Stereo ( Carl Zeiss , Oberkochen , Germany ) with a 10× objective lens and an Orca Flash 4 . 0T high-speed digital camera ( Hamamatsu Photonics , Hamamatsu City , Shizuoka , Japan ) . Confocal images were collected on a Leica TCS SP5 with a 63x objective lens . To assess motility , embryos were tapped gently with a whisker . ATG- and splice-blocking morpholinos to ttna ( conventional and alternative promoter ) and ttnb were designed with the help of Gene Tools staff ( http://www . gene-tools . com/ ) . Morpholino sequences were TAA ATG TTG GAG CTT GCG TTG ACA T ( ATG ) and AAT TAC CTT TGA TTG TGA CTT TGG T ( Seeley et al . , 2007 ) for ttnb and CCT GTG TTG TCA TGG TGG AAG GCA C ( conventional ATG ) , GTT CCA GGA GAC ACA GGT AAT CCC T ( internal ATG ) and TGC AAC TGA TAC TCA CCT TCT CCA C ( exon4-intron4 junction ) for ttna . Morpholinos were re-suspended in sterile water to a concentration of 1 mM and diluted to working concentration in sterilized water . Male and female wild-type adult zebrafish were housed and embryos bred using standard protocols ( Westerfield , 1993 ) . Morpholinos were introduced into the zebrafish yolk via microinjection at no later than the two-cell developmental stage . Injected embryos were then kept at 28 . 5°C in E3 solution ( 5 mM NaCl , 0 . 17 mM KCl , 0 . 33 mM CaCl2 , 0 . 33 mM MgSO4 ) . The conventional promoter ttna/ttnb ATG-morpholinos were highly toxic , even at low concentrations . We thus focused on using the ttna and ttnb splice-blocking morpholino’s as well as the internal promoter ttna ATG morpholino at working concentrations of 5 ng/nl . To further reduce toxicity , we co-injected the ttna splice-blocking morpholino with a previously described p53 morpholino at 5 ng/nl ( Robu et al . , 2007 ) . In terms of sample size , we typically injected 100–150 embryos for every experiment . High depth RNA-Sequencing of wild-type embryonic hearts was first performed to select constitutive exons for gRNA targeting . Embryo heart purification was carried out as described ( Geoffrey Burns and Macrae , 2006 ) . Briefly , 300–400 embryos of Tg ( cmcl2-GFP ) at 3 dpf were pooled for a single sequencing experiment . Embryo Dissociation Medium ( EDM ) was freshly prepared using Leibovitz’s L-15 Medium ( Life Technologies , 11415–064 ) containing 10% FBS ( Sigma , F-2442 ) , and kept ice-cold prior to use . After washes with EDM , embryos in 1 ml of EDM were drawn to a 3 ml syringe with 19G needle and expelled back to the tube for 30 times . Large embryo fragments were filtered out using a 100 um filter , and the filter was washed several times with more EDM . The flow-through was collected in a 60-mm glass petri dish on ice , and then applied to a 40 um filter . The filter was inverted and the retained fragments including hearts were washed off into a glass dish with 5x1ml EDM . GFP + hearts were sorted under fluorescent light using a Leica ( Leica Camera , Wetzar , Germany ) M205 FA fluorescence microscope . Trizol-extracted RNA was purified by column ( RNA Clean and Concentrator , Zymo Research , R1016 ) . Sequencing libraries were prepared using the SureSelect Strand Specific RNA-Seq kit ( Agilent Technologies , Santa Clara , CA , G9691A ) , following the manufacturer’s instructions . 101 bp paired end sequencing was performed on a HiSeq 2500 sequencer ( Illumina , San Diego , CA ) . For analysis of splicing of ttna mutants , hearts and skeletal muscle ( trunk ) were isolated from ttna mutant and wild-type zebrafish embryos ( 72 hrpf ) and adults using manual dissection after euthanasia . Pooled hearts and skeletal muscle from 10 fish were used for subsequent library preparation . The tissue was lysed in Trizol ( Life Technologies ) . RNA was recovered from the aqueous phase and purified purified by column ( RNA Clean and Concentrator , Zymo Research , R1016 ) according to manufacturer’s protocol . cDNA was prepared using the Ovation RNA-Seq System V2 ( NuGEN Technologies ) with sonication into 200bp fragments using an M220 sonicator ( Covaris , Woburn , MA ) . Sequencing libraries were prepared using the Ovation Ultralow Ultralow DR Library System ( NuGEN Technologies , San Carlos , CA ) . 101 bp single end sequencing was performed on a HiSeq 2500 sequencer in rapid run mode , with 7 samples per lane of the flow cell . Reads were mapped to the dr7 build of the zebrafish genome using TopHat ( Trapnell et al . , 2009 ) . Samples were normalized using the edgeR package ( Robinson et al . , 2010 ) , with trimmed mean of M values ( TMM ) normalization . Percent spliced in estimates were made using in-house Python scripts implementing previously published methods of splice isoform quantification using exon-exon junction reads ( Pervouchine et al . , 2013 ) , which avoid the estimation difficulties arising from non-uniform read coverage of exons ( Kakaradov et al . , 2012 ) . The heart from an E12 . 5 129X1/SvJ x C57BL/6 cross mouse was disrupted in Trizol with plastic pestle and passed through QIAshredder ( Qiagen , Hilden , Germany , 79654 ) for homogenization . RNA was isolated from Trizol using a 1:1 mixture with 70% ETOH followed by column purification and DNase treatment ( RNeasy Mini kit , Qiagen , 74104 ) . The cDNA was generated and amplified using the Ovation RNA-Seq System V2 ( NuGEN Technologies ) and the library was prepared using the Ovation Ultralow Ultralow DR Library System ( NuGEN Technologies ) . 101 bp single end sequencing was performed on a HiSeq 2500 sequencer . Reads were mapped to the mm9 build of the mouse genome using TopHat ( Trapnell et al . , 2009 ) . To estimate nonsense-mediated decay ( NMD ) rates , we evaluated allelic imbalance in heterozygote embryos . For each mutant of interest , total RNA was extracted from three heterozygote embryos . Primers , designed to span at least one exon-exon junction , were used to amplify ∼100 bp surrounding the mutation of interest . For each embryo , 2 independent PCR reactions - each with 10 amplification cycles - was performed . Next generation sequencing libraries were prepared using the NEBNext Ultra kit ( New England Biolabs , Ipswich , MA , E7370S ) with 10 cycles post-amplification . Samples were assessed by 100 bp single end sequencing on a MiSeq instrument ( Illumina ) . The ratio of mutant to wild-type reads was used as a measure of allelic imbalance . For each mutation , between 600 and 10 , 000 reads were used for ratio estimates . Forward primers mapping to either e115 or the Cronos isoform was used in combination with a reverse primer spanning the e116-e117 exon-exon junction . Similar primers were designed to detect Cronos vs . full-length Ttn levels in mouse . Standard curves using a cloned PCR product were used to compare relative primer efficiencies ( data not shown ) . Quantitative real time-PCR analyses were carried out with cDNA templates using SensiFast SYBR green ( Bioline Reagents , London , UK ) on the CFX384 Real Time PCR system ( BioRAD , Hercules , CA ) ( Heredia et al . , 2013 ) . Zebrafish samples for qPCR consisted of embryonic and adult hearts and trunk muscle , and were isolated as described above . Mouse heart and skeletal muscle samples were taken from C57BL/6 or CD1 mice purchased from Jackson laboratory . All mice were euthanized by CO2 inhalation and cervical dislocation prior to surgical dissection of relevant tissues . We downloaded fastq format files for DCM patients and healthy controls corresponding to GEO Accessions GSM1380718 , GSM1380719 , GSM1380722 , and GSM1380723 . Reads were mapped to the hg19 build of the human genome using TopHat ( Trapnell et al . , 2009 ) . PSI values were computed as described above . We downloaded bed format files corresponding to GEO Accessions GSM665817 ( fetal heart ) , GSM530661 ( fetal heart ) , GSM665817 ( fetal heart ) , GSM665824 ( fetal heart ) , GSM665830 ( fetal heart ) , GSM701533 ( fetal trunk ) , GSM772735 ( fetal heart ) , GSM906406 ( left ventricle ) , GSM1027322 ( pediatric heart ) , GSM1058781 ( fetal leg ) , GSM1160200 ( fetal trunk ) , and GSM530654 ( fetal heart ) . Depth of coverage analysis was performed using in-house R scripts and data plotted using the ggplot2 package ( Wickham , 2009 ) . 5’ RACE was performed using published protocols ( Picelli et al . , 2014; Matz et al . , 1999 ) . An oligonucleotide for template switching ( TSO , Univ_UMI_RACE , Supplementary file 1B ) was purchased from Exiqon ( Woburn , MA ) . Total RNA was obtained from three sources . Zebrafish RNA was isolated from zebrafish skeletal muscle , as described above . Mouse RNA was isolated from cardiac tissue of a euthanized adult mouse ( C57BL/6 ) using manual dissection of the heart followed by tissue homogenization , Trizol extraction , and RNA column purification ( RNA Clean and Concentrator , Zymo Research , R1016 ) . Fetal human heart ( 31 week old , male ) RNA was purchased ( BioChain Institute , Newark , CA , R1244122-50 ) . First strand synthesis was performed using a TTN gene-specific primer ( Supplementary file 1B ) and the TSO , with SmartScribe reverse transcriptase ( Clontech , Mountain View , CA , 639536 ) and with added SUPERase In RNAse inhibitor ( Life Technologies , AM2694 ) . Uracil-DNA Glycosylase ( New England Biolabs , M0280S ) was used to degrade any remaining TSO according to manufacturer’s instructions . PCR amplification of the resulting cDNA was performed using a nested gene-specific primer ( Supplementary file 1B ) and a universal 5’ primer ( Supplementary file 1B ) with a touch-down annealing temperature protocol . PCR products were gel purified ( Zymo Research , D4001 ) , blunt-end cloned using the pGEM-T Easy Vector system ( Promega , Madison , WI , A12360 ) and analyzed by Sanger sequencing . Whole-mount in situ hybridization with digoxigenin-labeled mRNA antisense probes to Cronos and ttna was performed as previously described ( Thisse and Thisse , 2008 ) . For Cronos , an 89nt digoxigenin-labeled riboprobe unique to this isoform was hybridized at 60°C . For ttna , a 453nt riboprobe corresponding to a constitutive region was hybridized at 68°C . In both cases embryos were developed in BM purple ( 11442074001 , Sigma-Aldrich ) . Briefly , euthanized zebrafish embryos were Dounce homogenized in lysis buffer ( 500 mM Tris pH 7 . 4 150 mmM NaCl , 1% NP-40 , 1 mM PMSF , 1x protease inhibitor cocktail [Sigma-Aldrich] ) on ice and centrifuged at 14 , 000 g for 10 min at 4C . The denaturing/loading buffer ( 8 M urea , 2 M thiourea , 3% SDS , 75 mM DTT , 0 . 1% bromophenol blue , 0 . 05% Tris-HCl pH 6 . 8 ) was mixed with the lysate at a 4:1 ratio and the resulting sample denatured at 60°C for 15 min . Electrophoresis was performed using a denaturing 1% agarose gel at 15mA for 3–6 hr ( 3 hr for Myosin and 6 hr for Titin ) on a Hoefer ( Holliston , MA ) SE 600 gel unit ( Steffen , et al . , 2007; Warren , et al . , 2003 ) . Gels were fixed for at least 1 hr after running , briefly washed in water and vacuum dried at 40C overnight . The following day , the gel was rehydrated and stained with SYPRO Ruby protein gel stain ( S4942 , Sigma-Aldrich ) according to manufacturer’s instructions . 199 healthy senior athletes with no prior history of cardiac disease were recruited between 2004 and 2012 at the Huntsman World Senior Games ( held in St George , Utah ) . DNA from whole blood was extracted using the Gentra Puregene DNA isolation kit ( Qiagen , 158445 ) . Targeted capture and variant calling were performed as previously described ( Deo , et al . , 2014 ) . Briefly , NimbleGen’s SeqCap EZ Choice technology was used to construct oligonucleotide probes complementary to the exons ( Ensembl GRCh37 ) of 160 genes involved or predicted to be involved in Mendelian forms of cardiac disease ( Supplementary file 1A ) . Genomic DNA libraries were sheared by sonication on a M220 sonicator ( Covaris ) , hybrdized to the capture probes , and libraries prepared from the captured material using the Kapa Biosystems ( Wilmington , MA ) Library Preparation Kit . Libraries were sequenced on an Illumina HiSeq 2500 sequencer with multiplexing of 24 samples per lane . Assembly was performed using bwa ( Li and Durbin , 2009 ) and variants called using the GATK ( McKenna , et al . , 2010 ) . Median coverage was 455-fold with 99 . 6% of exons found to be callable for the major cardiac N2BA isoform ( ENST00000591111 ) and 98 . 4% callable for the Inferred Complete isoform ( ENST00000589042 ) . Seven samples were captured with a prior cardiomyopathy gene panel of 116 genes ( Deo , et al . , 2014 ) . Coding sequence mutations resulting in a premature nonsense codon or frame-shift were considered to be truncating mutations . For mutations of canonical ( -1/-2 ) splice acceptor variants , which would most likely result in exon skipping , we further required that the downstream exon was not in the same codon phase as the skipped exon , as this would fail to produce a frame-shift and thus not result in a truncation ( such a situation is common in the I-band ) . Likewise , for mutations of canonical splice donor variants ( +1/+2 ) , which we predicted would result in intron read-through , we performed in silico translation of the resulting retained intron to ensure that a premature truncation was observed ( this was not always the case , especially for the I-band ) . All statistical analysis was performed in R ( 3 . 1 . 1 ) . A difference in proportion test was applied to the prevalence of TTN truncations in senior athletes vs . literature controls , as well as the fraction of TTN truncations that map to the Novex-3 exon . | The heart is able to beat partly because of a large protein called Titin that helps to give heart muscle its elasticity . Mutations that shorten the gene that encodes Titin can cause part of the heart to become enlarged and weakened , a condition called dilated cardiomyopathy . Some people with shortened copies of this protein have a mild form of cardiomyopathy and are able to lead relatively normal lives . Others develop more severe symptoms that prevent the heart from pumping blood effectively and may even cause the individual to need a heart transplant . Genetic studies have revealed that mutations that shorten the Titin protein by disrupting the portion of the gene corresponding to the latter two-thirds of the protein ( which encodes the so-called “C-terminal” end of the protein ) cause more severe symptoms than mutations that occur near the start of the gene . But it is not clear why the location of the mutation matters . To investigate this problem , Zou et al . used a gene-editing tool called CRISPR to create genetically engineered zebrafish . These fish had mutations at one of six different points in the gene that encodes the zebrafish version of Titin . Just as with humans , mutations near the C-terminal end of the gene caused more severe muscle problems in the fish . Specifically , Zou et al . found that the worst disease was associated with mutations that occurred at or after a “promoter” region within the gene and near this C-terminal end . Normally , the promoter produces an independent smaller form of the Titin protein , which helps to reduce the severity of muscle problems in zebrafish that have mutations near the start of the gene . However , mutations near the C-terminal end of the gene also damage this smaller form , preventing this failsafe from working , and so lead to more severe symptoms . Zou et al . also found this promoter to be active in both mouse and human hearts . Future work will focus on learning how this smaller form of Titin works to help muscle develop and withstand stress and determine whether increasing its production can overcome the more severe forms of disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"genetics",
"and",
"genomics"
] | 2015 | An internal promoter underlies the difference in disease severity between N- and C-terminal truncation mutations of Titin in zebrafish |
Mammalian cerebral cortex is accepted as being critical for voluntary motor control , but what functions depend on cortex is still unclear . Here we used rapid , reversible optogenetic inhibition to test the role of cortex during a head-fixed task in which mice reach , grab , and eat a food pellet . Sudden cortical inhibition blocked initiation or froze execution of this skilled prehension behavior , but left untrained forelimb movements unaffected . Unexpectedly , kinematically normal prehension occurred immediately after cortical inhibition , even during rest periods lacking cue and pellet . This ‘rebound’ prehension was only evoked in trained and food-deprived animals , suggesting that a motivation-gated motor engram sufficient to evoke prehension is activated at inhibition’s end . These results demonstrate the necessity and sufficiency of cortical activity for enacting a learned skill .
Execution of complex voluntary movements depends on many functions including choosing a behavior , specifying each step , enacting the movements , and learning to perform with increasing skill . Motor control is achieved by the coordinated activity of myriad neural structures , including the cerebral cortex , basal ganglia , cerebellum , and spinal cord . Neurophysiological recordings , inactivation experiments , and stimulation studies have been used to describe the role of cortex in motor control . Cortical neurons display dynamic activity patterns during movement planning and execution , but their functional contribution to skilled action cannot be assessed from recordings alone ( Lemon , 1993; Scott , 2003; Evarts , 2011 ) . Previous inactivation experiments have suggested the cortex tunes—rather than initiates and executes—motor programs to achieve dexterous movement ( Walker and Fulton , 1938; Lawrence and Kuypers , 1968; Castro , 1972; Passingham et al . , 1983; Martin and Ghez , 1991; Whishaw , 2000; Fogassi et al . , 2001; Peters et al . , 2014 ) . The classical perturbation methods used in these studies lack temporal specificity and allow for the emergence of compensatory mechanisms ( but see supplement of Peters et al . , 2014 ) . Electrical stimulation of cortex has been shown to be sufficient to elicit complex movements , but it remains difficult to resolve if evoked behaviors are caused by direct cortical neuron stimulation or antidromic activation of inputs to cortex ( Ferrier , 1873; Penfield and Boldrey , 1937; Penfield , 1954; Gottlieb et al . , 1993; Graziano et al . , 2002; Ramanathan et al . , 2006; Harrison et al . , 2012 ) . Thus the precise role of cortex in skilled movement remains unclear . Here we inhibited cortex with an optogenetic method that minimizes compensation , and selective stimulation of cortical neurons was achieved by the cessation of inhibition .
To test the role of the cortex for control of skilled motor tasks , we developed a multi-step , cued , prehension task in which head-fixed mice reach for a food pellet , grab it , and bring the pellet to the mouth for consumption ( Figure 1A , Video 1 ) . This prehension behavior was broken down into six sequential epochs ( Lift , Hand open , Grab , Supinate , At mouth , and Chew; Figure 1B and C ) that divide the complex movement into components that involve distinct muscle activations . Unlike previous paradigms , our task employed head fixation to reduce postural variability , allow examination of reaction time , and enable future physiological investigations ( Whishaw et al . , 1991; Osborne and Dudman , 2014 ) . All animals that underwent training learned the task . Reaching for the pellet could be achieved in 5–10 sessions and stable performance ( median success rate was 74% , n = 13 animals ) could be realized in approximately 15 sessions . We used high-speed video and machine-learning-based tracking to semi-automatically convert filmed movements into hand trajectories ( see Materials and methods , Figure 1D and E , Video 2 ) . Stereotypy of head-fixed prehension can be seen in sequential trajectories from a single animal during a behavioral session ( Figure 1E ) . 10 . 7554/eLife . 10774 . 003Figure 1 . Head-fixed , cued , multi-step , prehension task . ( A ) Schematic of behavioral arena . Mice reached for and grabbed a food pellet presented on a turntable after an auditory cue . High-speed , near-infrared-sensitive cameras captured video of behavior from perpendicular angles . ( B ) Timeline of prehension components . Prehension behavior consists of the following elements: lift hand from perch , open hand while reaching , grab pellet , supinate hand , elevate hand to mouth , and chew pellet . ( C ) Example video frames ( front or side views ) of each prehension component; times are relative to cue . ( D ) Visualization of iterations of the Cascaded Pose Regression algorithm used to track the forelimb . Each line corresponds to a random initialization ( n = 50 initializations ) and represents the pose estimates for 100 successive iterations of the tracking algorithm ( color indicates iteration ) . Red dots indicate the distribution of final pose estimates . Our algorithm finds a smooth trajectory over time with high density in this distribution across all frames . ( E ) Three-dimensional view of stereotyped prehension trajectories ( 50 trials ) of a well-trained animal extracted using the Cascaded Pose Regression algorithm . Behavioral components in the trajectories are color coded as indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 00310 . 7554/eLife . 10774 . 004Video 1 . Example of successful control prehension . Side and front views of head-fixed prehension task . Animal responded to auditory cue and pellet delivery by executing stereotyped components of prehension , acquiring food pellet in single attempt . Hand position ( circle ) and trajectory ( line ) as determined by a Cascaded Pose Regression algorithm ( color represents behavioral component as indicated ) . Playback at 50 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 00410 . 7554/eLife . 10774 . 005Video 2 . Visualization of the Cascaded Pose Regression algorithm used to track the forelimb for each frame of a video . In each frame , each line represents 1 of 50 random initializations ( in space ) , with the blue to red color coding representing the progression through the first 10 of 100 iterations of the current frame’s pose estimate . Playback at 50 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 005 While chronic cortical perturbations degrade prehension performance in freely moving animals , such experiments allow the animal sufficient time to learn behavioral strategies to circumvent affected circuits ( Whishaw et al . , 1991; Whishaw et al . , 1993 ) . To avoid this confound , we employed rapid onset , within-trial suppression by light-activating channelrhodopsin ( ChR2 ) -expressing inhibitory neurons ( expressing the vesicular GABA transporter [Slc32a1] ) in cortex with pulses of light ( Figure 2 ) ( Guo et al . , 2014 ) . This method presumes that activation of Slc32a1-COP4*H134R/EYFP-positive neurons selectively suppresses cortex , although other possibilities do exist ( for evidence against , see methods ) . Cortical inhibition was limited in space ( Figure 2—figure supplement 1 ) and time by regulating the position and duration of light exposure . To limit compensation , light was delivered on less than 20% of trials during a behavioral session . We first applied inhibition at cue onset to test the role of cortex for behavior initiation . The only effective area of cortical suppression was centered over contralateral forelimb motor cortex ( Figure 2B–E , Video 3 ) . Because forelimb motor and sensory cortices are overlapping at this resolution in the rodent , we will refer to this region as sensorimotor ( SM ) cortex ( Ayling et al . , 2009 ) . 10 . 7554/eLife . 10774 . 006Figure 2 . Initiation of head-fixed prehension behavior requires contralateral sensorimotor cortex . Optogenetic stimulation of GABAergic inhibitory neurons in Slc32a1-COP4*H134R/EYFP mice was used to suppress specific cortical areas , as depicted in the schematic . Laser emission ( 40 Hz sinusoidal train for 4 s , blue sinusoidal lines ) directed through a cleared skull over each cortical area . Depictions of representative ethograms and cumulative histograms for a representative animal . Cumulative histogram is the running total ( for all trials ) of the occurrence of first list . ( A ) Tone-cued prehension behavior under control conditions with no optogenetic stimulation . ( B–E ) Optogenetic inhibition of cortical regions at trial onset . ( B ) Contralateral sensorimotor cortex ( cSM , yellow shading ) . ( C ) Ipsilateral sensorimotor cortex ( iSM ) . ( D ) Visual cortex ( V1 ) . ( E ) Frontal cortex ( FrC ) . Optogenetic inhibition of cSM; but not iSM , V1 , or FrC; prevented prehension . Initiation rates over bracketed time period for control and laser trials are listed per condition . Color-coded bars denote behavioral components for ( A–E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 00610 . 7554/eLife . 10774 . 007Figure 2—figure supplement 1 . Quantification of optogenetic laser intensity at different neural depths . ( A ) Schematic of laser intensity measurements . A custom-made , quantum dot-based photodetector ( expanded on right , red is quantum dot fluorescence ) and a fiber-coupled 473 nm laser were positioned over a craniotomy above SM cortex . The photodetector was moved at 0 . 1 mm steps into the brain . ( B ) Relationship between depth from cortical surface and light intensity . Inset , representative spectrogram of quantum-dot fluorescence following 473 nm laser excitation at 0 . 7 mm below the dural surface ( corresponding to red circle in plot , left ) . Quantum dot emission at 653 nm ( normalized arbitrary units ) was used to measure intensity of 473 nm light . Data are the mean of three animals , bars are standard deviation . ( C ) Relationship between depth from cortical surface and light intensity on log10 scale . Laser intensity decreased more than 225-fold between surface and 2 . 0 mm deep ( red circle , corresponding to the dorsal border of striatum ) . Note the deviation from exponential laser decay between 1 . 5 and 2 . 0 mm depths , potentially caused by densely packed white matter tracks . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 00710 . 7554/eLife . 10774 . 008Figure 2—figure supplement 2 . Cortical inhibition does not affect cued licking behavior . Slc32a1-COP4*H134R/EYFP mice were trained to lick to obtain a food pellet after an auditory cue . ( A ) Image of animal performing the cued licking task . ( B ) Ethograms of cued licking ( gray bars ) . Gray box contains control trials , blue box contains laser stimulation trials . Blue sinusoidal line marks laser duration . ( C ) Comparison of initiation time ( left ) and success rates ( right ) for control ( gray ) and laser-stimulation ( blue ) trials . Activation of cortical inhibitory neurons ( even in cortical areas previously reported to be involved in licking tasks ) ( Guo et al . , 2014 ) did not disrupt lick initiation or execution ( n = 84 control trials , n = 32 laser trials , t-test with unequal variance , p>0 . 05 ) . The licking task may not be cortically ( unilaterally ) dependent because the movement is too simple ( one-step task ) or the task may lack errors that engage the cortex . cSM , contralateral sensorimotor cortexDOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 00810 . 7554/eLife . 10774 . 009Video 3 . Optogenetic inhibition of contralateral sensorimotor cortex prevents prehension initiation . Auditory cue and pellet delivery were paired with contralateral sensorimotor cortex inhibition ( 4 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice , preventing mice from initiating prehension . Prehension was initiated at the termination of cortical inhibition . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 009 Contralateral sensorimotor cortex suppression at the beginning of an individual trial eliminated initiation of the prehension behavior ( initiation only occurred in 35 of 459 trials during contralateral sensorimotor cortex inhibition , compared to initiation in 3675 of 3692 control trials , n = 8 animals , Figure 2A and B , Video 3 ) . Prehension reliably occurred after optogenetic cortical suppression was relinquished ( initiation in 436 of 467 trials , n = 8 animals ) , Figure 2B , Video 3 ) . Inhibition of other cortical areas did not affect behavioral initiation rates ( ipsilateral sensorimotor ( initiation in 124 of 124 trials , n = 3 animals , Figure 2C , Video 4 ) , visual ( initiation in 122 of 124 trials , n = 3 animals , Figure 2D ) , and frontal ( initiation in 124 of 124 trials , n = 3 animals , Figure 2E ) cortex ) . In contrast to prehension , voluntary grooming was not impacted by sensorimotor cortical suppression ( 11/11 trials , n = 4 animals , Video 5 ) . To rule out that initiation failure was due to an inability to process the cue , we activated cortical inhibitory neurons during a cued licking task in which a pellet of food positioned close to the mouth was captured by the tongue; this behavior was not affected by sensorimotor inhibition ( Figure 2—figure supplement 2 , Videos 6 and 7 ) . During this cued licking task , animals sometimes moved their forelimb from the perch to their mouth/table ( 2/3 animals , Video 6 ) . While these movements shared some features with the trained prehension behavior , they were not kinematically stereotyped . Contralateral sensorimotor cortical suppression did not prevent these forelimb movements during cued licking ( forelimb movement in 44/44 control trials and 20/20 laser trials for two animals , Video 7 ) . These results indicate that contralateral sensorimotor cortex activity is necessary for initiation of a learned , goal-directed , skilled prehension behavior . 10 . 7554/eLife . 10774 . 010Video 4 . Optogenetic inhibition of ipsilateral sensorimotor cortex does not affect prehension . Auditory cue and pellet delivery were paired with ipsilateral sensorimotor cortex inhibition ( 4 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice . Mouse performed normal prehension behavior without delay and successfully acquired pellet in single attempt during inhibition of ipsilateral sensorimotor cortex . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 01010 . 7554/eLife . 10774 . 011Video 5 . Optogenetic inhibition of contralateral sensorimotor cortex does not affect forelimb movements during grooming . Auditory cue and pellet delivery delivered during grooming . Contralateral sensorimotor cortex inhibition ( 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice did not prevent ongoing grooming . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 01110 . 7554/eLife . 10774 . 012Video 6 . Cued licking task . Set-up is similar to cued prehension task except food pellet was delivered in close proximity to the tongue . After the cue and delay , the animals used their tongue to acquire the pellet . Forelimbs and hands are used to assist in chewing . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 01210 . 7554/eLife . 10774 . 013Video 7 . Optogenetic inhibition of contralateral sensorimotor cortex does not affect cued licking and associated forelimb movements . Auditory cue and pellet delivery was paired with contralateral sensorimotor cortex inhibition ( 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice . Lick performance and associated arm movements were not affected by cortical suppression . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 013 We next tested the role of cortex during the execution of prehension by activating inhibitory neurons at different times after initiation ( Videos 8–16 ) . Sensorimotor cortical suppression during the Lift , Hand open , Grab , and Supination components of the movement resulted in a “freezing” of progression towards the target , after a short latency ( 5/5 animals; Figure 3A–D; Videos 8 , 9 , 10 , 12 and 16 ) . The latency to freezing ( between 72–170 ms for a representative animal ) implies that this is the timescale over which the cortex commands specific components of the movement ( Figure 3A4–D4 ) . Movement during the delay was not simply the result of momentum , since in many cases the movement progressed to the next distinct component before freezing ( Figure 3A3–E4 ) . Cortical suppression was not effective at impeding hand movements involved in chewing , suggesting these motor programs do not require sensorimotor cortex ( Figure 3E and F , Videos 11 and 13–15 ) , despite involving similar muscle groups . Prolonged suppression of sensorimotor cortex during freezing resulted in a retraction of the limb either to the perch or a characteristic midair position ( Figure 3A2–E2 , Videos 8 , 9 , 10 and 12 ) . These results indicate that ongoing contralateral sensorimotor cortex activity is necessary for progressing through the steps of a learned , goal-directed , skilled prehension behavior . 10 . 7554/eLife . 10774 . 014Video 8 . Prehension progression is blocked by optogenetic inhibition of contralateral sensorimotor cortex during Lift . Contralateral sensorimotor cortex inhibition ( 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice occurred during Lift . Progression to the pellet was blocked . Prehension was initiated at the termination of cortical inhibition . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 01410 . 7554/eLife . 10774 . 015Video 9 . Prehension progression is blocked by optogenetic inhibition of contralateral sensorimotor cortex during Hand open . Contralateral sensorimotor cortex inhibition ( 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice occurred during Hand open . Animal progressed to Grab and then froze . Prehension was initiated at the termination of cortical inhibition . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 01510 . 7554/eLife . 10774 . 016Video 10 . Prehension progression blocked by optogenetic inhibition of contralateral sensorimotor cortex during Grab . Contralateral sensorimotor cortex inhibition ( 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice occurred during Grab . The animal was not able to get the food into the mouth . Prehension was initiated at the termination of cortical inhibition . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 01610 . 7554/eLife . 10774 . 017Video 11 . Prehension progression impeded but not prevented by optogenetic inhibition of contralateral sensorimotor cortex during Grab . Contralateral sensorimotor cortex inhibition ( 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice occurred during Grab . Trajectory to the mouth was impeded , but interestingly the unaffected arm was able to facilitate task completion . Prehension was not initiated at the termination of cortical inhibition . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 01710 . 7554/eLife . 10774 . 018Video 12 . Optogenetic inhibition of contralateral sensorimotor cortex during Supinate can stop progression to At mouth . Contralateral sensorimotor cortex suppression ( 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice occurred during Supinate . The mouse was unable to deliver the pellet into the mouth . Prehension was initiated at the termination of cortical inhibition . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 01810 . 7554/eLife . 10774 . 019Video 13 . Optogenetic inhibition of contralateral sensorimotor cortex during Supinate can fail to stop progression to At mouth . Contralateral sensorimotor cortex inhibition ( 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice occurred during Supinate . Mouse was able to deliver pellet to mouth . Prehension was not initiated at the termination of cortical inhibition . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 01910 . 7554/eLife . 10774 . 020Video 14 . Prehension is not impeded by optogenetic inhibition of contralateral sensorimotor cortex during At mouth epoch . Contralateral sensorimotor cortex inhibition ( 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice occurred during At mouth . Animal was able to deliver pellet into mouth . Prehension was not initiated at the termination of cortical inhibition . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 02010 . 7554/eLife . 10774 . 021Video 15 . Chewing is not impeded by optogenetic inhibition of contralateral sensorimotor cortex . Contralateral sensorimotor cortex inhibition ( 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice occurred during chewing . Animal continued chewing . Prehension was not initiated at the termination of cortical inhibition . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 02110 . 7554/eLife . 10774 . 022Video 16 . Relief of brief inhibition of contralateral sensorimotor cortex while animal holding pellet initiates a new prehension attempt . Contralateral sensorimotor cortex inhibition ( 0 . 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice occurred while animal was holding the pellet . Despite this sensory information , the termination of the brief cortical inhibition caused the animal to drop the pellet and initiate a regrabbing at the target position . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 02210 . 7554/eLife . 10774 . 023Figure 3 . Contralateral sensorimotor cortex is required for the ongoing execution of head-fixed prehension behavior . ( A1 to E1 ) Optogenetic activation of contralateral sensorimotor cortical inhibitory neurons of Slc32a1-COP4*H134R/EYFP mice during the epochs of tone-cued prehension . Three example trials presented as ethograms for a representative animal . Laser duration marked by sinusoidal and dashed blue lines; behavioral components denoted by color-coded bars ( abbreviations indicated ) . Percentage ( top-right , pooled across 5 animals , rate during bracketed time period ) of trials where animals reinitiated the behavior upon cessation of cortical inhibition . ( A2 to E2 ) Hand trajectory before and during cortical inhibition of each prehension component . White line labels trajectory from trial start to the onset of inhibition . Colored line represents trajectory during the laser , blue to red color variation represents passing time . Open circles mark hand position at start of inhibition ( blue ) and end of inhibition ( red , small red circles mark laser off position for additional trials ) . ( A3 to E3 ) Histograms of epoch progression rate , averaged over 5 animals . Control trials are gray and laser trials are the color code of the epoch ongoing at laser onset . For each animal , the number of control trials plotted was equal to the number of laser trials . Control trials were selected based on achievement of at least the epoch ongoing at laser onset . X-axis defines epochs of progression , y-axis represents percent of trials where progression was successful . Progressions where number was not available were removed from the x-axis . Bars are standard deviation . Asterisks indicate statistical significance ( t-test with unequal variance , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . ( A4 to E4 ) Path deviations during cortical inhibition of a representative animal . Measurement is the median shortest distance between hand position and either the median Grab or At mouth position . Time is in reference to laser onset . Lines represent control trials ( black ) and trials with optogenetic cortical inhibition ( blue , see Materials and methods ) . Shaded area surrounding the line represents the 25th and 75th percentiles . Vertical dashed line marks the time when the 25th and 75th percentiles no longer overlap . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 023 Cortical suppression-induced behavioral freezing was rapidly reversible; immediately upon cessation of the light , animals again reached/grabbed for the pellet ( “rebound prehension” , 8/8 animals , Figure 2B , Figure 3A1–E1 , Videos 3 , 8–10 , 12 and 16 ) . The initiation of rebound prehension had shorter and less variable latencies when compared to control ( uninterrupted , tone-cued ) prehension ( Figure 2A and B , 4A ) . Once initiated , however , durations of the distinct phases of rebound prehension and the trajectories were not statistically different from control behavior ( Figure 4A–C; Figure 4—figure supplement 1 ) . End-point accuracies of control and rebound prehension attempts were similar ( 4/5 animals , Wilcoxon–Mann–Whitney rank sum , p>0 . 05; for 1/5 animal difference was significant but median end-point error was only 0 . 74 +/– 0 . 52 mm ( median +/– median absolute deviation ) ; Figure 4D and E ) . Interestingly , end-point accuracies were also similar for rebound prehension attempts either beginning from the perch or from a diversity of midair positions ( 5/5 animals , Wilcoxon–Mann–Whitney rank sum , p>0 . 05; Figure 4D and E ) . Therefore , motor programs comparable to natural forms of the behavior follow termination of inhibitory neuron activation in sensorimotor cortex . 10 . 7554/eLife . 10774 . 024Figure 4 . Termination of cortical inhibition is sufficient to generate normal prehension behavior . ( A ) Timing differences between control prehension behavior and prehension after laser termination ( “rebound prehension” ) , averaged across 5 Slc32a1-COP4*H134R/EYFP animals . Values include First Lift ( difference between First Lift and either the cue ( control ) or laser off ( rebound ) ) and the intervals between sequential behavioral components . Within individual animals , First Lift timing ( 3/5 animals ) and variance ( 5/5 animals ) were also significantly different ( Wilcoxon–Mann–Whitney rank sum , p<0 . 05 ) . ( B and C ) Comparison of the trajectories of control ( B ) and rebound prehension ( C ) Trajectories ( n = 20 trials ) color coded by behavioral component . Top row displays trajectories from Lift to Grab , bottom row represents trajectories from Grab to At mouth . Circles represent on perch ( gray ) , Grab ( green ) , and At mouth ( dark blue ) hand position . ( D ) Rebound prehension was accurately performed from a diversity of starting positions . Control prehension from the perch ( gray , n = 25 trials ) , rebound prehension from perch ( dark blue , n = 25 trials ) , and rebound prehension from mid-air starting position ( light blue , n = 18 trials ) for a representative animal . Lines connect starting position of the reach to Grab position . Color-coded circles represent Grab position . ( E ) End-point quantification for control , rebound prehension from perch , and rebound prehension from mid-air trials for a representative animal . Box plots ( whiskers mark minimum and maximum non-outliers ) of Grab error: Grab error is defined as the distance between Grab position of an individual trial and median Grab position ( control , success trials ) . No significant difference between control and rebound prehension from the perch or from midair positions ( 4/5 and 5/5 animals , respectively; Wilcoxon–Mann–Whitney rank sum , p>0 . 05 ) . ( F ) Extinguishing the laser over contralateral sensorimotor cortex elicited rebound prehension even in the absence of a cue and pellet . ( G ) Laser-off over ipsilateral sensorimotor cortex did not generate rebound prehension . ( F ) and ( G ) Behavioral components denoted by color-coded bars . Blue sinusoidal line marks laser duration . Cumulative histogram of First Lift is shown . ( H ) Effect of laser duration on the likelihood of rebound prehension , initiation latency , and associated jitter in the absence of cue and pellet . Statistically significant ( repeated-measures ANOVA , p<0 . 05 ) main effect of laser duration on rebound prehension initiation probability , delay , and jitter . ( I ) Rate of rebound prehension as a function of success and chewing status at laser off ( average of 5 animals , bars are standard deviation ) . ( J ) Control ( gray ) and rebound ( blue ) prehension rates as a function of food deprivation ( dep . ) state . Animals ( n = 4 ) were taken off food deprivation three times and the rates were averaged . Prehension rates plotted are the means of these averages , bars are standard deviation . Asterisks indicate statistical significance ( t-test with unequal variance , *p<0 . 05 , **p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 02410 . 7554/eLife . 10774 . 025Figure 4—figure supplement 1 . Trajectories of control and rebound prehension are similar . Comparison of the average trajectory of control ( white line ) and rebound ( blue line ) prehension of Slc32a1-COP4*H134R/EYFP mice . Squares on line mark average hand position at Lift , 1/3 of total reach , 2/3 of total reach , and Grab position . Side ( left column ) and front ( right column ) views of head-fixed mouse . Rows display clouds of hand position ( filled-circles ) for Lift , 1/3 of total reach , 2/3 of total reach , and Grab position . Open circles marks 2-sigma ellipse describing the sample mean and covariance of hand position . Number in upper right corner indicates the balanced nearest neighbor ( NN ) accuracy ( score of 0 . 5 reflects chance assignment to control or rebound prehension groups , see Materials and methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 02510 . 7554/eLife . 10774 . 026Figure 4—figure supplement 2 . Action potential firing rates during and after photostimulation in cortical neurons of awake , non-behaving Slc32a1-COP4*H134R/EYFP mice . ( A ) Silicon probe extracellular recordings of a putative pyramidal cortical neuron . Blue line represents 4 s of 473 nm light ( 7 mW ) . Increased spiking immediately at the termination of each light exposure . ( B ) Mean firing rates ( peristimulus time histograms , 100 ms bins ) for putative pyramidal ( spike width >0 . 45 ms ) and fast spiking ( spike width <0 . 35 ms ) cortical neurons around photostimulation . Columns , four different durations of photostimulation ( blue bars ) . Top row , mean of putative pyramidal neurons ( n = 77 units , 2 mice ) . Bottom row , mean of putative fast spiking neurons ( n = 10 units , 2 mice ) . Prominent rebound activity is seen with light exposure of 2 s and greater . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 02610 . 7554/eLife . 10774 . 027Figure 4—figure supplement 3 . Contralateral sensorimotor cortex is required for the ongoing execution of rebound prehension . ( A–F ) Termination of initial optogenetic activation ( laser 1 ) of sensorimotor cortical inhibitory neurons of Slc32a1-COP4*H134R/EYFP mice elicited rebound prehension; second laser ( laser 2 ) administration interrupts rebound prehension . Examples of laser exposure during each component of rebound prehension are shown . Laser duration is marked by sinusoidal and dashed blue lines and behavioral components are denoted by color-coded bars . Similar to control prehension , cortical inhibition froze progression of rebound prehension . Termination of second laser exposure elicited another rebound prehension . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 027 Two possibilities could underlie rebound prehension: the animals could make them voluntarily or they could be a direct consequence of the relief of inhibition . An involuntary nature of rebound prehension was supported by its occurrence even on cued trials when the animal had already grabbed the pellet ( Figure 3 , Videos 10 , 12 and 16 ) . To further test this , we removed both the pellet and the cue to disengage the animal from the task . During these uncued experiments , the animals remained motionless during contralateral sensorimotor cortical inhibition as expected . Surprisingly , immediately ( ~250 ms ) following light cessation , the animals reached and grabbed at the empty pellet tray ( initiation in 192/224 trials , n = 8 animals , Figure 4F , Videos 17 ) . Rebound execution did not occur in untrained animals ( 3/3 animals ) and was specific to prehension; no other type of movement was generated following light termination ( n = 1434 trials , n = 5 animals ) . If a pellet was in the target position but the animal was not alerted of its presence , cessation of inhibition frequently generated the full reach-grab-eat sequence ( Videos 18 ) . Taken together , these observations indicate that rebound prehension is not the result of ongoing sensory information or the animals’ remembered behavioral state . Rather , termination of inhibitory neuron activation in sensorimotor cortex , in combination with plastic changes induced by training , is sufficient to evoke a complex , learned motor act . 10 . 7554/eLife . 10774 . 028Video 17 . Even in the absence of a cue or a pellet , cessation of optogenetic inhibition of contralateral sensorimotor cortex evokes prehension . Contralateral sensorimotor cortex inhibition ( 2 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice was not paired with a cue or a pellet . During the laser , the animal remained motionless . Prehension to the remembered target was reliably initiated at the termination of inhibition . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 02810 . 7554/eLife . 10774 . 029Video 18 . In the absence of a cue but the presence of a pellet , cessation of optogenetic inhibition of contralateral sensorimotor cortex evokes the full prehension behavior . Contralateral sensorimotor cortex inhibition ( 1 s of light delivery ) in Slc32a1-COP4*H134R/EYFP mice was not paired with a cue but a pellet was unexpectedly available . During cortical suppression , the animal remained motionless . Prehension to the remembered target was reliably initiated at the termination of inhibition; the pellet was grabbed , delivered to the mouth and chewed . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 029 Rebound prehension showed dependence on a number of factors . Anatomically , no area outside of sensorimotor cortex contralateral to the utilized limb was effective in evoking the behavior in the no cue/no pellet condition ( ipsilateral sensorimotor , 0/68 trials , n = 3 animals , Figure 4G; frontal , 0/20 trials , n = 2 animals; visual , 0/21 trials , n = 3 animals ) . Duration of cortical inhibition was also a critical factor in initiating rebound prehension . At least 2 s ( median across animals ) of cortical inhibition was required to achieve rebound prehension rates above 90% ( n = 4 animals , Figure 4H ) . Longer durations ( 4 s was longest tested ) of cortical inhibition also reduced the latency variability of prehension initiation ( n = 2 animals , Figure 4H ) . Interestingly , long durations of photostimulation ( 2 s or more ) were specifically associated with prominent rebound spiking in cortical neurons ( Figure 4—figure supplement 2 ) . This raises the possibility that rebound spiking in the cortex is responsible for rebound prehension . Rebound prehension itself is also cortically dependent , as reactivation of sensorimotor inhibitory cortical neurons freezes progression of the rebound movement ( Videos 19 , Figure 4—figure supplement 3 ) . The presence of food in the mouth decreased rebound prehension rates ( 18/95 trials exhibited rebound; Figure 4I; Videos 11 , 13–15 ) . For nearly all of these ( 93/95 ) trials , animals were still chewing at light termination . To distinguish whether the reduced rate of rebound prehension was related to motor-program competition ( chewing versus prehension ) or to motivation ( due to consumption of the pellet ) , we assayed rebound rates for failure trials where animals were chewing without the pellet ( 29/256 trials , Figure 4I ) . Prehension generally occurred ( 21/29 trials ) if the light was terminated during pelletless chewing . To further test the role of hunger/satiety , we terminated food deprivation; animals on free food for one day exhibited significantly reduced rates of rebound prehension despite still engaging in the control version of the behavior ( Figure 4J ) . These results suggest that motivational drive , operating along both short ( eating ) and long ( sated ) time scales , gates rebound prehension . Therefore , evocation of the learned behavior depends not only on features of the inhibition but also on the satiety of the animal . 10 . 7554/eLife . 10774 . 030Video 19 . Optogenetic inhibition of contralateral sensorimotor cortex ( 3 s of light delivery twice , separated by 320 ms ) during Supinate of a post-inhibitory prehension attempt can stop progression to the mouth . Contralateral sensorimotor cortex inhibition in Slc32a1-COP4*H134R/EYFP mice occurred during Supinate of a post-inhibitory prehension attempt . Mouse unable to deliver the pellet into the mouth . Prehension was initiated again at the termination of the second cortical inhibition . Side and front views of head-fixed mouse . Playback at 100 ms/s . DOI: http://dx . doi . org/10 . 7554/eLife . 10774 . 030
By leveraging the rapid and reversible nature of optogenetic inhibition , we were able to probe the online role of cortex in a learned prehension task . Contrary to the view that cortex is only involved in achieving dexterity ( Kawai et al . , 2015 ) , we found cortex to be necessary for initiating and executing the dexterous and non-dexterous steps of this skilled movement . The necessity of cortex for initiation and execution is consistent with some interpretations of experiments where cortical stimulation delayed or interrupted voluntary movements in primates ( Penfield and Jasper , 1954; Day et al . , 1989; Lemon et al . , 1995; Churchland and Shenoy , 2007 ) . Apparent disagreements between our work and that described by Kawai et al . may be due to differences in the behavior and/or the perturbation methodology . The forelimb-lever press task in the study by Kawai et al . requires animals to learn a time interval but not the precise spatial position of an external target . In contrast , the prehension behavior studied here requires the animal to learn the position of the pellet and how to precisely execute a reach to that location . This difference suggests that cortex is not required for interval tasks but rather skilled interactions with objects in the environment . An outlier animal described by Kawai et al . provides support for this hypothesis; this animal was unique in that it made precise movements to the lever and showed a significant decrease in performance after cortical perturbation . Additionally , the two experiments also differ in the acuteness of the cortical manipulations; we used rapid onset optogenetic inhibition while Kawai et al . utilized pharmacological and surgical manipulations . The rapidity of optogenetic inhibition may impede compensation mechanisms that ameliorate resulting phenotypes . We found that termination of inhibitory neuron activation in sensorimotor cortex suffices to drive a full action sequence to the learned target . Previous activation studies of cortex have shown that electrical ( Ferrier , 1873; Penfield and Boldrey , 1937; Penfield , 1954; Gottlieb et al . , 1993; Graziano et al . , 2002; Ramanathan et al . , 2006; Bonazzi et al . , 2013 ) and optogenetic ( Harrison et al . , 2012 ) stimulation for behaviorally relevant durations produces motor sequences . Rebound prehension differs from these movements in three ways: rebound prehension is a coordinated sequence of movements ( reaching , grabbing , and eating ) accurately directed to an external goal regardless of starting position , depends on training , and is gated by motivation ( Ramanathan et al . , 2006; Harrison et al . , 2012; Bonazzi et al . , 2013 ) . These differences suggest that termination of inhibition evokes a motor engram specifying a learned goal/end-point of a trained behavior ( Bernstein , 1967; Todorov and Jordan , 2002 ) . Cortical synapses modified during the learning of skilled actions ( Rioult-Pedotti et al . , 1998; Xu et al . , 2009; Wang et al . , 2011 ) may be responsible for the formation of these motor engrams . Alternatively , the engram could be stored elsewhere , and relief of cortical inhibition could trigger or permit its activation in a downstream network . The ability to specifically elicit a post-manipulation transition of cortical state from quiescence to motor engram activation should enable functional characterization of the neural programs responsible for enacting skilled multi-step behavior .
Animal procedures were performed in accordance with protocols ( Protocol number: 13–99 ) approved by the Institutional Animal Care and Use Committee ( IACUC ) of the Janelia Research Campus . Animals were housed on a 12-hr light/dark schedule with ad libitum water . Mice undergoing behavioral training were food restricted to 80–90% of original body weight by limiting food intake to 2–3 g/day . Otherwise , mice had ad libitum food . Animals were monitored daily by veterinary staff , and animals recovering from surgery or on food restriction were given a Pain Assessment Score of 1–5 , based on IACUC guidelines . Animals with a Pain Assessment Score above 3 were temporarily removed from behavioral testing , given analgesia as determined by the veterinarian , and in some cases their food allotment was increased . Animals were returned to food restriction and behavioral testing after three consecutive days with a Pain Assessment Score below 3 . Slc32a1-COP4*H134R/EYFP ( Stock Number: 014548 ) were obtained from The Jackson Laboratory ( Bar Harbor , ME ) . Cortical neurons were optogenetically stimulated either through a cranial window or through a cleared skull . Transgenic mice ( 2–5 months old ) were anesthetized with isoflourane ( 2% ) and placed in a stereotactic frame ( Kopf Instruments; Tujunga , CA ) on a 37°C heating pad . Aseptic technique was used during surgical procedures . The scalp and periosteum over the skull were removed , a layer of UV-curing OptiBond adhesive ( Kerr; Orange , CA ) ( 22 ) was applied , and a custom-made headpost was affixed with cement . The clear skull preparation was used for experiments testing the role of different cortical areas . For this preparation , a final smooth layer of clear dental acrylic covered the skull ( Jet Repair Acrylic; Lang Dental Manufacturing; Wheeling , IL ) . Before optogenetic experimentation , the skull was polished using Acrypoints Acrylic Polishing Kit ( Pearson Dental; Sylmar , CA ) and a thin layer of clear nail polish ( Electron Microscopy Services; Hatfield , PA ) was applied to reduce glare . Cranial windows ( three 170 μm-thick panes of laser-cut glass , 2 mm diameter , glued together with UV adhesive ) were placed over sensorimotor cortex . Following surgery , injections of ketoprophen ( 5 mg/kg ) and buprenorphine ( 0 . 1 mg/kg; Henry Schein Animal Health; Melville , NY ) were administered subcutaneously . If animals exhibited signs of pain or distress following surgery , additional doses of either ketoprophen or buprenorphine were administered , as directed by veterinary staff . Mice recovered for at least 1 week following surgery and were given ad libitum food and water . Mice were habituated to head fixation in a custom-built apparatus ( adapted from Rodent In Vivo Electrophysiology Targeting System ) with forks to secure the head post , adjustable walls to restrain the body , and a perch for the hands ( Figure 1 ) ( Osborne and Dudman , 2014 ) . This apparatus was placed in a light tight , ventilated , soundproof 28-inch cubic behavioral box . A near-infrared-sensitive webcam was used to monitor animals . Mice were initially trained for approximately 30 min per day , until they started licking pellets ( 10 or 20 mg; Test Diet; St Louis , MO ) placed directly below their mouth . Food pellets arrived ~ 200 ms after the start of an auditory tone ( 5 kHz ) by rotating the turntable with a servomotor driven by custom-programmed Arduino software . Mice were initially ( 1–5 training sessions ) trained to retrieve a food pellet by licking and eating the pellet , often using their hand to guide the pellet into their mouth . After cued licking was learned , the turntable was moved progressively further away ( over 3–10 sessions ) to encourage mice to reach for the pellet after the cue . After mastering the task ( approximately 15 sessions ) , turntable placement was reproduced daily for each animal by aligning live images to a reference image . Inter-trial intervals ranged from 20 s to 1 min and approximately 50 trials were collected per day . Mice almost always started with hands on perch and trials where animals lifted the hand before the cue were discarded . Mice were trained each day for approximately 60 min until they routinely responded to the auditory cue ( within 1 s ) and grabbed the pellet . Reprieve of food deprivation was accomplished by administering food ad libitum for one day , prehension rates were tested , and then animals were returned to previous food deprivation conditions . Two high-speed , high-resolution monochrome cameras ( Point Grey Flea3; 1 . 3 MP Mono USB3 Vision VITA 1300; Point Grey Research Inc . ; Richmond , BC , Canada ) with 6–15 mm ( f/1 . 4 ) lenses ( C-Mount; Tokina , Japan ) were placed perpendicularly in front and to the right of the animal . A custom-made near-infrared LED light source was mounted on each camera . Cameras were synced to each other and captured at 500 frames/s at a resolution of 352 × 260 pixels . Video was recorded using custom-made software developed by the Janelia Research Campus Scientific Computing Department and IO Rodeo ( Pasadena , CA ) . This software controlled and synchronized all facets of the experiment , including auditory cue , turntable rotation , optogenetic lasers , and high-speed cameras . Fiji video editing software was used to label laser onset , termination , and timestamp in the videos . To excite ChR2 in cortical inhibitory interneurons through either a cranial window or cleared skulls of Slc32a1-COP4*H134R/EYFP mice , we used a 200 μm core , 0 . 39 numerical aperture fiber ( FT200UMT; Thorlabs Inc; Newton , New Jersey ) and a fiber-coupled 473 nm laser ( LuxX 473–80; Omicron Laserage; Rodgau-Dudenhofen , Germany ) . Fibers were positioned on or above ( 1–2 mm ) the skull at stereotactic coordinates for different cortical regions ( cM1: 0 . 5 mm anterior to bregma , 1 . 5 mm lateral of midline; iM1: 0 . 5 mm anterior to bregma , 1 . 5 mm lateral of midline; V1: 4 . 0 mm posterior to bregma , 3 . 0 mm lateral of midline; FrC: 3 . 1 mm anterior to bregma , 1 . 5 mm lateral of midline ) . 40 Hz sinusoidal stimulation with average power of 1–15 mW at the fiber’s end was delivered for 0 . 1–4 s , and power was reduced to 0 mW over the final 100 ms of stimulation . The approximate diameter of the light spot on the skull/window surface was 400 μm . Laser power of 10–15 mW has been shown to silence spike rates by 90% within 1 . 5 mm of the center of the laser spot ( Guo et al . , 2014 ) . We found no significant differences between prehension evoked with 2–4 s laser pulses , so they were grouped . To compensate for potential confounds of visible blue light during optogenetic stimulation , the behavioral box was illuminated with a blue LED throughout training and testing . During testing , trials with optogenetic perturbation were interleaved with control trials without laser stimulation ( < 20% laser trials , randomly interspersed ) . Optogenetic activation of cortical inhibitory neurons and subsequent local cortical suppression most parsimoniously explains the reported phenotypes; however , other mechanisms are possible using the technique employed in this paper . The Slc32a1-COP4*H134R/EYFP mouse line used in this study expresses ChR2 in a small set of striatal neurons ( Guo et al . , 2014 ) . This raises a possibility that blue light directed onto the cortical surface activates striatal neurons and this stimulation is responsible for the described phenotypes . We directly measured the transmission of blue light through the cortex into the striatum ( Figure 2—figure supplement 1 ) . At the depths ( +1 . 8mm ) of the striatum , the intensity of the 473 nm light is negligible using our experimental strategy . We also implanted fibers into striatum ( FT200UMT; Thorlabs Inc; Newton , New Jersey; coordinates: 1 . 7 mm lateral of midline; 0 . 5 mm anterior of bregma; 2 . 25 mm deep of brain surface ) of three trained Slc32a1-COP4*H134R/EYFP mice . Delivery of 473 nm light did not halt prehension behavior ( prehension initiation in 190/191 laser trials versus 271/271 control trials ) in any of these animals . Therefore , it is very unlikely that activation of striatal neurons is responsible for the behavioral phenotypes described in this paper . Another possibility is that long-range GABAergic inputs to the cortex are labeled in the Slc32a1-COP4*H134R/EYFP and antidromic activation of these inputs inhibits a non-cortical area that is necessary for prehension ( Saunders et al . , 2015 ) . However , using photostimulation , we were able to produce prehension deficits in mice ( n = 2 animals , mouse 1 exhibited prehension initiation in 1/24 laser trials versus 69/72 control trials , mouse 2 exhibited prehension initiation in 11/21 laser trials versus 67/67 control trials ) where viral injection limited ChR2 expression to sensorimotor cortical neurons ( AAV 2/1 or 2/7 Synapsin-FLEX-REV-ChR2 , 5 sites in sensorimotor ( forelimb ) cortex , depths of 300 and 600 μm ) in Gad2-IRES-Cre ( Stock number: 010802; The Jackson Laboratory; Bar Harbor , ME ) mice . Finally , long-range inhibitory cortical outputs have been reported more rostral of forelimb motor cortex ( Lee et al . , 2014 ) . Therefore , it is possible that the blue light activates such neurons in motor cortex resulting in inhibition of non-cortical areas that are responsible for the phenotype . However , evidence for GABAergic projection neurons does not exist for areas that produce the phenotypes described in this paper ( Lee et al . , 2014 ) . We also did not detect ChR2-positive axons in the external capsule or corpus callosum of Slc32a1-COP4*H134R/EYFP mice ( data not shown ) , further ruling out the possibility that long range cortical inputs and outputs mediate the reported phenotypes . A custom photodetector Bittner , et al . , 2015 was fabricated by inserting a small plug of 650 nm-emitting quantum dots ( Ocean Nanotech; Dunedin , FL ) inside the tip of a glass pipette , and sealing the tip in clear epoxy ( Henkel; Rocky Hill , CT ) . Quantum-dot fluorescence excited by laser light was collected by a high-index gel waveguide within the pipette taper and delivered to a 100 μm core optical fiber ( 0 . 22 numerical aperture ) . The collected quantum dot fluorescence was detected using a fiber-coupled QE65000 spectrometer ( Ocean Optics; Dunedin , FL ) . The photodetector and laser fiber were positioned in parallel directly over a craniotomy above sensorimotor cortex ( 0 . 7 mm anterior to bregma , 1 . 6 mm lateral to midline ) . For this near-parallel source-detector geometry , the photodetector pipette except the tip was painted black ( Liquitex; Cincinnati , OH ) to prevent scattering of 473-nm laser light from entering the back of the detector pipette . The photodetector was mounted on a Sutter MP-285 motorized manipulator ( Novato , CA ) and light intensity measurements were taken at 100 μm increments from the dural surface to 3 . 0 mm deep . Laser excitation was similar to that used in behavioral experiments: sinusoidal 40 Hz pulses of 1 . 2 mW peak amplitude for 3 s from a 473 nm fiber-coupled laser . To quantitatively compare tone-cued ( control ) and laser-evoked ( rebound ) prehension , we developed a machine-learning-based method to automatically track the mouse’s utilized hand . We manually labeled the position of the hand in a small subset of video frames , then used a modified version of Cascaded Pose Regression ( Burgos-Artizzu et al . , 2013 ) to learn a function that could input a frame and automatically predict the position of the hand in that frame . Cascaded Pose Regression operates directly on a single video frame , and learns a cascade of multiple regressors , which each iteratively bring the target position estimate closer to the ground-truth labels . Our modifications included a different feature-selection method and a multi-pass method for enforcing trajectory smoothness over time . In addition , we developed a new interface for automatically detecting frames in which the tracking might contain errors , and allowing a user to manually fix these . Using the two-dimensional ( 2D ) trajectories corresponding to the front and side view videos , we reconstructed the three-dimensional trajectories by calibrating the cameras using the Camera Calibration Toolbox for MATLAB ( J . -Y . Bouguet , http://www . vision . caltech . edu/bouguetj/calib_doc/ [2004] ) . A single calibration was initially performed , then fine-tuned per-mouse based on the 2D trajectories . Individual prehension component behaviors were manually annotated . Lift was defined as the interval from initial separation between hand/perch until the hand was halfway to table . Hand-open was defined as the interval from fingers beginning to separate from palm until fully extended . Grab was defined as the interval from hand moving downwards as digits close to supination . Supination was defined as the interval from the beginning of upward wrist rotation to the hand reached the mouth . At mouth was defined as hand in close proximity to mouth . Chew was defined as mastication after pellet in mouth . Chew without pellet was defined as mastication without successfully retrieving pellet . Lick was defined as visible tongue emerging from mouth directed towards pellet until tongue returned to mouth . Grooming was defined as using both hands to clean nose , head , or whiskers . First Lift was defined as initial frame of the first Lift after cue . Lift-Hand open was defined as the difference between first frames of Hand open and Lift . Lift-Grab was defined as the difference between first frames of Grab and Lift . Hand open-Grab was defined as the difference between first frames of Grab and Hand open . Grab-Supinate was defined as the difference between first frames of Supinate and Grab during the final prehension sequence . Supinate-At mouth was defined as the difference between first frames of At mouth and Supinate during the final prehension sequence . Grab-At mouth was defined as the difference between first frames of At mouth and Grab during the final prehension sequence . We used built-in and custom-made scripts within MATLAB ( MathWorks; Natick , MA ) to perform the following tests: repeated measures ANOVA , Fisher’s exact , linear mixed-effects , pairwise t-tests assuming unequal variance , and Wilcoxon–Mann–Whitney rank sum . In Figure 3 , we compare the distance to a goal after laser interruption . Per-mouse , we computed the location of the pellet and the mouth as the average position of the hand at the start of Grab and At mouth epochs , respectively , during tone-cued prehension . We separated laser trials by the epoch interrupted , defined as the last component initiated before the laser-on time point . We excluded trials for which the prehension behavior was not impeded . For Lift through Supinate epochs , lack of inhibition was detected as the occurrence of the Chew epoch during the laser-on period . For the At mouth and Chew epochs , lack of an effect was not detectable , thus we included all trials . We constructed a control comparison data set , for example , laser during Lift trials , as follows . For each tone-cued trial , we select the time interval starting t frames after a randomly chosen Lift , where t is the time between the last Lift preceding laser-on and the laser-on time point for a randomly selected laser during Lift trial . To measure how similar control and rebound trajectories were , we used a strategy based on nearest-neighbor classifiers . First , we aligned all trajectories in time based on the relative distance traveled in the trajectory . We then asked , for a given distance , how well we could predict whether a trial was a control or rebound trial based on the hand position using a 1-nearest-neighbor classifier . For all mice , the accuracy of this classifier was close to chance . We report the balanced accuracy: the average of the true positive and true negative rates . Extracellular spikes were recorded from Slc32a1-COP4*H134R/EYFP mice ( n = 2 ) using silicon probes ( A4 × 8-5 mm-100-200-177; NeuroNexus; Ann Arbor , MI ) ( Guo et al . , 2014 ) . Voltage signals ( 32 channels ) were multiplexed , digitized by a PCI6133 board at 312 . 5 kHz ( National Instruments; Austin , TX ) at 14 bit , demultiplexed ( sampling at 19531 . 25Hz ) . Brain movement was minimized by applying silicone gel ( 3–4680 , Dow Corning , Midland , MI ) over the craniotomy after the electrode was positioned in the brain . Recordings began several minutes thereafter to allow the brain to settle . Under awake , non-behaving condition , the mice remained idle during different photostimulation conditions . Recordings ( n = 2 per animal ) were made through the same craniotomy on subsequent days . The extracellular recording traces were band-pass filtered ( 300–6000 Hz ) . Events that exceeded an amplitude threshold ( 4 standard deviations of the background ) were subjected to manual spike sorting to extract single units ( Guo et al . , 2014 ) . Eighty-seven single units were recorded under awake , non-behaving conditions . For each unit , spike width was computed as the trough-to-peak interval in the mean spike waveform . Units with spike width <0 . 35 ms were defined as fast spiking neurons ( 10/90 ) and units with spike width >0 . 45 ms were defined as putative pyramidal neurons ( 77/90 ) ( Guo et al . , 2014 ) . Units with intermediate values of spike width ( 0 . 35–0 . 45 ms , 3/133 ) were excluded from our analyses . Effect of photoinhibition on activity was quantified in “average spike rate” across the population . The time course of photoinhibition and rebound activity was computed from averaged peristimulus time histogram ( Figure 4—figure supplement 2 ) . | Many of the movements that humans and other animals make every day are deceptively complex and only appear easy because of extensive practice . For example , picking up an object involves several steps that must be precisely controlled , including reaching towards the item and holding it using the right amount of pressure to not crush it or drop it . Part of the brain called the motor cortex is thought to be important for learning and controlling these skilled movements , but its exact role in these processes is not clear . A technique called optogenetics allows the roles of individual parts of the brain to be studied by rapidly altering their activity , whilst minimizing the likelihood that the brain will compensate for these changes . By genetically modifying animals to produce light-sensitive channel proteins in certain brain cells , the activity of particular regions of the brain can be controlled by shining light onto them . Guo et al . have now used optogenetics to control the motor cortex as the mice performed a task they had been trained to do – reaching for and picking up a food pellet . Suddenly shutting down the motor cortex at the start of a trial prevented the mice from starting the task , and shut down part way through the task caused the front limbs of the mice to freeze in midair . However , only the learned , skilled task was frozen by motor cortex shutdown; mice could still move their limbs normally if the motor cortex was instead shut down during routine movements . When the cortex was reactivated , the mice instantly resumed trying to pick up the food pellet . Unexpectedly , even during rest periods when there was no food pellet and the mice were just waiting for the experiment to begin , turning the motor cortex off and then back on again suddenly caused the mice to perform the complete grabbing motion . This implies that the cortical activity evoked at the end of inactivation acts to trigger the full movement sequence . This was particularly likely to occur if the animal had been deprived of food before the test or was particularly well trained , but did not depend on the position of the limb . Overall , Guo et al . ’s work opens the question of how the instructions that describe the learned movement are encoded within the motor cortex and its downstream networks . Future studies could also investigate how learning a set of movements affects the structure of cortical neurons and their connections , thus suggesting how these memories are stored . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Cortex commands the performance of skilled movement |
Exophilin-8 has been reported to play a role in anchoring secretory granules within the actin cortex , due to its direct binding activities to Rab27 on the granule membrane and to F-actin and its motor protein , myosin-Va . Here , we show that exophilin-8 accumulates granules in the cortical F-actin network not by direct interaction with myosin-Va , but by indirect interaction with a specific form of myosin-VIIa through its previously unknown binding partner , RIM-BP2 . RIM-BP2 also associates with exocytic machinery , Cav1 . 3 , RIM , and Munc13-1 . Disruption of the exophilin-8–RIM-BP2–myosin-VIIa complex by ablation or knockdown of each component markedly decreases both the peripheral accumulation and exocytosis of granules . Furthermore , exophilin-8-null mouse pancreatic islets lose polarized granule localization at the β-cell periphery and exhibit impaired insulin secretion . This newly identified complex acts as a physical and functional scaffold and provides a mechanism supporting a releasable pool of granules within the F-actin network beneath the plasma membrane .
Cells , including professional secretory cells , possess a peripheral microfilament web beneath the plasma membrane , referred to as the actin cortex , which maintains the cell’s shape and integrity ( Orci et al . , 1972; Aunis and Bader , 1988 ) . Secretory granules generated at the trans-Golgi network ( TGN ) must somehow pass through the actin cortex before they fuse with the plasma membrane and , as such , the actin cortex may act as a mechanical barrier to the exocytic site . In fact , pharmacological depolymerization of F-actin has been shown to potentiate granule exocytosis . However , such a massive disruption of the cytoskeleton may obscure physiological steps in the trafficking process . For example , cortical actin and its motor proteins , such as myosin-Va , have been suggested to function as a carrier to capture and/or transport granules to the vicinity of the plasma membrane in support of exocytosis ( Wu et al . , 1998; Lang et al . , 2000; Rudolf et al . , 2003; Giner et al . , 2005; Ivarsson et al . , 2005; Varadi et al . , 2005; Desnos et al . , 2007; Wollman and Meyer , 2012 ) . However , the molecular mechanisms by which granules link to the F-actin network and are then processed toward exocytosis is poorly understood . Thus , a fuller understanding of the role of the cortical F-actin network in exocytosis must be established . Exophilin-8 ( also known as MyRIP and Slac2-c ) is a candidate molecule for the anchoring of granules within the actin cortex . It exhibits affinities both to Rab27a/b on granule membrane and to F-actin and its motor proteins , myosin-Va and -VIIa ( El-Amraoui et al . , 2002; Fukuda and Kuroda , 2002 ) . In fact , overexpressed exophilin-8 has been shown to redistribute granules to the actin-rich cell periphery in the pancreatic β-cell line , MIN6 ( Mizuno et al . , 2011 ) , and in the enterochromaffin cell line , BON ( Huet et al . , 2012 ) . Further , knockdown of exophilin-8 decreases the number of granules beneath the plasma membrane ( Mizuno et al . , 2011; Huet et al . , 2012 ) and their cargo secretion ( Waselle et al . , 2003; Ivarsson et al . , 2005; Mizuno et al . , 2011 ) . These findings suggest that exophilin-8 potentiates exocytosis by locating granules within the actin cortex beneath the plasma membrane . The present study is the first to use exophilin-8-knockout mice to demonstrate its in vivo function in glucose tolerance . Exophilin-8-null pancreatic β-cells lose polarized granule location at the cell periphery and exhibit decreased insulin secretion . We further show that exophilin-8 directly interacts with RIM-BP2 , a binding protein to RIM ( Wang et al . , 2000 ) , and that this complex formation is essential for both peripheral accumulation and efficient exocytosis of granules . In contrast to the previous proposal that exophilin-8 captures granules within the F-actin network via its direct interaction with myosin-Va ( Desnos et al . , 2003; Huet et al . , 2012 ) , we found that exophilin-8 does so via an indirect interaction with a specific form of myosin-VIIa through RIM-BP2 . RIM-BP2 also associates with the L-type voltage-dependent Ca2+ channel ( VDCC ) mediating stimulus-induced Ca2+ influx , Cav1 . 3 , and the priming factors , RIM and Munc13-1 , in β-cells , as originally observed in neurons ( Hibino et al . , 2002; Südhof , 2013 ) . Thus , the exophilin-8–RIM-BP2–myosin-VIIa complex not only physically anchors granules to the actin cortex , but may also functionally assemble molecules involved in their exocytosis . These findings reveal a previously unknown molecular mechanism in the secretory processes that efficiently retain granules beneath the plasma membrane for exocytosis .
All previous studies of exophilin-8 ( encoded by the Myrip gene ) have been performed at the cellular or molecular levels . In the present study , we generated exophilin-8-knockout mice to examine its in vivo function ( Figure 1—figure supplement 1A–C ) . The mutant mice were viable and fertile , with no apparent abnormalities in general appearance or behavior . However , they showed slightly reduced body weight and significantly higher blood glucose levels after a glucose load , although their insulin sensitivity was not altered ( Figure 1A ) . Exophilin-8 was expressed in pancreatic islets , as well as in pituitary and brain ( Figure 1—figure supplement 1D ) . Further , its absence induced decreased insulin secretion in responses to glucose , potassium , or forskolin ( an activator of adenylate cyclase ) with glucose ( Figure 1B–D ) , but did not change secretion in response to phorbol-12-myristate-13-acetate ( PMA; a protein kinase C activator ) with glucose ( Figure 1E ) . Cortical F-actin-disrupting PMA ( Vitale et al . , 1995 ) might negate the function of exophilin-8 that is localized within the actin cortex ( Desnos et al . , 2003; Waselle et al . , 2003; Mizuno et al . , 2011 ) . 10 . 7554/eLife . 26174 . 003Figure 1 . Phenotypes of exophilin-8 null mice . ( A ) In vivo phenotypes of exophilin-8-knockout ( KO ) mice . Each measurement was performed in age-matched , wild-type ( WT; gray bars and diamonds ) and KO ( red bars and squares ) male mice: body weight ( left; 18-weeks-old , n = 10 ) ; blood glucose concentrations during an intraperitoneal glucose tolerance test ( IPGTT; middle; 19- to 23-week-old , n = 10 ) ; and percentage of starting blood glucose concentration during an intraperitoneal insulin tolerance test ( IPITT; right; 16- to 18-weesk-old , n = 8 ) . ( B–E ) Islets isolated from age-matched ( 16- to 25-week-old ) WT or KO male mice ( n = 6 for B and E , n = 5 for C and D ) were stimulated by 16 . 7 mM glucose for 30 min ( B ) , 60 mM KCl for 15 min ( C ) , or 16 . 7 mM glucose for 20 min ( horizontal black line ) in the continuous presence of either 10 μM forskolin ( D ) or 0 . 5 μM PMA ( E ) ( horizontal black and gray lines ) . They were then perifused with buffer containing 2 . 8 mM glucose . The amount of insulin secreted into each fraction was normalized by insulin content remaining in the cells , although the latter values were not significantly different between WT and KO islets . The area under the curve during stimulation was measured . Data are means ± SEM . *p values calculated using two-tailed unpaired t-test are 0 . 031 ( A , body weight ) , 0 . 0027 ( A , IPGTT 0 min ) , 3 . 0 × 10−7 ( A , IPGTT 15 min ) , 6 . 6 × 10−3 ( A , IPGTT 30 min ) , 2 . 8 × 10−4 ( A , IPGTT 60 min ) , 7 . 7 × 10−3 ( B ) , 3 . 1 × 10−3 ( C ) , and 3 . 3 × 10−4 ( D ) , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 00310 . 7554/eLife . 26174 . 004Figure 1—figure supplement 1 . Generation of exophilin-8-null mice . ( A ) Targeted disruption of the exophilin-8 ( Myrip ) gene on mouse chromosome 9 . The targeting vector contains a neomycin resistance gene driven by the pgk promoter ( pgk-neo ) and a diphtheria toxin A-fragment gene driven by the MC1 promoter ( MC1-DTA ) as positive and negative selection markers , respectively . Exon structures are vertically lined and shown from the fourth exon to the seventh exon . Homologous recombination results in insertion of pgk-neo in the genomic region of exon4 . ( B ) Genomic Southern hybridization analysis of the backcrossed progenies from a cross of chimeric mice with C57BL/6 mice . The location of the external probe is shown with horizontal closed lines in ( A ) . The probe hybridizes to AflIII fragments of 11 . 0 kb and 13 . 0 kb from wild-type ( WT ) and mutant knockout ( KO ) alleles , respectively . ( C ) PCR genotyping of the offspring . PCR with Exo8/Fow1 , Neo/Fow2 , and Exo8/Rev1 yields 525 bp and 686 bp products for WT and KO alleles , as shown in dark and light gray boxes in ( A ) , respectively . ( D ) An equal amount of protein ( 20 μg ) from the tissues of WT and KO mice was electrophoresed for immunoblotting with antibodies toward exophilin-8 and GAPDH . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 004 We then compared the distribution of insulin granules between wild-type and exophilin-8-null islets . We first coimmunostained insulin as a granule marker and Na+-K+ ATPase as a plasma membrane marker in isolated islets . Although the antibodies were accessible to only surface β-cells , insulin granules were preferentially polarized close to the cell edges in wild-type islets , whereas they were diffusely distributed in the perinuclear cytoplasm in exophilin-8-null islets ( Figure 2A ) . Electron microscopy revealed that exophilin-8-null β-cells have a significantly lower number of granules that have centers within 300 nm of the plasma membrane ( Figure 2B , C ) . Notably , however , they still exhibited granules directly attached to the plasma membrane ( see arrows in Figure 2B ) . 10 . 7554/eLife . 26174 . 005Figure 2 . Distribution of insulin granules in exophilin-8-null β-cells . ( A ) Islets isolated from WT and exophilin8-KO mice were coimmunostained with anti-insulin and anti-Na+-K+ ATPase antibodies . Note that the antibodies were accessible to only surface β-cells . Bars , 10 μm . Insets show details at a higher magnification . ( B ) The isolated islets were cultured overnight and incubated in 2 . 8 mM glucose buffer at 37°C for 1 hr . They were then fixed and processed in a standard fashion for electron microscopy . Bar , 1 μm . Squares in left panels are shown at a higher magnification in right panels . Black arrowheads indicate the position of the plasma membrane , whereas red arrows indicate granules directly attached to the plasma membrane . ( C ) The distributions of insulin granules were morphometrically analyzed by electron microscopy in a total of nine β cells from three , 20- to 22-week-old male WT ( gray columns ) or KO ( red columns ) mice ( three cells from each individual mouse ) . All granules with centers that resided within 500 nm of the plasma membrane were categorized at 100 nm intervals . Data are shown as a percentage of the total granule number , and are shown as means ± SEM . *p values calculated using two-tailed unpaired t-test are 0 . 00182 ( <100 nm ) , 0 . 01360 ( 100–200 nm ) , 0 . 021 ( 200–300 nm ) , and 1 . 4 × 10−3 ( ≥500 nm ) , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 005 To understand the molecular mechanisms by which exophilin-8 functions , we investigated its interacting proteins , using the tandem affinity purification approach based on the Myc-TEV-FLAG ( MEF ) tag , as previously described ( Ichimura et al . , 2005; Matsunaga et al . , 2009 , 2017 ) . Namely , we expressed exophilin-8 fused to the MEF tag at its amino terminus in MIN6 cells and recovered the bound proteins in successive purification steps . Among the protein bands specific for the MEF-exophilin-8 eluate ( Figure 3A ) , those with the highest molecular mass ( 150 ~ 190 kDa ) were identified as RIM-BP2 and myosin-VIIa by a liquid chromatography ( LC ) -tandem mass spectrometry ( MS/MS ) analysis . Because the interaction with myosin-VIIa was already known , we further investigated that with RIM-BP2 . We confirmed the presence of RIM-BP2 in the immunoprecipitate of MEF-exophilin-8 in MIN6 cells ( Figure 3B ) . We also identified the endogenous complex between the two proteins in another β-cell line , INS-1 823/13 ( Figure 3C ) . Although RIM-BP2 was downregulated in exophilin-8-null pancreatic islets ( Figure 3D ) , it was expressed in wild-type islets and in these β-cell lines at even higher levels than in brain ( Figure 3E ) , where it was initially identified ( Wang et al . , 2000; Hibino et al . , 2002 ) . 10 . 7554/eLife . 26174 . 006Figure 3 . Identification of RIM-BP2 as an exophilin-8-interaction protein . ( A ) MIN6 cells expressing MEF-exophilin-8 or control LacZ protein were lysed and subjected to MEF-tag based purification . Bound proteins were detected by SDS-PAGE and Oriole fluorescent gel staining . Six bands specific to MEF-exophilin-8 are numbered . The number 1 band was identified as RIM-BP2 by LC-MS/MS analysis . ( B ) MIN6 cells were infected by adenovirus encoding MEF-exophilin-8 or control GFP protein . The immunoprecipitates ( IP ) with anti-FLAG antibody , as well as 1/100 of the original extracts , were immunoblotted with anti-FLAG and anti-RIM-BP2 antibodies . ( C ) The immunoprecipitates with anti-exophilin-8 antibody or control immunoglobulin G ( IgG ) in INS-1 832/13 cells , as well as 1/100 of the original extract , were immunoblotted with anti-exophilin-8 and anti-RIM-BP2 antibodies . ( D ) The total islet protein lysates ( 20 μg ) from wild-type ( WT ) or exophilin-8-knockout ( KO ) mice were analyzed by immunoblotting with antibodies against exophilin-8 , RIM-BP2 , and α-tubulin . ( E ) The protein lysates ( 20 μg ) from wild-type mouse tissues or cultured β-cell lines were analyzed by immunoblotting with anti-RIM-BP2 and anti-α-tubulin antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 006 We next determined the binding domains responsible for the interaction between exophilin-8 and RIM-BP2 . RIM-BP2 has three separate SH3 domains and three contiguous fibronectin type III ( FNIII ) domains ( Figure 4A ) . When expressed in HEK293A cells , the first SH3 domain significantly , and the third SH3 domain strongly , interacted with exophilin-8 , although the second SH3 domain or the whole FNIII domains did not ( Figure 4B ) . Exophilin-8 can be divided into Rab27-binding domain ( RBD ) , myosin-binding domain ( MBD ) , and actin-binding domain ( ABD ) ( Fukuda and Kuroda , 2002 ) . The C-terminus of exophilin-8 containing ABD bound RIM-BP2 , although the N-terminus consisting of RBD and MBD did not ( Figure 4A , C ) . Exophilin-8 has two SH3 domain-interacting proline-rich sequences , RXXPXXP ( Mayer , 2001 ) , at residues 474–480 in the MBD and at residues 798–804 in the ABD ( Figure 4A ) . Consistent with the finding of the above binding experiments ( Figure 4C ) , the PA mutant replacing arginine and proline residues with alanine residues in the motif of ABD , but not that of MBD , disrupted the interaction with RIM-BP2 ( Figure 4D ) . We further confirmed that wild-type exophilin-8 , but not the PA ( ABD ) mutant , interacts with endogenous RIM-BP2 in MIN6 cells ( Figure 4E ) . Taken together , these findings indicate that the two proteins interact between the first and/or third SH domains in RIM-BP2 and the C-terminal RXXPXXP sequence in exophilin-8 . 10 . 7554/eLife . 26174 . 007Figure 4 . Protein domains responsible for the interaction between RIM-BP2 and exophilin-8 . ( A ) Schematic representation of mouse RIM-BP2 ( left ) , mouse exophilin-8 ( right ) , and their deletion or point mutants . ( B–D ) HEK293A cells were transfected with plasmids encoding the indicated proteins shown in ( A ) . The immunoprecipitates with anti-FLAG ( B , D ) or anti-HA ( C ) antibodies , as well as 1/30 of the original extracts , were immunoblotted with anti-FLAG , anti-GFP , and anti-HA antibodies . ( E ) MIN6 cells were infected with adenovirus encoding the indicated proteins . The immunoprecipitates with anti-FLAG antibodies , as well as 1/100 of the original extracts , were immunoblotted with anti-FLAG and anti-RIM-BP2 antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 007 Exogenously expressed exophilin-8 redistributes secretory granules to the cell corners and/or the tips of cell extensions ( Mizuno et al . , 2011; Huet et al . , 2012 ) . Immunostaining experiments have revealed that both RIM-BP2 and exophilin-8 are endogenously colocalized with insulin granules , especially those accumulated at the cell corners , in INS-1 832/13 cells ( Figure 5A , Figure 5—figure supplement 1A ) . Because the antibodies toward exophilin-8 and RIM-BP2 were both derived from rabbits , we investigated the colocalization of exogenously expressed exophilin-8 with endogenous RIM-BP2 . We found that wild-type exophilin-8 is colocalized with RIM-BP2 and accumulates both insulin granules and RIM-BP2 at the cell corners ( Figure 5B , C , Figure 5—figure supplement 1B , C ) . By contrast , the PA ( ABD ) mutant deficient in binding activity to RIM-BP2 was not polarized at the cell corners and dispersed insulin granules and RIM-BP2 diffusively in the cytoplasm . These findings indicate that exophilin-8 induces peripheral localization of insulin granules through the interaction with RIM-BP2 . 10 . 7554/eLife . 26174 . 008Figure 5 . Exophilin-8 mutant deficient in binding to RIM-BP2 fails to cluster RIM-BP2 and insulin granules at cell corners . ( A ) INS-1 832/13 cells were coimmunostained with anti-insulin and either anti-exophilin-8 ( upper ) or anti-RIM-BP2 ( lower ) antibodies . ( B , C ) INS-1 832/13 cells were infected with adenovirus encoding MEF-tagged , wild-type ( WT ) or PA ( ABD ) mutant exophilin-8 . They were coimmunostained with anti-FLAG and anti-insulin antibodies ( B ) or with anti-myc and anti-RIM-BP2 antibodies ( C ) . Bars , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 00810 . 7554/eLife . 26174 . 009Figure 5—figure supplement 1 . More images of INS-1 832/13 cells and those expressing wild-type or PA ( ABD ) mutant exophilin-8 . INS-1 832/13 cells and those expressing MEF-tagged , wild-type ( WT ) or PA ( ABD ) mutant exophilin-8 were coimmunostained with the indicated antibodies , as shown in Figure 5 . Bars , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 009 To reinforce the significance of the interaction , we performed rescue experiments in islets isolated from exophilin-8-null mice . Although wild-type exophilin-8 expressed at the endogenous protein level increased glucose-stimulated insulin secretion , the PA ( ABD ) mutant failed to do so ( Figure 6A , Figure 6—figure supplement 1A , B ) . Furthermore , overexpression of RIM-BP2 significantly enhanced insulin secretion in wild-type islets , but not in exophilin-8-null islets ( Figure 6B , Figure 6—figure supplement 1C ) . A similar secretion-promoting effect of RIM-BP2 was observed in MIN6 cells , but not in the exophilin-8-null β-cell line ( Figure 6C , Figure 6—figure supplement 1D ) , which was generated by a method similar to that by which the ‘wild-type’ MIN6 cell line was previously established ( Miyazaki et al . , 1990 ) . We then generated RIM-BP2 lacking all the three SH3 domains ( RIM-BP2ΔSH3 ) , and confirmed that this mutant loses the binding activity to exophilin-8 ( Figure 6D ) . When overexpressed in MIN6 cells , only wild-type RIM-BP2 , but not RIM-BP2ΔSH3 , enhanced insulin secretion ( Figure 6E , Figure 6—figure supplement 1E ) . These findings indicate that the secretion-promoting activity of exophilin-8 or RIM-BP2 requires the interaction with the other protein . 10 . 7554/eLife . 26174 . 010Figure 6 . Exophilin-8 and RIM-BP2 potentiate insulin secretion through their mutual interaction . ( A ) Pancreatic islets isolated from exophilin-8-knockout ( KO ) mice were infected with adenovirus encoding control GFP or MEF-tagged , wild-type ( WT ) or PA ( ABD ) mutant exophilin-8 , as shown in Figure 6—figure supplement 1A , B , and were cultured in a fresh medium for 48 hr followed by Krebs-Ringer bicarbonate ( KRB ) buffer containing 2 . 8 mM glucose for 30 min . The cells were then incubated in the same buffer for 30 min ( gray bars ) followed by the buffer containing 25 mM glucose for 30 min ( black bars ) . The ratios of insulin secreted into the medium to that remaining in the cells were normalized to those found in control cells stimulated by the secretagogue . ( B , C ) WT or KO islets ( B ) , and MIN6 or exophilin-8-null β-cell lines ( C ) were infected with adenovirus encoding either GFP or MEF-RIM-BP2 , as shown in Figure 6—figure supplement 1C and D , respectively , and were subjected to insulin secretion assays as in ( A ) stimulated by either 25 mM glucose ( B ) or 60 mM KCl ( C ) for 30 min . ( D ) INS-1 832/13 cells were infected with adenovirus encoding control GFP , wild-type MEF-RIM-BP2 , or MEF-RIM-BP2ΔSH3 lacking all the SH3 domains , and the immunoprecipitates with anti-FLAG antibody , as well as 1/100 of the original extracts , were immunoblotted with anti-FLAG , anti-exophilin-8 , and anti-myosin-VIIa antibodies . Note that RIM-BP2ΔSH3 lost binding activities to exophilin-8 and myosin-VIIa . ( E ) MIN6 cells were infected with adenovirus encoding control GFP , or wild-type or ΔSH3 RIM-BP2 , as shown in Figure 6—figure supplement 1E , and were subjected to insulin secretion assays as in ( C ) . All quantitative data are means ± SD ( n = 3 ) . *p values calculated using two-tailed unpaired t-test are as follows: ( A ) 0 . 03278 ( GFP ) , 0 . 03607 ( Exophilin-8 PA ) vs Exophilin-8 WT , ( B ) 0 . 04952 ( WT-islet MEF-RIM-BP2 ) , 0 . 03503 ( KO-islet GFP ) , 0 . 00015 ( KO-islet MEF-RIM-BP2 ) vs WT-islet GFP , ( C ) 9 . 4 × 10−5 ( MIN6 MEF-RIM-BP2 ) , 2 . 5 × 10−6 ( KO GFP ) , 1 . 3 × 10−7 ( KO MEF-RIM-BP2 ) vs MIN6 GFP , and ( E ) 1 . 7 × 10−3 ( MIN6 GFP ) , 0 . 014 ( MIN6 MEF-RIM-BP2ΔSH3 ) vs MIN6 GFP MEF-RIM-BP2 . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 01010 . 7554/eLife . 26174 . 011Figure 6—figure supplement 1 . Expression levels of exophilin-8 and RIM-BP2 . ( A , B ) Pancreatic islets isolated from exophilin-8-knockout ( KO ) mice were infected with adenovirus encoding control GFP or MEF-tagged , wild-type ( WT ) or PA ( ABD ) mutant exophilin-8 . The islet extracts , as well as those from noninfected WT mice , were immunoblotted with anti-β-actin and anti-exophilin-8 antibodies to compare the expression levels between either exogenous and endogenous exophilin-8 ( A ) or exogenous WT and PA ( ABD ) exophilin-8 ( B ) . ( C , D ) WT or KO islets ( C ) , and MIN6 or exophilin-8-null β-cell lines ( D ) were infected with adenovirus encoding either GFP or MEF-RIM-BP2 . They were lysed for immunoblotting with anti-FLAG , anti-exophilin-8 , and anti-β-actin ( C ) or anti-α-tubulin ( D ) antibodies . ( E ) MIN6 cells were infected with adenovirus encoding control GFP , or wild-type or ΔSH3 RIM-BP2 , and the cell extracts were immunoblotted with anti-FLAG and anti-β-actin antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 011 We then examined the silencing effects of exophilin-8 or RIM-BP2 on peripheral accumulation and exocytosis of insulin granules in INS-1 832/13 cells , where robust glucose-stimulated insulin secretion occurs ( Hohmeier et al . , 2000 ) . Although exogenous or endogenous RIM-BP2 expressed in β-cells exhibited a doublet in gels ( Figure 3E , Figure 6—figure supplement 1C , D ) , both bands were specifically downregulated by small interfering RNA ( siRNA ) against RIM-BP2 ( Figure 7A ) , suggesting that they represent differentially modified RIM-BP2 . Exophilin-8 or RIM-BP2 knockdown profoundly decreased glucose-stimulated insulin secretion ( Figure 7B ) . Furthermore , either type of knockdown induced diffusive redistributions of granules and the other protein in the cytoplasm ( Figure 7C–E , Figure 7—figure supplement 1 ) . Therefore , both exophilin-8 and RIM-BP2 are necessary for peripheral accumulation and efficient exocytosis of secretory granules . To substantiate this conclusion , we performed rescue experiments . When wild-type or ΔSH3 RIM-BP2 was expressed in RIM-BP2-knockdown cells , only the wild type rescued the decreased insulin secretion and restored the peripheral granule accumulation ( Figure 7F , G , Figure 7—figure supplements 2 and 3 ) , again supporting the importance of the complex formation between the two proteins . 10 . 7554/eLife . 26174 . 012Figure 7 . Silencing of exophilin-8 or RIM-BP2 decreases the cell-corner localization and exocytosis of insulin granules . ( A ) INS-1 832/13 cells were transfected with control siRNA duplexes , or siRNA duplexes against exophilin-8 or RIM-BP2 . The cell extracts were immunoblotted with anti- RIM-BP2 , anti-exophilin-8 , and anti-β-actin antibodies . ( B ) INS-1 832/13 cells treated with siRNA as shown in ( A ) were incubated for 2 hr in KRB buffer containing 2 . 8 mM glucose , and were then stimulated for 60 min in the same buffer ( gray bars ) or buffer containing 25 mM glucose ( black bars ) . The ratios of insulin secreted into the medium to that remaining in the cells were normalized to those found in control cells stimulated by 25 mM glucose . ( C–E ) The cells treated with siRNA were immunostained with anti-insulin antibody and either anti-RIM-BP2 ( D ) or anti-exophilin-8 ( E ) antibodies . In each experiment , a total of 100 cells were visually inspected , and the fraction of cells exhibiting a higher granule density in at least one cell corner or extension than in the cell center was manually counted ( C ) . Bars , 10 μm . ( F , G ) INS-1 832/13 cells treated with control siRNA or siRNA against RIM-BP2 were infected with adenovirus encoding control GFP , or wild-type or ΔSH3 MEF-RIM-BP2 , as shown in Figure 7—figure supplement 2 . They were subjected to insulin secretion assays ( F ) as in ( B ) , or were immunostained with anti-insulin antibody to examine granule localization ( G ) as in ( C ) . All quantitative data are means ± SD ( n = 5 for B , and n = 4 for C , and n = 3 for F ) . *p values calculated using two-tailed unpaired t-test are as follows: ( B ) 1 . 9 × 10−7 ( si Exophilin-8 ) , 7 . 3 × 10−6 ( si RIM-BP2 ) vs si Control , ( C ) 0 . 00010 ( si Exophilin-8 ) , 0 . 00025 ( si RIM-BP2 ) vs si Control , ( F ) 3 . 7 × 10−3 ( si RIM-BP2 , GFP ) , 8 . 2 × 10−3 ( si RIM-BP2 , MEF-RIM-BP2 ) vs si Control , GFP , and 0 . 033 ( si RIM-BP2 , GFP ) , 0 . 060 ( si RIM-BP2 , MEF-RIM-BP2ΔSH3; marked by # ) vs si RIM-BP2 , MEF-RIM-BP2 , and ( G ) 1 . 1 × 10−3 ( si RIM-BP2 , GFP ) , 5 . 5 × 10−3 ( si RIM-BP2 , MEF-RIM-BP2 ) vs si Control , GFP , and 7 . 7 × 10−3 ( si RIM-BP2 , GFP ) , 7 . 1 × 10−5 ( si RIM-BP2 , MEF-RIM-BP2ΔSH3 ) vs si RIM-BP2 , MEF-RIM-BP2 . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 01210 . 7554/eLife . 26174 . 013Figure 7—figure supplement 1 . More images of INS-1 832/13 cells treated with siRNA . INS-1 832/13 cells treated with control siRNA or siRNA against exophilin-8 or RIM-BP2 were immunostained with anti-insulin antibody and either anti-RIM-BP2 or anti-exophilin-8 antibodies , as shown in Figure 7D , E . Bars , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 01310 . 7554/eLife . 26174 . 014Figure 7—figure supplement 2 . Expression levels of endogenous and exogenous RIM-BP2 . INS-1 832/13 cells treated with control siRNA or siRNA against RIM-BP2 were infected with adenovirus encoding control GFP , or wild-type or ΔSH3 MEF-RIM-BP2 . The cell extracts were immunoblotted with anti-RIM-BP2 and anti-α-tubulin antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 01410 . 7554/eLife . 26174 . 015Figure 7—figure supplement 3 . Granule localization in INS-1 832/13 cells treated with siRNA against RIM-BP2 and then expressing wild-type or ΔSH3 RIM-BP2 . INS-1 832/13 cells were treated with control siRNA or siRNA against RIM-BP2 and then infected adenovirus expressing either GFP , or wild-type or ΔSH3 RIM-BP2 , as shown in Figure 7—figure supplement 2 . They were immunostained with anti-insulin antibody to examine granule localization . Bars , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 015 We then molecularly characterized the exophilin-8–RIM-BP2 complex . To reveal the exophilin-8-interacting proteins via RIM-BP2 , we compared the proteins coimmunoprecipitated with MEF-tagged , wild-type exophilin-8 with those from the PA ( ABD ) mutant , in the presence of hemagglutinin ( HA ) -tagged RIM-BP2 in INS-1 832/13 cells . As shown in Figure 8A , the wild type , but not the PA ( ABD ) mutant , coprecipitated RIM-BP2 , and concordantly interacted with RIM-BP2-interacting Cav1 . 3 and RIM , and RIM-interacting Munc13-1 , as has been observed in neurons ( Wang et al . , 2000; Betz et al . , 2001; Hibino et al . , 2002 ) . By contrast , both the wild type and the mutant interacted with previously known exophilin-8-interacting proteins , such as Rab27a , protein kinase A ( PKA ) , Sec6 ( Goehring et al . , 2007 ) , and actin . However , myosin-Va was not found in the immunoprecipitate of either wild-type or mutant exophilin-8 , but was easily detected in that of exophilin-3 ( also known as melanophilin and Slac2a; Figure 8B ) , another Rab27 effector that exhibits binding activity to myosin-Va ( Fukuda et al . , 2002; Nagashima et al . , 2002; Strom et al . , 2002; Wu et al . , 2002 ) . Instead , we identified myosin-VIIa in the immunoprecipitate , the tail domain of which has been reported to interact with exophilin-8 in both heterologous cells and in vitro ( El-Amraoui et al . , 2002; Fukuda and Kuroda , 2002 ) . Unexpectedly , however , the PA ( ABD ) exophilin-8 , despite the mutation outside of MBD , failed to interact with myosin-VIIa . This finding indicates that the same proline motif of exophilin-8 is involved in binding with myosin-VIIa as well as with RIM-BP2 , or that exophilin-8 indirectly interacts with myosin-VIIa through RIM-BP2 . To determine which was the case , we investigated whether the interaction between RIM-BP2 and myosin-VIIa occurs in the absence of exophilin-8 . We found that the interaction persists in the exophilin-8-null β-cell line and is preserved by the exogenous expression of wild-type exophilin-8 ( Figure 8C ) , indicating that RIM-BP2 , but not exophilin-8 , is primarily involved in the interaction with myosin-VIIa in cells . 10 . 7554/eLife . 26174 . 016Figure 8 . Exophilin-8 indirectly interacts with myosin-VIIa via RIM-BP2 in INS-1 832/13 cells . ( A , B ) INS-1 832/13 cells were infected with adenoviruses expressing HA-RIM-BP2 and either control GFP , or MEF-tagged , wild-type ( WT ) or PA ( ABD ) mutant exophilin-8 ( A , B ) , and with that expressing MEF-exophilin-3 ( B ) . The immunoprecipitates with anti-FLAG antibody , as well as 1/100 of the original extracts , were immunoblotted with the indicated antibodies to investigate the interacting proteins . The protein bands found in Input lanes of Munc13-1 are non-specific proteins , whereas those found in the immunoprecipitates of SNAP25 are immunoglobulin G . ( C ) Exophilin-8-null β-cell lines were infected with adenoviruses expressing GFP , HA-tagged wild-type exophilin-8 , and/or MEF-RIM-BP2 . The immunoprecipitates with anti-FLAG antibody were immunoblotted with the indicated antibodies as in ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 016 To further investigate the interactions , we simultaneously expressed these proteins in HEK293A cells . The One-STrEP-FLAG ( OSF ) -tag was attached to wild-type or PA mutant exophilin-8 ABD , and the binding proteins were pulled down using Strept-Tactin beads . As expected , wild-type , but not PA mutant , exophilin-8 ABD bound RIM-BP2 , strongly indicating the direct interaction because there were no other specific proteins ( Figure 9A , lanes 2 and 5 ) . Unexpectedly , however , both wild-type and PA mutant exophilin-8 ABD bound myosin-VIIa with or without RIM-BP2 expression ( Figure 9A , lanes 3 , 4 , 6 , and 7 ) , in contrast to the finding in INS-1 832/13 cells ( Figure 8B ) . Therefore , exophilin-8 seems to directly interact with myosin-VIIa at least in heterologous cells , consistent with the previous findings ( El-Amraoui et al . , 2002; Fukuda and Kuroda , 2002 ) , and furthermore , this interaction does not require the MBD or the proline-rich motif in the ABD of exophilin-8 . We then expressed OST-tagged RIM-BP2 and the binding proteins were examined . We confirmed that RIM-BP2 bound wild-type , but not PA mutant , exophilin-8 ABD ( Figure 9B , lanes 2 and 3 ) . However , RIM-BP2 could not bind myosin-VIIa without simultaneous expression of wild-type exophilin8 ABD ( Figure 9B , lanes 4–6 ) , indicating that RIM-BP2 cannot directly bind myosin-VIIa . We noticed that , although the approximate molecular mass by sodium dodecyl sulfate ( SDS ) -polyacrylamide gel electrophoresis ( PAGE ) of myosin-VIIa expressed in HEK293A cells is ~260 kDa corresponding to its calculated molecular mass ( Figure 9A , B ) , that of myosin VIIa interacting with RIM-BP2 in β-cell lines is ~170 kDa ( Figure 8B , C ) . As described in Figure 3A , myosin-VIIa was identified with RIM-BP2 from exophlin-8-interacting , 150 ~ 190 kDa protein bands in MIN6 cells by LC-MS/MS . Furthermore , in INS-1 832–13 cells , both minor 260 kDa and major 170 kDa protein bands were similarly downregulated by myosin-VIIa siRNA treatment ( Figure 9C ) . To further examine whether the 170 kDa protein represents myosin-VIIa specifically expressed in β-cells , we examined RIM-BP2-binding myosin-VIIa in INS1 832/13 cells expressing OST-tagged RIM-BP2 , using another myosin-VIIa antibody we generated . We found that only the 170 kDa protein was pulled down with RIM-BP2 , although the 260 kDa protein was also pulled down after exogenous expression of myosin-VIIa ( Figure 9D ) . These findings indicate that the 170 kDa protein is derived from myosin-VIIa endogenously expressed in β-cells . By contrast , the 260 kDa protein is derived from transfected myosin-VIIa cDNA and interacts with RIM-BP2 through exophilin-8 , because it can directly interact with exophilin-8 but not with RIM-BP2 ( Figure 9A , B ) . Finally , we investigated the endogenous complex formation in INS1 832/13 cells by sucrose density gradient centrifugation ( Figure 9—figure supplement 1 ) . Exophilin-8 ( 130 kDa ) and RIM-BP2 ( 150 kDa ) exhibited similarly wide distributions in fractions 3–10 , whereas both endogenous 170 kDa and 260 kDa myosin-VIIa showed relatively narrower distributions around fractions 6–10 . Judged from the positions of molecular mass markers , these findings are consistent with the ternary complex formation among these proteins in the cells , although each protein also appeared to exist as a monomer . 10 . 7554/eLife . 26174 . 017Figure 9 . The molecular mass of myosin-VIIa interacted with RIM-BP2 in INS-1 832/13 cells is 170 kDa , whereas that directly binds exophilin-8 in HEK293A cells is 260 kDa . ( A ) HEK293A cells were transfected with plasmids expressing RIM-BP2 , myosin-VIIa , and/or OSF-tagged , wild-type or PA mutant exophilin-8 ABD . Exophilin-8 ABD and the binding proteins were pulled down using Strep-Tactin beads , and were subjected to SDS-PAGE and Coomassie Brilliant Blue staining . ( B ) HEK293A cells were transfected with plasmids expressing wild-type or PA mutant exophilin-8 ABD , myosin-VIIa , and OSF-tagged RIM-BP2 . RIM-BP2 and the binding proteins were pulled down as in ( A ) . ( C ) INS-1 832/13 cells were transfected with control siRNA duplexes or siRNA duplexes against myosin-VIIa , and the cell extracts were immunoblotted with anti-myosin-VIIa and anti-α-tubulin antibodies . Note that both 260 kDa and major 170 kDa protein bands were downregulated by myosin-VIIa siRNA . ( D ) INS-1 832–13 cells were transfected with plasmids expressing myosin-VIIa and/or OSF-tagged RIM-BP2 . Because the transfection efficiency in these cells are poor compared with that in HEK293A cells , the binding proteins were detected by immunoblotting . The anti-myosin-VIIa used in this figure was that generated by our laboratory ( αMyo7 ) , in contrast to the commercially available antibody used in Figures 6D , 8 , 9C and 10 ( see Materials and methods ) . ( E ) Illustration of the cortical F-actin network with secretory granules and the plasma membrane , the exophilin-8–RIM-BP2–170 kDa myosin-VIIa complex found in the present study , and the previously known , exocytic protein interactions ( Südhof , 2013 ) . The exact molecular nature of the 170 kDa form of myosin-VIIa and whether it directly interacts with RIM-BP2 are currently unknown . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 01710 . 7554/eLife . 26174 . 018Figure 9—figure supplement 1 . Sucrose density gradient centrifugation analysis of the exophilin-8–RIM-BP2–myosin-VIIa complex . Molecular mass markers ( A ) , bovine serum albumin ( 67 kDa ) , aldorase ( 160 kDa ) , and thyroglobulin ( 660 kDa ) , or the cytosolic fraction of INS1 832/13 cells ( B ) were separated using sucrose density gradient centrifugation . Fractions were subjected by SDS-PAGE separated using sucrose density gradient centrifugation . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 018 We then investigated the silencing effects of the exophilin-8-interacting proteins . In pancreatic β-cells , the two isoforms of RIM , RIM1 and RIM2 , are expressed ( Iezzi et al . , 2000; Yasuda et al . , 2010 ) , and homozygous ablation of RIM2 or RIM-interacting Munc13-1 has been shown to affect insulin secretion profoundly ( Kang et al . , 2006; Yasuda et al . , 2010 ) . The L-type Cav1 . 3 subtype is a predominant VDCC mediating stimulus-induced Ca2+ influx to trigger insulin secretion ( Yang and Berggren , 2006 ) . Consistently , when these proteins were downregulated separately by specific siRNA ( Figure 10—figure supplement 1 ) , glucose-induced insulin secretion was markedly decreased ( Figure 10A ) . However , the peripheral accumulation of granules was not affected ( Figure 10B , C , Figure 10—figure supplement 2 ) , suggesting that these proteins function after granules are recruited to the cell periphery . Endogenous myosin-Va and -VIIa and their respective C-terminal tails expressed exogenously in INS-1 832/13 cells were all associated with insulin granules , including those accumulated in the actin-rich cell periphery ( Figure 10D , Figure 10—figure supplement 3A , B ) . Although silencing of either myosin-Va or -VIIa decreased insulin secretion , only the latter affected the peripheral accumulation of granules ( Figure 10A–C , Figure 10—figure supplement 2 ) , suggesting that the two motor proteins may have differential effects on secretory granule positioning . 10 . 7554/eLife . 26174 . 019Figure 10 . Myosin-VIIa clusters granules at cell corners . ( A–C ) INS-1 832–13 cells were transfected with control siRNA duplexes or siRNA duplexes against the indicated proteins ( Figure 9B , Figure 10—figure supplement 1 ) and were subjected to insulin secretion assays ( A ) as described in Figure 7B , or to insulin immunostaining ( B , C ) to examine the peripheral accumulation of granules as described in Figure 6C–E . ( D ) INS-1 832/13 cells were coimmunostained with anti-insulin antibody and either with anti-myosin-Va or -VIIa antibody . Bars , 10 μm . All quantitative data are means ± SD ( n = 4 ) . *p values calculated using two-tailed unpaired t-test are as follows: ( A ) 0 . 00029 ( si Myosin-VIIa ) , 0 . 01168 ( si Myosin-Va ) , 0 . 01881 ( si RIM1 ) , 0 . 00688 ( si RIM2 ) , 0 . 00045 ( si Cav1 . 3 ) vs si Control , and ( B ) 6 . 2 × 10−5 ( si Myosin-VIIa ) vs si Control . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 01910 . 7554/eLife . 26174 . 020Figure 10—figure supplement 1 . Silencing of myosin-Va , RIM , and Cav1 . 3 by siRNA . INS-1 832/13 cells were transfected with control siRNA duplexes or siRNA duplexes against the indicated protein . The cell extracts were immunoblotted with the indicated antibody to evaluate its downregulation . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 02010 . 7554/eLife . 26174 . 021Figure 10—figure supplement 2 . More images of INS-1 832–13 cells treated with siRNA . INS-1 832/13 cells treated with siRNA against the indicated protein were immunostained with anti-insulin antibody , as shown in Figure 10B . Bars , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 02110 . 7554/eLife . 26174 . 022Figure 10—figure supplement 3 . Localization of F-actin and its motor proteins . ( A ) INS-1 832/13 cells expressing the C-terminal tail of HA-tagged myosin-VIIa ( 1550–2215 amino acids ) or -Va ( 1444–1853 amino acids ) were coimmunostained with anti-insulin and anti-HA antibodies . ( B ) INS-1 832/13 cells were costained with anti-insulin antibody and rhodamine-labeled phalloidin . Bars , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 26174 . 022
In the present study , we showed that mice deficient in the Rab27a effector , exophilin-8 , exhibit glucose intolerance and impaired insulin secretion in vivo . We further identified a new multiprotein complex , in which exophilin-8 directly interacts with RIM-BP2 and then binds the actin-motor protein , myosin-VIIa , and the exocytic machinery , such as Cav1 . 3 and RIM . Disruption of the interaction by ablation or mutation of either exophilin-8 or RIM-BP2 leads to loss of polarized granule distribution and to a marked decrease in granule exocytosis . RIM-BP2 also has binding activities to the α1 subunits of L- , N- , and P/Q type VDCC ( Hibino et al . , 2002 ) . On the other hand , RIM , originally identified as an effector of Rab3 ( Wang et al . , 1997 ) , binds and activates a priming factor , Munc13-1 ( Deng et al . , 2011 ) , and also has binding activities to N- and P/Q type VDCC , but not to L-type VDCC ( Kaeser et al . , 2011 ) . Because RIM binds Rab3 , but not Rab27 ( Fukuda , 2003 ) , and because exophilin-8 binds Rab27 , but not Rab3 ( Fukuda and Kuroda , 2002 ) , the RIM-BP2–myosin-VIIa complex can theoretically associate with granules via either Rab27a–exophilin-8 or Rab3–RIM ( Figure 9E ) . Although an interaction of exophilin-8 with myosin-Va has been shown in heterologous cells and in vitro ( Fukuda and Kuroda , 2002; Desnos et al . , 2003 ) , we cannot find evidence that the two proteins act together in secretory cells: they do not form a complex and silencing of myosin-Va does not affect peripheral granule accumulation unlike that of exophilin-8 , although it is possible that the degree of myosin-Va depletion in our experimental condition is enough to decrease insulin secretion , but not to affect granule localization . Some question remains regarding the physiological and functional interaction between exophilin-8 and myosin-Va . For example , exophilin-8 does not interact with myosin-Va under physiological conditions in MIN6 cells , where only the brain-isoform of myosin-Va is expressed ( Brozzi et al . , 2012 ) . Furthermore , these two proteins show negligible physical and functional interactions in the melanocyte cell line , Melan-a , where the melanocyte-isoform of myosin-Va is expressed ( Kuroda and Fukuda , 2005 ) . We instead show that myosin-VIIa associates with secretory granules and indirectly interacts with exophilin-8 via RIM-BP2 . Furthermore , silencing of myosin-VIIa disrupts granule clustering in the actin-rich cell periphery and markedly decreases stimulus-induced granule exocytosis , as does the silencing of exophilin-8 . Myosin-VIIa appears to function in the anchoring , rather than in the active transport , of granules within the F-actin network , because exophilin-8-positive granules are markedly immobile in the resting state in living cells ( Desnos et al . , 2003; Mizuno et al . , 2011; Huet et al . , 2012 ) . It should be noted that myosin-VIIa identified in the exophilin-8–RIM-BP2 complex in β-cells shows a molecular mass ~170 kDa in gels , much smaller than its authentic mass ~260 kDa . Because this form does not appear to be generated from transfected myosin-VIIa cDNA in INS1 832/13 cells , it may represent an alternatively spliced form that specifically functions in secretory cells . In fact , such a small myosin-VIIa isoform has been reported in the UCSC Genome Browser ( https://genome . ucsc . edu/ ) . It has recently been shown that , among the four alternatively spliced isoforms of the minus-ended motor , myosin-VIa , the small insert isoform specifically tethers secretory granules to the cortical actin ( Tomatis et al . , 2013 ) , although its molecular link to granules , such as Rab and its effector , has not been identified . Although the exact molecular nature of the 170 kDa form of myosin-VIIa and its interaction mode with RIM-BP2 are unknown , our findings that either the PA ( ABD ) exophilin-8 or the ΔSH3 RIM-BP2 fails to restore the peripheral granule accumulation or to rescue the decreased insulin secretion in knockdown or knockout cells strongly indicate that the exophilin-8–RIM-BP2–170 kDa–myosin-VIIa complex formation is functionally critical for granule exocytosis . Currently , roles for myosin-VIIa are well-documented in actin-rich protrusion , such as stereocilia of cochlear and vestibular hair cells and microvilli in intestine brush borders and photoreceptor cells ( Yan and Liu , 2010; Lu et al . , 2014 ) , but no such roles have been identified in granule anchoring or trafficking . This may be because of the existence of parallel secretory pathways bypassing the complex formation: all the granules may not be captured within the cortical F-actin network , or some granules may be captured there by other mechanisms . In fact , exophilin-8-null islets preserve approximately half of the normal level insulin secretion in response to glucose or depolarization stimulation . Further research is required to elucidate the mechanism by which secretory cells differentially or redundantly engage multiple motor proteins , such as myosins-Va , -VI , and -VIIa . Under electron microscopy , exophilin-8-null pancreatic β-cells exhibit a decrease in the number of granules close to the plasma membrane , but still harbor docked granules whose limiting membrane is directly attached to the membrane . Using the same experimental protocol , we previously demonstrated that β-cells deficient in another Rab27 effector , granuphilin , exhibit a much greater decrease in the number of granules close to the plasma membrane , and almost complete loss of docked granules ( Gomi et al . , 2005 ) . It is notable that the granule number decrease in granuphilin-null β-cells occurs in a more restricted area , with granule centers residing within 200 nm of the plasma membrane , whereas the centers in exophilin-8-null β-cells are within 300 nm . This difference has a significant effect on the number of the granules directly attached to the plasma membrane , given that the diameter of insulin granules is 300 ~ 350 nm . Furthermore , when cultured β-cells overexpressing these proteins are observed under optical microscopy , granuphilin redistributes granules along the plasma membrane ( Torii et al . , 2004 ) , whereas exophilin-8 does so in a relatively broad peripheral area ( Mizuno et al . , 2011 ) . Therefore , it is likely that only a portion of exophilin-8 and its associated granules are linked to the Ca2+ channel and the exocytic machinery along the plasma membrane . These findings indicate that , compared with granuphilin , exophilin-8 plays a relatively minor role in granule docking , although it may indirectly contribute to the process by recruiting granules close to the plasma membrane . What is the physiological relevance of the exophilin-8-induced peripheral granule accumulation found in monolayer β-cells ? Although pancreatic β-cells do not display the typical epithelial polarity , they concentrate F-actin and exocytic machinery , including Cav1 . 3 , at cell edges and particularly at cell vertexes facing blood vessels within islets ( Geron et al . , 2015 ) . This polarity seems to be functionally important , because exocytosis in islets preferentially occurs in the vicinity of vessels , around which β-cells form rosettes ( Takahashi et al . , 2002 ) . The exophilin-8–RIM-BP2–myosin-VIIa complex can provide the molecular basis of these observations , because it interacts with both F-actin and Cav1 . 3 . In fact , we found that loss of exophilin-8 prevents polarized granule accumulation both in β-cells within islets and in those cultured in a monolayer . Therefore , even monolayer β-cells could display such a polarity in a cell-autonomous manner . Monolayer chromaffin cells have also been shown to cluster L- and P/Q type Ca2+ channels and exocytic machinery on the border of cortical F-actin cages ( Torregrosa-Hetland et al . , 2011 ) . How , then , do exophilin-8-positive granules captured within the F-actin network fuse upon secretagogue stimulation ? It has recently been shown that the cortical actomyosin II network acts like a casting net to drive granules towards the plasma membrane by undergoing relaxation in a secretagogue-dependent manner ( Papadopulos et al . , 2015 ) . Thus , the stimulus-induced Ca2+ influx not only triggers a final fusion reaction but also may help granules within the F-actin network make further access to the exocytic site . In summary , the novel exophilin-8–RIM-BP2–myosin-VIIa complex seems to replenish and to maintain a releasable pool of granules beneath the plasma membrane by locating granules both within the peripheral F-actin network and close to the Ca2+ channel and exocytic machinery . This may be the first to provide the molecular basis correlating the physical location of secretory granules with their functional pool , given that stable docking to the plasma membrane does not necessarily form a releasable granule pool ( Gomi et al . , 2005; Mizuno et al . , 2016 ) . Genetic or functional alterations of the complex may lead to type 2 diabetes and other secretory disorders in humans .
The exophilin-8 ( encoded by the Myrip gene ) knockout mice ( Accession No . CDB1100K: http://www2 . clst . riken . jp/arg/mutant%20mice%20list . html ) were generated as described elsewhere ( http://www2 . clst . riken . jp/arg/Methods . html ) . To construct a targeting vector , genomic fragments of the exophilin-8 locus were obtained from a BAC clone , RP24-276O9 ( BACPAC Resources ) . The exon4 of the exophilin-8 gene was disrupted by insertion of a loxP-flanked cassette of the neomycin resistance gene under the pgk promoter . Targeted TT2 ( derived from F1 of C57BL/6 and CBA ) embryonic stem cell clones ( Yagi et al . , 1993 ) were microinjected into eight-cell stage ICR embryos , and were then transferred into pseudopregnant ICR females . The resulting chimeras were bred with C57BL/6 mice , and heterozygous offspring were identified by Southern blotting and polymerase chain reaction ( PCR ) . The primers used for PCR were Myrip/Fow1 ( 5'-GATGGGTCCTGCTTCTCACC-3' ) and Myrip/Rev1 ( 5'-CTCCGCCCTCTTTCCAGAAC-3' ) for the wild-type allele , and Neo/Fow2 ( 5'-AGGACATAGCGTTGGCTACC-3 ) and Myrip/Rev1 for the targeted allele . Mutant lines were backcrossed with C57BL/6 mice . The protocol for animal experimentation was approved by the Institutional Animal Care and Use Committee of RIKEN Kobe Branch . All the animal experiments were conducted in accordance with the RIKEN institutional guideline and the rules and regulations of the Animal Care and Experimentation Committee , Gunma University . Wild-type and knockout males for the experiments were obtained by heterozygous mating between backcrossed progenies . Mice had free access to water and standard laboratory chow in an air-conditioned room with a 12 hr light/12 hr dark cycle . An intra-peritoneal glucose tolerance test ( 1 g glucose/kg body weight ) and an intra-peritoneal insulin tolerance test ( 0 . 75 U human insulin/kg body weight ) were performed as described previously ( Wang et al . , 2013 ) . Blood glucose levels were determined by a glucose oxidase method using Glutest sensor and Glutest Pro GT-1660 ( Sanwa Kagaku Kenkyujyo , Nagoya , Japan ) . Islet isolation by pancreatic duct injection of collagenase solution and insulin secretion assay in perifused islets were performed as described previously ( Wang et al . , 2013 ) . Briefly , isolated islets were perifused with standard low-glucose ( 2 . 8 mM ) Krebs-Ringer bicarbonate ( KRB ) buffer for 30 min . Thereafter , the collection of each fraction ( 1 ml/min ) was started , and an appropriate secretagogue was applied at 10 min after the start . Insulin was measured using an AlphaLISA insulin kit with an EnVision 2101 Multilabel Reader ( PerkinElmer , Waltham , MA ) . The examination of granule distribution by electron microscopy was performed as described previously ( Gomi et al . , 2005 ) . Mouse exophilin-8 cDNA ( Mizuno et al . , 2011 ) was subcloned into the pcDNA3-MEF , pcDNA3-FLAG , and pENTR3C-MEF vector ( Ichimura et al . , 2005 ) . Site-directed mutagenesis of exophilin-8 was performed using the following primers: 5’-CTGCAGGCGAAGGCCGCTAAGAACGCTGCAGTG-3’ and 5’-CACTGCAGCGTTCTTAGCGGCCTTCGCCTGCAG-3’ for PA ( MBD ) and 5’-CAGAGGGCGAAACTGGCTGCCCCTGCTGTGAAA-3’ and 5’-TTTCACAGCAGGGGCAGCCAGTTTCGCCCTCTG-3’ for PA ( ABD ) . The resulting PA ( MBD ) and PA ( ABD ) mutants carry substitutions of alanine for arginine and proline at amino acid positions 474 , 477 , 480 , and 798 , 801 , 804 , respectively . The cDNA fragment encoding 1–167 amino acids of exophilin-8 was subcloned into the pGEX4T-1 ( GE Healthcare , Little Chalfont , UK ) and pMAL-CRI ( NEB ) to generate glutathione S-transferase ( GST ) -fused and maltose-binding protein ( MBP ) -fused proteins , respectively . Mouse RIM-BP2 and myosin-VIIa cDNAs were amplified from the cDNA of MIN6 cells . Mouse wild-type RIM-BP2 resistant to siRNA 11 toward rat RIM-BP2 ( GE Dharmacon; Lafayette , CO ) was generated by site-directed mutagenesis using the following primers: 5’-GCGAATTCATGCGAGAGGCTGCTGAGC-3’ and 5’-AGGAGAACCAAGGCACTGATC-3’ , and 5’-CCACTCGAGCTAAGGGGGCTGGCTTACCC-3’ and 5’-GATCAGTGCCTTGGTTCTCCT-3’ . Deletion mutant of RIM-BP2ΔSH3 was made from the above siRNA-resistant RIM-BP2 using the primers: 5’-GCGAATTCATGCGAGAGGCTGCTGAGC-3’ and 5’-CCACTCGAGCTACTCCTCAGCACCAGGGTC-3’ , and 5’-TCCAAGCAAAGCAGCTCGAATGAGTCGCGGCTGGCT-3’ and 5’-AGCCAGCCGCGACTCATTCGAGCTGCTTTGCTTGGA-3’ . The full-length and deleted cDNAs were subcloned into the pcDNA3-3×HA , EGFP-C2 ( BD Biosciences , Franklin Lakes , NJ ) , pENTR3C-MEF , and pENTR3C-3×HA . They were also subcloned into the pCAG vector with or without an OSF tag ( Morita et al . , 2007; Matsunaga et al . , 2017 ) . Recombinant adenoviruses were prepared as described previously ( Wang et al . , 2013 ) . Rabbit anti-exophilin-8 antibody ( αExo8N ) was raised against GST-fused N-terminal exophilin-8 protein ( 1–167 amino acids ) . The sera were passed through a column containing the same N-terminal exophilin-8 protein fused with MBP . The affinity-purified antibodies were then eluted and concentrated . Rabbit anti-myosin-VIIa antibody ( αMyo7 ) was raised against MBP-fused , mouse myosin-VIIa protein ( 1560–1727 amino acids ) . Mouse anti-myc 9E10 monoclonal antibody was purified from the ascites fluid of a hybridoma-injected mouse . Guinea pig anti-porcine insulin serum was a gift from H . Kobayashi ( Gunma University ) . The following commercially purchased antibodies were also used: rabbit polyclonal antibodies toward FLAG ( F7425 , Sigma-Aldrich , St . Louis , MO; RRID:AB_439687 ) , myosin-Va ( LF-18 , Sigma-Aldrich; RRID:AB_260545 ) , RIM-BP2 ( 15716–1-AP , Proteintech , Rosemont , IL ) , Munc13-1 ( 55053–1-AP , Proteintech ) , HA ( 561 , MBL , Nagoya , Japan; RRID:AB_591839 ) , green fluorescent protein ( GFP; 598; MBL; RRID:AB_2313843 ) , Rab27a/b ( 18975; IBL , Fujioka , Japan; RRID:AB_494635 ) , PKA ( ab26322; Abcam , Cambridge , United Kingdom ) , myosin-VIIa ( ab3481; Abcam; RRID:AB_303841 ) , Cav1 . 3 ( ACC-005; Alomone Labs , Jerusalem , Israel; RRID:AB_2039775 ) , RIM1/2 ( 140203; Synaptic Systems , Goettingen , Germany ) , and VAMP2 ( 627724; Calbiochem , San Diego , CA; RRID:AB_212589 ) ; and mouse monoclonal antibodies toward glyceraldehyde-3-phosphate dehydrogenase ( GAPDH; 3H12; MBL ) , α-tubulin ( T5168; Sigma-Aldrich; RRID:AB_477579 ) , β-actin ( A5316; Sigma-Aldrich; RRID:AB_476743 ) , syntaxin1 ( HPC-1; Sigma-Aldrich; RRID:AB_592786 ) , Na+-K+ ATPase ( C464 . 6; Upstate Biotechnology , Lake Placid , NY; RRID:AB_309699 ) , Sec6 ( ADI-VAM-SV021; Assay Designs , Ann Arbor , MI ) , and SNAP25 ( 610366; BD Biosciences; RRID:AB_397752 ) Tissue extract preparation , immunoblotting , and immunoprecipitation were performed as described previously ( Matsunaga et al . , 2017 ) . Each image resulting from immunoblotting is representative of at least three independent experiments . All cells were cultured in a humidified incubator with 95% air and 5% CO2 at 37°C . MIN6 cells ( originally provided from Dr . Jun-ichi Miyazaki , Osaka University; Miyazaki et al . , 1990 ) were cultured in Dulbecco’s modified Eagle’s medium ( DMEM ) containing 15% fetal bovine serum ( FBS ) supplemented with 1 mM L-glutamine and 50 μM 2-mercapthoethanol . INS-1 832/13 cells ( originally provided from Dr . Christopher Newgard , Duke University; Hohmeier et al . , 2000 ) were cultured in RPMI1640 containing 10% FBS supplemented with 1 mM L-glutamine , 1 mM HEPES , 1 mM sodium pyruvate , and 50 μM 2-mercapthoethanol . HEK293A cells ( Invitrogen , Carlsbad , CA ) were cultured in DMEM containing 10% FBS supplemented with 1 mM L-glutamine . INS1 832/13 ( RRID: CVCL_7226 ) , MIN6 ( RRID: CVCL_0431 ) , and HEK293A ( RRID: CVCL_6910 ) cells were listed by NCBI Bio sample ( BioSample: SAMEA4104055 , SAMEA4168040 , and SAMEA4146837 , respectively ) . Exophilin-8-null β-cell lines were established from exophilin-8-null mice , by a method similar to that by which granuphilin-null β-cell lines were previously established ( Mizuno et al . , 2016 ) . These cell lines have tested for mycoplasma contamination by 4' , 6-diamidino-2-phenylindole staining . Plasmid transfections and adenovirus infections were performed as described previously ( Wang et al . , 2013 ) . The purification procedure was similar to that reported previously ( Ichimura et al . , 2005 ) , with minor modifications . Briefly , the extracts of MIN6 cells expressing MEF-exophilin-8 were subjected to immunoprecipitation with anti-myc antibody , cleavage by TEV protease , immunoprecipitation with an anti-FLAG antibody , and FLAG peptide-dependent elution . The final eluate was separated by SDS-PAGE and visualized by Oriole fluorescent gel staining ( BioRad , Hercules , CA ) . Specific bands were excised and digested in the gel with trypsin , and the resulting peptide mixtures were analyzed by nanoflow LC-MS/MS at Gunma University . All MS/MS spectra were searched against the non-redundant RefSeq protein sequence database at the National Center for Biotechnology Information using Mascot software ( Matrix Science , London , UK ) . INS-1 832/13 cells cultured on coverslips were fixed with 3% paraformaldehyde in phosphate buffered saline ( PBS ) for 30 min and permeabilized with 0 . 1% Triton X-100 in PBS for 30 min . Isolated pancreatic islets were cultured on 35-mm glass bottom dishes for 3 days , fixed with 3% paraformaldehyde in PBS for 60 min , and permeabilized with 0 . 2% Triton X-100 in PBS for 60 min . The cells or islets were then treated with 50 mM NH4Cl-PBS for 10 min at room temperature and blocked with PBS containing 1% bovine serum albumin for 15 min . They were incubated with primary antibodies ( diluted at 1:100 or 1:200 ) for 2 hr or overnight , washed three times with PBS , and then incubated with Alexa Fluor 488- or 568-conjugated secondary antibodies ( Invitrogen; diluted at 1:500 ) for 60 min . Finally , INS-1 832/13 cells were mounted using Slow Fade Gold reagent ( Invitrogen ) , whereas the islets were filled with PBS . Both samples were observed using laser scanning confocal microscopes , A1 ( Nikon , Tokyo , Japan ) equipped with a 100× oil immersion objective lens ( 1 . 49 NA ) or FV1000 ( Olympus , Tokyo , Japan ) equipped with 100× oil immersion objective lens ( 1 . 40 NA ) . The images were acquired by NIS elements ( Nikon ) or Fluoview ( Olympus ) , and were adjusted using Adobe Photoshop CS4 software ( Adobe Systems , San Jose , CA ) . Each image resulting from immunofluorescence is representative of at least three independent experiments . On-Target plus SMARTpool siRNA against rat exophilin-8 ( catalog no . 360034 ) , RIM-BP2 ( 266780 ) , RIM1 ( 84556 ) , RIM2 ( 116839 ) , myosin-Va ( 25017 ) , myosin-VIIa ( 266714 ) , and Cav1 . 3 ( 29716 ) , as well as the control On-Target plus non-targeting pool siRNA , were purchased from GE Dharmacon . INS-1 832/13 cells were plated at a density of 2 . 5 × 106 in a 6-well culture plate and were grown for 24 hr . Suspended cells after trypsinization were transfected twice with siRNAs using Lipofectamine RNAiMAX reagent ( Invitrogen ) , according to the manufacturer’s instructions . The second transfection was performed 72 hr later , and the cells were analyzed 36 hr thereafter . Sucrose density gradient centrifugation analysis was performed as described previously ( Fujita et al . , 2009 ) . Briefly , INS1 832/13 cells were homogenized in homogenization buffer ( 50 mM Tris-HCl pH 7 . 5 , 150 mM NaCl ) containing complete protease inhibitor cocktail ( Roche , Basel , Switzerland ) by repetitively passing through 1-ml syringe with a 27-gauge needle . The homogenate was centrifuged at 10 , 000 × g for 10 min , and the supernatant was centrifuged at 100 , 000 × g for 60 min . Resulting 600 μl of the supernatant ( cytosol fraction ) was loaded on the top of discontinuous sucrose gradients . The gradient was composed of 1 ml each of stepwise concentration of sucrose solutions ( 3 , 6 , 9 , 12 , 15 , 18 , 21 , 24 , 27 , 30% sucrose in homogenization buffer ) . The samples were centrifuged at 100 , 000 × g for 18 hr , and then 600 μl fractions were collected from the top of the gradient . No samples or animals were excluded from the analysis . No statistical methods were used to predetermine sample size and experiments were not randomized . Statistical significance was determined using a two-tailed unpaired t-test . The investigators were not blinded to allocation during experiments and outcome assessment . | The human body contains trillions of cells with hundreds of different jobs that must cooperate with each other . Many cells communicate using hormones and other chemical messengers that they release into the blood or tissues . These messengers are stored in containers called secretory granules , which are held just under the surface of the cell by a web of fibres made from a protein called actin . When a message needs to be sent , the granules fuse with the membrane that surrounds the cell , releasing their contents into the space outside . A protein called exophilin-8 helps granules to fuse with the membrane . This protein attaches to both the granules and actin bundles , but its precise role is not clear . Here , Fan et al . generated mutant mice that cannot make exophilin-8 to find out what happens when this protein is missing . The experiments show that the loss of exophilin-8 prevented granules from building up at the edges of cells and releasing their contents . This was accompanied by a decrease in the amount of insulin – a hormone that regulates blood sugar levels – released by cells in the pancreas . As a result , the mutant mice had higher levels of blood sugar than normal mice . Further experiments revealed that exophillin-8 associates with a group of other proteins that work together to catch the secretory granules and anchor them to the actin bundles near to the inner edge of the cell . If secretory granules do not fuse with the membrane properly , the chemical messages they contain are not transmitted , which can lead to disease . Since the loss of exophilin-8 affected the release of insulin from the pancreas it is possible that further work could open new avenues for diabetes research . A future challenge is to examine whether exophillin-8 also plays a similar role in the fusion of secretory granules in other cells such as nerve and immune cells , which also release a number of important chemicals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2017 | Exophilin-8 assembles secretory granules for exocytosis in the actin cortex via interaction with RIM-BP2 and myosin-VIIa |
The subthalamic nucleus ( STN ) is a small almond-shaped subcortical structure classically known for its role in motor inhibition through the indirect pathway within the basal ganglia . Little is known about the role of the STN in mediating cognitive functions in humans . Here , we explore the role of the STN in human subjects making decisions under conditions of uncertainty using single-neuron recordings and intermittent deep brain stimulation ( DBS ) during a financial decision-making task . Intraoperative single-neuronal data from the STN reveals that on high-uncertainty trials , spiking activity encodes the upcoming decision within a brief ( 500 ms ) temporal window during the choice period , prior to the manifestation of the choice . Application of intermittent DBS selectively prior to the choice period alters decisions and biases subject behavior towards conservative wagers .
Deep brain stimulation ( DBS ) is a remarkable therapy that has revolutionized the potential for treating neurological and neuropsychiatric illness by directly modifying neural function , though the underlying mechanism of action remains unknown . In its simplest form , DBS can be thought of as a pacemaker for the brain . Current DBS devices deliver continuous electrical stimulation to a targeted brain region to modify or reset abberant neural activity or synchrony ( Herrington et al . , 2016a ) . A major area of current research is in identifying more refined methods of stimulation delivery by exploring the timing of stimulation delivery , multi-site stimulation , and real-time sensing and stimulation . The subthalamic nucleus ( STN ) is a small almond-shaped nucleus in the basal ganglia classically known for its role in inhibiting motor responses as part of the indirect pathway ( Schmidt and Berke , 2017; Schmidt et al . , 2013 ) . More recently , a growing body of literature has begun to uncover a more nuanced role of the STN in higher-order cognitive processes such as , emotional processing ( Le Jeune et al . , 2009; Le Jeune et al . , 2008; Drapier et al . , 2006; Eitan et al . , 2013 ) , response inhibition ( Frank et al . , 2007; Cavanagh et al . , 2011; Isoda and Hikosaka , 2008 ) , and even psychiatric illness ( Mallet et al . , 2008 ) . The STN is also an important deep brain stimulation ( DBS ) target for the treatment of movements disorders such as Parkinson’s Disease ( PD ) . DBS surgery provides one of only a few opportunities to record neuronal responses in humans subjects engaged in cognitive tasks ( Zénon et al . , 2016; Herz et al . , 2016; Cavanagh et al . , 2011; Zaghloul et al . , 2012 ) . In addition , researchers can leverage implanted DBS electrodes as a neuromodulation tool to study the role of the STN in an extra-operative setting . As such , numerous researchers have utilized this approach to interrogate the function of the STN in conflict ( Frank et al . , 2007; Schroeder et al . , 2002 ) , decision-making ( Seymour et al . , 2016; Seinstra et al . , 2016; Wylie et al . , 2010; Zaehle et al . , 2017; Cavanagh et al . , 2011 ) , and emotional processing ( Le Jeune et al . , 2008; Le Jeune et al . , 2009 ) . Studies of STN stimulation using current generation DBS systems have been limited by the systems’ design to deliver continuous stimulation . This is a critical limitation to using DBS to explore dynamic aspects of cognition and might be an important source of variability on DBS effects reported in the literature . We hypothesized that targeting stimulation to specific temporal windows during the evolution of a cognitive process ( e . g . , decision-making ) , rather than delivering long periods of continuous stimulation , will be critical to understanding the cognitive function and developing new neuromodulation therapies going forward . In this study , we explore the role of the STN in making decisions under conditions of uncertainty . We employed single-neuronal recordings and intermittent electrical stimulation in human subjects while they engaged in a financial decision-making task ( Patel et al . , 2012 ) . We find that the STN is selectively activated during a brief window for high-uncertainty trials from single-neuronal data . To assess the role of the STN in decision-making during this brief temporal window , we built a custom device which allowed us to precisely deliver intermittent stimulation within short temporal windows . We found that a brief , high-frequency stimulation pulse delivered prior to the choice period promoted a reduction in risk-seeking behavior . To the best of our knowledge , this is the first study to apply intermittent DBS in humans actively engaged in a cognitive task and the first to demonstrate a reduction in risk-seeking behavior following STN DBS .
We collected behavioral and neurophysiological data from six subjects ( five men , one woman; 63 . 2±6 . 8 years old; mean ± Sc . D . ; Table 1 ) that underwent DBS surgery for PD . On average subjects performed 1 . 83 sessions of the gambling task with an average of 105 . 2 trials per session . The gambling task was designed such that on any given trial a positive outcome was probabilistically weighted by the subject’s card . As such , we expected an engaged participant to display longer reaction times for trials in which the outcome was unpredictable; whereas on predictable trials we expected behavior to converge to an optimal strategy resulting in shorter reaction times . We found such a trend ( F4 , 10=10 . 2 , p=4 . 0×10−4; ANOVA; Figure 1b ) . Specifically , 6-card trials had the highest average reaction time ( 1 . 16± . 19s and 1 . 33± . 61s; mean ± Sc . D . , respectively ) consistent with the unpredictable nature of the outcome ( i . e . an equal chance of winning and losing ) . Similarly , reaction times for the most predictabletrials were amongst the lowest . First , we examined behavior on trials in which subjects were dealt a 6-card . A behavioral deviation from a 50/50 betting strategy on these trials would indicate a risk-seeking or risk-averse bias . Overall , we found that subjects had a risk-averse bias placing a high wager only 24% of the time ( χ1 , 112=42 . 24 , p=1 . 44×10−5; Figure 1c ) . In this study , all intraoperative subjects were off dopaminergic medications at least 12 hr prior to surgery . The low-dopamine state may have contributed to subject’s risk-avoidant behavior ( St Onge et al . , 2011; Claassen et al . , 2011 ) . Current models suggest that STN activity inhibits responses during cognitively demanding situations ( Frank , 2006; Frank et al . , 2007 ) . This inhibition may serve to allow for additional time to process internal and environmental information before ultimately arriving at and executing a decision . To explore this hypothesis in our study we leveraged the intrinsic symmetry of the behavioral paradigm , and divided trials into low and high cognitive demand . The 10- and 2-cards are extreme situations in which the player is probabilistically likely or unlikely to win , respectively — we call these low-uncertainty trials . Conversely , on the 6-card trials the player has an equal probability of winning and losing and there is no optimal strategy — we call these high-uncertainty trials . We examined single-neuronal data from the STN using standard stereotactic and intraoperative microelectrode mapping procedures . We collected 27 well-isolated neurons with an average of 3 . 1±1 . 1 ( mean ± Sc . D . ) neurons per subject ( Figure 2—figure supplement 1 ) . All analyses were performed on normalized and pooled spiking data . We apriori selected a 500 ms window during the choice period based on previous findings ( Patel et al . , 2012 ) and explored the relationship between STN activity and the level of uncertainty on a given trial . To do this , we applied a regression model predicting z-scored spike counts as a function of the card value and wager . Interestingly , we found an interaction effect between card value and wager ( F9 , 1450=2 . 55 , p=0 . 02; ANOVA ) but no main effects for card value ( F9 , 1450=1 . 69 , p=0 . 14 ) or wager ( F9 , 1450=2 . 68 , p=0 . 10 ) . Further exploration revealed a significant effect on high-uncertainty trials ( t1450=−2 . 38 , p=0 . 01; xtbfFig . 2a; Figure 2—figure supplements 2 , 3 and 4 ) which was not present on low-uncertainty trials ( t1450=−1 . 02 , p=0 . 30; t1450=−0 . 16 , p=0 . 86; 2- and 10-cards respectively; Figure 2b ) . Interestingly , we found trending activity for the 4- and 8-trials ( t1450=−1 . 730 , p=0 . 08; t1450=−1 . 78 , p=0 . 07 ) which contain an intermediate degree of uncertainty . No other stimulus epoch correlated with subject behavior ( Table 2 ) . This signal is unlikely to represent an overt finger movement because our task design balances the presentation of the $five and $20 wagers equally to the left- and right-hand side of the screen . Also , we found no difference in activity between wagers centered on the button press ( F9 , 1450=0 . 24 , p=0 . 98; Figure 2a , b ) suggesting this signal was not movement-related . In addition , there was no relationship between the wager ( t23=0 . 09 , p=0 . 92 ) or the outcome ( t23=0 . 71 , p=0 . 48 ) on the previous trial . Lastly , we found that z-scored reaction times on high-uncertainty trials were longer when subjects placed a high vs . low wager ( t6=−3 . 28 , p=0 . 01; Figure 2c ) . There was no difference in reaction times on low-uncertainty trials for high vs . low wager ( t9=1 . 17 , p=0 . 27; Figure 2d ) . We have shown that STN activity within a brief temporal window during the choice period predicts the upcoming wager selectively for high but not low-uncertainty trials . Interestingly , previous human neurophysiology studies have described similar conflict signals arising earlier during the stimulus presentation epoch ( Zaghloul et al . , 2012; Sheth et al . , 2012 ) . To explore this discrepancy , we used intermittent DBS to test whether altering STN activity during this finite time window would alter the subject’s ultimate decision using intermittent DBS . We recruited 13 subjects ( 12 men , one woman; 62 . 6±7 . 4 years old; mean ± Sc . D . ; Table 3 ) who had previously undergone STN DBS surgery for PD . All subjects had completed surgery at least 6 months prior tithe study . Through patients’ implanted DBS electrodes we applied intermittent electrical stimulation to the STN while subjects were engaged in the same gambling task . Specifically , we applied one of three different stimulation categories randomly on 6-card trials , either giving: no stimulation , 1 s of stimulation during the fixation epoch , or 1 s of stimulation prior to the choice period . To control for observational effects of turning on/off the stimulator ( e . g . feeling a sensation when the stimulator turns on ) , we systematically lowered the voltage setting—blinded to the subject—to a sub-threshold level prior to each experimental session . In addition , we characterized the latency from the trigger to current delivery and found it to be 174 ±0 . 002 ms ( n=26; mean ± Sc . D . ; Figure 3—figure supplement 1 ) . All other settings ( e . g . electrode contacts , frequency , and pulse-width ) were unaltered from therapeutic levels and were returned to normal following the study . On average subjects performed 2 sessions of the gambling task with an average of 108 trials per session . Similar to the intraoperative experiment , we found that subjects demonstrated understanding of the underlying structure of the task ( F4 , 26=5 . 83 , p=0 . 0002; ANOVA; Figure 3a ) . The fastest reaction times were observed for low-uncertainty trials ( 1 . 19± . 76 seconds , 1 . 11± . 71 seconds; mean ± Sc . D . ; 2- and 10-cards respectively ) ; and on average , the high-uncertainty trials wreathe slowest ( 1 . 46±1 . 16 seconds; mean ± Sc . D . ) . Unlike during the intraoperative sessions , subjects were on their clinical regimen of dopamine replacement therapy during this experiment . We did not observe the same risk-averse behavior on 6-card trials ( χ1 , 272=34 . 13 , p=0 . 13; Figure 3b ) . Guided by our neurophysiological findings , we expected that modulation of intrinsic decision signaling prior to the choice period would selectively bias subject behavior . As such , we expected no difference when stimulation was delivered during the fixation period compared to when it was omitted . The data confirmed this hypothesis ( F2 , 28=2 . 93 , p=0 . 05; ANOVA ) . In contrast , when stimulation was delivered prior to the choice period , we found that on average subjects had a strong risk-averse bias and placed a high wager only 33 . 0 ±4 . 83% ( mean ± s . e . m . ) of the time , on average an absolute 15% less than the no stimulation group ( t28=2 . 77 , p=0 . 009; Figure 3c ) . Importantly , there was no difference between the omitted and fixation stimulation conditions ( t28=0 . 14 , p=0 . 88 ) , on which subjects placed a high wager on average of 48 . 3 ±5 . 92% and 49 . 2 ± 5 . 6% of the time ( mean ± s . e . m . ) , respectively . To further explore the effects of intermittent stimulation on decision-making we more closely examined the effects within individual subjects . To do so , we plotted each subject’s average high wager percentage when stimulation was omitted and delivered at the choice period ( Figure 3d ) . Overall , we found subjects spanned a large range in baseline tendency for placing high wagers , ranging from 11% to 100% . We found that 7 out of 11 subjects displayed a reduction in risk-seeking behavior ( Table 4 ) . Of the seven subjects the average magnitude of change was 12 . 8% ( range: 1–18% ) . Interestingly , we observed that the magnitude of the reduction in risk-seeking behavior correlated with their initial starting point ( t7=2 . 46 , p=0 . 05; Figure 3e ) . For the three subjects that showed an increase in risk-seeking behavior , the average magnitude of change was 12 . 1% ( range: 2–22% ) . The same correlation did not appear to exist in the this group ( t2=1 . 13 , p=0 . 46 ) , though the sample size is limited . One subject experienced no change in either direction from the stimulation . Lastly , we explored whether stimulation had an effect on subject’s reaction time performance . We found that there was no overall main effect of stimulation epoch ( F2 , 577=1 . 37 , p=0 . 25; ANOVA ) or wager ( F1 , 577=0 . 31 , p=0 . 57; ANOVA ) on reaction time . However , we did find an interaction effect between stimulation epoch and wager ( F2 , 577=4 . 15 , p=0 . 01; Figure 3f ) . Specifically , during the choice period stimulation condition , reaction times were faster when subjects placed a high wager compared with a low wager ( t22=3 . 72 , p=0 . 001 ) , supporting previous findings ( Frank et al . , 2007 ) . No similar differences were observed for the omitted and fixation stimulation conditions .
We used a multi-modal approach consisting of single-neuronal recordings and intermittent stimulation to characterize the neurophysiological role of the STN in decision-making under uncertainty . To do so , we used a financial decision-making task designed to interrogate risk-taking behavior . Using this task , we categorized trials into high and low-uncertainty . We defined high-uncertainty as trials in which the probability of a positive and negative outcome are equal . As a result there was no optimal behavioral strategy . Conversely , low-uncertainty trials were cases in which the outcome was heavily biased towards or against a positive outcome . On these trials , subject behavior was reliably stereotyped towards the most appropriate wager to maximize gains or minimize losses . We found that on high-uncertainty trials STN neural activity encoded the upcoming decision in a discrete 500 ms temporal window immediately before the choice period . In a recent functional imaging study , Fleming et al . found a bilateral increase in BOLD response localized to the STN selectively for high-uncertainty trials where subjects responded against a status-quo bias . Although their study uses perceptual decisions , we demonstrate that the same underlying mechanism may extend to value-based decisions made under conditions of uncertainty . Other studies have observed similar neural responses in the STN following conflict related encoding ( Cavanagh et al . , 2011; Zaghloul et al . , 2012 ) and control signal encoding ( Isoda and Hikosaka , 2008; Wiecki and Frank , 2013 ) . Unfortunately , the task design in the present study does not let us dissect the influence of conflict , control , and uncertainty on the observed neural responses reported here . This remains an open question within the STN literature body . It is worth noting the possibility that the observed STN neural response in this experiment is a combination of conflict and control . More specifically , a departure from a prepotent response ( i . e . placing a high wager ) induces STN activity and allows for the recruitment of control centers to mediate a new decision . This would be supported by computational models ( Wiecki and Frank , 2013 ) and experimental data ( Coulthard et al . , 2012 ) but require further investigation to tease apart . Furthermore , Cavanaugh et al . have previously shown that increases in local-field potential oscillations in the medial prefrontal cortex and STN correlate with trial-by-trial decision conflict and that continuous electrical stimulation through implanted DBS electrodes can prevent adjustments in decision thresholds ultimately resulting in rapid or impulsive decision-making ( Cavanagh et al . , 2011 ) . In contrast , in our data the application of intermittent DBS prior to the choice period resulted in an increasein risk-averse decisions and in reaction times for those decision . One potential explanation for these seemingly conflicting findings is that the effects of stimulation may vary depending on the duration of stimulation . It has previously been suggested that short bursts of high-frequency STN stimulation serve to increase local firing rates which are subsequently silenced with prolonged stimulation ( Lee et al . , 2009 ) . Continuous , high-frequency stimulation has also been proposed to act as an informational lesion , essentially overwriting the normal time-varying activity of the target ( Herrington et al . , 2016b ) . Our finding also appears to correspond to the observed neurophysiological data from this study , where a slight increase in overall STN activity during the choice period correlates with placing a low wager . An interesting limitation of the stimulation study is that stimulation was only delivered on the high-uncertainty trials , limiting our ability to understand the constraints of its effect on modifying behavior on medium- or low-uncertainty trials . We would hypothesize the effect would be limited or not present on low-uncertainty trials given that no differential encoding was observed , however this remains to be studied . Interestingly , our findings differ from other human neurophysiology studies in which conflict activity was observed during the stimulus presentation , as opposed to the choice period , in the dorsal anterior cingulate ( Sheth et al . , 2012 ) and the STN ( Zaghloul et al . , 2012 ) . To further explore the temporal dynamics of the observed signal , we performed a second experiment in which we applied intermittent STN stimulation through implanted DBS electrodes selectively during high-uncertainty trials . Stimulation was delivered either during the fixation period , choice period , or it was omitted . This technique is uniquely different than previous studies using DBS as a method to interrogate neural circuits because we implemented a system for rapidly turning on and off the implanted device , permitting us to time-lock delivery to specific task-epochs . This approach may further reduce confounding effects of long-term stimulation , such as carry-over effects . As a result , we found that intermittent stimulation prior to the choice period—the same interval during which we observed the neurophysiological decision signal from the first experiment—selectively altered subject behavior . No differences were observed in subject behavior when stimulation was omitted or delivered during the fixation period . We found that stimulation prior to the choice period interrupted subjects’ ability to appropriately slow responses when betting against their bias ( i . e . when they placed a high bet ) , resulting in a shortened reaction time , consistent with previous work ( Frank et al . , 2007; Cavanagh et al . , 2011 ) . Although we attempted to reduce confounding effects by demonstrating both physiological and stimulation evidence to support our claim , our experimental design has several fundamental limitations . In the first experiment , we perform intraoperative recordings in patients undergoing a neurosurgical procedure . Naturally , there are several limitations for performing studies in the operating room , such as the length of each experimental session . For this reason , the total number of trials and neurons we are able to record can often be limited . In this study , we focus our neurophysiological findings to population responses . Despite this limitation , however , we find the reported effects to be consistent across the population . In addition , we compensate for this limitation by developing a novel stimulation method to carefully test the relationship between our neurophysiological findings and subject behavior . Furthermore , the subjects in this study all suffer from advanced PD , a disease known to affect natural reward processing . For obvious reasons , these experiments are constrained to populations requiring neurosurgical treatment , and direct comparisons to a healthy population are limited to behavioral measures . In conclusion , we provide functional imaging and neurophysiological evidence in human subjects demonstrating the critical role of the STN in encoding decisions under conditions of uncertainty . Moreover , we demonstrate that electrical stimulation of the STN within a finite temporal window can selectively bias subject behavior towards more risk-averse decisions . Together , this provides evidence for the role of precision neuromodulation approaches and closed-loop deep brain stimulation for the advancement of neurological and neuropsychiatric therapies .
We recruited six subjects undergoing STN DBS for the treatment of Parkinson’s disease to participate in the intraoperative neurophysiology study . Each individual was evaluated and considered for surgery by a multidisciplinary team of neurologists , neurosurgeons , and psychiatrists . Once approved and scheduled for surgery an independent member of the research team approached each patient to describe the possibility of study inclusion . At that time risks and benefits were clearly addressed to each subject . All study subjects enrolled voluntarily and provided informed consent under guidelines approved by the Massachusetts General Hospital Institutional Review Board . Subjects were free to withdraw from the study at any time without consequence to operative approach or clinical care . This study was approved by the Massachusetts General Hospital Institutional Review Board ( protocol number 2001P000877 ) . For a more detailed description on performing cognitive studies with microelectrode recording during DBS , see ( Patel et al . , 2013 ) . A computer monitor was fixed to an adjustable arm and mounted to the operating bed and positioned comfortably within the viewing distance of the patient . A button box was similarly mounted to the operating bed and placed comfortably under the patient’s right hand . Subjects were in a comfortable reclined position . The behavioral task was presented using custom written software in Matlab ( Math works , Natick , MA ) , Monkey logic ( www . monkeylogic . org ) ( Asaad and Eskandar , 2008a; Asaad and Eskandar , 2008b; Asaad et al . , 2013 ) . The task is analogous to the classic card game , War . On each trial , the subject and computer are each dealt a card and the player with the higher card wins . To simplify the game the deck is limited to five cards: even cards from 2 through 10 from one suit . The rules were carefully explained to each subject prior to the study . Each trial requires the subject to evaluate his/her card , determine its value , and place a $five or $20 wager with the goal of maximizing profits . Thus , when the subject is dealt a 10-card , the optimal choice is to place a $20 wager as the outcome is likely positive or at worst a draw . Conversely , the optimal choice for a 2-card is to place a $five wager since the outcome is likely negative or at best a draw . There is no optimal strategy for the 6-card—the outcome is probabilistically equal . Each trial began with a fixation point presented at the center of the screen for 350 ms to indicate the start of trial ( Figure 1a ) . Next , the subject’s card and the back of the opponent’s card were displayed for 1000 ms . Two red circles then appeared , indicating the mapping of each button ( left and right buttons ) to its respective wagers ( $five and $20 ) . The button map was presented randomly such that the $five and $20 wagers are assigned to the left and right buttons equally . The presentation of the button map also serves as the choice period , indicating when to initiate a wager . The time it took the subject to press a button was considered the reaction time with a maximum of 5 s . Following the wager , there was a randomized delay period of 250–500 ms , which was immediately followed by the presentation of the subject’s and computer’s card for 1000–1250 ms . Lastly , feedback was given for 1000 ms by displaying an image of a $five or $20 bill with text indicating the outcome . In the case of a draw , only text is displayed . Subjects were monetarily rewarded following their participation in the study . For a detailed description , please see ( Patel et al . , 2013 ) . Intraoperative microelectrode recordings were performed using three Para-sagittal tungsten microelectrodes ( Figure 1b ) . The electrodes were advanced using a motorized Alpha Omega ( Alpha-Omega Engineering , Nazareth , Israel ) Microdrive . Intraoperative motor testing was performed at <1 mm increments throughout the dorsolateral-ventromedial axis of the STN to characterize the motor and non-motor compartments . Recordings were band-pass filtered between 300 Hz and 6 . 5 kHz by an Alpha Omega acquisition system . Data was recorded at 20 kHz by a PowerLinc 1401 acquisition system ( Cambridge Electronic Design , Cambridge , England ) and stored for post-hoc analysis . Offline , the neurophysiology data was sorted into individual neuronal records using a template clustering method ( Offline Sorter , Plexon , Houston , TX ) . Data from each electrode was sorted separately . All analyses were performed using a combination of iPython and R . Because of inter-subject variability in baseline motor performance , we explored reaction time differences by first z-scoring data using each session’s mean and variance reaction time . Normalized subject data and allowed for equal comparisons for group level analyses . We then applied either a one-way or two-way ANOVA on the z-scored reaction time data to assess statistical differences . Post-hoc analyses were performed using two-tailed t-tests . To visualize neural activity , the instantaneous firing rate was approximated by convolving a Gaussian kernel ( sigma = 150 ms ) with 1 ms binned spike trains . Because of the limitations in the number of trials recorded in each experimental session , statistical analyses at the individual cell level were rarely significant , and instead all analyses were performed at the population level . Statistical differences between population responses were assessed using two-tailed t-tests during pre-defined 500 ms windows: 500–1000 during the choice period and −250–250 centered on the button press based on a previous study ( Patel et al . , 2012 ) . To explore the relationship between neural activity and the decision , we applied a linear regression model of the form: Z=β0+βcC+βwW+βcwC×W , where Z is a vector of z-scored spike counts ( relative to each neuron ) in a 500 ms window , C is the card value , W is the wager , and C×W the interaction between the two terms . Both C and W are categorical variables and represented with dummy variables in the regression model . Coefficients were estimated through a least-squares approach . Thirteen study participants were recruited from STN DBS patients identified by their movement disorders neurologist to participate in the intermittent stimulation study . A study staff member contacted potential study participants by telephone to introduce the study and invite the patient to participate . On the day of the study , after obtaining written informed consent , the patient’s deep brain stimulator was turned off . Subsequently the stimulation voltage was lowered in small increments with the stimulator being turned on and off in a blinded fashion until a voltage threshold was reached at which the patient was unable to detect the stimulation . The stimulator controller was secured over the patient’s pulse generator , and after approximately 15 min with the stimulator off , the patient began playing the task . The task was conducted during the day in a quiet room . Patients were permitted to take short breaks as needed during the task . We used three different stimulation conditions on 6-card trials: 1 s of stimulation at the fixation epoch , 1 s of stimulation at the choice period , or no stimulation was delivered . This design allowed each subject to act as his own control , helping to account for variance due to differing disease , medication , and electrode location factors between patients and also allowed us to control for general versus time specific effects of stimulation . This study was approved by the Massachusetts General Hospital Institutional Review Board ( protocol number 2007P001806 ) . | Deep brain stimulation , or DBS for short , is used to treat movement disorders like Parkinson’s disease in patients who are responding inadequately to medications . It requires implanting an electrode into the brain and using electrical stimulation aimed at a specific cluster of brain cells to reduce unwanted symptoms . DBS helps to normalize abnormal brain activity similar to a pacemaker resetting an abnormal heart rhythm . Scientists are currently studying whether DBS might also help people with obsessive-compulsive disorder , depression , Alzheimer’s disease or other disorders that affect thinking . To alter human behavior and treat disorders that affect thinking , DBS will have to be delivered at precise time points as the brain processes information . One potential target is for DBS in both movement and thinking disorders is the subthalamic nucleus . This is a small almond-shaped cluster of brain cells that helps people stop movements . Recent studies suggest it also may play a role in processing emotions , controlling inappropriate responses , and psychiatric illnesses . Now , Patel et al . show that the subthalamic nucleus helps people decide what to do in the face of uncertainty and that targeting this brain structure with DBS can shift a person’s decision-making . In the experiments , patients with Parkinson’s disease who were awake and undergoing surgery to implant the DBS electrodes also played a computerized gambling game . Patel et al . recorded the electrical activity in the brain cells of the patient’s subthalamic nucleus during the game . The experiments showed that when patients were faced with a decision with 50/50 odds , the pattern of electrical activity in the cells of their subthalamic nucleus reveals their choice about 500 milliseconds before they act on it . After their surgeries , patients engaged in the same gambling game . This time , Patel et al . specifically targeted the decision-related activity in their subthalamic nucleus with DBS . This caused the patients to make fewer risky decisions in the game . The experiments show DBS can change decision-making behavior in humans . Newer DBS technology may be even more effective at treating brain disorders and cause fewer side effects . Further study into how the brain processes information will also help scientists to better target DBS and possibly treat a broader range of diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Intermittent subthalamic nucleus deep brain stimulation induces risk-aversive behavior in human subjects |
Exosomes are extracellular vesicles that are released when endosomes fuse with the plasma membrane . They have been implicated in various functions in both health and disease , including intercellular communication , antigen presentation , prion transmission , and tumour cell metastasis . Here we show that inactivating the vacuolar ATPase in HeLa cells causes a dramatic increase in the production of exosomes , which display endocytosed tracers , cholesterol , and CD63 . The exosomes remain clustered on the cell surface , similar to retroviruses , which are attached to the plasma membrane by tetherin . To determine whether tetherin also attaches exosomes , we knocked it out and found a 4-fold reduction in plasma membrane-associated exosomes , with a concomitant increase in exosomes discharged into the medium . This phenotype could be rescued by wild-type tetherin but not tetherin lacking its GPI anchor . We propose that tetherin may play a key role in exosome fate , determining whether they participate in long-range or short-range interactions .
Exosomes are extracellular vesicles that have been implicated in a wide range of functions , including intercellular communication , tumour cell migration , RNA shuttling , and antigen presentation . By definition , exosomes are derived from multivesicular endosomes or multivesicular bodies ( MVBs ) , which contain intralumenal vesicles ( ILVs ) . When the MVBs fuse with the plasma membrane , the ILVs are discharged , and the resulting extracellular vesicles are called exosomes . However , there are other types of extracellular vesicles , such as those that are produced by shedding from the plasma membrane , and at present there is no standard way of specifically purifying exosomes . Thus , some of the functions that have been attributed to exosomes may need to be reassessed , because of the possibility of contamination with other types of vesicles ( Raposo and Stoorvogel , 2013 ) . In a recent screen for regulators of clathrin-mediated endocytosis in HeLa cells , we observed that knocking down or inactivating the vacuolar ATPase ( V-ATPase ) caused the cells to produce clusters of extracellular vesicles ( Kozik et al . , 2013 ) . These vesicles had the characteristic appearance of exosomes , suggesting that when endosomes are unable to acidify , they have an increased tendency to fuse with the plasma membrane ( see Figure 6 in Kozik et al . , 2013 ) , a phenomenon also reported by others ( Alvarez-Erviti et al . , 2011; Danzer et al . , 2012 ) . Knocking down or inhibiting the V-ATPase also caused a block in clathrin-mediated endocytosis , and we proposed that this block was due to a redistribution of cholesterol from the plasma membrane to an endosomal compartment . Our hypothesis was supported by the finding that we could partially rescue the phenotype by adding exogenous cholesterol to the cells . One question raised by our study was why , if cholesterol-rich non-acidified endosomes fuse with the plasma membrane , does the plasma membrane not regain its cholesterol ? Indeed , when we measured cell surface-associated cholesterol by light microscopy , using filipin as a cholesterol probe , the loss upon V-ATPase knockdown or treatment with the V-ATPase inhibitor Bafilomycin A1 ( BafA1 ) was only partial ( ~50% ) . In contrast , others have shown that treating cells with methyl-β-cyclodextrin , which has a similar effect on clathrin-mediated endocytosis , removes nearly all of the plasma membrane cholesterol ( Rodal et al . , 1999 ) . We initiated the present study to try to answer some of the questions posed by our previous study . We started by quantifying the amount of plasma membrane cholesterol more precisely by developing a method for localising cholesterol at the electron microscope level . Next , we characterised the extracellular vesicles that are produced when V-ATPase is inactivated , by labeling for exosomal markers . Finally , we investigated why the vesicles remain aggregated and associated with the plasma membrane instead of diffusing away .
In our previous study , we concluded that in the absence of V-ATPase activity , cholesterol accumulates in endosomal compartments , based on immunofluorescence double labelling with the cholesterol probe , filipin , and various endosomal markers ( Kozik et al . , 2013 ) . To visualise these compartments at the ultrastructural level , we used correlative light and electron microscopy ( CLEM ) . BafA1-treated HeLa cells ( Figure 1—figure supplement 1A ) were stained with filipin , imaged by light microscopy , and then prepared for electron microscopy . The structures that stained most intensely with filipin were found to correspond to MVBs , packed full of ILVs ( Figure 1A ) . 10 . 7554/eLife . 17180 . 003Figure 1 . BafA1 treatment causes cholesterol to accumulate in intralumenal vesicles of multivesicular bodies and to be lost from the plasma membrane . ( A ) HeLa cells were treated with BafA1 ( 100 nM , 16 hr ) , then fixed , stained with the cholesterol probe filipin , and prepared for correlative light and electron microscopy ( CLEM ) . Scale bars: 10 μm ( upper ) and 1 μm ( lower ) . ( B ) BafA1-treated cells ( 100 nM , 16 hr ) were stained with filipin , then permeabilised and stained with TNM-BF . Scale bar: 20 μm . ( C ) Ultrathin cryosections of mock-treated and BafA1-treated HeLa cells ( 100 nM , 16 hr ) were labelled with TNM-BF and stained with rabbit anti-BODIPY followed by 10 nm protein A-gold . There is labelling both in endosomes ( upper panels ) and at the cell surface ( lower panels ) . Scale bars: 200 nm . ( D ) Intact mock-treated , BafA1-treated ( 100 nM , 16 hr ) , or MβCD-treated ( 10 mM , 30 min ) HeLa cells were labelled with TNM-BF followed by anti-BODIPY and protein A-gold , revealing surface cholesterol localisation . Scale bar: 500 nm . ( E ) Quantification of TNM/BF/anti-BODIPY gold labeling density for DMSO-treated , BafA1-treated ( 100 nM , 16 hr ) , or MβCD-treated ( 10 mM , 30 min ) cells . Graphs show mean ± S . E . M for at least 100 μm of the plasma membrane , over two independent experiments . See also Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 00310 . 7554/eLife . 17180 . 004Figure 1—figure supplement 1 . Controls for specificity . ( A ) LysoTracker or Magic Red staining of DMSO-treated or BafA1-treated HeLa cells ( 100 nM , 16 hr ) . Lysotracker labels acidic organelles , while Magic Red labels organelles containing active acid hydrolases . The loss of signal upon BafA1 treatment shows that organelle acidification has been abolished . Scale bar: 20 μm ( B ) DMSO-treated HeLa cells were stained with filipin , permeabilised and stained with TNM-BF . Scale bar: 20 μm . ( C ) Ultrathin cryosections of cells treated with MβCD ( 10 mM , 30 min ) , which extracts cholesterol from the plasma membrane , were stained with TNM-BF and labelled with anti-BODIPY followed by 10 nm protein A-gold . Arrows indicate the plasma membrane . Scale bars: 500 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 004 So far , the only published studies showing cholesterol localisation at the electron microscope level have been carried out using a cleaved and a biotinylated form of the toxin perfringolysin O ( Waheed et al . , 2001; Möbius et al . , 2002 , 2003; Kwiatkowska et al . , 2014 ) . The authors of these studies reported that in erythrocytes , lymphoblastoid cells , and platelets , cholesterol was mainly associated with the plasma membrane , MVBs ( especially ILVs ) , and tubular endosomes ( Möbius et al . , 2002 ) . Unfortunately , the reagent they used is no longer available , and no other techniques for EM localisation of cholesterol have been described . Recently , however , theonellamides ( TNMs ) labeled with fluorescent dyes , such as BODIPY , have been shown to be effective tools for visualizing sterols in fixed cells by fluorescence microscopy ( Nishimura et al . , 2013 ) . Because there are BODIPY antibodies available , we reasoned that BODIPY-conjugated TNM ( TNM-BF ) might be a suitable reagent for immuno-gold EM localization of cholesterol . Double labeling fluorescence microscopy with filipin and TNM-BF showed that the two probes have virtually identical patterns in HeLa cells , with labelling particularly concentrated in the juxtanuclear region ( Figure 1B , Figure 1—figure supplement 1B ) . For electron microscopy , we labelled cryosections of control and BafA1-treated cells with TNM-BF , followed by a commercial rabbit anti-BODIPY antibody and protein A coupled to colloidal gold ( Figure 1C ) . In both types of cells , we observed strong labelling of MVBs , with gold particles particularly abundant on the ILVs ( upper panels ) , consistent with previous studies using perfringolysin O ( Möbius et al . , 2002 ) . In control cells , we also saw labelling of the plasma membrane ( Figure 1C , arrowheads ) . However , in the BafA1-treated cells , the plasma membrane was virtually devoid of label , although there was label associated with extracellular vesicles ( Figure 1C , arrows ) . To look specifically at cell surface cholesterol , we performed pre-embedding labelling . Control and BafA1-treated cells were fixed and labelled with TNM-BF followed by anti-BODIPY without permeabilisation . This method showed even more dramatically that cholesterol is lost from the plasma membrane following BafA1 treatment , and also highlighted the strong labelling of extracellular vesicles ( Figure 1D , E ) . As a control for the specificity of labeling , we treated cells with methyl-β-cyclodextrin ( MβCD ) , which extracts cholesterol from the plasma membrane . We found a near-complete loss of surface labeling , although endosomes were still labeled ( Figure 1D , E , Figure 1—figure supplement 1C ) . These results are largely in agreement with our previous study , in which we used filipin as a cholesterol probe for light microscopy . In both cases , we found a ~50% loss of surface labeling in BafA1-treated cells . However , in our previous study , we were unable to distinguish between the plasma membrane and extracellular vesicles associated with the cell surface . The present study shows that there is in fact a 15-fold loss of plasma membrane cholesterol , with a concomitant rise in cholesterol-positive extracellular vesicles ( Figure 1E ) . Are the cholesterol-rich extracellular vesicles that we observe in BafA1-treated cells in fact exosomes , or could they be plasma membrane-derived vesicles ? We addressed this question in several ways . First , exosomes have the same diameter as ILVs , i . e . , 30–100 nm diameter , while other types of extracellular vesicles are much more heterogeneous in size , often up to 1 μm in diameter . The BafA1-induced vesicles are indistinguishable in appearance from ILVs ( Figure 2A , inset left ) . Second , to find out whether the vesicles come from endosomes , we carried out pulse-chase experiments using BSA conjugated to 5 nm colloidal gold as an endocytic tracer . The cells were allowed to endocytose BSA-gold for 10 min , washed , incubated for either 30 min or 4 hr to chase the gold into endosomes or lysosomes respectively , and then treated for 16 hr with BafA1 . Figure 2A shows that gold could be observed in association with extracellular vesicles in the cells that had been chased for 30 min , but not in the cells that had been chased for 4 hr , indicating that the vesicles are endosomal in origin . Third , we labelled non-permeabilised cells with an antibody against CD63 , which has been shown to be enriched on both ILVs and exosomes ( Pols and Klumperman , 2009 ) , and saw strong labelling of the extracellular vesicles in the BafA1-treated cells , but little or no surface labelling in control cells ( Figure 2B ) . Fourth , as shown in Figure 1 , both the extracellular vesicles and the ILVs of BafA1-treated cells are enriched in cholesterol , while cholesterol is largely absent from the plasma membrane . Together , these observations indicate that the extracellular vesicles that accumulate in BafA1-treated cells are indeed exosomes . 10 . 7554/eLife . 17180 . 005Figure 2 . The extracellular vesicles that accumulate in BafA1-treated cells are exosomes . ( A ) HeLa cells were incubated with BSA coupled to 5 nm gold for 10 min before being washed several times with PBS to remove any uninternalised label . The cells were then chased in full medium for either 30 min or 4 hr to load BSA-gold into endosomes or lysosomes respectively , then treated with BafA1 ( 100 nM , 16 hr ) , fixed , and prepared for conventional EM . Gold could be seen associated with extracellular vesicles from the cells chased for 30 min , but not from the cells chased for 4 hr . Insets: monomeric gold could be found within MVBs from cells chased for 30 min ( arrows ) , and aggregated within lysosomes from cells chased for 4 hr ( arrowhead ) . Scale bar: 200 nm . ( B ) DMSO- or BafA1-treated HeLa cells ( 100 nM , 16 hr ) were surface-labelled with an antibody against the CD63 lumenal domain to identify exosomes . Scale bar: 500 nm . ( C ) HeLa cells were treated with DMSO or BafA1 ( 100 nM , 16 hr ) before being fixed and prepared for conventional EM . Exosomes were often associated with clathrin-coated pits ( arrow ) . Scale bar: 500 nm ( D ) The number of exosomes per μm of plasma membrane was quantified . Data shown are means from three independent experiments , ± S . E . M . ***p<0 . 001 ( E ) Culture supernatants were collected from mock-treated and BafA1-treated HeLa cells ( 100 nM , 16 hr ) , and centrifuged first at 10 , 000xg and then at 100 , 000xg . Western blots of the pellets were probed with anti-CD63 . ( F ) Total internal reflection fluorescence microscopy ( TIRF ) was performed on HeLa cells transiently expressing both CLC-mCherry and CD63-GFP , following BafA1 treatment ( 100 nM , 16 hr ) . Representative stills are shown . Scale bar: 10 μm . ( G ) Conventional EM of BafA1-treated cells ( 100 nM , 16 hr ) , with arrowheads indicating proteinacious material on exosomes . Scale bars: 500 nm . See also Figure 2—figure supplement 1 , 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 00510 . 7554/eLife . 17180 . 006Figure 2—figure supplement 1 . Prolonged treatment with BafA1 induces exosome release . HeLa cells were treated with BafA1 ( 100 nM ) for 0 , 4 , 8 or 16 hr before being fixed and prepared for immunofluorescence , without any permeabilisation . Cells were stained with a lumenal anti-CD63 antibody to reveal antigens exposed at the cell surface . Scale bar: 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 00610 . 7554/eLife . 17180 . 007Figure 2—figure supplement 2 . Exosome-enriched preparations probed with antibodies against various extracellular vesicle markers . Culture supernatants were collected from mock-treated and BafA1-treated HeLa cells ( 100 nM , 16 hr ) , and centrifuged first at 10 , 000xg and then at 100 , 000xg . Western blots of the pellets were probed with antibodies against CD63 , Alix , Tsg101 , or CD9 . Bands were quantified using ImageJ and normalized to the 100 , 000xg + BafA1 bands . Representative blots are shown on the left; on the right are the means from three independent experiments , with the error bars showing S . E . M . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 007 When we quantified the number of exosomes at the plasma membrane , we found that they were almost non-existent in control HeLa cells , but were very abundant ( ~5–8 exosomes per μm plasma membrane ) after prolonged treatment with BafA1 ( Figure 2C , D , Figure 2—figure supplement 1 ) . We also used a biochemical approach to investigate exosome release , by collecting culture medium from control and BafA1-treated cells and centrifuging it first at 10 , 000xg ( which enriches for larger particles like plasma membrane-derived vesicles and apoptotic bodies ) and then at 100 , 000xg ( which enriches for exosomes ) . Western blots probed with an antibody against CD63 showed that the marker was barely detectable in the fractions from control cells , but extremely abundant in the 100 , 000xg pellet from BafA1-treated cells ( Figure 2E ) , consistent with our EM observations . We also probed the exosome-enriched preparations for other extracellular vesicle markers , including Alix , Tsg101 , and CD9 ( Figure 2—figure supplement 2 ) , and in most cases , we saw at least a slight effect of BafA1 treatment . However , Western blotting is not the most precise way of quantifying differences in protein concentration , and in future we intend to use mass spectrometry for more accurate and comprehensive analyses . The BafA1-induced exosomes are often in close proximity to the non-constricted clathrin-coated pits that we described in our previous study ( e . g . , see the arrow in Figure 2C ) , and we speculated that there might be a temporal relationship between exosome release and clathrin-coated pit formation , with frequent fusion events followed by a burst of clathrin recruitment . To investigate this relationship further , we cotransfected cells with GFP-tagged CD63 and mCherry-tagged clathrin light chain . Live-cell TIRF imaging of BafA1-treated cells showed that under these conditions , clathrin is in fact very static and fusion events are relatively rare . However , when fusions do occur , the CD63-GFP signal persists rather than diffusing into the medium , and the ventral surface of the cell is decorated with stable CD63-GFP puncta of varying intensities ( Figure 2F ) . The frequency of exosomes in thin sections of BafA1-treated cells , compared with the rarity of CD63-GFP fusion events , indicates that the exosomes are somehow tethered to the plasma membrane , rather than released as diffusible vesicles . We have previously hypothesised that ILVs are held together inside endosomes by an unknown material that can be observed by electron microscopy ( Edgar et al . , 2014 ) . Careful analysis showed that exosomes released from BafA1-treated cells display a similar material , which may crosslink them together ( Figure 2G ) . One candidate for a protein that might attach exosomes both to the plasma membrane and to each other is tetherin , also called Bst2 , CD317 , and HM1 . 24 . Tetherin is an interferon-inducible Type II transmembrane protein with a GPI anchor at its C terminus . It acts to inhibit the spread of certain enveloped viruses , including HIV , by cross-linking the virions and holding them together at the plasma membrane ( Neil et al . , 2008 ) . We speculated that tetherin might act in a similar manner on exosomal vesicles ( Figure 3A ) . 10 . 7554/eLife . 17180 . 008Figure 3 . Tetherin localises to exosomes and facilitates exosome tethering . ( A ) Schematic diagram of tetherin . ( B ) Mock-treated or BafA1-treated HeLa cells ( 100 nM , 16 hr ) were either permeabilised or left intact and stained with an anti-tetherin antibody . Scale bars: 20 μm . ( C ) Mock-treated or BafA1-treated HeLa cells ( 100 nM , 16 hr ) were surface-labelled using an anti-tetherin antibody followed by 10 nm protein A-gold . Scale bar: 500 nm . ( D ) The tetherin gene was knocked out using CRISPR/Cas9 , and the loss of tetherin in a clonal population was confirmed by Western blotting . ( E ) Wild-type or tetherin-knockout HeLa cells were treated with BafA1 ( 100 nM , 16 hr ) and processed for EM to analyse exosome frequency . Scale bar: 500 nm . ( F ) The frequency of exosomes per μm plasma membrane in BafA1-treated ( 100 nM , 16 hr ) wild-type or tetherin-knockout cells was calculated . The mean ± S . E . M are shown from three independent experiments . ***p<0 . 001 . ( G ) Exosome-enriched preparations were generated from the culture supernatants of both wild-type and tetherin-knockout cells , either with or without BafA1-treatment ( 100 nM , 16 hr ) , and Western blots were probed with anti-CD63 . See also Figure 3—figure supplements 1 , 2 , 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 00810 . 7554/eLife . 17180 . 009Figure 3—figure supplement 1 . Tetherin and CD63 colocalise in mock-treated and BafA1-treated cells . HeLa cells were treated with DMSO or BafA1 ( 100 nM , 16 hr ) , then fixed and labeled with antibodies against tetherin ( rabbit polyclonal ) and CD63 ( mouse monoclonal ) , under permeabilised ( A ) or non-permeabilised ( B ) conditions . Scale bars: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 00910 . 7554/eLife . 17180 . 010Figure 3—figure supplement 2 . Quantification of surface labeling . The density of surface gold labeling was calculated using an anti-tetherin antibody following 16 hr of DMSO or BafA1 ( 100 nM ) . Graphs show mean ± S . E . M for at least 100 μm of plasma membrane . Total lengths of plasma membrane measured were mock 102 μm for mock-treated cells , and 120 μm for BafA1-treated cells . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 01010 . 7554/eLife . 17180 . 011Figure 3—figure supplement 3 . Exosome-enriched preparations probed with antibodies against other extracellular vesicle markers . Exosome-enriched preparations were generated from the culture supernatants of both wild-type and tetherin-knockout cells , either with or without BafA1-treatment ( 100 nM , 16 hr ) , and Western blots were probed with antibodies against CD63 , Alix , Tsg101 , and CD9 . Bands were quantified using ImageJ and normalized to the WT + BafA1 bands . Representative blots are shown on the left; on the right are the means from three independent experiments , with the error bars showing S . E . M . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 011 Immunofluorescence labelling of permeabilised cells showed that endogenous tetherin in HeLa cells is localised to a juxtanuclear compartment , both under control conditions and after BafA1 treatment ( Figure 3B ) . This tetherin labelling colocalised with CD63 labelling , indicating that the juxtanuclear compartment is endosomal ( Figure 3—figure supplement 1A ) . In non-permeabilised cells , where the antibody was only able to access the cell surface , there was relatively little tetherin labelling under control conditions , but BafA1 treatment caused an increase in surface tetherin labelling ( Figure 3B ) . Again , there was excellent colocalisation between tetherin and CD63 ( Figure 3—figure supplement 1B ) . Pre-embedding EM labelling of BafA1-treated cells revealed that this surface labelling was concentrated on exosomes ( Figure 3C , Figure 3—figure supplement 2 ) . However , the presence of tetherin on exosomes after BafA1 treatment does not necessarily mean that tetherin stabilises the association of exosomes with the plasma membrane . Tetherin is known to be trafficked through the endocytic pathway ( Neil et al . , 2008; Habermann et al . , 2010 ) , and thus would be released following endosome-plasma membrane fusion even if it were not playing an active role in exosome aggregation . In order to determine whether tetherin actually holds exosomes together , we used CRISPR/Cas9-mediated genome editing to knock out the tetherin gene in HeLa cells . Clonal cell lines were assayed by Western blotting , and a cell line in which tetherin was no longer expressed was selected for further studies ( Figure 3D ) . To investigate the effect of tetherin loss on exosome distribution , EM was performed on control HeLa cells and on our knockout cell line , following BafA1 treatment . We observed fewer exosomes in the tetherin knockout cells , and those that we did find appeared to be less aggregated ( Figure 3E ) . Quantification of three independent experiments revealed a ~4-fold decrease in exosomes at the plasma membrane in the tetherin knockout cells compared with controls ( Figure 3F ) . We also probed Western blots of exosome-enriched preparations from the culture supernatant of control and tetherin knockout cells , treated with or without BafA1 , to see whether the decrease in exosomes at the plasma membrane could be correlated with an increase in the discharge of exosomes into the medium . There was a weak signal for CD63 in the knockout cells even without BafA1 treatment , and a strong increase in the signal from BafA1-treated tetherin knockout cells compared with BafA1-treated wild-type cells ( Figure 3G ) . Quantification of the signal from three experiments showed a ~4-fold increase , in agreement with our EM results . We also probed for three other extracellular vesicle-associated proteins , Alix , Tsg101 , and CD9 ( Figure 3—figure supplement 3 ) . We found that the tetherin knockout appeared to have a weak effect on Alix and Tsg101 , and a stronger effect on CD9 . Again , it will be important to confirm and extend these observations by mass spectrometry . To obtain a three-dimensional view of exosomes associated with the plasma membrane , we performed scanning electron microscopy on wild-type and tetherin knockout cells ( Figure 4 ) . In untreated wild-type cells , the plasma membrane was essentially devoid of extracellular vesicles , supporting our TEM analysis . Following BafA1-treatment , wild-type cells displayed extracellular vesicles that were not randomly distributed , but rather appeared in clusters . There were also extracellular vesicles associated with the plasma membrane of tetherin knockout BafA1-treated cells , but they were less abundant and the clusters were generally smaller . By combining scanning EM with immunogold labeling , we were able to visualize the surface distribution of both cholesterol ( using TNF-BF ) ( Figure 5A ) and tetherin ( Figure 5B ) . Both were found to be scattered evenly over the surface of untreated cells , but to localize to exosome clusters after BafA1 treatment . 10 . 7554/eLife . 17180 . 012Figure 4 . Scanning electron microscopy reveals clustering of exosomes on the cell surface . Wild-type and tetherin-knockout cells were treated with or without BafA1 ( 100 nM , 16 hr ) before being fixed and prepared for scanning electron microscopy . Scale bar: 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 01210 . 7554/eLife . 17180 . 013Figure 5 . Immuno-SEM reveals the localization of cholesterol and tetherin in 3D . HeLa cells treated with or without BafA1 ( 100 nM , 16 hr ) were fixed , surface-labelled , and viewed by scanning electron microscopy . The cells in A were probed for cholesterol and the cells in B were probed for tetherin . Small ‘bumps’ observed on the plasma membrane probably represent surface-bound antibody complexes . Scale bars: 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 013 Tetherin has been proposed to function by using its transmembrane domain and its GPI anchor to insert itself into two apposing membranes , the plasma membrane and the viral envelope , and removal of the GPI anchor has been shown to abolish its ability to sequester viruses ( Neil et al . , 2008 ) ( see Figure 3A ) . To investigate whether tetherin uses the same strategy to attach exosomes to the plasma membrane and to each other , we transiently transfected tetherin knockout HeLa cells with either wild-type tetherin or a △GPI mutant , both tagged with HA . Expression of the two constructs was confirmed by probing Western blots with an antibody against the HA tag ( Figure 6—figure supplement 1 ) . The wild-type tetherin construct was found to rescue the knockout phenotype , and the clusters of exosomes on the cell surface were heavily decorated with anti-HA . In contrast , the △GPI mutant failed to rescue the phenotype: exosomes were scarce , and the construct localised to the entire plasma membrane ( Figure 6A ) . In both cases , however , there was strong labeling of the lumen of MVBs , indicating that tetherin does not need its GPI anchor to be sorted into ILVs ( Figure 6B ) . 10 . 7554/eLife . 17180 . 014Figure 6 . Exosome tethering can be rescued by WT-tetherin , but not by ΔGPI-tetherin . ( A ) Tetherin-knockout cells were transiently transfected with WT-tetherin-HA or ΔGPI-tetherin-HA , treated with BafA1 ( 100 nM , 16 hr ) , and prepared for surface labelling EM . Cells expressing tetherin-HA constructs were identified by surface gold labelling using an anti-HA antibody . Scale bars: 500 nm . ( B ) Tetherin-knockout HeLa cells were transiently transfected with either WT-tetherin-HA or ΔGPI-tetherin-HA constructs . Cryosections of both were immunolabelled with anti-HA antibodies , revealing localisation to the ILVs of MVBs . Scale bars: 200 nm . See also Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 01410 . 7554/eLife . 17180 . 015Figure 6—figure supplement 1 . Western blotting of tetherin rescue cells . Tetherin-knockout HeLa cells were transiently transfected with either WT-tetherin-HA or ΔGPI-tetherin-HA constructs . Lysed cells were analysed by Western blotting using an anti-HA antibody . EF2 is included as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 015 Finally , to extend our findings to a more physiological system and to ensure that the tethering we observe in HeLa cells is not due to endosome acidification defects , we investigated EBV-transformed B cell lines . These cells have been shown to generate exosomes at steady state without a stimulus ( Raposo et al . , 1996 ) . We observed not only clustering of exosomes by conventional EM ( Figure 7A ) , but also tetherin labeling on the clustered exosomes and on ILVs by cryo immuno EM ( Figure 7B ) . 10 . 7554/eLife . 17180 . 016Figure 7 . EBV-transformed B cells display clusters of tetherin-positive exosomes . ( A ) 721 . 221 EBV-transformed B cells were prepared for conventional EM . Clusters of exosomes can be seen at the cell surface . Scale bars: 500 nm . ( B ) Immuno-EM reveals that the clustered exosomes and the ILVs of MVBs are positive for tetherin labeling . Scale bars: 500 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 17180 . 016
In the present study , we have shown that V-ATPase is a key regulator of both cholesterol trafficking and endosome fate , with a dramatic increase in plasma membrane-associated exosomes when V-ATPase is inhibited by BafA1 . We have also shown that these exosomes are attached to the plasma membrane and to each other by the antiviral protein tetherin . Our findings on cholesterol trafficking absolutely depended upon being able to localise cholesterol at the ultrastructural level , using TNM-BF as a probe . We found that there is practically no cholesterol associated with the plasma membrane after V-ATPase inhibition; instead , the cholesterol is associated with extracellular vesicles , a distinction that was not apparent by light microscopy . There are still a number of unanswered questions about the connection between V-ATPase and cholesterol , including precisely how endosomal pH affects cholesterol trafficking . The availability of TNM-BF as an EM-compatible cholesterol probe should allow these and other questions to be addressed . Exosomes are not the only type of extracellular vesicles produced by cells; other types include microvesicles , produced by plasma membrane shedding , and apoptotic bodies , produced by a controlled 'explosion' of cells undergoing apoptosis . Hence , there is some confusion in the literature because not all studies purporting to be about exosomes actually distinguished between different types of extracellular vesicles ( Raposo and Stoorvogel , 2013 ) . We used several criteria to confirm the identity of the vesicles we observed , including similarity in size to ILVs ( the other types of vesicles are generally larger and more heterogeneous ) , the presence of a previously endocytosed tracer , and enrichment of cholesterol and CD63 . The connection between V-ATPase inhibition and exosome production is at present unclear , and indeed , little is known about the machinery involved in endosome-plasma membrane fusion in general , in part because such fusion events are rare in most cell types . Our finding that endosome-plasma membrane fusion can be induced by blocking endosome acidification should facilitate the discovery of some of the players in this event . Although exosomes are generally thought of as free extracellular vesicles , our live cell imaging results and EM images , together with published micrographs of B cell exosomes ( Raposo et al . , 1996; Simons and Raposo , 2009 ) , suggested to us that exosomes can have an alternative fate: to stay closely associated with the cell that produced them . We decided to investigate the potential role of tetherin in this association , because we were struck by the similarity between our images of exosomes in BafA1-treated cells , and published images of HIV-1 budding from cells in which tetherin was overexpressed and the tetherin antagonist , Vpu , deleted from the viral genome ( Neil et al . , 2008 ) . We discovered that both endogenous and tagged tetherin localise to exosomes , that knocking out tetherin strongly reduces the number of exosomes associated with the plasma membrane with a concomitant increase in exosomes released into the medium , and that this phenotype can be rescued by wild-type tetherin but not by a tetherin construct lacking a GPI anchor , which is also incapable of tethering virus particles . One difference between exosomes and viruses is that whereas overexpression of tetherin in virally infected cells leads to a massive accumulation of virions at the plasma membrane ( Neil et al . , 2008 ) , overexpression of tetherin in BafA1-treated cells did not produce any marked increase in exosome frequency . This may be because the unit number of exosomes is limited by the number of ILVs per endosome , whereas cells infected with a virus such as HIV-1 are programmed to bud as many viruses as possible from the plasma membrane , and so are less confined by space and membrane availability . A single endosome can generate ILVs via distinct mechanisms , resulting in a heterogeneous population of vesicles ( Edgar et al . , 2014 ) . If the endosome then fuses with the plasma membrane , all of the vesicles will be discharged; however , the ultimate fate of each vesicle may depend upon whether or not it contains tetherin . The presence of a GPI anchor suggests that tetherin may preferentially insert into the most cholesterol-rich vesicles; however , in order for tetherin to tether vesicles to each other as well as to the plasma membrane , it must be able to insert into the vesicles both via its GPI anchor and via its transmembrane domain ( see Figure 3A ) ( Venkatesh and Bieniasz , 2013 ) . In any case , a non-homogenous distribution of tetherin would provide the cell with a mechanism for keeping some vesicles attached the cell surface while releasing others into the extracellular matrix or fluid . In addition , expression of tetherin is tightly regulated . It is most highly expressed in cells of the immune system ( Ishikawa et al . , 1995; Goto et al . , 1994; Blasius et al . , 2006 ) but even in cells where it is normally undetectable , its expression can be dramatically upregulated by interferon ( Goto et al . , 1994; Blasius et al . , 2006 ) . This tight control of tetherin expression may provide a means whereby the cell can regulate the fate of its exosomes , and thus control whether the exosomes engage in local interactions ( e . g . , antigen presentation via an immunological synapse [Raposo et al . , 1996] ) or long-distance interactions ( e . g . , supporting or inhibiting cell migration [Sung et al . , 2015] ) . Supporting this notion , tetherin has recently been shown to promote dendritic cell activation and antigen presentation via MHC class II ( Li et al . , 2016 ) , a pathway that has been shown to be mediated by exosomes ( Raposo et al . , 1996 ) . There is currently great interest in exosomes as vehicles for intercellular communication , and clinical studies are underway to exploit them both as diagnostic tools and as potential therapeutic agents . However , as several recent review articles have pointed out , we still don’t really know what exosomes contain , because there is no way of specifically purifying them . In previous work from our lab on clathrin-coated vesicles ( CCVs ) , we have shown that even with an impure preparation , we can deduce the protein composition by comparative proteomics ( e . g . , by preparing the CCV fraction from control vs . clathrin-depleted cells [Borner et al . , 2012] ) . The present study demonstrates that one can specifically enrich or deplete exosomes from the extracellular fluid by manipulating endosomal pH and/or tetherin levels . Thus , by comparing exosome-enriched fractions prepared from cells treated in different ways , it should be possible to determine which proteins in the fraction are actually exosome components and which are contaminants . Exosomes have been proposed to act as shuttles for the spread and propagation of prions ( Fevrier et al . , 2004 ) , beta amyloid ( Rajendran et al . , 2006 ) , and alpha-synuclein ( Danzer et al . , 2012 ) , all of which can lead to severe neurodegenerative disorders . A number of enveloped viruses , including HIV-1 , have developed ways of keeping tetherin off the plasma membrane , and one of the consequences of this downregulation may be that exosome interactions are altered in infected cells . Interestingly , HIV-1-infected patients display an increased propensity to develop neurodegenerative disorders ( HAND – HIV-1 Associated Neurodegeneration ) , as well as increased beta amyloid ( Achim et al . , 2009 ) and alpha-synuclein deposition ( Khanlou et al . , 2009 ) . Exosomes have also been shown to play a pathological role in cancer . Tumor cells frequently shed excessive amounts of exosomes , and these are thought to promote the growth , survival , and metastasis of the tumor in a number of ways ( Zhang and Wang , 2015; Zhang et al . , 2015; Ono et al . , 2014; Melo et al . , 2014; Anastasiadou and Slack , 2014; Zhao et al . , 2016 ) . Interferon-α is sometimes given as a chemotherapeutic drug , and it is tempting to speculate that one of its mechanisms of action may be to prevent exosome shedding , by upregulating tetherin expression in the tumor cells . Thus , the importance of tetherin in health and disease is likely to extend well beyond its role as an antiviral protein .
HeLa cells were a gift from Paul Lehner ( University of Cambridge , UK ) and were cultured in DMEM supplemented with 10% fetal bovine serum , L-glutamine , and Penicillin/Streptomycin , in 5% CO2 at 37°C . The CD63-GFP construct was a gift from Paul Luzio ( University of Cambridge , UK ) , the CLC-mCherry construct was made in-house , and the tetherin-HA constructs were as described ( Billcliff et al . , 2013 ) . Transient transfections were performed using TransitIT-HeLa MONSTER ( Mirus ) . Tetherin knockout HeLa cells were generated using the CRISPR/Cas9 method . Guide RNAs targeting a sequence within the first exon of the tetherin gene ( 5’-GCTCCTGATCATCGTGATTCTGG ) were cloned into pX330 ( Addgene plasmid #42230 ) . Monoclonal HeLa cell lines were generated and assayed for expression of the tetherin protein by immunofluorescence and Western blotting . An apparent tetherin negative line was selected , and the absence of wild-type tetherin sequence was confirmed by PCR-cloning and Sanger sequencing of the targeted region . The EBV-transformed B cell line ( 721 . 221 ) was a gift from Gillian Griffiths ( University of Cambridge , UK ) and were cultured in RPMI supplemented with 10% fetal bovine serum , L-glutamine , and Penicillin/Streptomycin , in 5% CO2 at 37°C . All cells were free of mycoplasma . Antibodies used for immunofluorescence included a mouse monoclonal anti-CD63 , IB5 ( a gift from Mark Marsh , UCL , UK ) , rabbit polyclonal anti-BODIPY ( A-5770 , Molecular Probes - RRID:AB_2536193 ) , rabbit polyclonal anti-Bst2/tetherin ( NIH AIDS reagent program ) , and a mouse monoclonal antibody against tetherin , HM1 . 24 ( Goto et al . , 1994 ) . For Western blotting , the following antibodies were used: a mouse polyclonal anti-tetherin antibody , B02P ( Abnova , Taiwan - RRID:AB_1137604 ) , a mouse monoclonal antibody against the HA tag ( Covance , 16B12 - RRID:AB_10064220 ) , a goat polyclonal anti-EF2 antibody ( Santa Cruz , C-14 - RRID:AB_640040 ) , a rabbit polyclonal anti-Alix antibody ( a gift from Harald Stenmark , University of Oslo , Norway ) , a mouse monoclonal anti-Tsg101 antibody ( GeneTex - RRID:AB_373239 ) , a rabbit monoclonal anti-CD9 antibody ( Abcam , ab92726 - RRID:AB_10561589 ) and a rabbit polyclonal anti-clathrin antibody ( made in-house [Simpson et al . , 1996] ) . A rabbit anti-mouse IgG antibody ( Dako ) was used as a bridging antibody for immuno-EM . Magic Red-Cathepsin B ( ICT938 , BioRad ) and Lysotracker Red DND-99 ( L7528 , ThermoFisher ) were used for live cell imaging . Filipin complex ( F9765 , Sigma ) was dissolved in DMSO , and used at a 25 μg/ml . TNM-BF was synthesized as described previously ( Nishimura et al . , 2010 ) . Protocols for labeling are detailed below . Methyl-β-cyclodextrin ( C4951 , Sigma-Aldrich ) was dissolved in water and used at a final concentration of 10 mM for 30 min . BafA1 ( B1793 , Sigma-Aldrich ) was dissolved in DMSO and used at a final concentration of 100 nM . Control experiments were performed using equal volumes of DMSO . Cell viability with and without bafilomycin treatment was assessed by trypan blue exclusion staining , and in both cases it was found to be >98% . Cells were fixed with 4% PFA/PBS and permeabilised using 0 . 1% saponin . Blocking and subsequent steps were performed with 1% BSA , 0 . 01% saponin in PBS . Cells were mounted on slides with mounting medium containing DAPI ( Invitrogen ) . Cells were fixed with 4% PFA . They were initially stained with filipin , before being permeabilised with 0 . 1% saponin , blocked with 1% BSA , 0 . 01% saponin in PBS , and stained with 1 μM TNM-BF for 30 min on ice . The cells were washed with PBS before being mounted on slides with mounting medium ( Invitrogen ) , and visualized using a Zeiss 880 confocal microscope and 63X objective . Western blotting for tetherin and for CD63 was performed in non-reducing conditions and tetherin blots using a sample buffer without bromophenol blue ( 10% SDS , 15% glycerol , 0 . 2 M Tris-HCl pH 6 . 8 ) as previously described ( Giese and Marsh , 2014 ) . Other Western blots were performed according to standard protocols . Signals were quantified using ImageJ . Exosome-enriched preparations were generated from cell culture supernatants as previously described ( Théry et al . , 2006 ) . Cells were grown in equal numbers on 245 mm x 245 mm dishes in 60 ml of media . Supernatants were collected and cells removed by a low speed spin ( 300xg , 10 min ) . Pellets were discarded and supernatants were then spun to remove large cell debris ( 2000xg , 10 min ) . Again pellets were discarded and supernatants were ultracentrifuged ( Type 70 Ti Rotor , Beckman Coulter ) to remove smaller cell debris ( 10 , 000xg , 30 min ) . Pellets were retained for controls and supernatants were again collected and ultracentrifuged ( Type 70 Ti Rotor , Beckman Coulter ) at 100 , 000xg for 70 min . Pellets were washed in PBS and repelleted at 100 , 000xg for 70 min ( Type 70 Ti Rotor , Beckman Coulter ) to give exosome-enriched fractions . Exosome-enriched pellets were resuspended in 100 μl sample buffer and equal volumes loaded for Western blotting . Quantification of Western blotting was performed using ImageJ . Cells were fixed with 4% PFA , 0 . 2% glutaraldehyde , 0 . 1 M phosphate buffer ( pH 7 . 4 ) . They were then pelleted in 12% gelatine and 70 nm sections cut at −120°C with a Leica ultracut UCT ultramicrotome . Sections were labelled with 10 nm protein-A gold ( Utrecht university , the Netherlands ) and stained as previously described ( Slot et al . , 1991 ) . Control or BafA1-treated HeLa cells ( 100 nM , 16 hr ) were fixed with 4% PFA , 0 . 2% glutaraldehyde , 0 . 1 M phosphate buffer ( pH 7 . 4 ) . Cells were pelleted in 12% gelatine and 70 nm sections cut at −120°C with a Leica ultracut UCT ultramicrotome . Sections were picked up in 2 . 3 M sucrose and grids were blocked with 2% gelatin . Prior to gold labeling , grids were washed with 15 mM glycine and then washed twice with PBS . Sections were incubated on drops of 1 μM TNM-BF for 30 min on ice , washed , blocked with 1% BSA in PBS , and then incubated with a rabbit polyclonal anti-BODPIY antibody ( Molecular Probes , Invitrogen , UK ) . Sections were labelled with 10 nm protein-A gold ( Utrecht university , the Netherlands ) and stained as previously described ( Slot et al . , 1991 ) . HeLa cells were grown on 13 mm diameter Thermanox coverslips ( Nalge Nunc International , Rochester , NY , ) and briefly fixed in their culture medium with an equal volume of double strength fixative ( 4% PFA , 5% glutaraldehyde , 0 . 1 M cacodylate buffer pH 7 . 4 ) , before being exchanged for single strength fixative ( 2% PFA , 2 . 5% glutaraldehyde in 0 . 1 M cacodylate buffer pH 7 . 4 ) . EBV-transformed B cells were briefly fixed in a double strength fixative ( 4% PFA , 5% glutaraldehyde , 0 . 1 M cacodylate buffer pH 7 . 4 ) at an equal volume to their culture medium before being spun into a pellet . Cells were post-fixed with 1% osmium tetroxide and incubated with 1% tannic acid to enhance contrast . They were dehydrated using increasing concentrations of ethanol before being embedded onto EPON stubs ( coverslips ) or EPON-filled molds ( pellets ) , and cured overnight at 65°C . Coverslips were removed using a heat-block . Ultrathin 70 nm conventional sections were cut using a diamond knife mounted to a Reichart ultracut S ultramicrotome . Sections were stained with lead citrate before being observed on a FEI Tecnai transmission electron microscope ( FEI , the Netherlands ) at an operating voltage of 80 kV . Cells were grown on gridded glass coverslips ( MatTek Corporation , Ashland , MA ) . Immunofluorescence staining was performed as described above , following which cells were refixed with 2% PFA , 2 . 5% glutaraldehyde , 0 . 1 M cacodylate and processed for conventional EM . Resin stubs were inverted over areas of interest and baked overnight . Coverslips were subsequently removed using liquid nitrogen . Cells were grown on Thermanox coverslips ( Nalge Nunc International , Rochester , NY ) and briefly fixed with 4% PFA/0 . 1 M cacodylate at a 1:1 ratio with culture medium . After 2 min , fixative was replaced with fresh 2% PFA/0 . 1 M cacodylate . Cells were blocked with 1% BSA , 0 . 1% Aurion BSA ( Aurion , the Netherlands ) before being incubated with mouse monoclonal antibodies against lumenal/extracellular antigens for 1 hr at room temperature , followed by incubation with a rabbit anti-mouse secondary antibody ( Dako ) . The cells were then labeled with 10 nm protein-A gold ( Utrecht university , the Netherlands ) , and re-fixed with 2% PFA , 2 . 5% glutaraldehyde / 0 . 1 M cacodylate , before being post-fixed with 1% osmium tetroxide and processed for conventional resin embedding as above . For cell surface cholesterol labeling , cells were fixed as above . Coverslips were then incubated with 1 μM TNM-BF for 30 min on ice , washed , and incubated with a rabbit polyclonal anti-BODIPY antibody followed by 10 nm protein-A gold as above . BSA conjugated to 5 nm colloidal gold was prepared as previously described ( Slot and Geuze , 1985 ) . Cells were incubated with BSA-gold for 10 min , followed by five washes with PBS to remove surface BSA-gold . The cells were then chased in fresh medium for either 30 min or 4 hr to load endosomes or lysosomes respectively , before addition of BafA1 ( 100 nM , 16 hr ) . Cells were then fixed for conventional electron microscopy ( as above ) . Exosomes were counted manually and lengths of plasma membrane calculated using ImageJ . Five consecutive photographs along random portions of plasma membrane were imaged , from at least five random cells – omitting areas where cells were within 5 μm of another cell . Only exosomes within 500 nm of plasma membrane were quantified . Three independent experiments were performed . The number of gold particles associated with plasma membrane or extracellular vesicles per image were counted . The length of plasma membrane ( μm ) per image was then calculated using ImageJ and the number of gold particles per μm were calculated for each image . Over 100 μm total plasma membrane was quantified per condition , over two separate experiments . Cells grown on glass coverslips were fixed with 2% PFA/ 2 . 5% glutaralydehyde , 0 . 1 M cacodylate and secondarily fixed with 1% osmium . After critical point drying , the cells were coated with 15 nm of carbon using a Quorum Q150 carbon coater and imaged with a FEI Verious 460 SEM . Cells were grown on glass coverslips and fixed with 4% PFA in PBS . Coverslips were labeled with gold as described above ( pre-embedding surface labeling of cell surface/exosomal antigens ) . After gold labeling , cells were refixed with 2% PFA , 2 . 5% glutaraldehyde , and sputter coated with a thin layer of carbon using a Quorum Q150 carbon coater . SEM images were collected using an FEI Various 640 SEM . A secondary detector was used to acquire cell surface images , immediately followed by acquisition using a backscatter detector to detect gold particles . | Cells generally communicate with each other over short distances by direct contact , and over long distances by releasing chemicals such as hormones . But there is also a third way that is less well understood – small capsules or “vesicles” called exosomes can transfer molecules from one cell to another . Exosomes are involved in the immune response and have been linked to a number of diseases , including cancer and neurodegeneration . However , scientists are still trying to understand how exosomes are made , what they contain and how they are released from cells . A common set of cells used in laboratory studies are known as HeLa cells . These cells are the descendants of cancerous cells taken from a patient called Henrietta Lacks in 1951 . When treated with a particular drug , HeLa cells produce vesicles that look like exosomes . Yet instead of moving freely like other exosomes , these structures stick together in clusters . This raises questions – are these cancer cell vesicles truly exosomes ? And if so , why and how are they tethered to the cell ? Using electron microscopy and biochemical tests , Edgar et al . confirm that the unusual vesicles produced by HeLa cells are exosomes . As well as sharing characteristics with other exosomes , the vesicles also show similarities with viruses like HIV , which attach themselves to cell surfaces and each other using a protein called tetherin . Using a technique called gene editing to remove tetherin from HeLa cells allowed the exosomes in the cluster to move apart . Further investigation revealed that some cells in the immune system also produce exosome clusters and that these clusters also contain tetherin . Edgar et al . propose that cells control whether exosomes are involved in short-range or long-range communication by controlling the amount of tetherin they produce . So far , studies into the roles that exosomes play in the body have been hampered by a lack of experimental tools . The study by Edgar et al . opens up new methods of investigation by providing ways of altering the number of exosomes released from a cell . This should help to clarify what exosomes do and how they work in a wide range of different cell types . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2016 | Tetherin is an exosomal tether |
Uromodulin is the most abundant protein in the urine . It is exclusively produced by renal epithelial cells and it plays key roles in kidney function and disease . Uromodulin mainly exerts its function as an extracellular matrix whose assembly depends on a conserved , specific proteolytic cleavage leading to conformational activation of a Zona Pellucida ( ZP ) polymerisation domain . Through a comprehensive approach , including extensive characterisation of uromodulin processing in cellular models and in specific knock-out mice , we demonstrate that the membrane-bound serine protease hepsin is the enzyme responsible for the physiological cleavage of uromodulin . Our findings define a key aspect of uromodulin biology and identify the first in vivo substrate of hepsin . The identification of hepsin as the first protease involved in the release of a ZP domain protein is likely relevant for other members of this protein family , including several extracellular proteins , as egg coat proteins and inner ear tectorins .
Uromodulin , also known as Tamm-Horsfall protein , is a 105 kDa glycoprotein exclusively expressed in the kidney by epithelial cells lining the thick ascending limb ( TAL ) of Henle’s loop . It is a glycosylphosphatidylinositol ( GPI ) -anchored protein mainly localised at the apical plasma membrane of TAL epithelial cells . Uromodulin is secreted into the renal tubule through a conserved proteolytic cleavage ( Santambrogio et al . , 2008 ) . In the urine it constitutes the most abundant protein under physiological conditions ( Serafini-Cessi et al . , 2003 ) and it is mainly found as high molecular-weight polymers . Uromodulin polymerisation is mediated by its Zona Pellucida ( ZP ) domain , a module present in several proteins that assemble into extracellular polymers , such as components of the egg coat ( including sperm receptors ZP3 and ZP2 ) and of the inner ear tectorial membrane ( Jovine et al . , 2005 ) . Specific cleavage is necessary for protein assembly into polymers , as it releases an inhibitory hydrophobic motif ( external hydrophobic patch , EHP ) distal to the ZP domain that prevents polymerisation ( Jovine et al . , 2004; Schaeffer et al . , 2009; Han et al . , 2010 ) . The biological function of uromodulin is complex and only partly understood . Studies in Umod-/- mice demonstrated that it protects against urinary tract infections ( Bates et al . , 2004 ) and renal stone formation ( Liu et al . , 2010 ) . Uromodulin has also been shown to modulate the activity of innate immunity cells via Toll-like receptor 4 signalling ( Säemann et al . , 2005 ) or via activation of the inflammasome ( Darisipudi et al . , 2012 ) and to regulate granulopoiesis and systemic neutrophil homeostasis ( Micanovic et al . , 2015 ) . Finally , it is involved in the regulation of NaCl transport in the TAL , based on its role in promoting surface expression of the renal outer medullary potassium channel ( ROMK2 ) ( Renigunta et al . , 2011 ) and of the sodium-potassium-chloride co-transporter ( NKCC2 ) ( Mutig et al . , 2011 ) . Recent data showed that uromodulin plays an important role in chronic diseases of the kidney ( Rampoldi et al . , 2011 ) . Mutations in the uromodulin gene ( UMOD ) cause an autosomal dominant tubulo-interstitial kidney disease ( Hart , 2002 ) . In addition , genome-wide association studies demonstrated that common variants in the UMOD gene are associated with increased risk for chronic kidney disease ( CKD ) and hypertension ( Köttgen et al . , 2009; Padmanabhan et al . , 2010 ) . Such an effect is due to higher UMOD expression driven by the presence of risk alleles in its gene promoter ( Trudu et al . , 2013 ) . Given the importance of polymerisation for uromodulin activity and the fact that this process depends on a specific protein cleavage , in this work we aimed at identifying the protease responsible for such cleavage and urinary secretion .
As for other ZP proteins , uromodulin cleavage at a specific site in the protein C-terminus releases the interaction between two hydrophobic motifs ( internal hydrophobic patch , IHP; external hydrophobic patch , EHP ) ( Figure 1A ) , leading to conformational activation of the ZP domain and protein polymerisation ( Jovine et al . , 2004; Schaeffer et al . , 2009; Han et al . , 2010 ) . 10 . 7554/eLife . 08887 . 003Figure 1 . MDCK cells as a model to study physiological uromodulin shedding . ( A ) Schematic representation of human uromodulin domain structure containing a leader peptide ( predicted to be cleaved at residue 23 ) , three EGF-like domains , a central domain with 8 conserved cysteines ( D8C ) , a bipartite Zona Pellucida ( ZP ) domain ( ZP-N/ZP-C ) and a glycosylphosphatidylinositol ( GPI ) -anchoring site ( predicted at position 614 ) . Internal ( IHP ) and External ( EHP ) Hydrophobic Patches ( Jovine et al . , 2004; Schaeffer et al . , 2009 ) , Consensus Cleavage Site ( CCS ) and seven N-glycosylation sites ( Ψ ) are also indicated . ( B ) Immunofluorescence analysis of non-permeabilised MDCK cells expressing uromodulin . Polymers formed by the protein are clearly detected on the cell surface . Scale bar , 50 µm . ( C ) Electron microscopy analysis of uromodulin polymers purified from the medium of MDCK cells . The arrows indicate the typical protrusions of uromodulin filaments spaced about 130 Å . Scale bar , 100 nm . ( D ) Representative Western blot analysis of N-deglycosylated uromodulin secreted by transfected MDCK cells or purified from urine . A single isoform is clearly seen in the urinary sample . An isoform with similar molecular weight is released by MDCK cells ( white arrowhead ) , which also secrete a longer and more abundant one ( black arrowhead ) . ( E ) Representative tandem mass-spectrometry ( MS/MS ) spectrum confirming the identity of the C-terminal peptide 572DTMNEKCKPTCSGTRF587 of the short uromodulin isoform released by MDCK cells and table of fragmented ions . The C-terminal residue F587 is identical to the one that we mapped in human urinary protein ( Santambrogio et al . , 2008 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 003 To understand the nature of such cleavage , we took advantage of a cellular system , Madin-Darby Canine Kidney ( MDCK ) cells , where transfected human uromodulin assembles extracellularly in filamentous polymers ( Figure 1B , C ) that are indistinguishable from the urinary ones ( Jovine et al . , 2002 ) . In these cells , uromodulin is secreted as two isoforms that can be separated on gel electrophoresis after enzymatic removal of protein N-glycans at about 72 and 77 kDa ( Figure 1D ) . Only the shorter isoform assembles into polymers , since it is generated by a cleavage that releases the inhibitory EHP motif , while the longer one is generated by a more distal cleavage and still retains the EHP ( Schaeffer et al . , 2009 ) . The short uromodulin isoform released by MDCK cells corresponds to the one found in the urine , as it shares the same molecular weight ( Figure 1D ) and the same C-terminal residue ( F587 [Santambrogio et al . , 2008] ) ( Figure 1E ) , demonstrating that uromodulin undergoes physiological cleavage in these cells . As mapping of the C-terminus of the short uromodulin isoform suggests proteolytic cleavage , we first treated uromodulin-expressing MDCK cells with a protease inhibitor cocktail ( PIC ) . This treatment led to significant reduction of uromodulin polymerisation on the surface of the cells ( Figure 2A ) that is not due to any alteration of protein expression or intracellular distribution ( Figure 2—figure supplement 1 ) , confirming that the polymerisation-competent uromodulin isoform is generated through the action of a protease . We then treated MDCK cells with the single PIC components , each targeting a specific catalytic type of protease , and observed that only inhibitors of serine proteases significantly reduced uromodulin polymerisation ( Figure 2A ) . Consistently , the expression of the Kunitz-type serine protease inhibitor Hepatocyte growth factor Activator Inhibitor-1 ( HAI-1 ) ( Shimomura et al . , 1997 ) essentially abolished uromodulin polymerisation on the surface of MDCK cells and reduced the secretion of the shorter uromodulin isoform ( Figure 2—figure supplement 2 ) . To explore the role of membrane-association for uromodulin cleavage , we expressed a soluble isoform of uromodulin , generated by truncation at the GPI-anchoring site ( S614X ) , that is efficiently trafficked and secreted in MDCK cells ( Schaeffer et al . , 2009 ) . Interestingly , this isoform was neither processed to the shorter variant nor it formed polymers ( Figure 2B , C ) , indicating that , as in the case of ZP3 ( Jovine et al . , 2002 ) , membrane association is necessary to allow uromodulin physiological cleavage and polymerisation . From these results we conclude that physiological cleavage of uromodulin depends on a serine protease , and that this enzyme is likely membrane associated . 10 . 7554/eLife . 08887 . 004Figure 2 . A serine protease is responsible for the release of polymerisation-competent uromodulin . ( A ) Immunofluorescence analysis showing uromodulin on the surface of MDCK cells treated with vehicle ( DMSO ) ( ctr ) , protease inhibitor cocktail ( PIC ) or single PIC components , as indicated . Scale bar , 50 µm . Quantification of the average surface of uromodulin polymers shows that only treatment with PIC or with specific serine protease inhibitors ( AEBSF , aprotinin and leupeptin ) significantly reduces uromodulin polymerisation on the surface of MDCK cells . Bars indicate average ± s . e . m . ***p<0 . 001 ( Mann-Whitney test ) . The graph represents mean ratios of 3 independent experiments ( Figure 2—source data 1 ) . ( B ) Immunofluorescence analysis of permeabilised ( left ) or non-permeabilised ( right ) MDCK cells expressing soluble uromodulin mutant S614X . Uromodulin polymerisation is abolished by preventing its association to the membrane . Scale bar , 50 µm . ( C ) Representative Western blot analysis of N-deglycosylated uromodulin secreted by MDCK cells expressing wild-type or soluble ( S614X ) uromodulin . Lack of plasma membrane anchoring does not affect uromodulin secretion but it abolishes its cleavage at the physiological site ( white arrowhead ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 00410 . 7554/eLife . 08887 . 005Figure 2—source data 1 . Quantification of the area of uromodulin polymers on the surface of MDCK cells after protease inhibitor treatment ( Figure 2A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 00510 . 7554/eLife . 08887 . 006Figure 2—source data 2 . Short cleavage inhibition by HAI-1 expression ( Figure 2—figure supplement 2C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 00610 . 7554/eLife . 08887 . 007Figure 2—figure supplement 1 . PIC treatment does not affect uromodulin intracellular distribution and expression . ( A ) Immunofluorescence analysis showing intracellular distribution of uromodulin in MDCK cells treated with vehicle ( DMSO ) ( ctr ) or protease inhibitor cocktail ( PIC ) . KDEL is a marker of the endoplasmic reticulum ( ER ) . Scale bar , 16 µm . ( B ) Representative Western blot analysis of uromodulin in lysates of MDCK cells under the same conditions as above . The upper band corresponds to the mature , fully glycosylated protein , the lower band corresponds to the immature protein carrying ER-type glycosylation ( Schaeffer et al . , 2012 ) . Alpha-tubulin is shown as a loading control . PIC treatment does not alter uromodulin intracellular distribution nor its expression . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 00710 . 7554/eLife . 08887 . 008Figure 2—figure supplement 2 . Expression of the serine protease inhibitor HAI-1 effectively reduces uromodulin cleavage at the urinary site . ( A ) Immunofluorescence analysis showing uromodulin on the surface of MDCK cells co-expressing the serine protease inhibitor HAI-1 ( Hepatocyte growth factor Activator Inhibitor-1 ) , as indicated . HAI-1 expression essentially abolishes uromodulin polymerisation on the cell surface . Scale bar , 50 µm . ( B ) Representative Western blot analysis showing uromodulin and HAI-1 expression in cell lysates of transfected MDCK cells , as indicated . Glyceraldehyde-3-phosphate dehydrogenase ( Gapdh ) is shown as a loading control . ( C ) Representative Western blot analysis of N-deglycosylated uromodulin secreted by MDCK cells co-expressing HAI-1 , as indicated . Densitometric analysis ( average ± s . d . of 3 independent experiments , Figure 2—source data 2 ) shows the ratio between the short and the long uromodulin isoforms in the absence or presence of HAI-1 co-expression , as indicated . The serine protease inhibitor HAI-1 strongly reduces the amount of the short uromodulin isoform released in the culturing medium by MDCK cells . *p<0 . 05 ( Student’s t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 008 We moved further starting from the interesting observation that uromodulin expressed in a different kidney-derived cell line , Human Embryonic Kidney 293 ( HEK293 ) , did not form polymers on the cell surface ( Figure 3A ) . Accordingly , only the longer polymerisation-incompetent uromodulin isoform was released by these cells ( Figure 3B ) . We hypothesised that this was due to lack of expression of the proper serine protease responsible for physiological uromodulin cleavage . Under this assumption , we compared the expression profiles available in Gene Expression Omnibus of known serine proteases ( about 220 enzymes ) between MDCK and HEK293 cells , to identify those exclusively expressed by the former . Among the resulting 23 candidates we then selected only membrane-bound enzymes that were found to be expressed in available TAL segment transcriptomes ( Yu et al . , 2009; Cheval et al . , 2011 ) . This strategy led to the identification of only four candidate enzymes: prostasin ( Prss8 ) , hepsin/TMPRSS1 ( Hpn ) , mucin 1 ( Muc1 ) and dipeptidyl-peptidase IV ( Dpp4 ) ( Figure 3C ) . The last two enzymes were excluded on the basis of their catalytic activity: Dpp4 sequentially removes N-terminal dipeptides from polypeptides , cleaving after a proline residue ( Mentlein , 1999 ) , whereas Muc1 contains a self-cleaving module rather than a classical peptidase S1 domain ( Levitin et al . , 2005 ) . Validation of transcriptome data confirmed differential expression of the two candidate proteases , hepsin and prostasin , in MDCK and HEK293 cells ( Figure 3—figure supplement 1A ) . Interestingly , both proteases are inhibited by the Kunitz-type serine protease inhibitor HAI-1 ( Fan et al . , 2005; Herter et al . , 2005 ) , in line with our results in MDCK cells ( Figure 2—figure supplement 2 ) . 10 . 7554/eLife . 08887 . 009Figure 3 . Hepsin and prostasin interact with uromodulin to induce its cleavage and polymerisation . ( A ) Immunofluorescence analysis ( extended focus image ) showing uromodulin on the surface of HEK293 cells . The protein is not assembled into polymers when expressed in this cellular system . Scale bar , 21 µm . ( B ) Representative Western blot analysis of N-deglycosylated uromodulin secreted by MDCK and HEK293 cells . HEK293 cells only secrete the longer polymerisation-incompetent uromodulin isoform ( black arrowhead ) , while MDCK cells also secrete the shorter one ( white arrowhead ) . ( C ) Schematic representation of the selection process employed to identify candidate enzymes for the secretion of the short uromodulin isoform . Only membrane-bound serine proteases specifically expressed by MDCK cells and by the TAL segment of the nephron , but not by HEK293 cells , were selected . ( D ) Representative Western blot analysis of N-deglycosylated uromodulin secreted by HEK293 cells expressing wild-type or catalytically inactive human hepsin or prostasin , as indicated . Only wild-type proteases promote the secretion of the short uromodulin isoform by these cells ( white arrowhead ) . ( E ) Immunofluorescence analysis ( extended focus image ) showing uromodulin on the surface of HEK293 cells expressing wild-type or catalytically inactive hepsin or prostasin , as indicated . Uromodulin polymerisation is induced only when wild-type proteases are expressed . Scale bar , 21 µm . ( F ) Representative Western blot analysis of uromodulin immunoprecipitation ( upper panels ) from lysates of HEK293 cells expressing hepsin or prostasin , as indicated . Both enzymes are co-immunoprecipitated when uromodulin is co-expressed in HEK293 cells ( lower panels ) . The arrowheads point at hepsin and prostasin specific bands . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 00910 . 7554/eLife . 08887 . 010Figure 3—figure supplement 1 . Hepsin and prostasin expression in MDCK and HEK293 cells . ( A ) RT-PCR analysis showing gene expression of candidate proteases hepsin ( HPN ) and prostasin ( PRSS8 ) in MDCK and HEK293 cells . Constructs containing coding sequences of the human proteases were used as PCR positive controls ( C+ ) . Expression of GAPDH is shown as a cDNA positive control . PRSS8 and HPN are exclusively expressed in MDCK cells , confirming data obtained from available transcriptomes . ( B ) Representative Western blot analysis of uromodulin , hepsin and prostasin in lysates of transfected HEK293 cells . Wild-type proteases as well as catalytically inactive enzymes were expressed in HEK293 cells , as indicated . Protein disulfide-isomerase ( Pdi ) is shown as a loading control . The arrowhead points at hepsin specific band . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 010 We validated our hypothesis by expressing either hepsin or prostasin in HEK293 cells . This induced short cleavage as well as polymerisation of uromodulin , an effect that was dependent on the catalytic activity of the enzymes , as it was abolished when catalytically inactive isoforms of the proteases were expressed ( Figure 3D , E; Figure 3—figure supplement 1B ) . Moreover , this effect is likely direct , as both enzymes could be specifically co-immunoprecipitated with uromodulin from the lysates of HEK293 cells ( Figure 3F ) . Based on this observation , we assessed whether uromodulin can be directly cleaved by hepsin and prostasin in vitro . To do so , we carried out cleavage assays by employing a truncated recombinant form of uromodulin that corresponds to the elastase-resistant fragment of uromodulin from human urine ( efUmod , lacking residues 27–294 ) ( Jovine et al . , 2002 ) . We focussed on this fragment as it primarily consists of the ZP polymerisation domain ( Figure 4A ) . Also , deletion of the N-terminal part of uromodulin greatly improved its purification efficiency ( data not shown ) , probably because it lacks several N-glycosylation sites . Importantly , this isoform undergoes proper proteolytic cleavage and forms polymers in MDCK cells ( Figure 4B ) . When incubated with purified efUmod , both hepsin and prostasin induced a mass shift of efUmod as well as loss of its C-terminal histidine tag ( Figure 4C ) . These effects were abolished by mutagenesis of the consensus cleavage site ( efUmod YAla , carrying the mutation 586RFRS589 > 586AYAA589 ) . Taken together , these findings indicate that hepsin and prostasin physically interact with uromodulin and cleave it at the physiological site . Interestingly , hepsin was more efficient than prostasin , since it completely digested the substrate despite being used at lower concentration ( see figure legend ) . 10 . 7554/eLife . 08887 . 011Figure 4 . Hepsin and prostasin directly cleave uromodulin in vitro . ( A ) Schematic representation of human uromodulin domain structure as shown in Figure 1A . The region not included in recombinant efUmod is shadowed . ( B ) The deletion of the elastase-sensitive fragment of uromodulin does not affect protein polymerisation on the surface of MDCK cells , as shown by immunofluorescence analysis ( efUmod wt ) . As for full-length uromodulin ( Schaeffer et al . , 2009 ) , this process depends on correct protein cleavage at the physiological site , since it is abolished when the consensus cleavage site is mutated ( efUmod 4Ala , carrying the mutation 586RFRS589 > 586AAAA589 ) . Scale bar , 50 µm . ( C ) Purified efUmod , either wild-type ( efUmod wt ) or mutated at the consensus cleavage site ( efUmod YAla , carrying the mutation 586RFRS589 > 586AYAA589 ) , was incubated with recombinant prostasin or hepsin , as indicated . Both proteases decrease the mass of wild-type efUmod ( white arrowheads in upper and middle panels ) and cause the loss of its C-terminal His-tag ( lower panel ) . Hepsin is more efficient than prostasin , as it drives complete digestion of the product , despite being used at 20x lower concentration ( picomolar ratio between protease and efUmod was 1:100 for hepsin and 1:5 for prostasin , see lanes 7 and 8 for comparison ) . The asterisk indicates His-tagged prostasin . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 011 The two candidate proteases could not be discriminated on the basis of their subcellular localisation in MDCK cells . Indeed , while endogenous proteases could not be detected , possibly because of low expression levels or low affinity for the employed antibodies ( which were raised against their human counterparts ) , transfected hepsin and prostasin both co-localised with uromodulin on the apical membrane of polarised MDCK cells ( Figure 5A ) . Hence , we designed short hairpin RNAs ( shRNAs ) to specifically silence either protease in this cellular system . We successfully obtained comparable expression knock-down for both enzymes ( Figure 5B ) , but only hepsin silencing led to a significant reduction of uromodulin polymerisation on the cell surface ( Figure 5C ) . These results demonstrate that hepsin is the enzyme responsible for uromodulin cleavage at the urinary site in MDCK cells . 10 . 7554/eLife . 08887 . 012Figure 5 . Hepsin is the protease mediating uromodulin polymerisation in MDCK cells . ( A ) Confocal immunofluorescence analysis showing uromodulin ( green ) , hepsin or prostasin ( red ) and E-cadherin ( blue ) ( basolateral membrane marker ) in polarised MDCK cells , as indicated . Upper panels represent the reconstruction on the xz axis of merged xy scans , for which a representative image is shown in lower panels . Both serine proteases co-localise with uromodulin on the apical plasma membrane of polarised MDCK cells . z stacks = 0 . 3 µm . A: apical , BL: basolateral . Scale bars , 5 µm . ( B ) Transcript levels of HPN and PRSS8 , as assessed by Real-Time qPCR in MDCK cells transfected with shRNA vectors , as indicated . Expression values ( normalised to glyceraldehyde-3-phosphate dehydrogenase , GAPDH ) are shown as relative to cells transfected with control vector . Expression of the proteases is specifically reduced in silenced cells . Bars indicate average ± s . e . m . **p<0 . 01 , ***p<0 . 001 ( Student’s t test ) . The graph represents mean ratios of 3 independent experiments ( Figure 5—source data 1 ) . ( C ) Immunofluorescence analysis showing uromodulin on the surface of MDCK cells transfected with control vector or with shRNA vectors targeting hepsin or prostasin , as indicated . Scale bar , 50 µm . Quantification of the average surface of uromodulin polymers shows that silencing of hepsin , but not of prostastin , substantially reduces uromodulin polymerisation on the membrane of MDCK cells . Bars indicate average ± s . e . m . ***p<0 . 001 ( Mann-Whitney test ) . The graph represents mean ratios of 3 independent experiments ( Figure 5—source data 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 01210 . 7554/eLife . 08887 . 013Figure 5—source data 1 . Transcript level of HPN and PRSS8 in MDCK cells after shRNA transfection ( Figure 5B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 01310 . 7554/eLife . 08887 . 014Figure 5—source data 2 . Quantification of the area of uromodulin polymers on the surface of MDCK cells after shRNA transfection ( Figure 5C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 014 To understand the physiological relevance of our findings , we then investigated the expression of endogenous hepsin in mouse kidneys . Remarkably , by microdissecting mouse nephron segments , we detected strong hepsin expression in the TAL ( Figure 6A ) . Moreover , hepsin co-localises with uromodulin on the apical membrane of TAL epithelial cells , suggesting a functional interaction between the two proteins ( Figure 6B ) . To assess the role of hepsin in uromodulin processing in vivo , we characterised uromodulin urinary secretion in a previously generated Hpn-/- mouse model , which does not show any gross phenotype or structural abnormality in major organs ( Wu et al . , 1998 ) . Notably , urinary uromodulin levels were significantly reduced in Hpn-/- mice ( Figure 6C ) . This was accompanied by marked accumulation of the mature , fully glycosylated form of uromodulin in kidney lysates ( Figure 6D ) . On the other hand , no difference was observed in the cellular distribution of uromodulin in TAL cells , showing typical apical membrane enrichment ( Figure 6E ) , and no change in Umod gene expression was detected ( Figure 6F ) . These data exclude that reduced levels of urinary uromodulin in Hpn-/- mice could be due to changes in protein expression or transport to the apical plasma membrane , strongly suggesting defective protein secretion . 10 . 7554/eLife . 08887 . 015Figure 6 . Defective uromodulin urinary secretion in mice lacking hepsin . ( A ) Transcript level of Hpn , as assessed by Real-Time qPCR on microdissected nephron segments ( normalised to Gapdh ) . Expression of Hpn is detected in proximal straight tubules ( PST ) , thick ascending limb ( TAL ) , proximal convoluted tubules ( PCT ) , distal convoluted tubules ( DCT ) and , to a lesser extent , in collecting ducts ( CD ) . Minimal expression of the protease is detected in glomeruli ( GI ) . Bars indicate average ± s . e . m . of 3 independent experiments ( Figure 6—source data 1 ) . ( B ) Immunofluorescence analysis of mouse kidney sections showing co-localisation of endogenous hepsin and uromodulin on the apical plasma membrane of TAL epithelial cells . Scale bar , 20 µm . ( C ) Representative Western blot analysis of urinary uromodulin secretion in Hpn-/- mice or control animals . Urinary protein loading was normalised to urinary creatinine concentration . Densitometric analysis shows reduced uromodulin urinary secretion in animals lacking hepsin ( average ± s . d . , n = 10/group , Figure 6—source data 2 ) . ***p<0 . 001 ( Student’s t test ) . ( D ) Western blot analysis of uromodulin in kidney lysates of Hpn-/- mice or control animals . Beta-actin is shown as a loading control . Densitometric analysis shows accumulation of uromodulin in kidney lysates from Hpn-/- mice ( average ± s . d . , n = 3/group , Figure 6—source data 3 ) . **p<0 . 01 ( Student’s t test ) . ( E ) Representative immunofluorescence analysis showing apical plasma membrane signal of uromodulin in kidney sections from Hpn-/- mice or control animals . Scale bar , 50 µm . ( F ) Transcript level of Umod , as assessed by Real-Time qPCR on total kidney extracts from Hpn-/- mice or control animals ( normalised to Hprt1 ) . Expression of Umod is comparable between the two groups of animals ( n = 3/group , Figure 6—source data 4 ) . Bars indicate average ± s . e . m . ( Student’s t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 01510 . 7554/eLife . 08887 . 016Figure 6—source data 1 . Transcript level of Hpn in mouse microdissected nephron segments ( Figure 6A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 01610 . 7554/eLife . 08887 . 017Figure 6—source data 2 . Quantification of urinary uromodulin secretion in Hpn-/- and control mice ( Figure 6C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 01710 . 7554/eLife . 08887 . 018Figure 6—source data 3 . Quantification of uromodulin levels in kidney lysates of Hpn-/- and control mice ( Figure 6D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 01810 . 7554/eLife . 08887 . 019Figure 6—source data 4 . Transcript levels of Umod in Hpn-/- and control mice ( Figure 6F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 019 After deglycosylation , urinary uromodulin from Hpn-/- mice appeared as two isoforms: a short one , with similar electrophoretic mobility as in wild-type urines , and a longer one , normally undetectable in urines from wild-type mice ( Figure 7A ) . Mass-spectrometry analysis on the short isoform showed that it lacks the physiological C-terminus ( F588 ) and is produced by a more distal cleavage occurring after residue R607 ( Figure 7B–D ) . The same analysis on the long isoform revealed an even more distal C-terminus , after residue K616 ( Figure 7—figure supplement 1 ) , consistent with its slower electrophoretic mobility . These data demonstrate that absence of hepsin abolishes uromodulin cleavage at the physiological site and induces its misprocessing . Interestingly , uromodulin from Hpn-/- urines was not sedimented in the pellet fraction of a biochemical polymerisation assay ( Jovine et al . , 2002 ) ( Figure 7E ) , indicating that the residual urinary protein is not assembled into polymers . This is consistent with the observation that both urinary isoforms in Hpn-/- mice are misprocessed and still retain the polymerisation-inhibiting EHP motif ( Figure 7D ) ( Jovine et al . , 2004; Schaeffer et al . , 2009 ) . 10 . 7554/eLife . 08887 . 020Figure 7 . Absence of hepsin in vivo abolishes physiological cleavage and polymerisation of uromodulin . ( A ) Representative Western blot analysis of N-deglycosylated urinary uromodulin secreted by Hpn-/- mice or control animals . Hpn-/- mice show the presence of two uromodulin isoforms: a short one with similar electrophoretic mobility as in wild-type urines ( white arrowhead ) , and a longer one that is absent in samples from wild-type mice ( black arrowhead ) ( n = 6/group ) . ( B ) Mass spectrometry ( MS ) sequence coverage ( 52% over the entire protein ) of trypsin-digested mouse uromodulin ( short isoform ) ( UniProt accession Q91X17 ) purified from urine of Hpn-/- mice . Matching peptides are shown in red , while the C-terminal peptide is shown in blue . This peptide ends at R607 , a distal C-terminal residue with respect to the one reported for mouse urinary uromodulin ( F588 [Santambrogio et al . , 2008] ) . ( C ) Representative tandem mass-spectrometry ( MS/MS ) spectrum , confirming the identity of the identified C-terminal peptide ( 599VLNLGPITR607 ) of the short uromodulin isoform released by Hpn-/- mice , and table of fragmented ions . ( D ) Schematic representation of mouse uromodulin domain structure as in Figure 1A . The blow-up shows that in the absence of hepsin , the cleavage generating the short uromodulin isoform is abolished and alternative ones at more C-terminal sites ( distal to R607 ) take place . ( E ) Representative Western blot analysis of uromodulin in supernatant ( SN ) and pellet ( P ) fractions from a polymerisation assay performed on urinary samples from Hpn-/- mice or control animals . Urinary uromodulin from control animals is precipitated in the pellet fraction , reflecting full engagement in polymeric structures , while the one from Hpn-/- mice is only detected in the supernatant ( n = 4/group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 02010 . 7554/eLife . 08887 . 021Figure 7—source data 1 . Transcript level of Prss8 in mouse microdissected nephron segments ( Figure 7—figure supplement 2A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 02110 . 7554/eLife . 08887 . 022Figure 7—source data 2 . Quantification of urinary uromodulin secretion in Prss8-/- and control mice ( Figure 7—figure supplement 2C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 02210 . 7554/eLife . 08887 . 023Figure 7—figure supplement 1 . Urinary uromodulin misprocessing in Hpn-/- mice . ( A ) Mass spectrometry ( MS ) sequence coverage ( 51% over the entire protein ) of trypsin-digested mouse uromodulin ( long isoform ) ( UniProt accession Q91X17 ) purified from urine of Hpn-/- mice . Matching peptides are shown in red , while the C-terminal peptide is shown in blue . This peptide ends at K616 , a distal C-terminal residue with respect to the one reported for mouse urinary uromodulin ( F588 [Santambrogio et al . , 2008] ) . ( B ) Representative tandem mass-spectrometry ( MS/MS ) spectrum , confirming the identity of the identified C-terminal peptide ( 608QGVQASVSK616 ) of the long uromodulin isoform released by Hpn-/- mice , and table of fragmented ions . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 02310 . 7554/eLife . 08887 . 024Figure 7—figure supplement 2 . Uromodulin secretion is not affected by lack of prostasin in vivo . ( A ) Transcript level of Prss8 , as assessed by Real-Time qPCR on microdissected nephron segments ( normalised to Gapdh ) . Expression of Prss8 is detected in proximal convoluted tubules ( PCT ) , proximal straight tubules ( PST ) and , to a lesser extent , in thick ascending limb ( TAL ) and collecting ducts ( CD ) . Bars indicate average ± s . e . m . of 3 independent experiments ( Figure 7—source data 1 ) . ( B ) Immunofluorescence analysis of mouse kidney sections shows strong signal of endogenous prostasin on the apical plasma membrane of proximal tubules , and weak signal on the apical plasma membrane of TAL epithelial cells where it co-localises with uromodulin . Scale bar , 20 µm . ( C ) Representative Western blot analysis of urinary uromodulin from control Prss8lox/loxor Prss8-/- mice . Urinary protein loading was normalised to urinary creatinine concentration . Densitometric analysis shows that uromodulin secretion is comparable between Prss8-/- mice and control Prss8lox/loxanimals ( average ± s . d . , n = 5/group , Figure 7—source data 2 ) ( Student’s t test ) . ( D ) Representative Western blot analysis of N-deglycosylated urinary uromodulin secreted by Prss8-/- mice or control animals . An isoform of identical molecular weight , corresponding to the short uromodulin isoform , is detected in urine samples of both genotypes ( n = 5/group ) . ( E ) Mass spectrometry sequence coverage ( 55% over the entire protein ) of AspN-digested mouse uromodulin ( UniProt accession Q91X17 ) purified from urine of Prss8-/- mice . Matching peptides are shown in red , while the C-terminal peptide is shown in blue . This peptide ends at F588 , the same C-terminal residue identified in urinary uromodulin of wild-type mice ( Santambrogio et al . , 2008 ) and control Prss8lox/loxanimals ( data not shown ) . ( F ) Representative MS/MS spectrum confirming the sequence of urinary uromodulin C-terminal peptide ( 573DSTSEQCKPTCSGTRF588 ) in Prss8-/- mice and table of fragmented ions . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 024 A role for hepsin in the activation of prostasin , whose zymogen is unable to auto-activate , was demonstrated in vitro ( Chen et al . , 2010 ) . To assess whether the results observed in Hpn-/- mice could be due to the lack of hepsin-mediated activation of prostasin , we first investigated the expression pattern of endogenous prostasin in mouse kidneys . Confirming data from available transcriptomes , we detected weak prostasin expression in TAL segments where it co-localises with uromodulin on the apical plasma membrane of epithelial cells ( Figure 7—figure supplement 2A , B ) . We then generated an adult nephron-specific Prss8-/- mouse model and analysed urinary uromodulin . Secretion of urinary uromodulin in these mice was comparable to controls ( Figure 7—figure supplement 2C , D ) . Moreover , mass-spectrometry analysis on the urinary protein confirmed that it shares the same C-terminal residue of control animals ( Figure 7—figure supplement 2E , F ) , excluding a role for prostasin in uromodulin processing in vivo .
Collectively , our results identify hepsin as the enzyme responsible for the physiological release of uromodulin in the urine . Hepsin , or TMPRSS1 , is a member of type II transmembrane serine proteases . It is expressed in several tissues and it has been previously implicated in the processing of different substrates; however , the physiological relevance of most of these observations remains to be established ( Chen et al . , 2010; Hsu et al . , 2012; Khandekar and Jagadeeswaran , 2014; Ganesan et al . , 2011 ) . By identifying hepsin as the enzyme responsible for the release of uromodulin in the renal tubule , our results reveal an important biological function of this type II transmembrane serine protease in the renal epithelium . At the same time , they represent a major advancement in understanding the biology of uromodulin , opening new avenues of research on the regulation and possibly modulation of its secretion . This could have impact for diseases such as urolithiasis where the urinary levels of uromodulin may be directly correlated with its protective function . Also , it might be relevant for CKD and hypertension where we demonstrated that UMOD risk variants are associated with increased uromodulin expression and urinary levels ( Trudu et al . , 2013; Olden et al . , 2014 ) . Lack of hepsin does not affect baseline renal function of Hpn-/- mice ( serum creatinine: wild-type 0 . 78 +/- 0 . 04 vs Hpn-/- 0 . 74 +/- 0 . 14 mg/dL , n = 5/group , p=0 . 59; blood urea nitrogen: wild-type 21 . 6 +/- 3 . 29 vs Hpn-/- 20 . 8 +/- 4 . 12 mg/dL , n= 5/group , p=0 . 76 ) . In this light , further studies in Hpn-/- mice with altered uromodulin processing and secretion are warranted to investigate if the mice develop a phenotype upon additional challenges . The results may provide new insight into uromodulin cellular and extracellular functions . It is well recognised that proteolysis plays a central role in uromodulin polymerisation . This step leads to the release of an inhibitory intramolecular interaction that prevents premature protein polymerisation ( Jovine et al . , 2004; Schaeffer et al . , 2009; Han et al . , 2010 ) . Considering the maturation steps of uromodulin along the secretory compartment , hepsin-mediated cleavage likely takes place at the plasma membrane where the two proteins co-localise , allowing the protein to traffic within the cell in a polymerisation-incompetent conformation , and to polymerise only once it is secreted ( Figure 8 ) . Interestingly , the C-terminus of human and mouse urinary uromodulin has been mapped to a phenylalanine residue ( F587 and F588 , respectively ) ( Santambrogio et al . , 2008 ) . Since in vitro studies showed that hepsin preferentially cleaves substrates after basic amino acids ( Herter et al . , 2005; Béliveau et al . , 2009; Owen et al . , 2010 ) , this could result from cleavage at the following arginine residue and subsequent C-terminal trimming . Post-cleavage trimming by carboxypeptidases is not uncommon and it has already been proposed for several secreted proteins , including urokinase ( Günzlerwa et al . , 1982 ) , and ZP subunits ( Kubo et al . , 1999; Boja et al . , 2003; Darie et al . , 2005 ) . 10 . 7554/eLife . 08887 . 025Figure 8 . Model of uromodulin shedding and polymerisation . Uromodulin is exclusively expressed by TAL tubular epithelial cells . The protein enters the secretory pathway and reaches the plasma membrane in a polymerisation-incompetent conformation . This is ensured by the interaction between two hydrophobic patches within and next to the ZP-C subdomain ( Internal Hydrophobic Patch , red circle , and External Hydrophobic Patch , dark green circle ) ( Jovine et al . , 2004; Schaeffer et al . , 2009; Han et al . , 2010 ) . Shedding by hepsin at the uromodulin consensus cleavage site ( red diamond ) , likely occurring at the plasma membrane , releases the hydrophobic interaction , generating polymerisation-competent species that are assembled into polymeric filaments within the tubular lumen . Pink circles indicate N-terminal EGF-like domains , yellow and green cylinders represent ZP-N and ZP-C subdomains . The orientation of uromodulin within polymers is hypothetical . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 025 Through the identification of hepsin as the protease releasing uromodulin in the urine we discover the first protease involved in the physiological cleavage of a ZP domain protein , which is a pre-requisite for polymerisation of this type of molecules . Notably , Hpn-/- mice were reported to have profound hearing loss likely due to inner ear defects , including a deformed and enlarged tectorial membrane ( Guipponi et al . , 2007 ) . This acellular structure , which overlays the organ of Corti , consists of filaments largely composed of ZP domain proteins α- and β-tectorin . The morphology of the tectorial membrane is similarly altered in mice carrying deafness-causing mutations in α-tectorin ( Legan et al . , 2014 ) that are likely affecting protein secretion ( Jovine et al . , 2002 ) . These findings , along with data presented in this manuscript , suggest that inner ear tectorins are also physiological substrates of hepsin . Moreover , release of glycoprotein-2 , a ZP domain protein highly homologous to uromodulin that is predominantly expressed in pancreas , is mediated by an apical protease sensitive to serine protease inhibitors ( Fritz and Lowe , 1996 ) , again suggesting the involvement of hepsin . In light of these observations and given the conserved function of the ZP domain and the high similarity of the consensus cleavage site of ZP domain proteins ( Jovine et al . , 2004 ) , our results are likely to be relevant for other members of this protein family .
Uromodulin vectors expressing wild-type HA- or FLAG-tagged protein , soluble mutant S614X and consensus cleavage site mutant 586RFRS589/586AAAA589 ( 4Ala mutant ) have already been described ( Schaeffer et al . , 2009; Schaeffer et al . , 2012 ) . A gene encoding the elastase-resistant fragment of uromodulin ( efUmod , lacking residues 27–294 ) was generated by overlapping PCRs ( PCR1: residues 1–26; PCR2: residues 295–640 ) from wild-type uromodulin cDNA and inserted in pcDNA 3 . 1 ( + ) ( Life Technologies , Carlsbad , CA ) for expression studies in MDCK cells . To generate soluble efUmod for in vitro cleavage experiments , a DNA fragment encoding uromodulin residues 295–610 was cloned in pIRES-hrGFP II ( Stratagene , Santa Clara , CA ) . Glycosylation site mutation N513Q and a C-terminal 6His-tag were inserted . The consensus cleavage site mutant 586RFRS589 > 586AYAA589 was also generated by site-directed mutagenesis ( QuikChange Lightning Site-Directed mutagenesis kit , Agilent Technologies , Santa Clara , CA ) . Human hepsin and prostasin cDNA ( OriGene , Rockville , MD ) were cloned in pcDNA3 . 1/Zeo ( + ) ( Life Technologies ) . Myc-tag was inserted at the C-terminus of hepsin by phosphorylating and annealing the following primers: forward 5’-CCGGTGAGCAAAAGCTGATTTCTGAGGAGGATCTGA-3’ and reverse 5’- CCGGTCAGATCCTCCTCAGAAATCAGCTTTTGCTCA -3’ . Catalytically inactive prostasin or hepsin were generated by mutating the serine residue of the catalytic triad ( S238A in PRSS8 , S353A in HPN ) using the QuikChange Lightning Site-Directed mutagenesis kit ( Agilent Technologies ) . Primer sequences for generation of all constructs are listed in Table 1 . 10 . 7554/eLife . 08887 . 026Table 1 . Primers ( 5'-3' ) used to generate UMOD and protease constructs . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 026efUMOD-PCR1 T7: TAATACGACTCACTATAGGG Reverse: ACACGTCCCCTCCACGTGGTGATGGTGATGATGACefUMOD-PCR2 Forward: CATCACCATCACCACGTGGAGGGGACGTGTGAGGA BGHrev: TAGAAGGCACAGTCGAGGefUMOD_sol Forward: GGACAGATCTACGTCGACGGGACGTGTGAGGAGTGCAG Reverse: CCGGAATTCCTAGTGATGATGGTGATGATGCTGGACACCTTTCCGTGTGefUMOD_AYAA Forward: CCTACCTGCTCTGGGACCGCATACGCAGCTGGGAGTGTCATAGATCAATCCC Reverse: GGGATTGATCTATGACACTCCCAGCTGCGTATGCGGTCCCAGAGCAGGTAGGefUMOD_N513Q Forward: CACCCAGTAGCCAGGCCACGGACCCCC Reverse: GGGGGTCCGTGGCCTGGCTACTGGGTGPRSS8_S238A Forward: GCCAGGGTGACGCTGGGGGCCCA Reverse: TGGGCCCCCAGCGTCACCCTGGCHPN_S353A Forward: CTGCCAGGGCGACGCCGGTGGTCCCTTT Reverse: AAAGGGACCACCGGCGTCGCCCTGGCAG Human HAI-1 cDNA ( Shimomura et al . , 1997 ) cloned in pcDNA3 . 1 ( + ) was a kind gift of Prof . Clara Camaschella ( San Raffaele Scientific Institute ) . Madin-Darby Canine Kidney ( MDCK ) cells and Human Embryonic Kidney 293 ( HEK293 ) cells were grown in DMEM supplemented with 10% foetal bovine serum ( Euroclone , Pero , Italy ) , 200 U/ml penicillin , 200 μg/ml streptomycin , and 2 mM glutamine at 37°C with 5% CO2 . They were transfected using Lipofectamine 2000 ( Life Technologies ) or Metafectene Pro ( Biontex , Munich , Germany ) respectively . Selection of stably transfected MDCK cells with G418 ( Life Technologies ) was started 24 hr after transfection and was pursued for 1–2 weeks . To obtain electrically tight monolayers ( >100 Ω cm-2 ) , MDCK cells were grown on filters ( Corning , Corning , NY ) for 4 days . Uromodulin secretion in MDCK or HEK293 cells was analysed after 16 hr incubation in Optimem ( Life Technologies ) . MDCK cells stably expressing uromodulin were grown on coverslip . After 24 h , the medium was replaced with Optimem supplemented with 0 . 1% protease inhibitor cocktail ( PIC ) ( P8340 , Sigma-Aldrich , Saint Louis , MO ) , 0 . 1% DMSO , 0 . 1 mM AEBSF , 0 . 08 μM aprotinin , 1 . 4 μM E64 , 4 μM bestatin , 2 μM leupeptin and 1 . 5 μM pepstatin A ( Sigma-Aldrich ) . Immunofluorescence analysis was performed after 24 hr of treatment . Sfold ( http://sfold . wadsworth . org/cgi-bin/index . pl ) and siDESIGN center ( http://www . thermoscientificbio . com/design-center/ ) were used to identify siRNAs targeting canine hepsin and prostasin . Selected oligos ( sense sequence ) are listed in Table 2 . Oligonucleotides for shRNA strategy were designed combining the siRNA sequence ( sense ) followed by a loop ( sequence 5’-3’: TTCAAGAGA ) , the reverse complement of the siRNA sequence ( antisense ) and RNA polymerase III terminator ( sequence 5’-3’: TTTTTTGGAA ) . Oligos were phosphorylated , annealed and inserted in pENTR/pTER+ , a gift from Eric Campeau ( plasmid # 17453 , Addgene , Cambridge , MA ) ( Campeau et al . , 2009 ) . MDCK cells stably expressing uromodulin were co-transfected with pENTR/pTER+ vector and EGFP-expressing vector pcDNA3x ( + ) MyEGFP ( Life Technologies ) . Cells were collected 24 hr after transfection and sorted using MoFlo sorter ( Beckman Coulter , Brea , CA ) . Recovered EGFP-positive cells were maintained in culture for 48 hr . 10 . 7554/eLife . 08887 . 027Table 2 . Oligonucleotide sequences used to generate shRNAs . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 027Target geneTarget position from ATG codonsiRNA oligo ( 5’-3’ ) HPN_siRNA#1122-140CCTTCCTACTCAAGAGTGAHPN_siRNA#21195-1213TGGATCTTCCAGGCCATAAPRSS8_siRNA#1258-276GAAGGAAGACTATGAGGTAPRSS8_siRNA#2593-611TGTACAACATCAACGCTAAControl siRNANATAGTGAGATTCGTTAAGAT A list of human serine proteases was derived from the MEROPS database ( http://merops . sanger . ac . uk/ ) . The expression profile of these enzymes in MDCK and HEK293 cells was analysed in Gene Expression Omnibus DataSets database ( http://www . ncbi . nlm . nih . gov/gds ) using baseline replicates of the following series: GSD3267 ( Hellman et al . , 2008 ) , GSE20193 ( Sharma et al . , 2010 ) , GSE18739 ( Abd Alla et al . , 2010 ) , GSE15575 ( Abd Alla et al . , 2009 ) , GSE1309 ( Zagranichnaya , 2005 ) , GSE1364 ( Jack et al . , 2006 ) , GSE51118 and GDS1852 ( Elkon et al . , 2005 ) . Membrane association of the enzymes was derived from Uniprot database ( http://www . uniprot . org/ ) . Protease expression in TAL cells was analysed by including series GSE13672 ( Yu et al . , 2009 ) and GSE25223 ( Cheval et al . , 2011 ) . Wild-type , Hpn-/- ( Wu et al . , 1998 ) , Prss8lox/lox ( Rubera et al . , 2002 ) and Prss8-/- animals are in C57BL/6J background . Adult nephron-specific Prss8-deficient ( Prss8-/- ) mice were obtained by interbreeding of Prss8lox/lox with Pax8-rtTA/LC1tg/+ ( Traykova-Brauch et al . , 2008 ) . Briefly , 28 day-old Prss8-/- and Prss8lox/lox ( control ) littermates were treated with doxycycline ( 2 mg/ml ) in 2% sucrose in drinking water for 15 days . Genotyping by PCR was performed as previously described ( Malsure et al . , 2014; Ronzaud et al . , 2013 ) . All animal procedures were carried out at the Lerner Research Institute , Cleveland Clinic , USA; at the University of Lausanne , Lausanne , Switzerland and at the University of Zurich , Zurich , Switzerland , according to protocols approved by the Institutional Animal Care and Use Committee at the Lerner Research Institute and by the Swiss Cantonal Veterinary Authority . MDCK and HEK293 cells , or mouse tissues were solubilised in lysis buffer ( 50 mM Tris-HCl , pH 7 . 4 , 150 mM NaCl , 60 mM octyl-β-D-glucopyranoside and 0 . 1% PIC ) . Cell-conditioned medium or urinary proteins were precipitated with acetone and resuspended in PBS . When needed , samples were N-deglycosylated with PNGase F ( New England Biolabs , Ipswich , MA ) . Lysates were quantified with the Bio-Rad Protein Assay ( Bio-Rad , Hercules , CA ) . Precleared lysates of HEK293 cells ( 300 µg ) were loaded onto protein G Sepharose beads ( Sigma-Aldrich ) pre-conjugated with 2 µg of mouse anti-HA antibody ( Covance , Princeton , NJ ) and incubated 16 hr at 4°C . After washes in PBS-triton 0 . 5% , beads were resuspended in Laemmli Buffer . Protein lysates , immunoprecipitated proteins , urinary or secreted proteins were separated on reducing 8% SDS-PAGE gel and transferred onto nitrocellulose membrane ( GE Healthcare , Little Chalfont , United Kingdom ) . Western blotting ( WB ) was performed following standard protocols . Quantification was performed using the gel analysis option of ImageJ software ( http://rsbweb . nih . gov/ij/ ) ( Schneider et al . , 2012 ) . Kidneys for immunofluorescence or microdissection of nephron segments were collected from 2–3-month-old male wild-type control mice . Nephron segment microdissection was performed and validated as previously described ( Glaudemans et al . , 2014 ) . Plasma , urine and kidneys from wild-type and Hpn-/- mice were collected from 2 . 5-month-old male mice . Urine and kidneys from Prss8lox/lox and Prss8-/- were collected from 1 . 5-month-old mice . Animals were housed in a light- and temperature-controlled room with ad libitum access to tap water and standard chow . Urine collections ( 16 h ) were obtained at baseline using individual metabolic cages , after appropriate training . Urinary creatinine quantification was performed using creatinine assay kit ( Cayman Chemical , Ann Arbor , MI ) . Blood samples were collected from the vena cava . Serum creatinine levels were measured by Consolidated Veterinary Diagnostics ( West Sacramento , CA ) . Polymerisation assay on urinary samples was performed as previously described ( Jovine et al . , 2002 ) . Supernatants were precipitated with acetone , while pellets were solubilised in Laemmli Buffer . Immunofluorescence ( IF ) of cells grown on coverslip was carried out as previously described ( Schaeffer et al . , 2009 ) . Cells grown on filters were fixed 20 min in 4% paraformaldehyde ( VWR International , Radnor , PA ) and permeabilised 15 min at 37°C with 0 . 7% gelatin and 0 . 016% saponin . Cells were incubated for 1 hr at 37°C with primary antibody and subsequently with the appropriate AlexaFluor-labeled secondary antibody ( Life Technologies ) . Nuclei were stained with 4’ , 6-diamidino-2-phenylindole ( DAPI , Life Technologies ) and slides were mounted using fluorescent mounting medium ( Dako , Agilent Technologies ) . Slides were visualised under a DM 5000B fluorescence upright microscope ( Leica , Wetzlar , Germany ) or under an UltraVIEW ERS spinning disk confocal microscope ( PerkinElmer , Waltham , MA ) . Quantification of mean polymeric area was performed using ImageJ software , as previously described ( Schaeffer et al . , 2012 ) . Polymers were quantified on 5–9 independent images for each condition taken at 10x magnification . Cell and polymer reconstructions were carried out by collecting stacks of consecutive confocal images at 0 . 5 μm intervals . Extended focus 2D images were obtained by using Volocity 3D Image Analysis Software version 6 . 3 ( PerkinElmer , Waltham , MA ) . Kidneys for immunofluorescence were collected from mice anaesthetised and perfused with 3% paraformaldehyde and snap-frozen in liquid propane . Kidney sections ( 4–5 μm thick ) were reactivated in PBS for 30 min and blocked with 10% goat serum for 30 min . They were incubated overnight at 4°C with with relevant primary antibodies ( anti-hepsin , -prostasin or –uromodulin ) . For co-localisation with uromodulin , sections were blocked again for 30 min and incubated for 2 hr with the anti-uromodulin primary antibody . Slides were incubated with the appropriate AlexaFluor-labelled secondary antibody diluted in PBS containing DAPI and mounted with Prolong Gold anti-fade reagent ( Invitrogen , Life Technologies ) . Slides were visualised with a Leica SP5 Confocal microscope . We used the following primary antibodies: mouse anti-HA ( MMS-101P , 1:1000 for WB and 1:500 for IF , Covance ) , rabbit anti-FLAG ( F7425 , 1:500 for IF , Sigma-Aldrich ) , goat anti-Myc ( NB600-335 , 1:500 for IF , Novus Biologicals , Littleton , CO ) , sheep anti-uromodulin ( T0850 , 1:1000 for WB , US Biological , Salem , MA ) , sheep anti-uromodulin ( K90071C , 1:200 for IF , Meridian Life Science , Cincinnati , OH ) , goat anti-uromodulin ( 55140 , 1:1000 for WB and 1:500 for IF , MP Biomedicals , Santa Ana , CA ) , rabbit anti-hepsin ( 100022 , 1:1000 for WB and 1:50 for IF , Cayman Chemical ) , sheep anti-prostasin ( AF4599 , 1:1000 for WB and 1:200 for IF , R&D System , Minneapolis , MN ) , rabbit anti-prostasin ( kind gift of Prof . Carl Chai , University of Central Florida College of Medicine , FL; 1:200 for IF ) ( Chen , 2006 ) , rabbit anti-HAI-1 ( H-180 , 1:1000 for WB , Santa Cruz Biotechnology , Santa Cruz , CA ) , rabbit anti-PDI ( H-160 , 1:1000 for WB , Santa Cruz Biotechnology ) , mouse anti-E-cadherin ( 610182 , 1:500 for IF , BD Biosciences , San Jose , CA ) , mouse anti-KDEL ( ADI-SPA-827-D , 1:500 , Enzo Life Sciences , Farmingdale , NY ) , mouse anti-GAPDH ( 6C5 , 1:5000 for WB , Santa Cruz Biotechnology ) , mouse anti-β actin ( A2228 , 1:16000 for WB , Sigma-Aldrich ) , mouse anti-α tubulin ( SC-8035 , 1:1000 for WB , Santa Cruz Biotechnology ) and mouse anti-5His ( 34660 , 1:1000 for WB , Qiagen , Venlo , The Netherlands ) . Total RNA was extracted from confluent cells or from mouse kidneys using TRIzol reagent ( Life Technologies ) and reverse-transcribed using iScript kit ( Bio-Rad ) . Expression of target genes was analysed by RT-PCR ( @Taq , Euroclone ) or by Real-Time qPCR on LightCycler 480 ( Roche , Basel , Switzerland ) using qPCR Core kit for SYBR Assay ( Eurogentec , Seraing , Belgium ) . Hpn and Prss8 expression in microdissected nephron segments were assessed by Real-Time qPCR with a CFX96TM Real-Time PCR Detection System ( Bio-Rad ) using iQTM SYBR Green Supermix ( Bio-Rad ) . The relative mRNA expression of genes of interest was calculated following the ΔΔCt method ( Pfaffl , 2001 ) . Primer sequences are listed in Table 3 and Table 4 . 10 . 7554/eLife . 08887 . 028Table 3 . Primers used for gene expression analysis ( RT-PCR ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 028GenePCR product lengthPrimer ( 5’-3’ ) HPN ( Dog / Human ) 353 bpForward: GCTGCGAGGAGATGGGCTTC Reverse: CGGGAAGCAGTGGGCGGCTGPRSS8 ( Dog / Human ) 547 bpForward: CCTGGCAGGTCAGCATCACC Reverse: CCAGAGTCACCCTGGCAGGCGAPDH ( Human ) 314 bpForward: CCACCCAGAAGACTGTGGAT Reverse: GTTGAAGTCAGAGGAGACCACCGAPDH ( Dog ) 289 bpForward: CTCTGGGAAGATGTGGCGTGAC Reverse: GTTGAAGTCACAGGAGACCACC10 . 7554/eLife . 08887 . 029Table 4 . Primers used for gene expression analysis ( Real-Time qPCR ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08887 . 029GenePCR product lengthPrimer ( 5’-3’ ) HPN ( Dog ) 164 bpForward: TGGTCCACCTGTCCAGCCCC Reverse: GACTCGGGCCTCCTGGAGCAHpn ( Mouse ) 152 bpForward: CTGACTGCTGCACATTGCTT Reverse: GGGTCTCGAAAGGGAAGGTAPRSS8 ( Dog ) 156 bpForward: TCCGGACTTGCCTTCTGCGGT Reverse: AGCTGAGAGCACCCACTGCTCAGAPDH ( Dog ) 289 bpForward: CTCTGGGAAGATGTGGCGTGAC Reverse: GTTGAAGTCACAGGAGACCACCPrss8 ( Mouse ) 147bpForward: ATCACCCACTCAAGCTACCG Reverse: AGTACAGTGAAGGCCGTTGGUmod ( Mouse ) 207 bpForward: ATGGACCAGTCCTGTCCTG Reverse: CCGTCTCGCTTCTGTTGAGTHprt1 ( Mouse ) 162 bpForward: ACATTGTGGCCCTCTGTGTG Reverse: TTATGTCCCCCGTTGACTGA Comparisons between groups were performed using two-tailed unpaired t-test or Mann-Whitney test . We expressed continuous measures as the mean ± the standard deviation ( s . d . ) ; averages of measurements as the mean ± the standard error of the mean ( s . e . m . ) . We set significance level to p<0 . 05 . Chinese Hamster Ovary ( CHO ) DG44 cells were grown in α-MEM medium supplemented with 10% foetal bovine serum ( Life Technologies ) and stably transfected with pIRES-hrGFP_efUMOD . Stable cells were cultured in HyQ SFM4CHO-utility medium ( Thermo Scientific , Waltham , MA ) . efUmod secreted by CHO cells was purified by batch IMAC with Ni-NTA Superflow ( Qiagen ) , followed by SEC using a HiLoad 26/60 Superdex 200/HiPrep 26/60 Sephacryl 200 double column system ( GE Healthcare ) equilibrated against 10 mM Tris-HCl pH 8 . 0 at 4°C , 50 mM NaCl , and concentration . Recombinant efUmod ( 25 ng/µl ) was incubated with human serine proteases hepsin ( Enzo Life Sciences , Farmingdale , NY ) or prostasin ( R&D System ) in 50 mM Tris , 20 mM NaCl , pH 7 . 8 at 37°C for 10 hr . Picomolar ratio between protease and efUmod was 1:100 for hepsin and 1:5 for prostasin . MDCK cell conditioned media were centrifuged 30 min at 18 , 000 x g at 4°C , resuspended in ultrapure water and applied to glow-discharged 400 mesh copper grids with a carbon support film . After 5 min the grids were washed with ultrapure water for 2 min and then negatively stained with 2% uranyl acetate . Samples were analysed using a CM120 electron microscope ( Philips , Eindhoven , The Netherlands ) equipped with a LaB6 electron source . Images were recorded on Kodak SO163 electron film ( Eastman Kodak , Rochester , NY ) and digitised using an Epson Perfection 4990 PHOTO flatbed scanner ( Epson , Suwa , Japan ) . MDCK cell-conditioned media were centrifuged at 3 , 000 x g for 5 min at 4°C . Supernatants were concentrated using Amicon Ultra-3K ( Millipore , Billerica , MA ) at 4 , 000 x g at 4°C and depleted of the most abundant proteins using IgY-12 columns ( Genway , San Diego , CA ) . Proteins were precipitated with 60% TCA ( trichloroacetic acid , Sigma-Aldrich ) for 1 hr at -20°C and washed with 90% acetone . Two hundred µg were diluted in 130 µl of a buffer containing 5 M urea ( Sigma-Aldrich ) , 2 M thiourea ( Sigma-Aldrich ) , 2% CHAPS ( Sigma-Aldrich ) , 2% Zwittergent ( GE Healthcare ) , 100 mM DeStreak ( GE Healthcare ) and 0 . 5% IPG buffer pH 3–10 NL ( GE Healthcare ) and loaded on Immobiline Dry strip pH 3–10 NL , 7 cm ( GE Healthcare , total focusing run 50 , 000 Vh ) . IPG strips were sequentially reduced and alkylated prior to the second dimension electrophoresis . SDS-PAGE 8% gels were Coomassie stained and spots of interest were excised and in-gel digested as previously reported ( Shevchenko et al . , 1996 ) . For nLC-MS/MS analysis , peptide mixtures were acidified up to 1% formic acid and analysed on an API QStar PULSAR ( AB-Sciex , Framingham , MA ) mass spectrometer as previously reported ( Magagnotti et al . , 2013 ) . All MS/MS samples were analysed using MASCOT engine ( version 2 . 2 . 07 , Matrix Science , London , United Kingdom ) and X ! Tandem ( within Scaffold software , v . 3 . 6 . 4 , Proteome Software , Portland , OR ) to search the UniProt_CP_Human 2013_05 database . Peptide mass tolerance of 200 ppm and 0 . 6 Da for precursor and fragment ions were selected respectively . Searches were performed with: semi-Lys-C or semi-Asp-N specificity; carbamidomethylation of cysteine as fixed modification and oxidation of methionine as variable modification . Scaffold was used to validate MS/MS based peptide and protein identifications with protein thresholds set to 99% , 2 peptides minimum and peptide thresholds set to 95% minimum . Uromodulin was purified from 1 ml of mouse urinary samples as previously described ( Santambrogio et al . , 2008 ) . PNGase F-treated and alkylated samples were separated on 8% SDS–PAGE gel . Bands of interests were excised from Coomassie stained gels , reduced , alkylated and finally digested overnight with trypsin or Asp-N ( Roche ) ( Santambrogio et al . , 2008 ) . After acidification , peptide mixtures were concentrated and desalted on homemade Stagetips µC18 ( Rappsilber et al . , 2007 ) and injected in a capillary chromatographic system ( EasyLC , Proxeon Biosystems , Thermo Scientific ) . Peptide separations occurred on a homemade fused silica capillary column ( 75 μm i . d . × 25 cm ) , packed with 3 μm ReproSil-Pur C18-AQ ( Dr . Maisch , Ammerbuch-Entringen , Germany ) . A gradient of eluents A ( H2O with 2% acetonitrile , 0 . 5% acetic acid ) and B ( 80% acetonitrile with 0 . 5% acetic acid ) was used to achieve separation , from 4% to 70% B ( in 65 min , 0 . 15 μL/min flow rate ) . The LC system was connected to an LTQ-Orbitrap mass spectrometer ( Thermo Scientific ) equipped with a nanoelectrospray ion source ( Proxeon Biosystems ) . MS and MS/MS spectra were acquired selecting the ten most intense ions per survey spectrum acquired in the orbitrap from m/z 300–1750 with 60 , 000 resolution . Target ions selected for the MS/MS were fragmented in the ion trap and dynamically excluded for 120 sec . Target values were 1 , 000 , 000 for survey scan and 100 , 000 for MS/MS scan . MS/MS spectra were converted into peaklist ( . msm files ) and analysed using MASCOT and Scaffold searching against the UniProt_CP_Mus 2012_10 database . Searches were performed with: semi-Asp-N or semi-trypsin specificity; parent ion tolerance of 5 . 0 ppm and fragment ion mass tolerance of 0 . 60 Da; N-ethylmaleimide or carbamidomethyl of cysteine as fixed modifications; oxidation of methionine , asparagine deamidation and acetylation of the N-terminus of proteins as variable modifications . | Several proteins in humans and other animals contain a region called a 'zona pellucida domain' . This domain enables these proteins to associate with each other and form long filaments . Uromodulin is one such protein that was first identified more than fifty years ago . This protein is known to play a role in human diseases such as hypertension and kidney failure , but uromodulin’s biological purpose still remains elusive . Uromodulin is only made in the kidney and it is the most abundant protein in the urine of healthy individuals . Uromodulin also contains a so-called 'external hydrophobic patch' that must be removed before the zona pellucida domain can start to form filaments . This hydrophobic patch is removed when uromodulin is cut by an unknown enzyme; this cutting releases the rest of the uromodulin protein from the surface of the cells that line the kidney into the urine . Brunati et al . have now tested a panel of candidate enzymes and identified that one called hepsin is able to cut uromodulin . Hepsin is embedded in the cell membrane of the cells that line the kidney . When the level of hepsin was artificially reduced in cells grown in the laboratory , uromodulin remained anchored to the cell surface , its processing was altered and it did not form filaments . Brunati et al . next analysed mice in which the gene encoding hepsin had been deleted . While these animals did not have any major defects in their internal organs , they had much lower levels of uromodulin in their urine . Furthermore , this residual urinary protein was not cut properly and it did not assemble into filaments . Thus , these findings reveal that hepsin is the enzyme that is responsible for releasing uromodulin in the urine . This discovery could be exploited to alter the levels of uromodulin release , and further studies using mice lacking hepsin may also help to understand uromodulin’s biological role . Finally , it will be important to understand if hepsin , or a similar enzyme , is also responsible for the release of other proteins containing the zona pellucida domain . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2015 | The serine protease hepsin mediates urinary secretion and polymerisation of Zona Pellucida domain protein uromodulin |
Tarantula toxins that bind to voltage-sensing domains of voltage-activated ion channels are thought to partition into the membrane and bind to the channel within the bilayer . While no structures of a voltage-sensor toxin bound to a channel have been solved , a structural homolog , psalmotoxin ( PcTx1 ) , was recently crystalized in complex with the extracellular domain of an acid sensing ion channel ( ASIC ) . In the present study we use spectroscopic , biophysical and computational approaches to compare membrane interaction properties and channel binding surfaces of PcTx1 with the voltage-sensor toxin guangxitoxin ( GxTx-1E ) . Our results show that both types of tarantula toxins interact with membranes , but that voltage-sensor toxins partition deeper into the bilayer . In addition , our results suggest that tarantula toxins have evolved a similar concave surface for clamping onto α-helices that is effective in aqueous or lipidic physical environments .
Protein toxins from the venom of poisonous organisms target a wide variety of ion channel proteins , and they have been powerful tools for studying the physiology and molecular mechanisms of these crucial signaling proteins . For example , scorpion toxins that physically block the pore of K+ channels ( MacKinnon and Miller , 1988 ) were initially used to identify pore-forming regions of the channel ( MacKinnon and Miller , 1989 ) , and to determine the stoichiometry of subunits ( MacKinnon , 1991 ) , inactivation gates ( MacKinnon et al . , 1993 ) , as well as accessory subunits ( Morin and Kobertz , 2008 ) . Protein toxins can also interact with ion channels and allosterically modify the process by which the channel opens in response to the stimulus that activates the channel . For example , both tarantula and scorpion toxins can bind to voltage-sensing domains in voltage-activated ion channels and either retard or facilitate activation in response to changes in voltage ( Rogers et al . , 1996; Swartz and MacKinnon , 1997a , 1997b; Cestèle et al . , 1998; Milescu et al . , 2013 ) . This second class of toxins have been valuable tools to explore the mechanism of voltage sensing ( Phillips et al . , 2005; Alabi et al . , 2007; Campos et al . , 2007 , 2008 ) and the roles of individual voltage sensors in the mechanisms of activation and inactivation ( Campos et al . , 2007; Bosmans et al . , 2008; Campos et al . , 2008 ) . Tarantula toxins have also been discovered that activate transient receptor potential ( TRP ) channels ( Bohlen et al . , 2010 ) , and both tarantula toxin and snake toxins have been identified that activate acid-sensing ion channels ( ASIC ) ( Chen et al . , 2006; Salinas et al . , 2006; Bohlen et al . , 2011 ) . For both the TRPV1 channel and ASICs , these toxins were used to stabilize specific conformations for structure determination by cryo-electron microscopy and X-ray diffraction ( Baconguis and Gouaux , 2012; Dawson et al . , 2012; Baconguis et al . , 2014 ) . Tarantula toxins such as hanatoxin and GxTx-1E are thought to target voltage-activated potassium ( Kv ) channels by partitioning into the lipid membrane and binding to their voltage-sensing domains to modify how the channel opens in response to changes in membrane voltage ( Swartz and MacKinnon , 1997a , 1997b; Li-Smerin and Swartz , 2000 , 2001; Lee and MacKinnon , 2004; Wang et al . , 2004; Phillips et al . , 2005; Alabi et al . , 2007; Milescu et al . , 2007 , 2009 , 2013; Schmidt and MacKinnon , 2008; Bosmans et al . , 2011a; Tilley et al . , 2014 ) ( Figure 1A , right ) . The strong interaction of these tarantula toxins with lipid membranes is mediated by a prominent amphipathic surface on the toxins ( Figure 1F ) ( Takahashi et al . , 2000; Lee et al . , 2004 , 2010 ) , which facilitates their entry into the membrane ( Lee and MacKinnon , 2004; Milescu et al . , 2007 , 2009; Jung et al . , 2010; Mihailescu et al . , 2014 ) . Although many aspects of this inhibitory mechanism have been well-studied ( Swartz and MacKinnon , 1997a , 1997b; Li-Smerin and Swartz , 1998 , 2000; Lee et al . , 2003; Wang et al . , 2004; Phillips et al . , 2005; Alabi et al . , 2007; Milescu et al . , 2007 , 2009; Bosmans et al . , 2011a; Milescu et al . , 2013; Mihailescu et al . , 2014; Tilley et al . , 2014 ) , no structures of complexes between tarantula toxins and Kv channels have been solved , limiting our understanding of how tarantula toxins dock onto voltage sensors and influence their conformational dynamics . However , a structurally related tarantula toxin named PcTx1 was recently crystalized in complex with ASIC1a ( Baconguis and Gouaux , 2012; Dawson et al . , 2012 ) ( Figure 1A , left ) , revealing that the toxin clamps onto the thumb helix-5 ( Figure 1G ) and inserts an Arginine finger into the subunit interface to interact with residues involved in proton activation of the channel . PcTx1 modulates the gating properties of ASIC1 channels by enhancing the apparent affinity of these channels for protons , thereby promoting open and desensitized states of the channel ( Escoubas et al . , 2000 , 2003; Chen et al . , 2005; Salinas et al . , 2006; Baconguis and Gouaux , 2012; Dawson et al . , 2012 ) , an allosteric mechanism related to that employed by tarantula toxins targeting Kv channels . However , in contrast to the present model for tarantula toxins binding to Kv channels within the membrane ( Figure 1A , right ) , the structures of PcTx1 bound to ASIC show that the toxin binds to the protein in aqueous solution far from the transmembrane domains of the channel ( Figure 1A left ) . Interestingly , the overall fold of PcTx1 and voltage sensor toxins are quite similar ( Figure 1H ) and PcTx1 has significant amphipathic character ( Figure 1D ) , similar to that seen in the structure of voltage-sensor toxins ( Figure 1E ) . 10 . 7554/eLife . 06774 . 003Figure 1 . Structural comparison between tarantula toxins targeting ASIC and Kv channels . ( A ) Crystal structure of PcTx1 bound to the ASIC1a channel ( left , PDB 4FZ0 ) and of the Kv1 . 2-Kv2 . 1 paddle chimera channel ( right , PDB 2R9R ) viewed from within the membrane . Voltage-sensor paddles and pore domain of the Kv1 . 2-Kv2 . 1 paddle chimera was colored purple and gray , respectively . ( B , C ) NMR structures of PcTx1 ( B , PDB 2KNI ) and GxTx-1E ( C , PDB 2WH9 ) oriented by superimposing the toxin backbones . NT denotes N terminus . ( D , E ) Surface profiles of PcTx1 ( D ) and GxTx-1E ( E ) colored using the Hessa-Heijne hydrophobicity scale in kcal mol−1 ( Hessa et al . , 2005 ) . The orientation of PcTx1 is identical to those shown in B , G and H , whereas the structure of GxTx-1E was rotated 90° about the indicated axis to better illustrate its amphipathic character . ( F ) Sequence alignment of PcTx1 and GxTx-1E . Conserved residues are highlighted yellow . ( G ) Structure of PcTx1 bound to helix-5 of ASIC1a ( PDB 4FZ0 ) . ( H ) NMR Structures of PcTx1 and GxTx-1E superimposed based on their backbone folds . Side chains of conserved residues are shown as sticks . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 003 The structural similarities of tarantula toxins targeting Kv and ASIC channels , when considered in the context of the unrelated structures of Kv and ASIC channels , as well as the distinct environments in which these toxins bind to their target ion channels , prompted us to undertake comparative studies of the membrane interaction properties and channel binding surfaces of the two classes of tarantula toxins . Does the amphipathic surface of PcTx1 enable the toxin to interact with membranes even though it binds to ASIC in solution ? If so , how does the interaction of PcTx1 with membranes compare with that observed for toxins targeting voltage sensors ? Do the two classes of structurally related toxins use distinct surfaces for binding to Kv channels within the membrane and to ASIC channels in solution ? Our results demonstrate that both classes of tarantula toxins interact with membranes , but that PcTx1 interacts more superficially when compare to toxins targeting voltage sensors . In addition , our results suggest that PcTx1 and the voltage-sensor toxins use related surfaces for binding to helices within their target channels even though these events occur in very different physical environments .
Previous studies have established that tarantula toxins targeting ion channel voltage sensors interact favorably with lipid membranes ( Lee and MacKinnon , 2004; Suchyna et al . , 2004; Jung et al . , 2005; Phillips et al . , 2005; Milescu et al . , 2007 , 2009; Swartz , 2007; Jung et al . , 2010; Mihailescu et al . , 2014 ) . We conducted experiments to assess whether PcTx1 partitions into membranes similar to voltage sensor toxins . One common approach to measuring membrane partitioning is Trp fluorescence , where the emission spectra of Trp residues on the toxins are shifted to lower wavelengths ( blue shifted ) when the toxin moves from an aqueous environment to a more constraining and hydrophobic environment of the lipid membrane ( Ladokhin et al . , 2000 ) . Pronounced blue shifts in Trp fluorescence are observed with GxTx-1E when lipid vesicles containing a 1:1 mix of POPC and POPG are added to aqueous solutions of the toxin ( Figure 2A ) , consistent with a previous study ( Milescu et al . , 2009 ) . Plotting of fluorescence intensity of GxTx-1E in the blue shifted region ( 320 nm ) of the emission spectra as a function of lipid concentration and fitting of a partition function ( Figure 2C ) yields a mole fraction Kx of 4 . 6 ± 0 . 8 × 106 , indicating favorable interactions of the toxin with membranes . To investigate whether PcTx1 can interact with membranes , we performed similar experiments and observed detectable blue shifts in the emission spectra of the toxin upon the addition of lipid vesicles ( Figure 2B ) . The blue shifts in the case of PcTx1 were very small ( maximal shift of ∼ 2 nm ) , however , they varied as a function of lipid concentration and saturated at high lipid concentrations , suggestive of an interaction between PcTx1 and lipid membranes . 10 . 7554/eLife . 06774 . 004Figure 2 . Interaction of GxTx-1E ( Nle ) and PcTx1 with lipid vesicles detected with intrinsic Trp fluorescence . ( A , B ) Fluorescence emission spectra of GxTx-1E ( Nle ) and PcTx1 in the absence ( black ) or presence of lipid vesicles composed of a 1:1 mix of POPC:POPG ( blue ) . Lipid concentration was 1 . 0 mM . ( C , D ) Fluorescence intensity at 320 nm plotted as a function of available lipid concentration for GxTx-1E ( Nle ) or PcTx1 . Smooth curve corresponds to a partition function with Kx = ( 4 . 6 ± 0 . 8 ) × 106 and F/F0max = 2 . 3 ± 0 . 05 for GxTx-1E ( Nle ) . ( E , F ) Stern-Volmer plots for acrylamide quenching of GxTx-1E ( Nle ) and PcTx1 in solution ( ∼2 µM , black diamonds ) and in the presence of lipid vesicles ( 1 . 0 mM , blue circles ) . In all cases data points are the mean ± SEM ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 004 Given the small blue shift observed for PcTx1 , we further investigated its interactions with membranes by determining whether lipid vesicles could protect the toxin from quenching by acrylamide , a water soluble quencher of Trp fluorescence . In the case of GxTx-1E in aqueous solution , we observed that the Trp fluorescence of the toxin is quenched by the addition of acrylamide in a concentration-dependent manner ( Figure 2E ) , and fitting of the quenching relationship with a linear function yields a slope ( Stern–Volmer constant; Ksv ) of 37 . 7 M−1 . Addition of lipid vesicles decreased the Ksv to 9 . 1 M−1 ( Figure 2E ) , indicating that membranes can shield the Trp residues in GxTx-1E from the soluble quencher . When these experiments were performed with PcTx1 , we again observed strong quenching by acrylamide in aqueous solution ( Ksv of 29 . 2 M−1 ) and robust protection of quenching by the addition of lipid vesicles ( Ksv of 11 . 2 M−1; Figure 2F ) , confirming that PcTx1 can interact with membranes . Varying the concentration of lipid vesicles in the presence of a fixed concentration of acrylamide also resulted in robust and concentration-dependent protection of acrylamide quenching for both GxTx-1E ( Figure 3A , C ) and PcTx1 ( Figure 3B , D ) . When the increase in maximal fluorescence intensity was plotted against the available lipid concentration and a partition function fit to the data , a Kdx value of ( 1 . 1 ± 0 . 2 ) × 107 was obtained for GxTx-1E , in reasonable agreement with the mole fraction partition coefficient for the toxin determined using blue-shifts in Trp fluorescence . In the case of PcTx1 , we obtained a value of ( 4 . 3 ± 0 . 6 ) × 106 , demonstrating that PcTx1 interacts strongly with membranes . 10 . 7554/eLife . 06774 . 005Figure 3 . Interaction of GxTx-1E ( Nle ) and PcTx1 with lipid vesicles using acrylamide dequenching and quenching with brominated lipids . ( A , B ) Fluorescence emission spectra of GxTx-1E ( Nle ) and PcTx1 in the absence ( black ) or presence of lipid vesicles composed of 1:1 mix of POPC:POPG ( blue ) in a solution containing 0 . 2 M acrylamide . Lipid concentration was 1 . 0 mM . Fluorescence intensity was normalized to that measured for the toxin in control solution in the absence of quencher . ( C , D ) Maximum fluorescence intensity plotted as a function of available lipid concentration for GxTx-1E ( Nle ) and PcTx1 . Smooth curves correspond to dequenching functions with Kdx = ( 1 . 1 ± 0 . 2 ) × 107 and F/F0max = 3 . 5 ± 0 . 02 for GxTx-1E ( Nle ) and Kdx = ( 4 . 3 ± 0 . 6 ) × 106 and F/F0max = 2 . 7 ± 0 . 08 for PcTx1 ( D ) . ( E , F ) Depth-dependent quenching of tryptophan fluorescence by brominated ( diBr ) phosphatidylcholines labeled at different positions on the acyl chain of POPC . Fluorescence emission spectra of GxTx1E ( Nle ) and PcTx1 in the absence ( black ) or presence of vesicles comprised of unlabeled ( gray ) or brominated lipids ( blue ) present at a concentration of 1 . 2 mM . Vesicles comprised of either unlabeled or brominated lipids contained a 1:1 mix of POPC:POPG , and the brominated lipid was POPC . In all cases data points are the mean ± SEM ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 005 The results thus far reveal that both classes of tarantula toxins interact quite favorably with lipid membranes . Although the structure of PcTx1 shows the presence of a clear hydrophobic cluster of residues on one face of the toxin , it is smaller than seen in structures of voltage-sensor toxins ( Figure 1D , E ) . For example , the solvent accessible surface area for the hydrophobic surface of PcTx1 is 380 Å2 , which compares with a value of 703 Å2 for GxTx-1E . In addition , the hydrophobic surfaces on the two toxins are situated on distinct surfaces ( Figure 1D , E ) . These differences between PcTx1 and GxTx-1E raise the possibility that the two classes of toxin might exhibit distinct depths of penetration into membranes . To estimate the depth to which these toxins penetrate lipid membranes , we compared the extent of quenching of Trp fluorescence by bromine atoms attached at different positions along lipid acyl chains ( Ladokhin , 1997 , 1999 , 2014 ) . In previous studies , the analysis of quenching profiles for lipids brominated at three positions along the acyl chain of POPC ( from most superficial to deepest; 6 , 7-diBr , 9 , 10-diBr or 11 , 12-diBr ) have shown the strongest quenching for bromines attached near the middle of the lipid tail ( 9 , 10-diBr ) , and analysis of these profiles suggested that Trp residues in hanatoxin , SGTx ( a close relative of hanatoxin ) and VSTx are positioned 8–9 Å from the center of the bilayer ( Phillips et al . , 2005; Swartz , 2007; Jung et al . , 2010; Mihailescu et al . , 2014 ) . When we performed quenching experiments with these three brominated lipids and GxTx-1E , we observed strong quenching that is similar to what has been reported previously for other voltage-sensor toxins ( Figure 3E ) ( Phillips et al . , 2005; Swartz , 2007; Jung et al . , 2010; Mihailescu et al . , 2014 ) , confirming an intimate interaction of the toxin with the hydrophobic core of the membrane . However , we observed only relatively small differences between the three brominated lipids , precluding further quantitative analysis of the profiles to obtain a specific depth of penetration . We suspect that this unique pattern of quenching by brominated lipids for GxTx-1E might be caused by the presence of three Trp residues in the toxin , each of which will likely be in distinct positions when the toxin interacts with membranes ( see below ) . Interestingly , when we undertook similar experiments with PcTx1 , we observed only very weak quenching by brominated lipids , again with only relatively small differences between the three lipids ( Figure 3F ) . The measurable quenching observed for PcTx1 confirms that this toxin interacts with membranes , however , the much weaker quenching when compared to voltage-sensor toxins indicates that this toxin interacts considerably more superficially with membranes . The X-ray structure of PcTx1 bound to ASIC shows that the toxin interacts primarily with residues on one face of thumb helix-5 , with an interface comprised of both hydrophobic and polar residue interactions ( Figure 1G ) ( Baconguis and Gouaux , 2012; Dawson et al . , 2012 ) . PcTx1 also contains an Arg finger motif composed of R26 , R27 and R28 that is required for activation of ASIC through interactions with channel residues situated in a crevice between subunits . In the case of voltage-sensor toxins such as hanatoxin and GxTx-1E , the S3b helix within the voltage-sensing domains of Kv channels contains the most influential determinants of toxin binding , and these also involve a combination of hydrophobic and polar residue interactions ( Swartz and MacKinnon , 1997b; Li-Smerin and Swartz , 2000 , 2001; Alabi et al . , 2007; Milescu et al . , 2009 ) . Given the common helical targets of PcTx1 and voltage-sensor toxins , and the finding that PcTx1 can interact with membranes , we wondered whether it might be possible to transfer sensitivity to PcTx1 into a Kv channel by exchanging the S3b helix with helix-5 from ASIC . We did not attempt to transfer channel residues interacting with the arginine finger motif because these residues and helix-5 are noncontiguous in the ASIC sequence . PcTx1 has nM affinity for ASIC1a ( Chen et al . , 2006; Dawson et al . , 2012 ) , but even at very high concentrations ( 5 µM ) has no effect on the Kv2 . 1 channel ( Figure 4 , Figure 4—figure supplement 1B ) . Previous chimera studies have shown that the voltage-sensor paddle , a helix-turn-helix motif comprised of S3b and S4 helices , can be transplanted between different channels that contain S1–S4 domains without disrupting voltage sensor function ( Alabi et al . , 2007; Bosmans et al . , 2008 , 2011b; Kalia and Swartz , 2013 ) , making us optimistic that helix-5 transfer might be possible . 10 . 7554/eLife . 06774 . 006Figure 4 . Determination of apparent affinity for mutants of GxTx-1E . ( A ) Families of macroscopic ionic currents elicited by test depolarizations before ( black ) and after ( red ) addition of 454 nM GxTx-1E ( Nle ) for an oocyte expressing the Kv2 . 1 channel . ( B ) G-V relations obtained from tail current amplitudes before and after addition of toxin . ( C ) Time course of inhibition of the Kv2 . 1 channel by 454 nM GxTx-1E ( Nle ) . Steady-state currents were measured at the end of test depolarizations to −10 mV , elicited every 10 s . Holding voltage was −100 mV . ( D ) Concentration-dependence for inhibition of the Kv2 . 1 channel by GxTx-1E ( Nle ) and representative toxin mutants . Fraction unbound ( Fu ) was measured using steady state current values before and after addition of toxin from time course of toxin inhibition as illustrated in C using weak depolarizations ( −20 mV to +10 mV ) . Data points are mean + SEM for 3 to 13 cells at each concentration . E1A and K10A are examples of mutants that produce weak perturbations in Kd ( ∆∆G < 1 kcal mol−1 ) , while W8A and W28A are examples of mutants that have larger perturbations ( ∆∆G > 2 kcal mol−1 ) . See Table 1 for all Kd and ∆∆G values . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 00610 . 7554/eLife . 06774 . 007Figure 4—figure supplement 1 . Influence of PcTx1 on chimeras between Kv2 . 1 and ASIC1a . ( A ) Sequence alignment of S3 helices in Kv2 . 1 ( blue ) replaced with helix-5 from ASIC1a ( red ) in the chimeric constructs . ( B–G ) Normalized G-V relations in the absence ( black circles ) or presence ( red circles ) of 5 µM PcTx1 . Conductance was measured using tail currents . In all cases data points are the mean ± SEM ( n = 3 ) . ( H–Q ) Models of PcTx1 bound to Kv2 . 1-ASIC1a chimeras viewed from a transmembrane ( left; H–L ) or extracellular ( right; M–Q ) perspective . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 007 We chose to make chimeras between ASIC1a and Kv2 . 1 by transferring helix-5 of ASIC1a into the Kv2 . 1 , a channel that is inhibited by hanatoxin and GxTx-1E ( Swartz and MacKinnon , 1997a; Lee et al . , 2003 , 2010; Phillips et al . , 2005; Milescu et al . , 2009; Tilley et al . , 2014 ) and has been a successful recipient of transplanted peptide toxin binding sites ( Alabi et al . , 2007; Bosmans et al . , 2008 , 2011b; Kalia and Swartz , 2013 ) . We generated a total of five chimeras where nine residues of helix-5 replaced nine residues in the S3b helix in five different frames starting at the N-terminal end of the S3b helix ( Figure 4—figure supplement 1A ) . All five chimeras gave rise to robust voltage-activated K+ currents and displayed voltage-activation relations that were shifted to more depolarized voltages when compared to the wild-type Kv2 . 1 channel ( Figure 4—figure supplement 1C–G ) , consistent with disruption of the interactions between S3b and S4 helices that have been shown to stabilize voltage sensors in an activated state ( Xu et al . , 2013 ) . Although these relatively radical chimeras were all functional , extracellular application of high concentrations of PcTx1 had no discernible effect on the activity of any of them ( Figure 4—figure supplement 1C–G ) . We also constructed models of PcTx1 bound to the five chimeras by transplanting PcTx1 bound to helix-5 from the complex structure ( Baconguis and Gouaux , 2012 ) into the X-ray structure of the Kv1 . 2/2 . 1 paddle chimera ( Long et al . , 2007 ) ( Figure 4—figure supplement 1H–Q ) . As expected , several of the models predict steric clashes between PcTx1 and neighboring helices within the S1–S4 domain , whereas in others the toxin could be readily accommodated when bound to helix-5 . Even in this latter category of models , however , PcTx1 would need to partition relatively deeply into the membrane in order to bind to helix-5 in the chimeras , a possibility that seems unlikely given the relatively superficial interaction of the toxin we inferred with brominated lipids . To further compare how PcTx1 and GxTx-1E interact with helices in ASIC and Kv channels , we alanine-scanned GxTx-1E to identify residues that are most critical for interaction with the Kv2 . 1 channel . We generated Ala mutants for 29 out of 36 residues of GxTx-1E ( excluding one native Ala and 6 Cys residues ) using solid phase chemical synthesis , and successfully folded all but one in mg quantities ( see ‘Materials and methods’ ) . To determine the apparent affinity of toxin mutants for the resting state of Kv2 . 1 , we obtained conductance ( G ) -voltage ( V ) relationships from negative holding voltages in the absence and presence of different concentrations of each mutant ( Figure 4A , B ) and estimated fractional occupancy of the channel from the fractional inhibition for relatively weak depolarizations ( −20 to 0 mV ) where toxin bound channels do not open , as previously described ( Swartz and MacKinnon , 1997a , 1997b; Li-Smerin and Swartz , 2000; Phillips et al . , 2005; Milescu et al . , 2009 ) . We obtained estimates of the apparent affinity for each mutant , and observed a wide range of different phenotypes , from small changes in affinity to almost 400-fold weakening of toxin affinity for Kv2 . 1 ( Figure 4D; Table 1 ) . When perturbation energies ( ∆∆G values ) for each mutant were mapped onto the NMR structure of GxTx-1E , a well-defined surface of the toxin was identified where mutations having dramatic effects on the apparent affinity of the toxin cluster together ( Figure 5A ) . This surface contains many aromatic residues ( F7 , W8 , W9 , Y22 and W28 ) , one aliphatic hydrophobic residue ( L30 ) and several polar residues ( K10 , K15 , S25 , K27 and N32 ) . The mutations producing the most dramatic perturbations on this surface were for mutations of aromatic residues , whereas those for the polar residues were relatively modest ( Figure 5A; Table 1 ) . 10 . 7554/eLife . 06774 . 008Table 1 . Affinities and perturbation energies for mutants of GxTx-1EDOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 008ToxinKd ( nM ) Kdmut/Kdwt∆∆G ( kcal mol−1 ) Gxtx-1E ( Nle ) 224 ± 251 . 0–E1A414 ± 381 . 80 . 36G2A297 ± 501 . 30 . 17E3A125 ± 620 . 6−0 . 35G5A13 , 652 ± 32060 . 92 . 44G6A379 ± 511 . 70 . 31F7A89 , 033 ± 9413397 . 43 . 55W8A9641 ± 50443 . 02 . 23W9A5837 ± 185026 . 11 . 93K10A984 ± 904 . 40 . 88G12A21 , 755 ± 108697 . 12 . 71S13A545 ± 562 . 40 . 53G14A461 ± 892 . 10 . 43K15A823 ± 1273 . 70 . 77P16A22 , 967 ± 2812102 . 52 . 75P20A652 ± 732 . 90 . 63K21A440 ± 822 . 00 . 40Y22A35 , 337 ± 7464157 . 73 . 00V23A516 ± 602 . 30 . 50S25A1169 ± 645 . 20 . 98P26A813 ± 993 . 60 . 76K27A763 ± 493 . 40 . 73W28A44 , 298 ± 2471197 . 73 . 14L30A2474 ± 27811 . 01 . 42N32A1228 ± 1665 . 51 . 01F33A491 ± 182 . 20 . 47P34A348 ± 601 . 60 . 26Nle35A306 ± 19 . 81 . 40 . 19P36A250 ± 221 . 10 . 0710 . 7554/eLife . 06774 . 009Figure 5 . Comparison of residues on PcTx1 and GxTx-1E involved in receptor binding . ( A ) GxTx-1E residues colored by perturbation in apparent affinity ( ΔΔGo in kcal mol−1 ) . Several side-chains with the largest perturbation energies are labeled . P16 was not colored since its mutation to Ala likely perturbs the structure of the toxin . ( B ) PcTx1 residues colored by changes in solvent accessible surface area ( ΔSASA in Å2 ) of the toxin upon binding to ASIC1a . SASA was calculated using UCSF-Chimera with a 1 . 4 Å sphere for the structure of PcTx1 alone or in complex with ASIC 1a ( PDB 4FZ0 ) . Residues with the largest ΔSASA ( W7 , W24 , R27 and R28 ) are labeled . ( C ) ΔSASA plotted against PcTx1 residue number . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 00910 . 7554/eLife . 06774 . 010Figure 5—figure supplement 1 . Interaction of tryptophan mutants of GxTx-1E ( Nle ) with lipid vesicles detected with intrinsic Trp fluorescence . ( A–D ) Fluorescence emission spectra of GxTx-1E ( Nle ) , W8A , W9A and W28A in the absence ( black ) or presence of lipid vesicles composed of a 1:1 mix of POPC:POPG ( blue ) . The lipid concentration was 1 . 0 mM . ( E–H ) Fluorescence intensity at 320 nm plotted as a function of available lipid concentration for GxTx-1E ( Nle ) , W8A , W9A and W28A . Smooth curves correspond to partition functions with Kx and F/F0max value of ( 4 . 6 ± 0 . 8 ) × 106 and 2 . 3 ± 0 . 05 for GxTx-1E ( Nle ) , ( 1 . 3 ± 0 . 2 ) × 106 and 1 . 8 ± 0 . 01 for W8A , ( 7 . 6 ± 1 . 5 ) × 106 and 2 . 5 ± 0 . 05 for W9A and ( 5 . 5 ± 0 . 9 ) × 106 and 1 . 5 ± 0 . 01 for W28A , respectively . ( I ) λmax values for Trp fluorescence of GxTx-1E ( Nle ) and three Trp mutants in control solution ( white bar ) or in the presence of lipid vesicles at a concentration of 1 mM . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 01010 . 7554/eLife . 06774 . 011Figure 5—figure supplement 2 . Depth-dependent quenching of tryptophan fluorescence by brominated ( diBr ) phosphatidylcholines . ( A–D ) Fluorescence emission spectra of GxTx-1E ( Nle ) W8A , W9A and W28A in the absence ( black ) or presence of unlabeled ( gray ) or brominated ( blue ) lipids present at a concentration of 1 . 2 mM . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 01110 . 7554/eLife . 06774 . 012Figure 5—figure supplement 3 . Comparison of the effects of GxTx-1E and SGTx1 mutation on apparent affinity and membrane partitioning . ( A ) Sequence alignment between GxTx-1E and SGTx1 . Conserved residues are highlighted in yellow . ( B ) GxTx-1E residues colored by perturbation in apparent affinity ( ΔΔGo ) . ( C and D ) SGTx1 residues colored by perturbation in apparent affinity ( ΔΔGo; C ) and membrane partitioning ( ΔΔGp; D ) . Apparent affinity values and membrane partitioning values are from Wang et al . ( 2004 ) and Milescu et al . ( 2007 ) , respectively . ( E ) SGTx1 residues colored by ΔΔGo − ΔΔGp values to highlight positions where mutations have much larger effects on apparent affinity compared to membrane partitioning . In all right panels structures were rotated 180° about the indicated axes in B and C . GxTx-1E and SGTx1 structures were oriented by superimposing backbone folds of the two toxins . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 012 Because voltage-sensor toxins bind to Kv channels within the membrane , it is possible that some of the mutations we have identified influence the apparent affinity of the toxin by weakening the interaction of the toxin with membranes rather than exclusively weakening the protein–protein interaction . To verify that the surface we identified is involved in binding to Kv2 . 1 , we investigated membrane interactions for each of the three Trp residues ( W8 , W9 and W28 ) , positions where mutations have some of the largest perturbation energies . Even though the apparent affinities of these mutants for Kv2 . 1 ranged from 26 to almost 200-fold lower ( Figure 4D; Table 1 ) , we obtained Kx values that were within a factor of two for the wild-type toxin ( Figure 5—figure supplement 1 ) . We also measured quenching of the three Trp mutants by brominated lipids and found that each exhibited robust quenching ( Figure 5—figure supplement 2 ) . It is notable that the blue-shifts observed with membrane partitioning vary for the three Trp mutants , as does the extent of quenching by brominated lipids ( Figure 5—figure supplement 2 ) , suggesting that the three Trp residues reside in non-equivalent environments within the membrane . Taken together , these collective data with GxTx-1E mutants strongly support the notion that the active surface identified by mutagenesis ( Figure 5A ) is involved in binding to the voltage-sensing domain . To compare this surface of GxTx-1E with the binding surface of PcTx1 , we calculated the change in solvent accessible surface area ( ΔSASA ) of PcTx1 upon binding to ASIC1a , and compared it to the binding surface identified in GxTx-1E ( Figure 5A–C ) . The comparison reveals an extensive overlap between the active surfaces of the two toxins responsible for binding to ASIC1a and Kv2 . 1 channels , indicating that the two classes of toxins use related surfaces for interacting with helices within structurally unrelated ion channel proteins .
The objective of this study was to compare membrane interactions and channel binding surfaces of the voltage-sensor toxin GxTx-1E and the ASIC toxin PcTx1 . The recent X-ray structures of complexes between PcTx1 and ASIC1a indicates that the tarantula toxin binds to the channel in aqueous solution ( Baconguis and Gouaux , 2012; Dawson et al . , 2012 ) , whereas a growing body of work supports that idea that voltage-sensor toxins bind to voltage sensors within the lipid bilayer ( Lee and MacKinnon , 2004; Wang et al . , 2004; Phillips et al . , 2005; Alabi et al . , 2007; Milescu et al . , 2007 , 2009 , 2013; Schmidt and MacKinnon , 2008; Bosmans et al . , 2011a ) . The results of our experiments demonstrate that both classes of toxins interact with membranes ( Figures 2 , 3 ) , a finding that is consistent with the amphipathic character evident in the solution structures of these toxins ( Figure 1 ) . However , our results also show that the interaction of PcTx1 with membranes is different than what has been observed for voltage-sensor toxins . Fluorescence spectroscopy results show that Trp residues on PcTx1 exhibit only very small blue shifts when interacting with membranes ( Figure 2 ) , and that the Trp fluorescence of the toxin is only weakly quenched by bromine atoms on the acyl chains of lipids ( Figure 3 ) . In contrast , voltage-sensor toxins typically exhibit strong blue shifts in Trp fluorescence and this is robustly quenched by brominated lipids ( Figures 2 , 3 ) . Although these properties have been consistently observed for other voltage-sensor toxins ( Jung et al . , 2005; Phillips et al . , 2005; Milescu et al . , 2007 , 2009; Swartz , 2007; Jung et al . , 2010; Mihailescu et al . , 2014 ) , the stark differences observed here for PcTx1 demonstrate that tarantula toxins of similar structure do not necessarily have similar membrane interacting properties . The membrane interactions of voltage-sensor toxins are unique in how far they can penetrate the lipid bilayer , further supporting the idea that this class of toxins interacts with voltage sensors within the lipid bilayer . In addition to providing a valuable comparison with voltage-sensor toxins , the interaction between PcTx1 and membranes may be interesting in its own right . Although a complex structure of PcTx1 and ASIC shows that the PcTx1 binding site is located about 45 Å above from the lipid bilayer , superficial adsorption of the toxin to the membrane would be predicted to enhance the rate of complex formation due to a reduction in dimensionality of diffusion ( Axelrod and Wang , 1994 ) . Interestingly , membrane interactions have also been reported for charybdotoxin ( Ben-Tal et al . , 1997 ) , a strongly cationic scorpion toxin that blocks the external pore of some potassium channels ( Miller , 1995; Banerjee et al . , 2013 ) . The interaction of charybdotoxin with membranes is heavily influenced by electrostatic interactions , suggesting that the toxin interacts relatively superficially , similar to what we observed with PcTx1 . The superficial nature of the interaction between either PcTx1 or charybdotoxin and lipid membranes could be important because it would allow these toxins to departition rapidly so that it can bind to its receptor above the membrane , or possibly to explore the protein surface as an extension of the membrane surface . The voltage-sensor toxins , in contrast , must bind to helices embedded within the membrane , and thus must venture deeper into the lipid bilayer . The inhibitor cystine knot ( ICK ) fold is found in many toxins and is thought to provide a structurally stable scaffold for presentation of surfaces to bind to receptors and ion channel proteins ( Pallaghy et al . , 1994; Norton and Pallaghy , 1998 ) , and in principle many unique surfaces on these toxins could be used for this purpose . However , our results show that GxTx-1E uses a surface to bind to the S3b helix within voltage sensors that is similar to that which PcTx1 uses to bind to helix-5 in ASIC ( Figure 5 ) , even though the former binds within a membrane environment and the latter in aqueous solution . The underlying rationale for using common surfaces is likely related to the fact that both GxTx-1E and PcTx1 bind directly to α-helices . The X-ray structure of PcTx1 in complex with ASIC shows that the toxin clamps onto a solvent-exposed lateral surface of helix-5 ( Figure 1 ) ( Baconguis and Gouaux , 2012; Dawson et al . , 2012 ) , using a somewhat concave surface of the toxin formed by residues in loop 1 and loop 3 , including both hydrophobic ( W7 , W24 , F30 , and P35 ) and polar residues ( K8 , K25 , R26 , S29 and T37 ) . The equivalent surface of GxTx-1E is also formed by residues in loop 1 and loop 3 ( Figure 5 ) , which contain a similar variety of hydrophobic ( F7 , W8 , W9 , Y22 , W28 and L30 ) and polar residues ( K10 , K15 , S25 , K27 and N32 ) where mutations weaken the apparent affinity of the toxin . This surface is also similar to that identified on SGTx1 ( Wang et al . , 2004; Milescu et al . , 2007 ) , a tarantula toxin related to hanatoxin and GxTx-1E that inhibits the Kv2 . 1 channel . Although the apparent affinity of the most critical mutants on SGTx1 could not be accurately determined because the affinity of the toxin is quite low ( Wang et al . , 2004 ) , residues having disproportionately large effects on the apparent affinity of the toxin compared to the strength of membrane partitioning cluster together on a surface that is similar to what we identified here for GxTx-1E ( Figure 5—figure supplement 3 ) ( Milescu et al . , 2007 ) . Given the similar binding faces for α-helices on 3 peptide toxins that share little commonality besides an tri-cystine scaffold , we suggest that this face of the ICK fold forms a shape complement ( Jones , 2012 ) for α-helices that may be exploited by many other peptide ligands . To further explore the structure of GxTx-1E bound to voltage-sensors , we constructed models of GxTx-1E bound to the X-ray structure of the activated/open Kv1 . 2/Kv2 . 1 paddle chimera by superimposing GxTx-1E with PcTx1 bound to ASIC and then transposing the toxin-helix interactions into the paddle chimera . We constructed five models where the toxin-helix interaction occurs in different registers in the S3b helix , similar to what we did in making the helix-5/paddle chimera models , and identified three models where the toxin could be accommodated without steric clashes with the channel ( Figure 6 , Figure 6—figure supplements 1 , 2 ) . Although all three models position critical residues on GxTx-1E where they can interact with the S3b helix , our preferred model ( Figure 6; Figure 6—figure supplements 1E , 2E ) positions the largest number of critical residues in GxTx-1E where they can interact with residues in the S3b helix . In addition , in this model the toxin interacts with many of the most critical residues within the S3b helix that have been identified using mutagenesis ( Figure 6 ) ( Swartz and MacKinnon , 1997b; Li-Smerin and Swartz , 2000 , 2001; Milescu et al . , 2009 ) , with a placement of the toxin within the membrane that is compatible with the depth of partitioning inferred using spectroscopic and neutron diffraction approaches ( Phillips et al . , 2005; Jung et al . , 2010; Mihailescu et al . , 2014 ) . Voltage-sensor toxins like GxTx-1E bind tighter to and stabilize the resting/closed state of Kv channels , however , they can remain bound when the voltage sensors activate and the channel opens ( Phillips et al . , 2005; Tilley et al . , 2014 ) . Thus , although our preferred model of GxTx-1E bound to the activated/open Kv1 . 2/Kv2 . 1 paddle chimera represents the toxin bound to a lower affinity activated state of the Kv channel , it nicely illustrates how GxTx-1E could interact with the S3b helix in ways that closely resemble what is seen in the X-ray structure of PcTx1 bound to ASIC . In the future it will be interesting to refine models of the complex by attempting to form bridges between toxin and channel , and ultimately solving structures of these fascinating complexes . 10 . 7554/eLife . 06774 . 013Figure 6 . Model of GxTx-1E bound to the S3b helix of the Kv1 . 2/2 . 1 paddle chimera . The preferred model of GxTx-1E bound to the X-ray structure of the voltage-sensing domain of the Kv1 . 2/2 . 1 paddle chimera generated as described in ‘Materials and methods’ . The model maintains a relatively superficial position for the toxin relative to the membrane ( see Figure 6—figure supplements 1 , 2 ) . The complex is viewed from a transmembrane perspective ( top ) and from the extracellular perspective ( bottom ) . GxTx-1E side-chains are colored by free energy perturbation scale ( ΔΔGo in kcal mol−1 ) . F271 and E274 of the channel , where in Kv2 . 1 , mutation to alanine decreases apparent Kd values for GxTx-1E ( Milescu et al . , 2009 ) , are represented by blue colored side-chains . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 01310 . 7554/eLife . 06774 . 014Figure 6—figure supplement 1 . Model structures of GxTx-1E bound to the voltage-sensing domain of a Kv channel . ( A ) Side chains of GxTx-1E ( left; PDB 2WH9 ) and the voltage-sensing domain of the Kv1 . 2/Kv2 . 1 paddle chimera ( right; PDB 2R9R ) were colored by free energy perturbation values obtained by alanine scanning mutagenesis of the toxin ( Table 1 ) or channel ( Milescu et al . , 2009 ) . ( B–E ) Models of GxTx-1E bound to the paddle chimera using Rosetta . Side chains are colored by Rosetta binding energy calculated from in slico alanine scans for GxTx-1E ( left ) and voltage-sensing domain ( right ) . Models were generated as described in ‘Materials and methods’ . The five models shown are for five registers for placement of the toxin relative to the S3b helix and each differ in register by one residue ( 100° rotation of the toxin relative to the helical axis and a 5 . 4 Å translation ) . The model in B is the deepest positioning of the toxin and the model in F is the most superficial . All models are viewed from a transmembrane perspective . Residues beyond a distance threshold of 12 Å from any binding partner are colored gray . Steric clashes between the toxin and channel occur in the models in C and D . The model in E corresponds to that shown in Figure 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 01410 . 7554/eLife . 06774 . 015Figure 6—figure supplement 2 . Effects of GxTx-1E and Kv2 . 1 mutations on the energetics of toxin-channel interactions . ( A ) Perturbations in the free energy of GxTx-1E interactions with the Kv2 . 1 channel assessed from alanine scans of the toxin ( Table 1 ) and channel ( Milescu et al . , 2009 ) . Gray circles in the results of the alanine scan of GxTx-1E represent mutation of either Proline or Glycine , which likely influence the local structure of the toxin . ( B–F ) Calculated perturbations in ROSETTA binding energy between GxTx-1E and Kv2 . 1 resulting from alanine scans of the toxin and S3b helix of the S1–S4 domain for the five models illustrated in Figure 6—figure supplement 1 . Energies were not calculated for residues beyond a distance threshold of 12 Å from any binding partner . DOI: http://dx . doi . org/10 . 7554/eLife . 06774 . 015
Peptide synthesis was conducted on an Applied Biosystems model 433A peptide synthesizer as previously described ( Lee et al . , 2010 ) . The linear precursor was synthesized using solid-phase methodology with Fmoc chemistry , starting from Fmoc-Pro-Wang resin ( Fmoc-Thr-Wang resin for PcTx1 ) using a variety of blocking groups for the protection of the amino acids . All the GxTx1E peptides were synthesized with norleucine ( Nle ) in place of Met35 to avoid oxidation of the toxin . A 4 mol excess of Fmoc amino acid , DIC , and Cl-HOBt were used for amino acid activation . After trifluoroacetic acid cleavage , the crude linear peptide was extracted with 2 M acetic acid and diluted to a final concentration of 25 µM in a solution containing 0 . 1 M ammonium acetate , 1 M guanidine-HCl and 2 . 5 mM reduced/0 . 25 mM oxidized glutathione ( pH 8 . 0 with aqueous NH4OH ) and stirred slowly at 4°C for 3 days . The folding reaction was monitored with RP-HPLC and the crude oxidized product was purified by preparative RP-HPLC with a C18 silica column . The purity of the synthetic PcTx1 , GxTx1E and the Ala mutants was confirmed by analytical RP-HPLC and MALDI-TOF-MS measurements . The concentration of the toxin was determined by measuring absorbance at 280 nm using calculated extinction coefficients ( Gill and von Hippel , 1989 ) . Chimeras were generated using sequential PCR with the Kv2 . 1∆7 channel ( Li-Smerin and Swartz , 1998 ) as template . Primers encoding the 9-amino acid sequence of helix-5 from ASIC1a were utilized for overlap PCR . The DNA sequence of all constructs and mutants was confirmed by automated DNA sequencing . Complementary RNA ( cRNA ) was synthesized using T7 polymerase ( mMessage mMachine kit; Ambion ) after linearizing the DNA with appropriate restriction enzymes . Oocytes from Xenopus laevis were removed surgically and incubated with agitation for 1 hr in a solution containing ( in mM ) 82 . 5 NaCl , 2 . 5 KCl , 1 MgCl2 , 5 HEPES , pH 7 . 6 ( with NaOH ) , and collagenase ( 2 mg/ml; Worthington Biochemical ) . Defolliculated oocytes were injected with cRNA encoding the Kv2 . 1 channel constructs and incubated at 17°C in a solution containing ( in mM ) 96 NaCl , 2 KCl , 1 MgCl2 , 1 . 8 CaCl2 , 5 HEPES , pH 7 . 6 with NaOH , and gentamicin ( 50 μg/ml , GIBCO-BRL ) , for 16–72 hr before electrophysiological recording . Oocyte membrane voltage was controlled using an OC-725C oocyte clamp ( Warner Instruments ) . Data were filtered at 1 kHz and digitized at 10 kHz . Microelectrode resistances were 0 . 1–1 MΩ when filled with 3 M KCl . The external recording solution for all recordings with GxTx1E and its analogues contained 50 mM RbCl , 50 mM NaCl , 10 mM HEPES ( increased to 20 mM at 10 μM and higher concentrations of F7A toxin mutant ) , 1 mM MgCl2 and 0 . 3 mM CaCl2 at pH 7 . 6 ( with NaOH ) . All experiments were performed at room temperature ( ∼22°C ) . Voltage-activations relationships in the absence and presence of different concentrations of the synthesized toxin were recorded . The inhibitory activity of the synthetic toxins were examined against the Kv2 . 1 channel by taking steady-state current measurements following weak depolarizations at 10-s intervals before , during and after addition of toxin to the recording chamber until the effect of the toxin on the channel reached equilibrium . At weak depolarizations , leak and endogenous current contributions to steady-state current levels were accounted for by using clampfit to subtract current before activation of the channels at those voltages . Experiments with PcTx1 were carried out with external solution containing 50 mM KCl . We examined the toxin occupancy of closed or resting channels at negative holding voltages where open probability is low and we estimated the fraction of unbound channels ( Fu ) using depolarizations that are too weak to open toxin-bound channels , as previously described ( Swartz and MacKinnon , 1997a , 1997b; Li-Smerin and Swartz , 2000; Lee et al . , 2003; Phillips et al . , 2005; Alabi et al . , 2007; Bosmans et al . , 2008; Milescu et al . , 2009 ) . For all toxins , we calculated the ratio of steady-state currents ( I/I0 ) at weak depolarizations before addition ( I0 ) of different concentrations of toxin and after the toxin effects reached equilibrium ( I ) . Voltage-activation relations before and after addition of toxin indicated that the values of I/I0 at these weak depolarizations were in the plateau phase where toxin-bound channels do not open . The apparent Kd for each toxin was calculated assuming four independent toxin-binding sites per channel , with single occupancy being sufficient to inhibit opening in response to weak depolarizations: Kd = ( ( 1/ ( 1 − Fu1/4 ) ) − 1 ) [Toxin] . Large unilamellar vesicles ( LUVs ) were prepared by drying the phospholipids from a chloroform solution under a nitrogen stream . The dried lipid film was rehydrated in buffer containing 10 mM HEPES , 1 mM EDTA , pH 7 . 6 ( HEB ) . The resulting dispersions were extruded through 100 nm pore size polycarbonate filters ( Millipore Corp . ) till the solution became clear . All fluorescence measurements were performed in quartz cuvettes with 1 cm path length . Fluorescence spectra ( averaging three spectra ) were recorded between 300 and 450 nm ( 5 nm band pass , 0° polarizer ) using an excitation wavelength of 280 nm ( 5 nm band pass , 90° polarizer ) ( SPEX FluoroMax 3 spectrofluorometer ) and corrected for vesicle scattering . For lipid partitioning experiments , LUVs composed of a mix of 1:1 molar ratios of POPC ( 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine ) and POPG ( 1-palmitoyl-2-oleoyl-sn-glycero-3-[phospho-rac- ( 1-glycerol ) ] ) were added to a solution of toxin ( ∼2 μM final concentration ) , maintained at 25°C with continuous stirring in a total volume of 2 ml . Mole-fraction partitioning coefficients ( Kx ) were calculated by measuring the fluorescence intensity ( F ) at 320 nm and normalizing to the zero lipid fluorescence intensity ( F0 ) ( Ladokhin et al . , 2000; Milescu et al . , 2007 ) . Kx was calculated based on the best fits of the following equation to the data: F/F0 ( L ) = 1 + ( F/F0max − 1 ) Kx[L]/ ( [W] + Kx[L] ) , where F/F0 ( L ) is the change in fluorescence intensity for a given concentration of lipid , F/F0max is the maximum fluorescence increase at high lipid concentrations , [L] is the average available lipid concentration ( 60% of total lipid concentration ) and [W] is the molar concentration of water ( 55 . 3 M ) . Protection of tryptophan fluorescence from acrylamide quenching was examined in the absence and presence of lipids ( toxin: lipid = 1:500 ) by titration with increasing concentrations of acrylamide ( Milescu et al . , 2007 ) . The Stern–Volmer quenching constant ( KSV ) was calculated based on the best fits of the following equation to the data: F0/F = 1 + KSV[Q] , where F0 and F are fluorescence of the toxin in the absence and presence of acrylamide , and [Q] is the concentration of acrylamide . Dequenching of tryptophan fluorescence by addition of lipids to a pre-quenched solution of toxin and acrylamide is a sensitive and reliable method to determine the strength of lipid interactions for molecules which do not exhibit large spectroscopic changes in response to lipid binding ( Posokhov et al . , 2007 ) . A solution containing ∼ 2 µM toxin and 0 . 3 M acrylamide was stirred continuously with addition of increasing concentrations of LUVs composed of 1:1 mixture of POPC and POPG . Fluorescence spectra ( averaging three spectra ) were recorded between 300 and 450 nm ( 5 nm band pass , 0° polarizer ) using an excitation wavelength of 280 nm ( 5 nm band pass , 90° polarizer ) ( SPEX FluoroMax 3 spectrofluorometer ) and corrected for vesicle scattering . For calculating mole–fraction partitioning coefficients ( Kdx ) , the maximal fluorescence intensity ( F at λmax ) was measured and normalized to the zero lipid fluorescence intensity ( F0 ) . Kdx was calculated based on the best fits of the following equation to the data: F/F0 ( L ) = 1 + ( F/F0max − 1 ) Kdx[L]/ ( [W] + Kdx[L] ) , where F/F0 ( L ) is the change in maximum fluorescence intensity for a given concentration of lipid , F/F0max is the maximum fluorescence increase at highest lipid concentration , [L] is the average available lipid concentration ( 60% of total lipid concentration ) and [W] is the molar concentration of water ( 55 . 3 M ) . For depth-dependent fluorescence quenching experiments with brominated lipids , LUVs contained a 1:1 molar ratio of POPG and POPC 6 , 7- , 9 , 10- , or 11 , 12-1-palmitoyl-2-stearoyl ( dibromo ) -sn-glycero-3-phosphocholine ( diBr; Avanti Polar Lipids , Alabaster , AL ) . Fluorescence spectra were recorded between 300 and 450 nm using an excitation wavelength of 280 nm , corrected for vesicle scattering and normalized to the zero lipid fluorescence intensity . Structural modeling of GxTx-1E—Kv1 . 2-Kv2 . 1 chimera complexes was performed using ROSETTA ( Gray et al . , 2003; Rohl et al . , 2004; Bradley et al . , 2005; Wang et al . , 2007; Raman et al . , 2009; Tyka et al . , 2011; Conway et al . , 2014 ) . For each Kv2 . 1-ASIC1a chimera expressed , residues of S3b segment from the crystal structure of a Kv1 . 2-Kv2 . 1 channel ( PDB 2R9R ) ( Long et al . , 2007 ) were aligned in register with residues of ASIC1a helix-5 from the ASIC1a:PcTx1 crystal structure ( PDB 4FZ0 ) ( Baconguis and Gouaux , 2012 ) . Then a GxTx-1E NMR structure ( PDB 2WH9 ) ( Lee et al . , 2010 ) was aligned with PcTx1 structure to minimize RMSD between cystine α-carbons of the two toxins . Each GxTx-1E—Kv1 . 2-Kv2 . 1 chimera complex model was the relaxed using ROSETTA to identify the lowest energy binding models . ROSETTA binding energies at the toxin—channel interface ( ΔΔG ) were calculated from in silico alanine scans as described previously ( Kortemme and Baker , 2002; Kortemme et al . , 2004 ) . | Venomous animals like tarantulas or scorpions inject their prey with toxins to disable them . Some of these toxins work by altering the activity of proteins called ion channels , which are found within membranes in cells . These channels can allow potassium ions and/or other ions to pass through the membrane and have many important roles . For example , ion channels are involved in heart muscle contraction and allow information to travel between brain cells . Researchers have used some of the toxins as tools to study how ion channel proteins operate . For example , a toxin produced by tarantulas called psalmotoxin binds to a type of ion channel called the acid-sensing ion channels ( ASIC ) . Researchers have previously been able to visualize the three-dimensional structure of psalmotoxin attached to ASIC , which revealed how the toxin binds to and alters the activity of the ion channel . Another tarantula toxin called guangxitoxin is very similar to psalmotoxin , but it binds to a different type of potassium ion channel . It is thought that guangxitoxin binds to a site on these ion channels that is deep within the membrane , but it is not clear how this works . Here , Gupta , Zamanian , Bae et al . compare some of the structural and chemical properties of the two toxins . The experiments show that both toxins interact with the membrane to enable them to bind with their target ion channels . However , guangxitoxin moves deeper into the interior of the membrane . Also , Gupta , Zamanian , Bae et al . 's findings suggest that both toxins use a similar surface that curves inwards to clamp onto their target ion channels . This common structural feature will be useful for designing experiments to visualize the three-dimensional structure of guangxitoxin bound to a potassium ion channel . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | Tarantula toxins use common surfaces for interacting with Kv and ASIC ion channels |
Genome sequences diverge more rapidly in mammals than in other animal lineages , such as birds or insects . However , the effect of this rapid divergence on transcriptional evolution remains unclear . Recent reports have indicated a faster divergence of transcription factor binding in mammals than in insects , but others found the reverse for mRNA expression . Here , we show that these conflicting interpretations resulted from differing methodologies . We performed an integrated analysis of transcriptional network evolution by examining mRNA expression , transcription factor binding and cis-regulatory motifs across >25 animal species , including mammals , birds and insects . Strikingly , we found that transcriptional networks evolve at a common rate across the three animal lineages . Furthermore , differences in rates of genome divergence were greatly reduced when restricting comparisons to chromatin-accessible sequences . The evolution of transcription is thus decoupled from the global rate of genome sequence evolution , suggesting that a small fraction of the genome regulates transcription .
A long-standing question in biology is what fraction of the genome regulates transcription ( ENCODE Project Consortium , 2012; Graur et al . , 2013; Niu and Jiang , 2013; Kellis et al . , 2014 ) . Recent studies of chromatin structure have implicated half of the human genome in regulatory interactions ( ENCODE Project Consortium , 2012 ) . Comparative genomic studies , however , have shown that less than 10% of the human genome is evolutionarily conserved ( Siepel et al . , 2005 ) , suggesting that many of the experimentally-detected interactions are not functional ( Graur et al . , 2013 ) . Recent studies have measured the association between sequence changes and changes in transcript levels , epigenetic modifications or binding of transcription factors regulating specific gene sets ( gene-specific transcription factors , GSTF ) ( Cookson et al . , 2009; McVicker et al . , 2013; Kasowski et al . , 2010; 2013; Heinz et al . , 2013; Villar et al . , 2014; Wong et al . , 2015; Brem et al . , 2002; Chan et al . , 2009; Shibata et al . , 2012 ) . These experiments demonstrated that genomic sequences can influence transcription even in the absence of evolutionary conservation . For instance , some repetitive elements previously thought to be 'junk' DNA have been shown to effectively regulate gene expression ( Rebollo et al . , 2012 ) . The rapid evolution of repetitive and other rapidly-evolving sequences could cause pervasive rewiring of transcriptional networks through creation and destruction of regulatory motifs ( Villar et al . , 2014 ) . Such rapid transcriptional evolution would set mammals apart from other metazoans like birds or insects , whose genomes contain far fewer repetitive elements ( Taft et al . , 2007 ) and tend to be more constrained ( Siepel et al . , 2005; Zhang et al . , 2014 ) . A few studies have attempted to assess whether transcriptional networks evolve more rapidly in mammals than in insects from the fruit fly genus Drosophila . These studies have reached conflicting conclusions . When examining the evolution of GSTF binding , chromatin immuno-precipitation ( ChIP ) studies in mammalian livers have generally described faster divergence rates than similar studies in fly embryos ( Villar et al . , 2014; Stefflova et al . , 2013 ) . However , divergence rates were estimated with different analytical methods in the different ChIP studies ( Supplementary file 1 ) ( Villar et al . , 2014; Bardet et al . , 2012 ) . Another study found that gene expression levels may diverge at a slower rate in mammals than in flies , by comparing genome-wide correlations of mRNA abundances estimated by RNA sequencing ( RNA-seq ) for mammals but by a mixture of technologies for flies including microarrays ( Coolon et al . , 2014 ) . Although the inconsistencies between these conclusions may indicate that the evolution of transcriptional networks is fundamentally different in mammals and insects , they may also reflect a sensitivity of evolutionary rate estimations to technical methodology . Here , we jointly examined the evolution of gene expression levels and the underlying genome-wide changes in GSTF binding and cis-regulatory sequences using consistent methodologies both within and across various animal lineages .
We assembled a comparative genomics platform encompassing >40 publicly available datasets spanning >25 organisms representative of the Mammalia ( mammals ) , Aves ( birds ) and Insecta ( insects ) phylogenetic classes ( Figure 1—figure supplement 1 ) . We designed a statistical framework to objectively compare the rates of divergence of these various datasets across lineages . In brief , an exponential model describing evolutionary divergence under a common , lineage-naïve rate was evaluated against a lineage-aware model , accounting for both statistical significance and effect size ( Figure 1; Materials and methods ) . We assessed the power of this statistical framework using simulations and found that it could detect differences in divergence rates with high sensitivity ( Materials and methods; Figure 1—figure supplement 2 ) . As a baseline , we first performed a comparative analysis of the evolution of genome sequences . We randomly sampled genomic segments from designated reference genomes: Mus musculus domesticus ( C57BL/6 ) for mammals , Gallus gallus for birds and Drosophila melanogaster for insects . The rates at which genomic segments that retained homologs with the other species within each lineage accumulate nucleotide substitutions were then estimated and compared using our statistical framework . Segments retaining homologs displayed high sequence conservation across all three lineages , although our framework detected a slightly but significantly faster divergence in insects than in mammals or birds ( P<0 . 05; Figure 2—figure supplement 1 ) . Next , we compared the rates at which randomly sampled genomic segments lost homology with the other species within each lineage . We observed a much larger difference in evolutionary rates across lineages using this measure ( P<0 . 05; Figure 2; Figure 2—figure supplement 2 ) . For instance , after 100 million years ( Myrs ) of evolution , only ~30% of mammalian segments retained homology , whereas >60% of bird and insect segments did . These findings recapitulated previous observations according to which genome sequences are less constrained in mammals than in insects ( Siepel et al . , 2005 ) or birds ( Zhang et al . , 2014 ) . 10 . 7554/eLife . 11615 . 003Figure 1 . Statistical framework to evaluate differences in evolutionary rates of change . Throughout this study we frequently evaluated whether the rate of evolutionary divergence of a given layer of transcriptional regulation differs between lineages . Our approach is equivalent to asking: if the lineage labels were hidden , would one be able to tell that the data points correspond to several lineages or would they seem equally likely to belong to a common distribution ? ( a , b ) Depict an example of statistically indistinguishable evolutionary rates . Without lineage labels ( a ) , the similarity data are modeled by an exponential decay as well as with lineage labels ( b ) . Adding lineage labels does not significantly improve the fit . ( c , d ) Depict an example of statistically different evolutionary rates . Adding lineage labels ( d ) significantly improves the fit of an exponential decay model over unlabeled data ( c ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 00310 . 7554/eLife . 11615 . 004Figure 1—figure supplement 1 . Comparative genomics platform for studying transcriptional network evolution across three metazoan lineages . The phylogenetic trees indicate the evolutionary relationships between the organisms included in this study . The trees are not drawn to scale . The numbers at each branch split represent the evolutionary distance in Myrs separating the organisms at the end of the lower branch from the reference species , whose names are bolded . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 00410 . 7554/eLife . 11615 . 005Figure 1—figure supplement 2 . Power of the statistical framework to evaluate differences in evolutionary rates . ( a-c ) Depict the sensitivity of our statistical framework to detect differences in 1000 simulations . The initial rates of one clade was fixed to either -0 . 007 ( a ) , -0 . 005 ( b ) or -0 . 003 ( c ) , and data were simulated by modeling an exponential decay where samples were drawn from a Gaussian distribution with standard deviation fixed to 0 . 5% or 5% . The second clade’s rate was modeled according to the absolute difference in rates with steps shown in the x axis and sampled similarly as for the first clade . Simulated data were used as input to our statistical framework and the frequency of detecting a significant difference is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 00510 . 7554/eLife . 11615 . 006Figure 2 . Genomic sequences evolve more rapidly in mammals than in birds and insects . The evolutionary retention of 5000 randomly sampled 75 bp segments was averaged over 20 trials . Organisms compared to reference species are as follows: M . musculus domesticus ( AJ ) , M . musculus castaneus , M . spretus , rat , guinea pig , rabbit , human , chimpanzee and dog for Mammalia; turkey , zebrafinch and flycatcher for Aves; D . simulans , D . erecta , D . yakuba , D . ananassae , D . pseudoobscura , D . virilis , D . willistoni and D . grimshawi for Insecta . Colored dashed lines: lineage-specific exponential fits , here and in all following displays . The trends were robust to variations in segment length and sequence similarity filters ( Figure 2—figure supplement 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 00610 . 7554/eLife . 11615 . 007Figure 2—figure supplement 1 . Genomic segments retaining homologs are highly conserved at the nucleotide level . The genomic segments found to retain homologs in Figure 2 were aligned to their homologous regions . The average nucleotide identity of the corresponding ungapped alignment is shown here . Evolutionary rates are slightly but significantly different among lineages ( P < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 00710 . 7554/eLife . 11615 . 008Figure 2—figure supplement 2 . Retention of genomic segments is robust to changes in sampled region size and sequence identity threshold . ( a ) Following the same procedure as in Figure 2 but varying the segment length to 150 bp and ( b ) increasing the LiftOver minMatch parameter to 0 . 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 008 We then studied the evolution of gene expression levels , using exclusively RNA-seq datasets . In mammals and birds , these datasets were generated from adult livers; in insects , they were from whole bodies of adult female fruit flies ( Materials and methods; Figure 3—source data 1 ) . After determining expression levels for each gene in each species using a common data processing pipeline , we correlated the expression levels of genes in the reference species with the expression levels of their one-to-one orthologs in all other species within the same lineage ( Materials and methods ) . We found that correlations of gene expression levels decreased over time at similar rates that were statistically indistinguishable: a lineage-naïve model describing the evolution of gene expression levels under a common rate fitted the data as well as a lineage-aware model ( Figure 3 ) . This result was robust to changes in correlation metrics or inclusion/exclusion of poorly expressed genes ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 11615 . 009Figure 3 . Gene expression levels diverge at a common rate in mammals , birds and insects . Gene expression levels were derived independently from two RNA-seq experiments for each reference species and then correlated against each other and against gene expression levels derived from individual experiments in other species within the same lineage . Black dashed line: lineage-naïve exponential fit of all the data , without differentiating the lineages , here and in all following displays . Organisms compared to reference species are as follows: M . musculus castaneus , M . spretus , rat , human and gorilla for Mammalia; turkey , duck and flycatcher for Aves; D . simulans , D . yakuba , D . ananassae and D . pseudoobscura for Insecta . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 00910 . 7554/eLife . 11615 . 010Figure 3—source data 1 . Accession numbers used in RNA-seq analyses . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 01010 . 7554/eLife . 11615 . 011Figure 3—figure supplement 1 . The common evolutionary rate of gene expression levels presented in Figure 3 is robust to changes in correlation metrics or expression threshold . ( a , b ) The gene expression levels used in Figure 3 were correlated with alternative correlation metrics , Kendalls τ ( a ) and Pearson’s r ( b ) . The resulting evolutionary rates remained statistically indistinguishable . ( c ) The gene expression level of all genes were analyzed rather than excluding the values below 5 TPM as was done in Figure 3 . The resulting evolutionary rates remained statistically indistinguishable . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 011 Several lines of evidence suggest that gene expression levels can remain relatively stable even as the genomic locations bound by GSTFs change rapidly over time ( Wong et al . , 2015; Chan et al . , 2009; Paris et al . , 2013 ) . Therefore , we next examined the evolution of GSTF binding patterns . We considered all GSTFs that were profiled using ChIP followed by massively parallel sequencing ( ChIP-seq ) in at least three related species , where separate ChIPs were performed per species . GSTFs meeting these requirements were Twist and Giant in fruit fly embryos , and CEBPA , FOXA1 and HNF4A in mammalian livers ( Materials and methods; Figure 4—source data 1; Supplementary file 1 ) . We aimed to measure cross-species similarity in GSTF occupancy with a unified analytical method across all of these datasets . Despite the widespread use of ChIP-seq , there is no consensus on the appropriate analytical method ( Wilbanks and Facciotti , 2010 ) . ChIP-seq analysis pipelines typically discretize continuous occupancy profiles into a set of occupied segments ( 'peaks' ) , but this step requires choosing a signal processing algorithm ( a peak caller ) and associated parameters ( Figure 4a ) . Further comparison of occupied segments across species requires additional analytical choices ( Figure 4a ) , some of which can strongly influence downstream findings ( Bardet et al . , 2012 ) . 10 . 7554/eLife . 11615 . 012Figure 4 . GSTF occupancy diverges at a common rate in mammals and insects . ( a ) Estimating shared GSTF occupancy across species requires multiple parameter choices . This diagram summarizes the main steps involved in comparing GSTF-occupied segments across species , showing a representative sample of choices at each step ( steps represented by purple shapes , specific choices by the first letter bolded ) . The detailed methods and specific choices illustrated here and implemented in panels b–d are described in Materials and methods . ( b , c ) An example of different analytical choices leading to different results despite starting from the same underlying data . Organisms compared to reference species are as follows: M . musculus domesticus ( AJ ) , M . musculus castaneus , M . spretus , rat , human and dog for Mammalia; D . simulans , D . erecta , D . yakuba , D . ananassae and D . pseudoobscura for Insecta . ( d ) Most combinations of choices yield indistinguishable evolutionary rates of GSTF binding patterns across lineages . The comparison of Twist and CEBPA is enlarged to show the color labels corresponding to the statistical interpretation regarding relative evolutionary rates . ( e ) A genome-wide comparison of GSTF occupancy profiles at single-nucleotide resolution shows indistinguishable evolutionary rates for CEBPA , HNF4A and FOXA1 in mammals , and for Twist and Giant in insects . PCC: Pearson correlation coefficient . ( f ) CTCF occupancy is highly conserved in mammals . Transparent points and lines are identical as in panel e . Hexagons correspond to cross-species correlations of CTCF occupancy at single-nucleotide resolution . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 01210 . 7554/eLife . 11615 . 013Figure 4—source data 1 . Accession numbers used in ChIP-seq analyses . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 01310 . 7554/eLife . 11615 . 014Figure 4—source data 2 . 648 segment-based ChIP analyses . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 01410 . 7554/eLife . 11615 . 015Figure 4—figure supplement 1 . Measured GSTF binding divergence rates are influenced by parameter choices . ( a ) The pie chart on the left shows the frequency at which either the mammalian or insect GSTF was found to evolve faster for 648 comparisons using different combinations of analytical choices . The majority of comparions showed indistinguishable rates . The stacked histograms indicate how often a parameter was used when a difference in divergence rate was detected . For instance , 106/150 cases where Mammalia factors decayed significantly faster used MACS2 as a peak caller , whereas 49/59 cases of Insecta GSTF decaying faster used SPP . Interestingly , asymmetric quality filters showed an enrichment for Mammalia GSTFs decaying faster ( 84/150 ) as well as for Insecta GSTFs decaying faster ( 33/59 ) . ( b ) Boxplots showing general influence of parameter choices on individual decay rates of Insecta ( top ) and Mammalia ( bottom ) . Only instances when a significant fit was detected are considered . For example , for mammalian GSTFs , stringent quality filters yielded slightly faster decay rates than asymmetric or lenient quality filters . Summary of all parameter choices and the results are shown in Figure 4—source data 2 . These parameters are further elaborated in Figure 4 and Methods . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 015 To explore the impact of these choices , we processed all ChIP-seq data using systematic combinations of parameters representative of , and expanding from , previous studies ( Supplementary file 1 ) ( Landt et al . , 2012 ) . In total , we executed 108 analytical pipelines to compare divergence rates across 6 pairs of GSTFs ( 2 in insects each compared with 3 in mammals ) , the occupancy profiles of which were examined in 3–7 species per lineage ( Materials and methods ) . The values of the estimated rates varied greatly from one combination of parameters to the next ( Figure 4b , c ) . However , in the majority of cases ( 56–78% over the 6 comparisons ) , GSTF binding patterns diverged at statistically indistinguishable rates in mammals and insects ( Figure 4d; Figure 4—source data 2 ) . Although the computed divergence rates were sensitive to technical methodology ( Figure 4—figure supplement 1 ) , for a given method the results were generally similar across lineages for all of the five GSTFs investigated . To substantiate these findings , we devised a method to compare genome-wide occupancy profiles at single-nucleotide resolution without discretization . We correlated occupancy profiles between pairs of species across all nucleotides where genomes aligned , after accounting for the differences in sequencing depth , read length and fragment size across datasets ( Materials and methods ) . Again , we found indistinguishable divergence rates , regardless of which GSTF or lineage was examined ( Figure 4e ) . After 100 Myrs of evolution , the correlation of GSTF occupancy profiles was 0 . 10 in mammals and 0 . 13 in insects . As a control , we also applied this method to CTCF , a pleiotropic DNA-binding protein that acts as chromatin insulator and looping factor ( Ohlsson et al . , 2010 ) . In mammals , patterns of DNA occupancy have been shown to be more conserved for CTCF than for GSTFs using unified analytical methods ( Schmidt et al . , 2012 ) . In contrast , CTCF DNA occupancy was shown to diverge rapidly in insects , perhaps due to the existence of other insulator proteins ( Villar et al . , 2014; Ni et al . , 2012 ) . Our analysis successfully recapitulated this difference ( Figure 4f ) , demonstrating that the common evolutionary rate observed among GSTFs ( Figure 4e ) was not an artifact of our method for profile correlation . The similarity of divergence rates observed across lineages for gene expression levels ( Figure 3 ) and GSTF binding patterns ( Figure 4 ) was unexpected given the rapid evolution of genomic sequences in mammals relative to insects ( Siepel et al . , 2005 ) or birds ( Zhang et al . , 2014 ) ( Figure 2 ) . We therefore further examined these trends at the level of cis-regulatory sequences . First , we considered the DNA sequence motifs thought to be specifically recognized by the mammalian and insect GSTFs included in the previous ChIP-seq analysis ( Figure 4 ) . We identified locations with significant matches to these motifs throughout the genomes of the reference species and estimated how frequently these loci retained the same motifs relative to background expectations ( Materials and methods ) . We found similar , indistinguishable retention rates in mammals and insects ( Figure 5a ) . Next , we studied the evolution of a broader set of motifs corresponding to GSTFs shared between M . musculus and D . melanogaster . We found that these motifs were retained at similar rates across lineages relative to background expectations in 8 out of 12 cases ( one example shown in Figure 5b; all other cases in Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 11615 . 016Figure 5 . Regulatory sequences diverge at similar rates across lineages . ( a ) The motifs for CEBPA , HNF4A and FOXA1 in mammals and for Twist and Giant in insects are retained at a common rate . Organisms compared to reference species are the same as Figure 4 . ( b ) The motifs for GSTFs shared in mammals and insects are retained at common rates . One example is shown here for the motifs corresponding to PHO ( FBgn0002521 ) in D . melanogaster and YY1 ( ENSMUSG00000021264 ) in M . musculus , which are orthologous GSTFs . Eleven other cases of motif evolution for shared GSTFs conserved in mammals and insects are shown in Figure 5—figure supplement 1 . Organisms compared to reference species are as in Figure 4 . ( c , d ) Chromatin-accessible sequences are retained at similar rates in mammals , birds and insects . Analyses were performed as in Figure 2 , limiting sampling to the inaccessible ( c ) and accessible ( d ) portions of the intergenic regions . Organisms compared to reference species are the same as Figure 2 . The trends were robust to variations in segment length and sequence similarity filters ( Figure 5—figure supplement 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 01610 . 7554/eLife . 11615 . 017Figure 5—figure supplement 1 . Conservation of cis-regulatory motifs for GSTFs conserved across insects and mammals . ( a ) Seven shared GSTFs whose motifs are retained at indistinguishable rates in mammals and insects . The evolution of these motifs behave similarly to that of the example shown in Figure 5 . ( b ) Four conserved GSTFs whose motifs are retained at different rates in mammals and insects . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 01710 . 7554/eLife . 11615 . 018Figure 5—figure supplement 2 . Retention of intergenic genomic segments in accessible and inaccessible chromatin is robust to changes in sampled region size and sequence identity threshold . ( a , b ) Repeating the procedure used in Figure 5c and Figure 5d sampling segments of different length ( 150 bp ) and ( c , d ) increasing the LiftOver minMatch ( 0 . 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11615 . 018 Most active cis-regulatory sequences are located in genomic regions with accessible chromatin ( Hesselberth et al . , 2009 ) . A recent study showed that chromatin-accessible sequences were significantly more conserved between human and mouse than expected by chance ( Yue et al . , 2014 ) . We expanded this analysis to a wide range of species by using chromatin-accessible sequences identified by DNAse I hypersensitivity in M . musculus livers , D . melanogaster embryos and G . gallus MSB-1 cells ( Materials and methods ) . We performed the segment sampling procedure described previously ( Figure 2 ) , after excluding genes and promoter regions since they typically are highly conserved ( Materials and methods ) . Whereas inaccessible segments lost homology much faster in mammals than in insects and birds ( P<0 . 05; Figure 5c ) , accessible segments retained homologs at more similar rates in the three lineages ( Figure 5d; Figure 5—figure supplement 2 ) . We still detected statistically significant differences across lineages ( P<0 . 05 ) , but the effect sizes were considerably smaller than for inaccessible segments . For instance , ~60% of segments retained homology after 100 Myrs in birds and insects , independently of accessibility , whereas ~50% of chromatin-accessible segments and only ~20% of inaccessible segments did so in mammals .
To our knowledge , the analyses presented here represent the most comprehensive study conducted to date on the evolution of transcriptional networks across animal lineages . By applying unified analytical methods to data from different lineages , we were able to glean novel insights into the evolution of transcription in animals . We observed that gene expression levels , GSTF binding patterns , regulatory motifs and chromatin-accessible sequences each diverged at rates that were similar across mammals , birds and insects . These unexpected results reconcile previously conflicting findings ( Villar et al . , 2014; Coolon et al . , 2014 ) , highlighting the importance of unified study methodologies and providing evidence for a common evolutionary rate in metazoan transcriptional networks . Most functional genomics studies have focused on humans and model organisms such as D . melanogaster or M . musculus , which are distantly related to each other . However , data on closely related species , like those which we collected in this study , are needed to investigate the dynamics of molecular network evolution . Unfortunately , such data remain scarce , leading to important limitations of our work . We only investigated three lineages and six to twelve organisms per lineage with non-uniform coverage over evolutionary time . In addition , we only examined a small number of tissues for each lineage and a total of five GSTFs ( none in birds ) . The generalizability of our observations thus remains to be further evaluated as more data becomes available . Despite these limitations , our finding that transcriptional networks evolve at a common rate per year across animal lineages was strikingly robust across data layers . The underlying mechanisms responsible for this concordance of evolutionary rates are unclear . Mammals , birds and insects exhibit wide differences in the features that are traditionally associated with evolutionary rates , such as generation times and breeding sizes . Populations with small breeding sizes , such as mammals , are thought to be more prone to genetic drift ( Ohta , 1992 ) . This theory accounts for the abundance of repetitive elements and the rapid evolution of genomic sequences in mammals relative to insects , which have much larger breeding sizes . If the same theoretical principles also governed the evolution of transcriptional networks , we would have expected that transcription would evolve more rapidly in mammals than in insects . Instead , our results show that the evolution of transcriptional networks , whether slow ( e . g . transcript levels ) or fast ( e . g . GSTF binding ) , is decoupled from the lineage-specific features that govern genome sequence evolution . One potential model could be that repetitive and rapidly-evolving sequences , which make up the majority of the mammalian genome ( Siepel et al . , 2005; Taft et al . , 2007 ) , play a negligible role in the global regulation of gene expression . Rather , chromatin-accessible regions may represent the only portion of the mammalian genome that effectively regulates transcription . We observed that chromatin-accessible regions diverge much more slowly than other non-coding sequences in mammals , consistent with previous findings ( Yue et al . , 2014 ) . These differences in divergence rates , however , were not found in birds and insects . As a result , chromatin-accessible regions in mammals are conserved at levels similar to those in birds and insects , in contrast to the genome as a whole . According to this model , the similar rates of evolution of chromatin-accessible sequences would constrain the dynamics of transcriptional evolution to be similar across lineages . The regulatory potential of repetitive and other rapidly-evolving elements could be rendered functionally inconsequential by silencing , or could be concentrated on controlling the expression of genetic elements that we did not investigate , such as non-coding RNAs or species-specific genes ( Sundaram et al . , 2014 ) . An alternative model could be that the sequences that control transcriptional regulation in birds and insects evolve particularly rapidly within otherwise stable genomes . In these organisms , transcriptional networks would diverge under the action of natural selection , through specific single nucleotide substitutions resulting in rapid compensatory turnover ( He et al . , 2011a ) . In mammals , transcriptional networks would diverge in a largely neutral fashion driven for instance by transposable elements ( Sundaram et al . , 2014 ) . In this case , similar rates of transcriptional divergence across lineages would arise through very different evolutionary processes . Importantly , none of the aforementioned models account for the differences in generation times between lineages . Evolutionary changes occurring based on chronological time and not generation time has also been observed for many protein-coding sequences . Observations such as these led to the molecular clock theory ( Kumar , 2005 ) . The mechanisms through which environmental forces entrain these chronological evolutionary clocks remain to be elucidated ( Kumar , 2005 ) .
We downloaded genome sequences for organisms belonging to three metazoan lineages: mammals , birds and insects . The mammalian and insect genome sequences were downloaded from the UCSC Genome Bioinformatics website ( Rosenbloom et al . , 2015 ) : mm9 for Mus musculus domesticus , rn5 for Rattus norvegicus and hg19 for Homo sapiens; dm3 for Drosophila melanogaster , droSim1 for Drosophila simulans , droEre2 for Drosophila erecta , droYak2 for Drosophila yakuba , droAna3 for Drosophila ananassae and dp4 for Drosophila pseudoobscura . Genomes for mice strains and species not available from the UCSC Genome Bioinformatics site ( M . musculus domesticus [AJ] , M . musculus castaneus and M . spretus ) were downloaded from ( Stefflova et al . , 2013 ) . We downloaded bird genome sequences from Ensembl version 80 BioMart ( Cunningham et al . , 2015 ) : galGal4 for Gallus gallus , Turkey_2 . 01 for Meleagris gallopavo , taeGut3 . 2 . 4 for Taeniopygia guttata and FicAlb_1 . 4 for Ficedula albicollis . Protein-coding gene names and symbols along with associated transcripts sequences were obtained from FlyBase ( dos Santos et al . , 2015 ) for insect species ( dmel-r5 . 46 , dsim-r1 . 4 , dere-r1 . 3 , dyak-r1 . 3 , dana-r1 . 3 and dpse-r2 . 30 ) , from Ensembl version 80 BioMart for bird species and from Ensembl version 59 BioMart for mammalian species ( Cunningham et al . , 2015 ) . For M . spretus and M . musculus castaneus , we used the same transcript annotations as for M . musculus . Within the genomes of our designated reference organisms ( M . musculus domesticus , G . gallus and D . melanogaster ) , we defined promoters as the region 0-2 kb upstream of transcription start site and delineated intergenic regions as regions that did not overlap annotated genes or promoters . Chromatin accessibility tracks used in Figure 5c , d and Figure 5—figure supplement 2 were downloaded from the UCSC bioinformatics website ( Rosenbloom et al . , 2015 ) for M . musculus domesticus and D . melanogaster , and obtained from ( He et al . , 2014 ) for G . gallus . We restricted our analyses to the sequences or annotations in , or homologous to , the well-defined chromosome scaffolds of the reference organism . Specific reference chromosomes analyzed are as follows: G . gallus ( 1–28 , Z , W ) , D . melanogaster ( 2L , 2R , 3L , 3R , 4 , X ) and M . musculus ( 1–19 , X , Y ) . We obtained orthology relationships between protein-coding genes using Ensembl COMPARA ( Vilella et al . , 2009 ) , matching the Ensembl versions used for protein coding genes for each species described above . These relationships were used in Figure 3 , Figure 3—figure supplement 1 , Figure 5b and Figure 5—figure supplement 1 . Homology between genomic segments was assigned using the LiftOver tool ( Rosenbloom et al . , 2015 ) , for all analyses presented in Figures 2 , 4 and 5 and associated figure supplements , with the exception of the nucleotide-resolution analysis of GSTF occupancy profiles presented in Figure 4e , f . We used pre-computed chain files from UCSC matching the genome versions listed above when chains were readily available ( Rosenbloom et al . , 2015 ) . When chain files were not available , we built chain files to map the UCSC M . musculus C57BL/6 mm9 to the genomes of M . musculus domesticus AJ , Mus musculus castaneus and Mus spretus , as well as to map the Ensembl 80 galGal4 to the genomes of M . gallopavo , F . albicollis and T . guttata ( Figure 1—figure supplement 1 ) . These chains were constructed by following the steps recommended by UCSC ( Supplementary file 2 ) ( http://genomewiki . ucsc . edu/index . php/Whole_genome_alignment_howto ) . For the nucleotide-resolution analysis of GSTF occupancy profiles , we assigned homology relationships using the chain files , or , in the case of mice strains , using genome mapping tables from ( Stefflova et al . , 2013 ) . We filtered the chain files to obtain one-to-one unambiguous mappings by retaining only highest scoring alignment for each position . These filtered mappings were then used to transfer data to from any organism onto the corresponding reference genome . Regions in the reference species genome lacking one-to-one unambiguous mappings were excluded from analysis . To define evolutionary distances separating species in Myrs , we chose published estimates generated as homogenously as possible within each lineage using a combination of sequence alignments and fossil records . All distances between insect species were taken from ( Tamura et al . , 2004 ) ; all distances between bird species were taken from ( Lu et al . , 2015 ) ; distances between mammalian species were taken from ( Stefflova et al . , 2013 ) and TimeTree ( Hedges , 2009 ) . For RNA-seq analyses ( Figure 3; Figure 3—figure supplement 1 ) , sequencing data for the reference species corresponding to two experiments performed independently by different research groups , and , when possible , representing different genotypes , were downloaded from public repositories . For M . musculus domesticus , we used data from ( Goncalves et al . , 2012; Sugathan and Waxman , 2013 ) , for G . gallus we used data from ( Brawand et al . , 2011 ) and ( Coble et al . , 2014 ) , for D . melanogaster we used data from ( ENCODE Project Consortium , 2012; Chen et al . , 2014 ) . Other species included were M . musculus castaneus ( Goncalves et al . , 2012 ) , M . spretus ( Wong et al . , 2015 ) , R . norvegicus ( Gong et al . , 2014 ) , H . sapiens ( ENCODE Project Consortium , 2012; Lin et al . , 2014 ) , G . gorilla ( Brawand et al . , 2011 ) , D . simulans ( Chen et al . , 2014 ) , D . yakuba ( Chen et al . , 2014 ) , D . ananassae ( Chen et al . , 2014 ) , D . pseudoobscura ( Chen et al . , 2014 ) , M . gallopavo ( Monson et al . , 2014 ) , A . platyrhynchos ( Huang et al . , 2013 ) and F . albicollis ( Uebbing et al . , 2013 ) . Specific accession numbers are listed in Figure 3—source data 1 . For ChIP-seq analyses ( Figure 4 ) , we downloaded data for FOXA1 in M . musculus domesticus ( C57BL/6 ) ( Stefflova et al . , 2013 ) , M . musculus domesticus ( AJ ) ( Stefflova et al . , 2013 ) , M . musculus castaneus ( Stefflova et al . , 2013 ) , M . spretus ( Stefflova et al . , 2013 ) and R . norvegicus ( Stefflova et al . , 2013 ) ; HNF4A and CEBPA in M . musculus domesticus ( C57BL/6 ) ( Stefflova et al . , 2013 ) , M . musculus domesticus ( AJ ) ( Stefflova et al . , 2013 ) , M . musculus castaneus ( Stefflova et al . , 2013 ) , M . spretus ( Stefflova et al . , 2013 ) , R . norvegicus ( Stefflova et al . , 2013 ) , H . sapiens ( Schmidt et al . , 2010 ) and C . familiaris ( Schmidt et al . , 2010 ) ; Twist in D . melanogaster ( He et al . , 2011b ) , D . simulans ( He et al . , 2011b ) , D . erecta ( He et al . , 2011b ) , D . yakuba ( He et al . , 2011b ) , D . ananassae ( He et al . , 2011b ) and D . pseudoobscura ( He et al . , 2011b ) ; Giant in D . melanogaster ( Paris et al . , 2013; Bradley et al . , 2010 ) , D . yakuba ( Bradley et al . , 2010 ) and D . pseudoobscura ( Paris et al . , 2013 ) . We also gathered data for CTCF in M . musculus domesticus ( C57BL/6 ) ( Schmidt et al . , 2012 ) , R . norvegicus ( Schmidt et al . , 2012 ) , H . sapiens ( Schmidt et al . , 2012 ) , C . familiaris ( Schmidt et al . , 2012 ) , D . melanogaster ( Ni et al . , 2012 ) , D . simulans ( Ni et al . , 2012 ) , D . yakuba ( Ni et al . , 2012 ) and D . pseudoobscura ( Ni et al . , 2012 ) . Accession numbers corresponding to the specific experimental replicates and control samples are listed in Figure 4—source data 1 . For motif analyses ( Figure 5a , b; Figure 5—figure supplement 1 ) , we gathered known position-weight matrixes from the JASPAR database ( Mathelier et al . , 2014 ) and the Fly Factor survey ( Zhu et al . , 2011 ) . We focused on the motifs corresponding to Twist and Giant in D . melanogaster , to CEBPA , HNF4A and FOXA1 in M . musculus domesticus , and on a set of 12 other motifs corresponding to GSTFs conserved across mammals and insects . This set was constructed by downloading all Core A vertebrata motifs from JASPAR ( Mathelier et al . , 2014 ) , identifying those corresponding to conserved GSTFs with one-to-one orthologs between M . musculus domesticus and D . melanogaster using COMPARA ( Vilella et al . , 2009 ) , and filtering the list down to those 12 instances where a position-weight matrix was also described in Fly Factor ( Zhu et al . , 2011 ) and were not already analyzed . We developed a statistical framework to compare evolutionary rates between lineages , and implemented it in R ( Development Core Team , 2011 ) . This framework takes as inputs: measures of pairwise cross-species similarity ( e . g . correlation of gene expression or sequence conservation ) , pairwise cross-species evolutionary distances and lineage labels . Conceptually , the framework estimates both a statistical significance and an effect size to determine whether rates of evolutionary divergence are indistinguishable or different between lineages ( Figure 1 ) . In practice , we model evolutionary divergence by an exponential decay in log-linear space . First , the nls function in R is applied to the log-transformed cross-species similarity data as a function of evolutionary distances to derive the following linear models: Second , an R function written in-house to handle nls model structures estimates the significance level of an ANOVA with a likelihood ratio test comparing the lineage-naïve and the lineage-aware model . Third , we define the effect size as the predicted absolute difference in similarity between lineage pairs after 100 Myrs of divergence as estimated from the lineage-specific models . We consider that the framework detected a difference between evolutionary divergence rates when the significance level is <0 . 05 and the effect size is >5% . We chose to use an exponential decay function because it is the simplest evolutionary model that fit all our input measures of cross-species similarity reasonably well . We chose to model the exponential decay in log-linear space because we noted that a simple exponential decay in linear space failed to capture the conservation observed between distant species ( mouse versus human at 91 Myrs and dog at 97 . 4 Myrs ) when analyzing the evolutionary dynamics of GSTF binding ( Figure 4 ) and motif retention ( Figure 5 ) . We hypothesize that these data layers likely follow a more complex decay model , but we did not want to explore this with our current data set to avoid over-fitting . The power of our statistical framework was assessed by simulating data for two lineages with measure of cross-species similarity decaying exponentially at different rates over time ( Figure 1—figure supplement 2 ) . We fixed one lineage to decay at set rates: −0 . 007 , −0 . 005 and −0 . 003 . We fixed the second lineage to be faster by a range of given differences . Over 1000 simulations , we sampled two values from a normal distribution centered on the expected values from the set exponential decay rates corresponding to the evolutionary distances shown in Figure 4b , with standard deviations set at 0 . 5% or 5% . Our framework detected an absolute rate difference of 0 . 001 in 39 . 3% of simulations and an absolute rate difference of 0 . 003 in 88 . 9% of simulations when the standard deviation was high ( 5% ) . When the standard deviation was low ( 0 . 5% ) , our framework detected an absolute rate difference of 0 . 001 in 25 . 7% of simulations and an absolute rate difference of 0 . 003 in 100% of simulations . Analysis of gene expression evolutionary rates was performed in four steps . First , we preprocessed the raw RNA sequencing data downloaded for public data sources . Second , we quantified the abundance of all annotated transcripts corresponding to protein-coding genes . Third , we estimated cross-species similarity by correlating transcript abundances at the genome-scale . Finally , we used these cross-species similarity estimates as input to our statistical framework to evaluate a common model against a lineage-aware model . RNA sequencing data were first preprocessed using FastQC ( www . bioinformatics . babraham . ac . uk/projects/fastqc/ ) and Trimmomatic ( Bolger et al . , 2014 ) . In order to quantify transcript abundances , we then used the program Sailfish ( Patro et al . , 2014 ) ( 1 ) to build transcriptome indices for each species using the transcriptome sequences described above , using the parameters '-p 8 -k 20'; and ( 2 ) to quantify transcript abundance using the transcriptome indices with the parameters '-p 8 -l "T=PE:O=><:S=U"' for samples with paired-end reads and '-p 8 –l "T=SE:S=U”' for samples with single-end reads . The bias-corrected transcripts per million ( TPM ) abundances estimated by Sailfish were then summed over the transcripts corresponding to the same gene locus . To estimate cross-species similarities in gene expression levels , for each lineage , we used R ( Development Core Team , 2011 ) to build a matrix containing the gene expression values for all the protein-coding genes of the reference organism and their one-to-one orthologs across other organisms within each lineage . We discarded instances where the abundance of a particular gene locus was less than or equal to 5 TPM . We then calculated the Spearman’s rank correlation for the expression of all genes between the reference and all other organisms within each lineage and plotted these correlations as against the evolutionary distance separating each organism pair ( Figure 3 ) . We also repeated the calculations using Kendall’s rank correlation coefficient and Pearson’s product-moment correlation on log2-transformed expression values ( Figure 3—figure supplement 1a , b ) . Finally , we calculated Spearman’s correlations among all genes including those with less than 5 TPM ( Figure 3—figure supplement 1c ) . All these scenarios were evaluated using our statistical framework . None indicated that a lineage-aware model described the data better than a common model . The first step of all our occupancy analyses was to align the ChIP-seq reads to the corresponding genomes in order to obtain occupancy profiles ( Figure 4a ) . For each accession ( Figure 4—source data 1 ) , the sequencing reads were aligned to reference genomes using Bowtie2 version 2 . 2 . 4 ( Langmead and Salzberg , 2012 ) with the parameters '-very-sensitive -N 1 . ' Reads containing the 'XS:' field ( multi-mappers ) were removed . Reads having the same start site were presumed to be PCR duplicates and removed using the 'rmdup' command of SAMtools version 1 . 1 ( Li et al . , 2009 ) . The filtered reads were then converted to tagAlign format . The tagAlign files corresponding to CEBPA , HNF4A , FOXA1 , Twist and Giant were then processed using 108 different segment-resolution methods and one nucleotide-resolution method; the tagAlign files corresponding to CTCF were only processed using the nucleotide-resolution method . The nucleotide-resolution method is described below and relates to Figure 4e , f . The aim of our segment-resolution analyses was to examine how robust the evolution of GSTF binding patterns was across 108 different analysis pipelines ( Figure 4a–d ) . We implemented all these pipelines , which follow the same general framework and differ only in the choice of 5 parameters , described and underlined below . First , the occupancy profiles in the tagAlign files were discretized into candidate occupied segments using a peak caller algorithm that aims at identifying segments where the ChIP sample is enriched in reads relative to the control sample . We implemented two peak callers: MACS version 2 ( M ) ( Zhang et al . , 2008 ) and SPP ( S ) ( Kharchenko et al . , 2008 ) . The occupied segments were then selected from the candidate set using a quality filter: stringent ( S ) , lenient ( L ) or asymmetric ( A ) . When using MACS2 ( Zhang et al . , 2008 ) as a peak caller , lenient segments were called using a p-value cutoff of 10−5 ( default ) and merged across replicates when available using the merge function in BEDTools ( Quinlan and Hall , 2010 ) . Stringent segments were called using a p-value cutoff of 10−22 and intersected across replicates when replicates were available . The intersection procedure , inspired from ( Stefflova et al . , 2013 ) , used BEDTools ( Quinlan and Hall , 2010 ) to implement the following two steps: ( 1 ) merge the two replicates and ( 2 ) select the merged segments corresponding to at least one segment in each original replicate . When using SPP ( Kharchenko et al . , 2008 ) as a peak caller , lenient segments were called using a q-value of 10−2 ( default ) , and merged across replicates when available ( Quinlan and Hall , 2010 ) . Stringent segments were called by selecting all candidate segments assigned to the lowest possible q-value in the sample , then intersected across replicates when available using the same intersection procedure . The asymmetric quality filter , inspired by ( Bardet et al . , 2012; He et al . , 2011b ) , indicates that segments were called stringently in the reference species and leniently in the other organism . The coordinates of the occupied segments called in the reference organism were projected onto the other organism’s genome using the LiftOver tool from the UCSC genome browser ( Rosenbloom et al . , 2015 ) and specifying a sequence similarity filter through the minMatch parameter . We used 3 different minMatch thresholds: stringent ( S: 0 . 95 default ) , lenient ( L: 0 . 5 ) , and none ( N: 0 . 001 ) . After cross-species coordinate projection , a reference subset was chosen to define the set of reference-occupied segments that would be further analyzed . Three choices were implemented: all reference-occupied segments independently of whether they map to any other species ( A ) ; for each pair of species , only reference-occupied segments with a homolog in the second species ( P ) ; only reference-occupied segments that had homologs across all the other species considered within the lineage ( S ) . The projected coordinates of the reference subset were then overlapped with the coordinates of the occupied segments in the other species using the intersect function in BEDTools ( Quinlan and Hall , 2010 ) . The overlap requirement was either lenient ( L; default parameter of 1 bp ) or stringent ( S; required a reciprocal overlap of half of the segments length: '-f 0 . 5 -r' ) . We systematically executed all combinations of the aforementioned 2 peak callers , 3 quality filters , 3 sequence similarity filters , 3 reference subsets , and 2 overlap requirements , yielding a total of 108 pipelines . The output of each pipeline was the fraction of reference subset segments that overlapped segments occupied in the others species ( i . e . segments retaining occupancy between the two species ) . This output was used as a cross-species similarity measure for GSTF binding patterns . We analyzed these similarity measures for 6 pairs of GSTFs ( Twist and Giant were each compared to FOXA1 , CEBPA and HNF4A ) using our statistical framework . Two GSTFs were considered to diverge differently from each other over time when 1 ) the significance of the test was less than 0 . 05 and 2 ) the effect size was greater than 5% . In summary we found that the choice of parameters greatly influenced what the evolutionary dynamics of a given GSTF looked like ( Figure 4b , c ) but that in general the rate of divergence of mammal and insects GSTFs were statistically indistinguishable ( Figure 4d ) . The results of these tests for all GSTF pairs considered across 108 pipelines are reported in Figure 4—source data 2 and summarized as pie-charts in Figure 4 . Observations about general trends of parameters and evolutionary divergence are further elaborated in Figure 4—figure supplement 1 . As a control we also conducted an analysis between FOXA1 and CEBPA since FOXA1 lacks data past 20 Myrs of evolutionary divergence , whereas for all others GSTFs we have broader coverage across the 100 Myrs range . We applied the same statistical framework to the within-lineage comparison between FOXA1 and CEBPA and detected that FOXA1 evolves faster than CEBPA in 74/108 instances . We believe that most of these detected differences are artifacts because the conservation of binding patterns for FOXA1 and CEBPA is in fact highly correlated throughout all combinations of parameters when restricting analyses to data points up to 20 Myrs ( Pearson’s r = 0 . 96 ) . We suspect that this type of artifact also affects the results of comparing FOXA1 with Twist or Giant ( Figure 4d ) . In order to compare occupancy profiles directly without discretizing them into occupied segments and unoccupied segments , we correlated sets of imputed fragment density vectors across species . The inputs to this method were the tagAlign files described above . To generate these vectors we first estimated the mean fragment size using a method adapted from ( Kharchenko et al . , 2008 ) , whereby the mean fragment size is computed as the number of base pairs of offset between the positive and negative strands that maximizes the Pearson’s correlation coefficient of their mapped read density . We used a modified approach that considered only the density of 5' read start sites on each strand , rather than the density of the entire read . The first peak of the cross-correlation values was identified by approximating the first derivative by the finite difference method , smoothing the derivative values with a Gaussian kernel of bandwidth 10 , and identifying the first downward zero-crossing of the curve . This position was used as the estimated mean fragment size L . We created imputed fragments by extending each read start site by L base pairs in the 3' direction . We then calculated a fragment density vector for each chromosome as the number of such imputed fragments that overlap each genomic position . When multiple replicates were available , replicates were merged by adding the fragment density vectors . In order to minimize bias introduced by the presence of unmappable regions , we implemented a masking scheme that adaptively normalizes each dataset depending on the read length and estimated fragment size of each sequencing run . First , all possible error-free reads of a given length were generated synthetically and aligned back to the genome using Bowtie2 2 . 2 . 4 with the following parameters: '-r -N 0 -D 0 -R 0 --dpad 0 --score-min “C , 0 , -1”' . Any multi-mapping reads with the ‘XS:’ flag were removed and the 5’ and 3’-most positions of the remaining read alignments recorded . The imputed fragment densities computed from the ChIP data were then normalized by dividing the density at each position by the fraction of positions within L base pairs upstream that were covered by the start site ( 5’ for positive-strand density and 3’ for negative-strand density ) of a uniquely-mapped genomic read . Positions with 0 uniquely-mappable read start sites within L base pairs upstream were excluded from further analysis . In order to compare between species , we transferred data from query organisms to the reference genome using the one-to-one filtered chain files described previously , and calculated the Pearson’s correlation between the concatenated chromosome vectors of reference and reference-mapped query data . The evolution of the correlation was modeled and compared using the statistical framework described above . We calculated the percentage of randomly sampled segments retaining homology . Within the genomes of the reference species , we delineated the boundaries of the regions from which to sample: whole genome ( Figure 2; Figure 2—figure supplement 1 ) , intergenic regions in accessible chromatin and intergenic regions in inaccessible chromatin ( Figure 5; Figure 5—figure supplement 2 ) . We used the BEDTools shuffle command ( Quinlan and Hall , 2010 ) to randomize the locations of 5000 segments of 75 bp length within the delineated boundaries using the option '-noOverlapping' . The resulting 5000 shuffled segments were then mapped across species using the LiftOver tool with minMatch parameter 0 . 001 ( Rosenbloom et al . , 2015 ) . We then calculated the percentage of segments that were successfully mapped ( i . e . retained homology ) , excluding segments that mapped to a region longer than 1000 bp . The entire simulation was repeated 20 times , starting each time with different sets of 5000 segments . The percentages of segments retaining homology were recorded for each of the 20 simulations , and averaged for each pair of species . These averages were plotted and used as inputs for our statistical framework . Varying the minMatch parameter of the LiftOver tool to 0 . 5 and segment length to 150 bp allowed us to verify that the observed trends were robust to sequence similarity thresholds and length sampled ( Figure 2—figure supplement 2; Figure 5—figure supplement 2 ) . The nucleotide sequences of the genomic segments from Figure 2 that retained enough homology to undergo a pairwise alignment were extracted using the getfasta function of BEDTools ( Quinlan and Hall , 2010 ) . These sequences were then pairwise aligned using EMBOSS suite’s implementation of Smith-Waterman local alignment ( Rice et al . , 2000 ) . Default values for gap open penalty ( 10 ) , gap extend penalty ( 0 . 5 ) and scoring matrix ( EDNAFULL ) were used to dynamically choose the best local alignment between reference and query sequences . For each cross-species comparison , we calculated the average percent identity of the ungapped alignments of all the segments across 20 randomizations . This procedure yielded values similar to those described previously for the mouse / human ( Waterston et al . , 2002 ) and D . melanogaster / D . pseudoobscura comparisons ( Richards et al . , 2005 ) . The average percent identity of ungapped alignments were used as inputs for our statistical framework , revealing that a model that incorporates lineage labels significantly improved fit to the data relative to a common model ( P<0 . 05; Figure 2—figure supplement 1 ) . Using the FIMO tool ( Grant et al . , 2011 ) in the MEME suite ( Bailey et al . , 2009 ) , the genomes of D . melanogaster and M . musculus domesticus were scanned for matches to experimentally-determined position-weight matrixes corresponding to the GSTFs of interest . Motif matches were called significant according to the default threshold of FIMO , P<10−4 . The genomic coordinates of significant motif matches were mapped to the other species within the same lineage using LiftOver ( minMatch 0 . 001 ) . The corresponding coordinates ( Mapped ) were then extended by 50 bp , and the resulting segments were scanned for motif occurrence ( Mappedwithmotif ) . In order to estimate background expectation , we randomly shuffled the locations of the Mapped segments and scanned these shuffled segments for motifs ( ShuffledMappedwithmotif ) . The percentage of motifs retained relative to background was calculated as: F=Mappedwithmotif−ShuffledMappedwithmotifMapped*100 The percentages F were then used as measures of cross-species similarity to estimate whether a lineage-aware model would describe the evolution of DNA binding motifs better than a common model ( Figure 5—figure supplement 1 ) . | The genetic information that makes each individual unique is encoded in DNA molecules . Cells read this molecular instruction manual by a process called transcription , in which proteins called transcription factors bind to DNA in specific places and regulate which sections of the DNA will be expressed . These 'transcripts' are active molecules that determine the cell’s – and ultimately the individual’s – characteristics . However , it is not well understood how alterations in the DNA of different individuals or species can lead to changes in where the transcription factors bind , and in which transcripts are expressed . Carvunis , Wang , Skola et al . set out to determine if there is a relationship between how often DNA changes and how often transcription changes during the evolution of animals . The experiments examined the abundance of transcripts in the cells of a variety of animal species with close or distant evolutionary relationships . For example , the house mouse was compared to a close relative called the Algerian mouse , to another species of rodent ( rat ) and to humans . The experiments show that the changes in transcript abundances are happening at similar rates in mammals , birds and insects , even though DNA changes at very different rates in these groups of animals . This similarity was also observed for other aspects of transcription , such as in changes to where transcription factors bind to DNA . The next challenges are to find out what makes transcription evolve at such similar rates in these groups of animals , and whether these findings extend to other species and to other processes in cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"short",
"report",
"computational",
"and",
"systems",
"biology"
] | 2015 | Evidence for a common evolutionary rate in metazoan transcriptional networks |
Genetic variants identified by genome-wide association studies explain only a modest proportion of heritability , suggesting that meaningful associations lie 'hidden' below current thresholds . Here , we integrate information from association studies with epigenomic maps to demonstrate that enhancers significantly overlap known loci associated with the cardiac QT interval and QRS duration . We apply functional criteria to identify loci associated with QT interval that do not meet genome-wide significance and are missed by existing studies . We demonstrate that these 'sub-threshold' signals represent novel loci , and that epigenomic maps are effective at discriminating true biological signals from noise . We experimentally validate the molecular , gene-regulatory , cellular and organismal phenotypes of these sub-threshold loci , demonstrating that most sub-threshold loci have regulatory consequences and that genetic perturbation of nearby genes causes cardiac phenotypes in mouse . Our work provides a general approach for improving the detection of novel loci associated with complex human traits .
Genome-wide association studies ( GWAS ) hold the promise of identifying genetic loci that drive complex disease , however realizing this goal has been challenging due to the modest effect sizes of most common variants that require extremely large cohorts to detect with significance . The recent demonstration that disease-associated single nucleotide polymorphisms ( SNPs ) reside preferentially in enhancer elements provides a unique opportunity to leverage epigenomic maps of regulatory elements for understanding the function of known GWAS loci and for prioritizing new loci missed in current studies ( Ernst et al . , 2011; Cowper-Sal·lari et al . , 2012; Maurano et al . , 2012; Trynka et al . , 2013 ) . Despite increasingly large GWAS cohort sizes , the current catalog of genome-wide significant loci still explains only a modest proportion of the heritability for any given trait , with an excess of low p-value loci still below the genome-wide significance threshold ( Arking et al . , 2014 ) . These observations suggest that many more signals with 'sub-threshold' significance remain to be identified , however , the recognition of biologically relevant sub-threshold loci is hindered by a higher false positive rate ( Hindorff et al . , 2009; Maher , 2008; Altshuler et al . , 2008 ) . Thus , new computational approaches that integrate genetic data with genome-wide epigenomic profiles are needed to use existing cohorts to discover new loci and genes that influence complex traits and diseases . Here , we use epigenomic maps of 127 tissues from the Roadmap Epigenomics Project as a guide to systematically identify biologically relevant sub-threshold variants ( Roadmap Epigenomics Consortium , 2015 ) . As proof of concept , we focused on two cardiac traits with clinical significance: electrocardiographic QT interval reflecting myocardial repolarization and QRS duration reflecting cardiac conduction . These two traits have a clear tissue of origin and published GWASs have reported over a hundred QT/QRS loci , making these traits ideal for testing variants with sub-threshold significance ( Supplementary file 1 ) ( Hindorff et al . , 2009; Maher , 2008; Altshuler et al . , 2008 ) . In particular , variation within QT interval length plays an important role in human disease , where extreme QT prolongation is associated with sudden cardiac death and can occur as an unintended side effect of many non-cardiac medications ( Rabkin et al . , 1982; Heist and Ruskin , 2010 ) . We combine genome-wide maps of cardiac enhancer activity with the results from a large study of QT interval duration to identify dozens of novel QT loci with sub-threshold statistical significance . We provide multiple lines of evidence to show that these sub-threshold loci can alter enhancer activity , and we implicate specific genes through which these loci act to influence QT interval length . Importantly , we demonstrate that epigenetic signals can distinguish true biological signals from noise , thus bypassing the higher false positive rate that has previously hindered study of sub-threshold loci . We expect our work will uncover new genes involved in cardiac electrophysiology , aid in the identification of patients at risk for sudden cardiac death , and enable development of new treatments for susceptible individuals . More broadly , our work demonstrates the power of integrating epigenomics with existing GWAS to discover sub-threshold genetic loci and novel genes associated with complex human disease .
We compiled a list of 112 QT/QRS loci from the NHGRI GWAS database ( accessed July 2013 , Supplementary file 1 ) and identified SNPs in strong linkage disequilibrium ( r2>0 . 8 ) using genotype data from the 1000 Genomes Project ( Phase 1 , CEU population ) ( 1000 Genomes Project Consortium , 2010 ) . We also collected GWAS loci from a later meta-analysis of QT interval studies , published in June 2014 by Arking et al . , which we held out from the aforementioned 112 QT/QRS loci as a validation dataset for subsequent analyses ( Arking et al . , 2014 ) . Because only 22 of 112 loci ( 20% ) harbor SNPs that overlap exons , we examined whether QT/QRS variants are enriched in predicted enhancer elements across the genome using chromatin maps across 127 tissues generated by the Roadmap Epigenomics Project including adult left ventricle ( LV ) , adult right ventricle ( RV ) , fetal heart ( FH ) and adult right atrium ( RA ) ( Roadmap Epigenomics Consortium , 2015 ) . QT/QRS variants have greatest overlap with predicted enhancers ( as defined by high levels of H3K4me1 and low H3K4me3 using ChromHMM ) from the four cardiac tissues compared to the other 123 non-cardiac tissues ( red circles , Figure 1b , Supplementary file 1 ) ( Ernst et al . , 2011 ) . This enrichment persists over randomly sampled sets of control loci with matched genetic properties such as minor allele frequency , number of SNPs in LD , distance to nearest gene , number of nearby genes , and presence on an Affymetrix 660W genotyping array ( Figure 1a , Materials and methods ) . Enhancers from the LV showed the strongest enrichment of the four cardiac tissues ( z-scores=7 . 67 , empirical p<1x10−5 , 105 permutations ) , demonstrating that an unbiased analysis can resolve the causal tissue with high precision , as QT interval and QRS duration are primarily reflective of myocardial repolarization in the ventricles . 10 . 7554/eLife . 10557 . 003Figure 1 . GWAS repolarization loci preferentially overlap cardiac enhancers . ( a ) Enrichment of human left ventricle enhancers in 112 QT/QRS loci . The number of loci that contain a SNP overlapping an enhancer are computed for the 112 QT/QRS loci , and compared against 100 , 000 permutations of randomly sampled control loci matched for LD block size ( number of SNPs ) , MAF , distance to nearest gene , number of nearby genes , and presence on genotyping array . ( b ) Top , Enrichment of enhancers from 127 human tissues in QT/QRS loci . Bottom , Enrichment of enhancers from non-cardiac tissues for QT/QRS loci is substantially weaker following removal of enhancers active in any of the four cardiac tissues . ( c ) Top , QT/QRS SNPs are more likely to disrupt motifs corresponding to expressed TFs compared to 100 , 000 sets of matched control loci . Bottom , Weaker enrichment was observed between repolarization and matched control loci when the sequence of the TF motif was randomly shuffled and re-mapped to the genome ( 10 , 000 permutations ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 00310 . 7554/eLife . 10557 . 004Figure 1—figure supplement 1 . 112 QT/QRS loci overlap enhancers more significantly than other genomic regions in adult left ventricle . Comparison of H3K4me1-enhancers defined by a 15-state model of ChromHMM against other ChromHMM states including protein-coding and non-coding genes and their promoters as well as DNase I hypersensitive ( DHS ) peaks that broadly mark regulatory regions . The left panel shows the enrichment of features in the 112 GWAS loci compared to randomly sampled control loci; the right panel shows the total number of the 112 GWAS loci overlapped by each feature . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 00410 . 7554/eLife . 10557 . 005Figure 1—figure supplement 2 . QT/QRS loci overlap enhancers more significantly than other genomic regions in non-LV cardiac tissue . Comparison of enhancers defined by H3K4me1 against other ChromHMM states in fetal heart , adult right ventricle and adult right atrium . The left panel shows the enrichment of features in the 112 GWAS loci compared to randomly sampled control loci; the right panel shows the total number of the 112 GWAS loci overlapped by each feature . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 00510 . 7554/eLife . 10557 . 006Figure 1—figure supplement 3 . Enrichment of cardiac enhancers remains high in the subset of 53 genome-wide significant ( p<5x10-8 ) QT/QRS loci . The number of loci that contain a SNP overlapping an enhancer are computed for the 112 QT/QRS loci , and compared against 100 , 000 permutations of randomly sampled control loci . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 006 Focusing on the left ventricle , we analyzed the enrichment of both coding annotations using GENCODE and non-coding annotations using individual chromatin marks and chromatin states defined by ChromHMM as well as DNase I hypersensitivity ( DHS ) maps available in heart tissue ( Ernst et al . , 2011; Harrow et al . , 2012; Thurman et al . , 2012 ) . We observed that intergenic enhancers are the most strongly enriched annotated genomic region ( z-score > 7 . 5 ) in QT/QRS loci , followed by gene transcription regions ( z-score between 3 and 6 ) ( Figure 1—figure supplements 1 and 2 ) . This enrichment increased significantly ( z-score from 7 . 67 to 9 . 31 for left ventricle ) when restricting the analysis to 'strong' enhancers ( i . e . H3K4me1 enhancers that are also marked by H3K27ac ) . Our results indicate that predicted enhancers are highly informative for annotating trait-associated variants compared to other classes of genomic regions . 10 . 7554/eLife . 10557 . 007Figure 2 . Enhancers overlapping QT/QRS loci differ in functional characteristics from all enhancers . Several functional characteristics were compared between enhancers overlapping QT/QRS loci ( red ) and non-GWAS left ventricle enhancers ( blue ) . Fold change represents fold change between median values for the two groups , and p-values were calculated using the Mann-Whitney U test . See Materials and methods for comparison methodology between GWAS QT/QRS enhancers and non-GWAS enhancers for each functional or epigenomic feature . For primate conservation , LV enhancers ( blue ) were size-matched ( +/-1 kb ) to GWAS enhancers to control for skewed enrichments driven by larger GWAS enhancer size . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 007 We next asked whether LV enhancers that overlap QT/QRS loci have features that distinguish them from putative LV enhancers identified by ChromHMM that do not overlap QT/QRS loci ( Figure 2 ) . First , we considered the density of H3K27ac marks , as the co-enrichment of H3K4me1 and H3K27ac correlates with strong enhancer activity ( Creyghton et al . , 2010; Rada-Iglesias et al . , 2011 ) . We found that the 65 enhancers overlapping 45 QT/QRS loci have a 3 . 1-fold higher H3K27ac density compared to non-GWAS LV enhancers ( p=1 . 53x10--4 ) . In fact , incorporating H3K27ac into ChromHMM enhancer predictions resulted in substantially greater enrichment of QT/QRS loci ( z-score = 9 . 31 vs . 7 . 67 for left ventricle ) ; 44 of the 45 QT/QRS loci overlap an H3K27ac-defined 'strong' enhancer . QT/QRS LV enhancers are also more likely to be marked by either H3K4me1 or H3K27ac in at least one of the other three heart tissues ( fetal , right atrium , right ventricle ) compared to non-GWAS LV enhancers ( p-values between 0 . 004 and 0 . 04 , Figure 2 ) and less likely to be active in non-cardiac tissues ( p=9x10-3 , Figure 2 ) . Left ventricular QT/QRS enhancers are significantly more hypomethylated than predicted LV enhancers not overlapping QT/QRS loci ( hypomethylation p=1 . 07x10-6 , hypermethylation p=0 . 60 , Figure 2 ) . Similar to H3K27ac , CpG hypomethylation correlates with increased enhancer activity , possibly through modulation of TF binding site accessibility ( Hon et al . , 2013; Stadler et al . , 2011 ) . Consistent with this idea , 22 of the 45 GWAS loci contain an enhancer SNP that alters a predicted motif for a cardiac-expressed TF ( empirical p=0 . 002 , 105 permutations ) ( Figure 1c ) . Moreover , QT/QRS GWAS enhancers are enriched for DHS and Cap Analysis Gene Expression ( CAGE ) signals in human fetal heart , both of which are marks of greater enhancer activity ( Figure 2 ) ( Thurman et al . , 2012; Andersson et al . , 2014 ) . Finally , QT/QRS left ventricular enhancers show significant evolutionary conservation across the primate lineage compared to non-GWAS LV enhancers ( p=6 . 82x10-5 compared to 105 size-matched sets of LV enhancers ) , suggesting that perturbation of these enhancers is under stronger negative selection . Taken together , QT/QRS loci preferentially overlap conserved enhancers that show cardiac-restricted activity , suggesting that common variants associated with these loci play roles in regulating cardiac functions that drive human phenotypes . Current GWAS loci collectively explain only a small fraction of the estimated heritability of a complex trait in part due to strict Bonferroni thresholds for multiple hypothesis testing ( p<5x10-8 ) and the limited statistical power of existing studies to discover variants with modest effect sizes ( Maher , 2008; Yang et al . , 2011 ) . We hypothesized that knowledge of the genomic properties associated with existing GWAS loci can guide the search for additional genetic signals that cannot be detected without increasing GWAS cohort sizes , and that these loci with weaker 'sub-threshold' p-values ( i . e . 0 . 05>p>5x10-8 ) might reveal novel genes and biological pathways that contribute to complex disease . To test this idea , we used SNP summary statistics from the Arking et al . ( 2014 ) QT interval GWAS study we had earlier held out as a validation dataset ( Arking et al . , 2014 ) . These summary statistics include the 112 QT/QRS loci identified by prior GWASs ( red dots , bottom , Figure 3 ) , as well as loci that reach genome-wide significance in the larger meta-analysis cohort but were not discovered in any previous GWAS ( and therefore were not included in the 112 QT/QRS loci used for enrichment analyses above , gold dots , bottom , Figure 3 ) . We observed that active LV enhancers are strongly enriched for loci harboring SNPs with p-values between 1x10-4 and 5x10-8 ( Figure 3a , black line ) . Furthermore , the combination of functional features identified for above-threshold QT/QRS enhancers ( Figure 2 ) substantially improves sub-threshold locus enrichment across a wide range of p-value thresholds ( Figure 3a , colored lines , Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 10557 . 008Figure 3 . Cardiac enhancers harbor additional sub-threshold QT loci . ( a ) Top , Enhancer characteristics learned on above-threshold QT/QRS loci from Figure 2 are predictive for additional sub-threshold loci ( colored lines ) . Each point on a curve represents the fold difference in proportion of SNPs with p-value below the cutoff in the enhancer set versus the whole genome . Bottom , Manhattan plot of p-values for all SNPs from Arking et al . ( 2014 ) QT interval GWAS . 112 QT/QRS loci and all SNPs within 1 Mb are highlighted in red . Genome-wide significant loci newly discovered by Arking et al . and not in the 112 QT/QRS loci are highlighted in yellow . ( b ) Top , Enrichment signals for sub-threshold SNPs in left ventricle enhancers persists following removal of the 112 GWAS loci and nearby SNPs ( +/- 1 Mb ) . Bottom , Manhattan plot of p-values for all SNPs from Arking et al . ( 2014 ) QT interval GWAS following removal of 112 QT/QRS loci and all SNPs within 1 Mb . Genome-wide significant loci newly discovered by Arking et al . and not in the 112 QT/QRS loci are highlighted in yellow . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 00810 . 7554/eLife . 10557 . 009Figure 3—figure supplement 1 . High density of fetal heart DNase I hypersensitivity reads in LV enhancers is robustly informative for identifying enriched sets of sub-threshold loci . Top: Enrichment of DHS reads in GWAS enhancers . Middle: Example comparison of sub-threshold locus enrichment in active LV enhancers vs . active LV enhancers with high DHS read density . Bottom: Y-axis of graphs corresponds to fold enrichment of sub-threshold loci in enhancers taken at three p-value cutoffs ( 10-4 , 10-5 and 10-6 ) . X-axis represents enrichments plotted for different subsets of enhancers chosen by varying DHS read density cutoffs . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 00910 . 7554/eLife . 10557 . 010Figure 3—figure supplement 2 . Enrichment in the sub-threshold significance range can be observed using only SNPs nearby known above-threshold loci . The foreground consisted of only SNPs within +/- 1 Mb of the 112 QT/QRS loci , and was compared against a background of all SNPs in the genome . Enrichment analyses were performed as described for Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 01010 . 7554/eLife . 10557 . 011Figure 3—figure supplement 3 . Sub-threshold loci associated with QT interval length are enriched in H3K4me1-defined left ventricle enhancers . Enrichment was calculated by comparing the number of loci that overlap an enhancer against 100 , 000 sets of randomly sampled control loci matched for genetic properties ( LD block size , MAF , distance to nearest gene , number of nearest genes and presence on genotyping array ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 01110 . 7554/eLife . 10557 . 012Figure 3—figure supplement 4 . Enhancers harbor additional sub-threshold loci associated with Alzheimer’s disease and LDL cholesterol . Top , Enhancer characteristics learned for QT/QRS loci ( e . g . H3K27ac , CpG hypomethylation ) are also effective for enrichment of LDL cholesterol-associated sub-threshold loci in human adult liver enhancers . Bottom , Enrichment of Alzheimer’s disease-associated sub-threshold SNPs in enhancers from peripheral blood monocytes . Tissue type was chosen using results by Gjoneska et al . ( 2015 ) computing enrichment of genome-wide significant Alzheimer’s disease loci across enhancers from Roadmap Epigenomics tissues ( Gjoneska et al . , 2015 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 012 Whether the enrichment of SNPs in the sub-threshold significance range represents linkage disequilibrium with existing above-threshold GWAS SNPs or novel biologically relevant loci remains an unresolved question ( Maurano et al . , 2012 ) . In fact , an enrichment analysis using only SNPs nearby above-threshold GWAS loci produced a strong enrichment signature in the sub-threshold significance range ( Figure 3—figure supplement 2 ) . To distinguish between the two possibilities , we took a conservative approach and removed all SNPs within 1Mb of the initial 112 QT/QRS loci . Remarkably , the enrichment for LV enhancers persists and increases in the sub-threshold range ( i . e . p=1x10-4 to 5x10-8 , Figure 3b ) , likely due to removal of nominally significant SNPs that are in LD with above-threshold QT/QRS loci and do not represent true association signals . The enrichment for active LV enhancers in sub-threshold loci is not driven by biases in MAF , LD block size , distance to nearest gene , number of nearby genes , or presence on a SNP genotyping array ( Figure 3—figure supplement 3 ) . In total , we identified 2075 SNPs with p<1x10-4 that are independent of the 112 published QT/QRS loci , of which 208 SNPs overlap LV enhancers ( Supplementary file 2 ) . Because the enrichment of sub-threshold SNPs in cardiac enhancers suggests that epigenetic prioritization can be used as a starting point for more in-depth investigations of sub-threshold signals from GWAS , we sought to directly test the molecular hypothesis that these sub-threshold loci impact the transcriptional regulation of cardiac genes ( Figure 4a ) . We grouped all 2075 sub-threshold SNPs using linkage disequilibrium data ( minimum r2=0 . 2 ) to identify 287 independent sub-threshold loci in the genome ( Materials and methods ) . We prioritized loci where a sub-threshold SNP overlapped an active LV enhancer and either ( i ) also overlapped a fetal heart DNase I hypersensitivity peak or ( ii ) was an expression quantitative trait locus ( eQTL ) for a nearby gene . In total , we cloned allele-specific enhancer fragments from 22 cardiac enhancers that overlap SNPs from 18 independent sub-threshold loci , and performed quantitative luciferase assays in human iPSC-derived cardiomyocytes to determine whether the sub-threshold SNP genotypes influence enhancer activity ( Materials and methods ) . We observed that 13 of 18 sub-threshold loci ( 72 . 2% ) contain an enhancer that drives luciferase activity in an allele-specific manner ( Figure 4b , d , Figure 4—figure supplement 1 ) . Moreover , we estimate that between 51 . 1%-89 . 8% ( 95% Bayesian confidence interval ) of prioritized sub-threshold loci show allele-specific activity on transcription , suggesting that the majority of sub-threshold loci identified by epigenomic prioritization do in fact have an impact on transcriptional enhancer activity . 10 . 7554/eLife . 10557 . 013Figure 4 . Sub-threshold loci prioritized by epigenomics alter enhancer activity . ( a ) Model detailing how sub-threshold SNPs overlapping enhancers can affect QT interval . Green text: methods used to test mechanistic step in model . ( b ) Summary of luciferase enhancer reporter experiments . Left , luciferase enhancer reporter construct . Right , number of loci tested in panel d that exhibits significant allelic activity ( p<0 . 05 between two haplotypes ) . ( c ) , Left , schematic of a 3-D enhancer-promoter chromatin interaction detectable by 4C-seq . Right , number of loci tested in panel d where an enhancer-promoter interaction is observed in human iPS-derived cardiomyocytes by 4C-seq . ( d ) Experimental evidence that sub-threshold SNPs alter enhancer activity and that sub-threshold enhancers interact with gene promoters . Fold below GWS column represents degree to which sub-threshold locus is below genome-wide significance ( 5x10-8 ) ; Luciferase reporter column colored green if significant allelic difference in activity ( p<0 . 05 , Figure 4—figure supplement 1 ) ; Enhancer-promoter interactions column colored green if there is a detectable enhancer-promoter interaction by 4C-seq ( Figure 4—figure supplement 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 01310 . 7554/eLife . 10557 . 014Figure 4—figure supplement 1 . Sub-threshold SNP alleles affect enhancer activity . For each sub-threshold locus , enhancers carrying one of two haplotypes were cloned upstream of a minimal promoter and firefly luciferase reporter gene . Blue: enhancer carrying reference allele; red: enhancer carrying alternate allele . Error bars represent standard error of the mean . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 01410 . 7554/eLife . 10557 . 015Figure 4—figure supplement 2 . Luciferase enhancer reporter assay for sub-threshold SNPs outside enhancers . ( a ) Four of five sub-threshold loci outside enhancers do not drive significant allele-specific enhancer activity ( p>0 . 05 ) . The 1 kilobase window surrounding the sub-threshold SNP was cloned into luciferase enhancer reporter construct . Blue: enhancer carrying reference allele; red: enhancer carrying alternate allele . Error bars represent standard error of the mean . ( b ) Allele frequency of rs9504919 in human populations . Most human individuals are homozygous for the less active C allele of rs9504919 , suggesting that a potential enhancer at rs9504919 could be missed in epigenome profiling studies . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 01510 . 7554/eLife . 10557 . 016Figure 4—figure supplement 3 . 4C-seq interactions with 10 enhancers in 8 sub-threshold loci . Height of blue bars represents interaction strength with 4C viewpoint . Red curves indicate enhancer-promoter interactions called within an annotated GENCODE promoter ( up to 2 . 5 kb upstream of TSS ) at a threshold of 5 . 0 . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 016 We also performed chromosome conformation capture combined with high-throughput sequencing ( 4C-seq ) to experimentally test whether predicted enhancers in sub-threshold loci can form contacts with promoters , and to identify potential target genes of sub-threshold enhancers . We used 4C-seq to test ten predicted enhancers from eight sub-threshold loci in human iPSC-derived cardiomyocytes ( van de Werken et al . , 2012 ) . Eight enhancers in six loci formed enhancer-promoter interactions in the proximal 500 kb region ( Figure 4c , Figure 4 — figure supplement 3 ) . These analyses provides evidence that the novel QT loci enhancers have regulatory activity and that the sub-threshold SNPs identified in our analyses can alter the activity of cardiac enhancers . We next tested whether epigenomic prioritization can distinguish biologically relevant sub-threshold loci by comparing properties of sub-threshold loci that do or do not overlap cardiac enhancers . From the 287 independent sub-threshold loci in the genome , we selected two subsets to compare: 60 loci that contain sub-threshold SNPs directly overlapping predicted active LV enhancers , and as a negative control , 129 sub-threshold loci that do not contain any SNPs ( r2>0 . 2 ) overlapping a cardiac enhancer . Only a very small number of above-threshold GWAS loci , including SORT1 for LDL cholesterol , the FTO/IRX3 locus for obesity , and the SCN5A/SCN10A locus for QRS duration , have been investigated in detail ( Musunuru et al . , 2010; van den Boogaard et al . , 2012; van den Boogaard et al . , 2014; Arnolds et al . , 2012; Smemo et al . , 2014 ) . These studies all identified SNPs within non-coding regulatory elements that disrupt expression of a nearby gene that plays a critical role in controlling a human phenotype . In contrast , no sub-threshold locus has been experimentally studied or validated to date . We selected one locus on chromosome 6 where our results from Figures 4 and 5 suggest that sub-threshold SNPs disrupt enhancer activity and therefore expression of a gene involved in cardiac electrophysiology . We set out to investigate whether this locus can serve as an example for future investigations of other sub-threshold loci . 10 . 7554/eLife . 10557 . 017Figure 5 . Epigenomic prioritization distinguishes biologically relevant sub-threshold loci ( a ) Left , Sub-threshold loci overlapping enhancers have significantly stronger association signals than loci outside enhancers in the QT interval GWAS . Middle , Loci overlapping enhancers have significantly more likely to be newly genome-wide significant in the held-out QT interval GWAS , than loci outside enhancers . Right , Sub-threshold loci overlapping enhancers are significantly more likely to be nominally significant ( p<0 . 05 ) in QRS GWAS than sub-threshold loci not overlapping enhancers . ( b ) Genetic perturbation of genes with predicted links to 60 enhancer-overlapping sub-threshold loci are significantly more likely to result cardiac conduction or contractility phenotypes than genes linked to all LV enhancers and genes nearby non-enhancer overlapping sub-threshold loci . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 017 The sub-threshold locus on chromosome 6 contains 8 SNPs with reported p-values less than 1x10-4 and another 2 SNPs in LD that do not have calculated p-values . We focused on the 3 SNPs in this locus that overlap active LV enhancers: rs1743292 ( p=6 . 48x10-5 ) and rs112332323 ( p-value not available ) that both overlap a 3 . 6 kb predicted enhancer , and rs1772203 ( p=5 . 87x10-5 ) that overlaps a 2 . 8 kb predicted enhancer ( Figure 6a , b ) . We cloned fragments corresponding to both enhancers upstream of a minimal promoter driving the luciferase gene , and compared luciferase activity between constructs carrying either the major or minor haplotypes at each site ( rs1743292 enhancer: Figure 6c , rs1772203 enhancer: Figure 6e , SNPs differing between cloned constructs listed at bottom of Figure 6b ) . We observed that the activity of both enhancers is dependent on the sub-threshold haplotype: at the rs1743292 enhancer , the major haplotype has 45% greater activity ( p=2 . 99x10-12 ) , while the minor haplotype is 28% more active in the rs1772203 enhancer ( p=1 . 79x10-5 ) . In the fetal human heart , rs1743292 overlaps a strong DNase I hypersensitivity peak marking a local region of open chromatin signifying potential transcription factor binding ( DHS track , Figure 6b ) ( Neph et al . , 2012 ) . Thus , to provide evidence that the rs1743292 locus alters enhancer activity in humans , we re-aligned the DHS sequencing reads from heterozygous human individuals in an allele-specific manner to assess the difference in the number of reads that map to either allele ( Maurano et al . , 2012 ) . In fetal heart tissue from one individual sequenced to high depth , rs1743292 shows a significant allelic imbalance for DHS reads with 97 reads mapping to the major C allele and 300 reads mapping to the minor T allele ( left , Figure 6d , p=3 . 1x10-25 , binomial test ) . This trend is consistent in all five additional human individuals heterozygous at rs1743292 sequenced at lower depth ( right , Figure 6d ) , suggesting that rs1743292 can affect enhancer activity potentially through altering chromatin accessibility or transcription factor binding . Moreover , using motif analysis , we observed that rs1743292 alters a predicted binding site for the cardiac-expressed nuclear factor NF-I family ( Figure 6f ) , which contains a family member ( NF-1a ) that itself has been associated by GWAS with cardiac electrophysiology ( Ritchie et al . , 2013 ) . We used 4C-seq to identify genes that could be regulated by the rs1743292 or rs1772203 enhancers . We observed that both enhancers form interactions with promoters of the upstream popeye-domain containing ( POPDC ) family members BVES/POPDC1 and POPDC3 , and with predicted enhancers situated within introns of the downstream PREP gene ( Figure 6a , g ) . This suggests that both enhancers may contribute to regulating the gene expression of BVES and POPDC3 , of note because the POPDC protein family of transmembrane proteins has recently reported roles in cardiac pacemaking ( Froese et al . , 2012; Kirchmaier et al . , 2012 ) . We sought to investigate the roles of the three candidate target genes ( BVES , POPDC3 , PREP ) of the rs1743292/rs1772203 locus in regulating myocardial repolarization . Consistent with the genetic association between this locus and QT interval length , we found that mice homozygous for loss-of-function copies of BVES exhibit cardiac conduction and pacemaker defects ( Figure 6h ) ( Bello et al . , 2015; Froese et al . , 2012 ) . In contrast , POPDC3 and PREP mouse loss-of-function models have no reported cardiac abnormalities , and instead show altered body fat , suggesting that this genetic locus alters QT interval length through the BVES gene ( Bello et al . , 2015 ) . Strengthening our evidence implicating BVES in QT interval , we observed that across 59 human tissues , BVES is most highly expressed in human left ventricle , whereas POPDC3 has much lower expression in cardiac tissue than skeletal muscle , and PREP is constitutively expressed across a wide range of tissues ( Figure 6—figure supplement 1 ) . We also used antisense morpholino oligonucleotides to knockdown transcripts from the BVES , POPDC3 and PREP orthologs in zebrafish , observing that bves knockdown leads to a reproducible shortening of the zebrafish ventricular action potential duration ( APD ) , the cellular correlate of the QT interval , ( p=0 . 002 and 0 . 09 for two independent morpholino sequences ) , whereas there is no reproducible difference in ventricular APD following loss of popdc3 or prep transcripts ( Figure 6—figure supplement 2 ) . Collectively , these data from multiple organisms provide evidence that SNPs within the rs1743292/rs1772203 locus alter QT interval duration through disruption of BVES expression . 10 . 7554/eLife . 10557 . 018Figure 6 . The rs1743292/rs1772203 sub-threshold locus disrupts activity of cardiac enhancers that interact with BVES , a gene important for cardiac electrophysiology . ( a ) Overview of rs1743292/rs1772203 sub-threshold locus . Gold rectangles represent predicted active LV enhancers , blue and green lines represent enhancer promoter interactions from the rs1743292 and rs1772203 enhancers , respectively ( see panel g ) . ( b ) Detailed view of cardiac enhancers overlapping rs1743292 ( left ) and rs1772203 ( right ) . ( c ) rs1743292 haplotypes differing at 6 SNPs ( listed at bottom of panel b ) affect activity of cardiac enhancer in human iPSC-derived cardiomyocytes , n=24 per haplotype . Error bars represent standard error of the mean . ( d ) , Left , rs1743292 alters level of DNase I hypersensitivity in a heterozygous human fetal heart sample . Right , Allelic imbalance of DHS reads at rs1743292 observed for 5 of 5 human individuals . ( e ) rs1772203 allele affects activity of cardiac enhancer in human iPSC-derived cardiomyocytes , n=16 per allele . Error bars represent standard error of the mean . ( f ) rs1743292 SNP overlaps a predicted nuclear factor I ( NF-I ) motif . ( g ) 4C-seq analysis of the rs1743292 ( blue ) and rs1772203 ( green ) enhancers identifies enhancer-promoter interactions with nearby BVES , BVES-AS1 and POPDC3 genes , and additional enhancer-enhancer interactions within introns in PREP . ( h ) Genetic perturbation of Bves , but not Popdc3 or Prep leads to cardiac electrophysiological defects in mouse models . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 01810 . 7554/eLife . 10557 . 019Figure 6—figure supplement 1 . Expression patterns of BVES , POPDC3 and PREP across 59 human tissues . Red bars correspond to chambers of the adult human heart ( LV = left ventricle , RV = right ventricle , RA = right atrium ) , orange corresponds to skeletal muscle ( SM = skeletal psoas muscle ) . Sample labels ( E## ) correspond to labels assigned by the Roadmap Epigenomics Consortium anrd available on the Roadmap Epigenomics website <http://egg2 . wustl . edu/roadmap/web_portal/meta . html> ( Roadmap Epigenomics Consortium , 2015 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 01910 . 7554/eLife . 10557 . 020Figure 6—figure supplement 2 . Knockdown of bves in zebrafish leads to ventricular repolarization defects . Effect of gene knockdown using two independent morpholino sequences ( red , blue ) on ventricular action potential duration compared against control scrambled morpholino ( black ) . Top: Sample optical voltage mapping traces from one matched morpholino and control knockdown pair . Bottom: Differences in APD80 between control and antisense morpholino oligonucleotide-mediated knockdown zebrafish . * corresponds to p<0 . 05 from unpaired two-tailed Student’s t-test , n=19 for control scrambled morpholinos , n=20 for each morpholino targeting BVES , POPDC3 , or PREP transcripts , error bars represent standard error of the mean . DOI: http://dx . doi . org/10 . 7554/eLife . 10557 . 020 These results provide evidence that cardiac enhancers can be used to identify novel sub-threshold loci and genes associated with cardiac traits . As demonstrated with the luciferase enhancer reporter assays , and specifically the rs1743292/rs1772203 locus , sub-threshold loci harbor SNPs that affect enhancer activity and regulate genes involved in QT interval . In the current QT interval GWAS , rs1743292 had an effective sample size of 68 , 900 individuals with 12 . 76% power to detect the locus at genome-wide significance . To detect rs1743292 at genome-wide significance with 80% power would require a GWAS cohort of 146 , 700 individuals . Thus , our study demonstrates that genome-wide enhancer maps are a powerful tool for identifying sub-threshold loci with bona fide roles in human cardiovascular physiology that would have remained otherwise unrecognized from existing GWAS cohorts .
A major limitation in the human genetics field is the inability to ascribe function to the vast majority of non-coding SNPs associated with complex human traits . Using enhancer annotations from hundreds of cell types and tissues , we find ~50% of QT/QRS GWAS loci overlap enhancers , and that these enhancers share common characteristics , including H3K27ac marks , CpG hypomethylation , and greater evolutionary conservation . The high density of common variation we observed in non-coding enhancers may be due to weaker evolutionary selection against the subtle phenotypes that arise from disruption of transcriptional regulatory units compared to the more severe disruption of protein-coding sequences commonly observed in rare Mendelian diseases . Studies of genetic heritability have indicated that many additional loci lie below the genome-wide significance threshold ( Yang et al . , 2011 ) . Our study contributes fundamental insights to overcoming the difficult problem of discovering the biologically relevant sub-threshold genetic signals that are orders of magnitude weaker than discovered by traditional GWAS . Three prior studies have observed the general enrichment of either sub-threshold SNPs or SNPs that explain a disproportionately high amount of heritability in cell type-specific regulatory elements ( Maurano et al . , 2012; Finucane et al . , 2015; Gusev et al . , 2014 ) . However , our study is unique in demonstrating the advantage of combining different epigenomic features to produce greater enrichments of sub-threshold loci . Critically , no previous study to our knowledge has implicated any specific sub-threshold locus in any complex human trait , whereas we establish that 13 of the 18 sub-threshold loci tested in this study are capable of altering enhancer activity . We also leverage GWAS summary statistics and genetic perturbations in mouse to demonstrate that epigenetic marks can discriminate true positive sub-threshold signals from noise , a key problem that , until now , has prevented the study of these loci . Finally , we perform an in-depth molecular dissection of the rs1743292/rs1772203 sub-threshold locus and implicate the popeye-domain containing family of transmembrane proteins in regulating myocardial repolarization . The study of above-threshold GWAS loci is generating more biological insights on new causal genes contributing to human disease , however there remains a wealth of untapped signals in the sub-threshold region . The work presented here represents a first step towards deciphering this signal and opens the door for the discovery of greater numbers of disease loci , genes , and pathways . Our study focused on QT interval and QRS duration due to their clear tissue of origin and a wealth of existing GWAS data , however we believe our approach could generalize to any well-powered GWAS on any trait . To this end , we chose two recently published , well-powered GWASs that relate to human diseases affecting large segments of the population: LDL cholesterol levels and Alzheimer’s disease . For both traits , we observed the enrichment of SNPs well into the sub-threshold significance range , that the enrichment signature persists following removal of all above-threshold loci , and that functional features that improve enrichment of QT-associated sub-threshold loci are also effective when applied to sub-threshold loci associated with LDL cholesterol and Alzheimer’s disease ( Figure 3—figure supplement 4 ) . These results suggest that epigenomics can be applied more broadly to identify new loci with sub-threshold statistical significance from GWAS of many complex human diseases . One important future extension of this work would be to build a formal machine learning classifier that can be first trained on above-threshold GWAS loci before being applied to quantitatively rank sub-threshold loci by predicted biological relevance . Finally , investigating the differences between above-threshold and sub-threshold loci to elucidate the factors that drive loci to different degrees of association with a trait will be an important area of future investigation . Many reported genome-wide significant loci have been discovered by GWAS despite low power , likely due to the existence of many other variants of similar effect that go undetected , termed the 'winner’s curse' , and thus this difference could be driven in part by random chance . However , we also hypothesize that sub-threshold loci with weaker effect sizes may act in different pathways from loci with stronger effect sizes , and that sub-threshold variants could have weaker effects on gene expression . In summary , our results provide a critical roadmap for the systematic analysis and re-analysis of genome-wide association studies to prioritize novel biologically relevant loci with weak association signals . As demonstrated with the rs1743292/rs1772203 locus , these loci would otherwise require substantially greater cohort sizes to reach statistical significance . Thus , we expect that this approach can be exploited to broadly improve the understanding of the biological pathways that contribute to complex human traits and disease .
We compiled a list of all SNPs associated with electrocardiographic QT interval ( reflecting myocardial repolarization ) or QRS duration ( reflecting cardiac conduction ) from the NHGRI GWAS catalog of published GWAS ( accessed on July 09 , 2013 ) , and removed loci identified from studies with small sample sizes ( <5000 individuals ) . As the GWAS catalog reports SNPs with p<1x10-6 , we performed a sensitivity analysis using only loci with p<5x10-8 to demonstrate that two different cut-offs does not meaningfully affect enrichment results for left ventricle ( Figure 1—figure supplement 3 ) . We used genotype data from the 1000 Genomes project to identify all SNPs in LD ( r2>0 . 8 , CEU population ) with the lead SNPs . For cases where two lead SNPs were in LD with each other ( i . e . different studies reported different SNPs from the same haplotype block ) , we merged the resulting loci . To avoid over-counting , if the sets of LD SNPs from two independent lead SNPs overlapped , we randomly assigned each of the shared LD SNPs to only one of the two lead SNPs . We used genomic features annotated by combinations of histone modifications ( e . g . enhancers and promoters using ChromHMM by the Roadmap Epigenomics Project ) or by GENCODE ( e . g . protein-coding exons ) . Previous studies have compared the number of GWAS SNPs overlapping a feature against the number expected for a randomly chosen region of similar size ( Maurano et al . , 2012; Hnisz et al . , 2013 ) . However , this approach does not control for biases associated with the location of GWAS SNPs . We controlled for these biases by following the Variant Set Enrichment approach where we generate a background distribution for genomic feature enrichment in loci around sets of 112 randomly sampled control lead SNPs ( Cowper-Sal·lari et al . , 2012 ) . We chose control lead SNPs from a genome-wide genotyping array ( Affymetrix 660W ) matched for size of the LD block ( +/- 5 SNPs ) , minor allele frequency of the lead SNP ( +/- 0 . 1 ) , distance to the nearest gene ( +/- 25 kb if outside gene ) , and number of nearby genes within a +/- 500 kb interval ( +/- 3 genes ) . We also considered differences in local GC content ( +/-25 nt ) but did not observe a strong difference between GWAS and control lead SNPs ( p=0 . 06 ) . To calculate enrichment of genomic regions in GWAS loci , we compared the number of GWAS loci that overlapped an enhancer to 100 , 000 sets of equally sized randomly sampled control lead SNPs . The 112 GWAS SNPs compiled from the NHGRI GWAS catalog includes 57 loci with p-values between 1x10-6 and 5x10-8 that have a higher false positive rate . In a sensitivity analysis , we examined the subset of 55 loci that met the more stringent 5x10-8 statistical threshold and found that sets of cardiac enhancers ( specifically fetal heart and adult left ventricle ) were also most highly enriched in these loci compared to the 123 non-cardiac tissues ( Figure 1—figure supplement 3 ) . We obtained TF motif instances in the human genome ( hg19 ) for 651 human motifs from the ENCODE project ( ENCODE Project Consortium , 2012 ) , and filtered these to only consider 287 motifs that correspond to TFs expressed in the left ventricle ( >1 RPKM by RNA-seq ) . We quantified the number of QT/QRS loci containing a SNP that disrupted an enhancer motif corresponding to an expressed TF in the left ventricle , and compared this against randomly sampled sets of control loci matched for MAF , LD block size , distance to the nearest gene and presence on the Affymetrix 660 W genotyping array . Summary GWAS data for LDL cholesterol was obtained from Willer et al . ( 2013 ) , and summary GWAS data for Alzheimer’s disease ( AD ) was obtained from Lambert et al . ( 2013 ) ( Lambert et al . , 2013 ) . Enrichment analyses were performed as described above for QT interval . For enrichment of Alzheimer’s disease-associated SNPs , the region encompassing the HLA locus was excluded ( chr6:24 , 182 , 924–34 , 537 , 546 in hg19 ) , as this region contained approximately 25% of all low p-value SNPs ( p<1x10-5 ) in the genome therefore and could skew enrichment results . The liver tissue was chosen for LDL cholesterol enrichment based on biological relevance . Tissue choice for AD SNPs was made using genome-wide enrichment analyses performed by Gjoneska et al . ( 2015 ) . For this analysis , we chose the second-most enriched tissue from Gjoneska et al . ( peripheral blood monocytes , with most significant p-value ) instead of the most enriched tissue ( peripheral blood mononuclear cells , PBMCs , with second-most significant p-value ) because the enrichment of AD SNPs in PBMC enhancers was substantially weaker than peripheral blood monocytes following removal of SNPs within the HLA locus . For AD GWAS , removal of SNPs within +/- 1 Mb of above-threshold loci was performed using 13 above-threshold loci with p<5x10-8 ( Stage 1 analysis ) listed in Table 2 of Lambert et al . For LDL cholesterol analyses , we first attempted to remove all SNPs within +/- 1 Mb of above-threshold loci reported in Supplementary files 2 & 3 of Willer et al . , however many SNPs with p<5x10-8 remained . Therefore , we performed LD pruning ( r2>0 . 2 from CEU population ) on summary-level p-value data from Willer et al . to define above-threshold loci and then removed 68 unique genomic intervals from the analysis . Enhancer functional characteristics applied to the enhancer sets were chosen based on the availability of additional data for the chosen tissue . DNase I hypersensitivity data not available for human liver , and genome-wide CpG methylation data was not available for peripheral blood monocytes . To assess whether QT sub-threshold loci overlapping enhancers are more likely to represent true biological signals , we queried the p-values of these loci in a related GWAS of QRS duration . In total , the QT GWAS we used to identify the sub-threshold loci consisted of 76 , 061 individuals , while the QRS GWAS queried consisted of 60 , 255 individuals . We compared the total sizes of each cohort used in the two studies and calculated that a minimum total of 46 , 452 individuals must be different between the two studies . Specifically , there are at least 31 , 129 individuals present in QT GWAS that are not present in the QRS GWAS , and at least 15 , 323 individuals present in the QRS GWAS that are not present in the QT GWAS . We used summary-level p-value data from the QRS GWAS testing four clinically applied QRS traits: Sokolow-Lyon , Cornell , 12-lead-voltage duration products , and QRS duration . For each SNP , the assigned p-value represented the minimum p-value across these four traits . For each sub-threshold locus , we identified all SNPs in strong LD ( r2>0 . 8 , CEU population from 1000 Genomes project ) , and assigned the p-value as the minimum of all p-values for LD SNPs in the QRS GWAS data . From the Roadmap Epigenomics Project , we were able to obtain matching 'strong' enhancer annotations and RNA-seq data for 59 of the 127 tissues , including LV . For each LV enhancer , we considered all genes with expression ≥1 RPKM in LV and in vitro differentiated human cardiomyocytes and distance within +/-500 kb as potential targets . We then split the RNA-seq data for the 59 tissues into two groups , depending on whether the enhancer is present or absent in each tissue , and applied a one-sided Mann-Whitney U test to ask whether each potential target gene showed significantly greater expression in tissues where the enhancer was active . Genes differentially expressed between tissues with active and inactive enhancers ( p<0 . 05 ) were considered computationally-determined potential target genes . For determining targets of sub-threshold enhancers , we first filtered our set of sub-threshold enhancers to remove those unlikely to be associated with QT interval . To do this , we excluded sub-threshold SNPs if the -log ( p-value ) was lower than 80% of the -log ( p-value ) of the most statistically significant SNP in LD ( r2>0 . 2 ) , as these are unlikely to be causal . For sub-threshold loci overlapping enhancers , and the set of all active LV enhancers , we identified nearby genes using the enhancer-gene linking method described above . This methodology was not applicable to the 129 sub-threshold loci that do not overlap enhancers , and therefore we identified the two nearest genes within 1 Mb using GREAT v2 . 0 . 2 and selected only genes with expression in adult human left ventricle data ( >1 RPKM ) . Mouse orthologs of human genes were identified using the Ensembl Genes 79 database through BioMart , and all queries of the MGI mouse phenotypes database were made between April 26 , 2015 and May 6 , 2015 . We used three search terms relevant to QT interval: 'ventricle muscle contractility' , 'cardiac contractility' and 'conduction' ( excluding non-cardiac conduction terms ) . We used DNase I hypersensitivity and digital genomic footprinting data from the ENCODE and Roadmap Epigenomics Projects because samples were sequenced to a greater depth than the chromatin modification ChIP-seq data , and there were data available from more individuals Roadmap Epigenomics Consortium , 2015 . To quantify allelic imbalance , we mapped DHS/DGF reads to a version of the human genome ( hg19 ) downloaded from the UCSC genome browser with all SNPs ( dbSNP141 ) masked by ambiguous nucleotides ( N’s ) using Bowtie2 ( v2 . 2 . 0 , flags: -N 1 , --sensitive , --end-to-end , --no-unal ) . As genotypes were not available , we considered a sample heterozygous at a particular SNP if reads from the hg19-defined reference and alternate alleles each mapped to 3 or more unique positions . Using this methodology , we observed the median difference in reads mapping to the reference versus alternate alleles to be 0 . In total , reads mapped to the reference allele more often than alternate at 6537 of 13 , 3826 heterozygous SNPs , and vice versa at 5884 of 13 , 826 heterozygous SNPs , with equal numbers of reads mapping to both alleles at the remaining 1405 SNPs . To quantify statistical significance of allelic imbalance at SNPs , we followed Maurano et al . ( 2012 ) and considered only SNPs with more than 21 reads . We performed a binomial test under the null hypothesis where reads map to both alleles at equal frequency , followed by Benjamini-Hochberg multiple testing correction across all heterozygous enhancer-overlapping SNPs . Human iPSC-derived cardiomyocytes ( iCMs ) ( Cellular Dynamics , Catalog #CMC-100-010-001 ) were thawed according to manufacturer’s instructions and diluted to a final plating density of 0 . 2x106 cells per mL with plating medium ( Cellular Dynamics , Catalog#CMM-100-110-001 ) . After 7 days in culture , iCMs were homogenized using a douncer , cross-linked and further processed as 4C template using DpnII as the first restriction enzyme and Csp6I as the second enzyme following the procedure outlined in van de Werken et al . ( 2012 ) . The median spacing between GATC fragments ( recognized by DpnII ) in the hg19 human genome is 264 nt . Sequencing of the 4C-Seq library was performed on an Illumina HiSeq 2000 , and sequencing reads were aligned to a reduced genome consisting of sequences that flank DpnII restriction sites . Primer sequences used for sequencing the 4C-seq library are listed in Supplementary file 3 . The human genome ( hg19 ) was used as reference genome for mapping 4C sequence captures . Non-unique sequences that flank a restriction site were removed from the analysis . To map 4C-seq reads to the genome , we first binned reads according to the reading primers used in each lane . We allow a single mismatch in the reading primer that overlaps the primary restriction cut site ( DpnII ) . The binned sequences were mapped to an in silico library of potential fragment ends generated based on the restriction enzymes used for the 4C template preparation . We did not allow any mismatch in the fragment-end , and for analysis we focused on the unique fragends only ( excluding repetitive fragment ends ) . As biases from sequencing yield or restriction cutting may be introduced by 4C-seq , we computed 4C-seq coverage in a genomic region by averaging mapped reads in running windows of 21 4C-seq fragment-ends . For peak-calling in a single 4C experiment , we perform explicit background modeling of the up- and downstream genomic regions independently . We assume that in a completely unstructured chromatin fiber the contact probability monotonically decreases as a function of the distance to the viewpoint . We model this by performing monotonic regression of the 4C signal as a function of the distance to the viewpoint . For this we use the R package isotone , which implements the monotonic regression ( Mair et al . , 2009 ) . We then compare the observed 4C signal to the predicted value from the background model and call the extremes that reach a significance threshold as peaks . For a given threshold q and a distribution F of residuals from the background model , every observation greater than Q3 ( F ) +q*IQR ( F ) , where Q3 is the third quartile of F and IRF ( F ) the inter-quartile range , is considered significant . Sub-threshold loci were considered candidates for testing by the luciferase reporter assay if the sub-threshold SNP overlapping the active LV enhancer either ( i ) overlaps a fetal heart DNase I hypersensitivity site , or ( ii ) is an eQTL in the left ventricle ( i . e . the SNP genotype is associated with differential expression of a nearby gene ) . We generated allele-specific enhancer constructs using two strategies outlined below: ( i ) PCR from genotyped heterozygous individuals , or ( ii ) direct synthesis of enhancer fragments . ( i ) Enhancer cloning from heterozygous individuals: We designed primer sequences to clone the entire predicted enhancer sequence defined by ChromHMM , and appended a 5’CACC sequence to forward primers to permit directional TOPO cloning . We designed primer sequences to clone fragments of up to 3 kb . For enhancers annotated as larger than 3 kb , we either selected a 3 kb fragment centered at the region of greatest histone modification density ( H3K4me1 , H3K27ac ) , or generated multiple fragments spanning the enhancer . Primer sequences and samples for human genomic DNA ( Coriell Cell Repositories ) are listed in Supplementary file 3 . We PCR amplified enhancers from human genomic DNA using Q5 High-Fidelity DNA Polymerase ( NEB , Catalog #M0491S ) and purified fragments corresponding to the correct length using a QIAquick Gel Extraction Kit ( Qiagen , Catalog #28706 ) . ( ii ) Direct synthesis of enhancer fragments: Enhancer fragments up to 1 kb in size were chosen so that the fragment covers both the sub-threshold SNP as well as peak within the DNase I hypersensitivity signal , and a 5’CACC sequence was appended to permit directional TOPO cloning . Fragments were synthesized using the gBlocks Gene Fragments service from Integrated DNA Technologies ( sequences are listed in Supplementary file 3 ) . Enhancer fragments from both methods were cloned into Gateway-compatible entry vectors using a pENTR/D-TOPO Cloning Kit ( Life Technologies , Catalog # K2400 ) and transformed into TOP10 E . coli bacteria following manufacturers guidelines . We used Sanger sequencing to verify that purified entry vectors carried enhancers with the correct insertion orientation and no mutations beyond the expected polymorphisms . Entry vectors were then Gateway-cloned using LR Clonase II Plus ( Life Technologies , Catalog # 12538-120 ) into a Gateway-converted pGL4 . 23 destination vector ( Promega , Catalog # E8411 ) for luciferase assays in human cell lines ( Fisher et al . , 2006 ) . We used Sanger sequencing to confirm a second time the correct enhancer orientation and sequence inside the destination vectors . Human iCMs ( Cellular Dynamics , Catalog #: CMC-100-010-001 ) were thawed according to manufacturer’s instructions and diluted to a final plating density of 0 . 2x106 cells per mL with plating medium ( Cellular Dynamics , Catalog#: CMM-100-110-001 ) . 96-well tissue culture treated plates were coated with 0 . 1 mL of 0 . 1% ( w/v ) gelatin per well and incubated at 37°C for at least two hours . The gelatin solution was aspirated off and wells rinsed with 100 uL of PBS , aspirated , and let sit in the tissue culture hood . Using a multichannel pipette , 100 uL of cells were seeded per well to obtain a target density of 20x103 iCMs . The plates were kept on a flat bench at room temperature for 10-15 minutes to allow for cells to settle down uniformly , followed by incubation at a tissue culture incubator set at 37°C and 7% CO2 . 48 hours post-seeding , the iCM plating medium was replaced with 100 uL of Maintenance Medium ( Cellular Dynamics , Catalog #:CMM-100-120-001 ) . The Maintenance Medium was replaced every other day . 3-4 days post-plating , iCMs began beating spontaneously and 7 days post-plating , they formed electrically connected syncytial layers that beat simultaneously . At this stage , the cells were transfected with the appropriate Luciferase reporter constructs and controls for downstream analyses . Media was replaced an hour before transfections . For each well , 95 ng of enhancer firefly Luciferase reporter ( cloned into pGL4 . 23 , Promega ) and 5 ng of Renilla Luciferase transfection control vector ( pGL4 . 73 , Promega ) was mixed with 10 ul of Opti-MEM ( Life Technologies , Catalog #:51985-034 ) . 0 . 2 uL of Viafect transfection reagent was added to the DNA/Opti-MEM mixture . After mixing , the transfection cocktail was incubated at room temperature for 5 min and 10 ul dispensed into the well with iCMs and plates transferred to 37°C . Media was changed 24 hr after transfection . 8 independent wells of iCMs were transfected per construct to account for variability in plating and transfection efficiencies . A mammalian expression vector , pEF-GFP ( Addgene , Plasmid 11154 ) , was used to visually monitor transfection efficiency . At least 65–70% of the population of iCMs expressed GFP 24 hr hours post-transfection . Luciferase activity was measured 24 hrhr after transfections using the Dual-Luciferase Reporter Assay System ( Promega , Catalog#:E1980 ) . After aspirating media , cells were rinsed with PBS once , and lysed with 20 uL of 1X Passive lysis buffer in the Luciferase assay kit . 15 min minutes after gentle shaking on an orbital shaker and complete lysis , the plate was stored at -80°C until further processing . Samples were prepared and luminescence measured according to Manufacturer’s Assay protocol for 96-well plates using the Varioskan Flash Multimode Reader ( Thermo Scientific ) . Zebrafish ( TuAB strain ) were cared for according to standard techniques . All animal experiments were approved by the Partners Subcommittee on Research Animal Care ( SRAC ) and were conducted in compliance with the regulations published in the US National Institute of Health Guide for the Care and Use of Laboratory Animals . At the single cell stage , fertilized oocytes were injected with standardized concentrations and volumes of antisense morpholino oligonucleotides ( 5’CAATAGATGGCGCTGTGTACCTGTC3’ and 5’AGAGCAGCCTGAAAGACAATAAAGA3’ for bves , 5’GGTTAATCCACTCACCTGCCTGAAA3’ and 5’CCGTCACTCGTATCCTGTTTTAGTG3’ for popdc3 , 5'3' and 'AGAAGTGTTTGCTCAGGTCACCTGT3' for prep , 5’GTTCAATTGTTTCTCACCTGCCAGA3’ and 5’CTAATCCTGTGAAAGCAGAAGATCC3’ for popdc2 ) dissolved in Danieau’s solution ( 58 mM NaCl , 0 . 7mM KCl , 0 . 4 mM MgSO4 , 0 . 6 mM Ca ( NO3 ) 2 , 5 . 0 mM HEPES pH 7 . 6 ) . Controls were injected with an equivalent dose of non-targeting morpholino of equal length but differing nucleotide composition ( 5’ATCCTCTTGAGGCGAACAAAGAGTC3’ ) . RNA was harvested at 72 hr using TRIzol ( Life Technologies ) according to the manufacturer’s instructions , cDNA synthesized by iScript reverse transcriptase ( Bio-Rad Laboratories , Hercules , CA , Catalog #1708840 ) and semi-quantitative PCR was used to assess relative percentage of gene knockdown . All studies of morpholino efficacy are a result of samples obtained from three independent injections . For evaluation of ventricular action potential duration , embryo hearts were microdissected at 72 hr hours of development and stained with di-8-ANEPPS ( Invitrogen , Catalog # ) . Cardiac contraction was arrested with 15 uM blebbistatin ( Sigma-Aldrich ) . Hearts were then field paced at 2Hz and imaged at 1000 frames per second . Analysis of action potential durations was performed using an in-house developed MatLab program . The action potential duration at 80% repolarization was utilized for all analyses . A minimum n of 9 embryos was required for all ventricular action potential studies , based on power calculations for effect size ( Cohen’s d ) of 1 . 5 at p=0 . 05 . No animals were excluded from analyses unless ventricular depolarization could not be induced at 120 paces per minute . No randomization of samples or blinding of investigators was utilized during these protocols . Statistical comparisons were performed using one-way ANOVA with Fisher’s Least Significant Difference testing with all comparisons being to clutchmate controls . All distributions were normal , and variances between control and experimental groups were not statistically significant . | Most complex traits are governed by a large number of genetic contributors , each playing only a modest effect . This makes it difficult to identify the genetic variants that increase disease risk , hindering the discovery of new drug targets and the development of new therapeutics . To overcome this limitation in discovery power , the field of human genetics has traditionally sought increasingly large groups , or cohorts , of afflicted and non-afflicted individuals . Studies of large cohorts are a powerful approach for discovering new disease genes , but such groups are often impractical and sometimes impossible to obtain . However , it has become possible to complement the genetic evidence found in disease association studies with biological evidence of the effects of disease-associated genetic variants . Wang et al . focus specifically on genetic sites , or loci , that do not affect protein sequence but instead affect the non-coding control regions . These are known as enhancer elements , as they can enhance the expression of nearby genes . These loci constitute the majority of disease regions , and thus are extremely important , but their discovery has been hindered by our relatively poor understanding of the human genome . Chemical modifications known as epigenomic marks are indicative of enhancer regions . By studying the factors that affect heart rhythm , Wang et al . show that specific combinations of epigenomic marks are enriched in known trait-associated regions . This knowledge was then used to prioritize the further investigation of genetic regions that genome-wide association studies had only weakly linked to heart rhythm alterations . Wang et al . directly confirmed that genetic differences in “sub-threshold” regions indeed alter the activity of these regulatory regions in human heart cells . Furthermore , mutating or perturbing the predicted target genes of the sub-threshold enhancers caused heart defects in mouse and zebrafish . Wang et al . have demonstrated that epigenome maps can help to distinguish which sub-threshold regions from genome-wide association studies are more likely to contribute to a disease . This allows for the discovery of new disease genes with much smaller cohorts than would be needed otherwise , thus speeding up the development of new therapeutics by many years . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"genetics",
"and",
"genomics"
] | 2016 | Discovery and validation of sub-threshold genome-wide association study loci using epigenomic signatures |
Resistance to targeted cancer therapies is an important clinical problem . The discovery of anti-resistance drug combinations is challenging as resistance can arise by diverse escape mechanisms . To address this challenge , we improved and applied the experimental-computational perturbation biology method . Using statistical inference , we build network models from high-throughput measurements of molecular and phenotypic responses to combinatorial targeted perturbations . The models are computationally executed to predict the effects of thousands of untested perturbations . In RAF-inhibitor resistant melanoma cells , we measured 143 proteomic/phenotypic entities under 89 perturbation conditions and predicted c-Myc as an effective therapeutic co-target with BRAF or MEK . Experiments using the BET bromodomain inhibitor JQ1 affecting the level of c-Myc protein and protein kinase inhibitors targeting the ERK pathway confirmed the prediction . In conclusion , we propose an anti-cancer strategy of co-targeting a specific upstream alteration and a general downstream point of vulnerability to prevent or overcome resistance to targeted drugs .
The inhibition of key oncogenes with target-specific agents elicits dramatic initial response in some cancers such as prostate cancer and melanoma ( Bollag et al . , 2010; Clegg et al . , 2012 ) . Even for the most successful single-agent targeted therapies , however , drug resistance eventually emerges leading to rapid progression of metastatic disease ( Garraway and Janne , 2012 ) . The mechanism of drug resistance may involve selection of resistant sub-clones , emergence of additional genomic alterations , and compensating interactions between alternative signaling pathways ( Choi et al . , 2007; Huang et al . , 2011; Wagle et al . , 2013 ) . One potential solution to overcome drug resistance is to combine targeted drugs to block potential escape routes ( Fitzgerald et al . , 2006 ) . Therefore , there is currently a need for systematic strategies to develop effective drug combinations . Targeted therapy has been particularly successful in treatment of melanoma . BRAFV600E gain-of-function mutation is observed in ∼50% of melanomas ( Davies et al . , 2002 ) . Direct inhibition of BRAFV600E by the RAF inhibitor ( RAFi ) vemurafenib and inhibition of the downstream MEK and ERK kinases have yielded response rates of more than 50% in melanoma patients with this mutation ( Chapman et al . , 2011; Flaherty et al . , 2012b ) . At the cellular level , inhibition of the ERK pathway leads to changes in expression of a set of critical cell cycle genes ( e . g . , CCND1 , MYC , FOS ) and feedback inhibitors of ERK signaling ( e . g . , DUSP , SPRY2 ) ( Pratilas et al . , 2009 ) . Resistance to vemurafenib emerges after a period of ∼7 months in tumors that initially responded to single-agent therapy ( Chapman et al . , 2011; Sosman et al . , 2012 ) . Multiple RAFi and MEKi ( e . g . , PD-0325901 , Trametinib ) resistance mechanisms , which may involve alterations in NRAS/ERK pathway ( e . g . , NRAS mutations , switching between RAF isoforms ) or parallel pathways ( e . g . , PTEN loss ) , have been discovered in melanoma ( Johannessen et al . , 2010; Nazarian et al . , 2010; Poulikakos et al . , 2010; Xing et al . , 2012 ) . The alterations associated with drug resistance may pre-exist alone , in combinations , or emerge sequentially and vary substantially between patients ( Van Allen et al . , 2014 ) . Effective drug combinations may target diverse resistance mechanisms . Despite anecdotal success , conventional methods and combinatorial drug screens generally fail to come up with effective combinations due to the genomic complexity and heterogeneity of tumors ( Zhao et al . , 2014 ) . In order to more effectively nominate drug combinations , we propose to employ system-wide models that cover interactions between tens to hundreds of signaling entities and can describe and predict cellular response to multiple interventions . There have been prior attempts to construct such signaling models . De novo and data-driven quantitative models were able to cover only a few signaling interactions and therefore had limited predictive power ( Nelander et al . , 2008; Bender et al . , 2011; Klinger et al . , 2013; Oates et al . , 2014 ) . Qualitative or discrete models can cover more interactions but typically lack the capability of generating quantitative predictions ( Saez-Rodriguez et al . , 2009; Breitkreutz et al . , 2010; Saez-Rodriguez et al . , 2011 ) . Detailed physicochemical models derived using generic biochemical kinetics data can be quite comprehensive and quantitative but typically lack sufficient cell-type specificity required for translationally useful predictions ( Chen et al . , 2009 ) . We construct comprehensive , cell-type specific signaling models that quantitatively link drug perturbations , ( phospho ) proteomic changes , and phenotypic outcomes ( Figure 1 ) . The models capture diverse signaling events and predict cellular response to previously untested combinatorial interventions . In order to generate the training data for network modeling , we first perform systematic perturbation experiments in cancer cells with targeted agents . Next , we profile proteomic and phenotypic response of cells to the perturbations . The cell-type specific response data serve as the input for network inference . In this study , we also incorporate prior pathway information from signaling databases to narrow the parameter search space and improve the accuracy of the models . For this purpose , we developed a computational tool ( Pathway Extraction and Reduction Algorithm [PERA] ) that automatically extracts priors from the Pathway Commons signaling information resource ( Demir et al . , 2010; Cerami et al . , 2011 ) . 10 . 7554/eLife . 04640 . 003Figure 1 . Quantitative and predictive signaling models are generated from experimental response profiles to perturbations . Perturbation biology involves systematic perturbations of cells with combinations of targeted compounds ( Box 1 , 2 ) , high-throughput measurements of responses ( Box 2 ) , automated extraction of prior signaling information from databases ( Box 3 , 4 ) , construction of ordinary differential equation ( ODE ) -based signaling pathway models ( Box 5 and Equation 1 ) with the belief propagation ( BP ) based network inference algorithm ( Box 6 ) , and prediction of system responses to novel perturbations with the models and simulations ( Box 7 ) ( Molinelli , Korkut , Wang et al . , 2013 ) . The newly developed ‘Pathway Extraction and Reduction Algorithm’ ( PERA ) ( Box 3 ) generates a qualitative prior model ( Box 4 ) , which is a network of known interactions between the proteins of interest ( i . e . , profiled ( phospho ) proteins ) . This is achieved through a search in the Pathway Commons information resource , which integrates biological pathway information from multiple public databases . In the quantitative network models , the nodes represent measured levels of ( phospho ) proteins or cellular phenotypes , and the edges represent the influence of the upstream nodes on the time derivative of their downstream effectors . This definition corresponds to a simple yet efficient ODE-based mathematical description of models ( Box 5 ) . Our BP-based modeling approach combines information from the perturbation data ( phosphoproteomic and phenotypic ) with prior information to generate network models of signaling ( Box 6 ) . We execute the resulting ODE-based models to predict system response to untested perturbation conditions ( Box 7 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04640 . 00310 . 7554/eLife . 04640 . 004Figure 1—figure supplement 1 . BP-guided decimation algorithm . The algorithm is used to construct executable , individual network model solutions from BP-generated probability distributions for each edge strength value ( Wij ) . The algorithm chart depicts one round of BP-guided decimation to generate a single model solution . Consecutive runs of BP-guided decimation algorithm lead to construction of a network model solution ensemble . DOI: http://dx . doi . org/10 . 7554/eLife . 04640 . 004 Even in the presence of large training data and priors , network inference is a difficult problem due to the combinatorial complexity ( i . e . , exponential expansion of the parameter search space with linear increase of parameters ) . For example , to infer a network model with 100 nodes using Monte Carlo-based methods , we in principle would need to cover a search space that includes ∼2 ( 100 × 100 ) network models—a computationally impossible task . To circumvent this problem , we previously had developed a network modeling algorithm based on belief propagation ( BP ) , which replaces exhaustive one-by-one searches over many individual network models by a search over probability distributions representing sets of low-error network models ( Molinelli , Korkut , Wang et al . , 2013 ) . The algorithm enables us to construct models that can predict response of hundreds of signaling entities to any perturbations in the space of modeled components . Here , we improved the perturbation biology method through automated incorporation of prior information ( signaling interactions from databases ) to obtain more accurate network models . The prior information is extracted from the signaling databases using the newly developed PERA tool . To improve predictive power and preserve cell-type specificity , we use prior information as soft restraints on search space through use of a probabilistic error model of priors , that is , the algorithm rejects interactions that do not conform to the experimental training data and predicts novel interactions not sampled in the priors ( see Equations 2–7 ) . To derive richer and more informative network models , we also scaled up the implementation to deal with a larger number of proteomic and phenotypic nodes ( see ‘Materials and methods’ for details and comments on potential future opportunities in perturbation biology ) . To quantitatively predict cellular response to combinatorial perturbations , we simulate the fully parameterized network models with in silico perturbations until the system reaches steady state ( Figure 1 ) . The steady-state readout for each proteomic and phenotypic entity ( i . e . , system variables ) is the predicted response to the perturbations . In this study , we improved and applied the perturbation biology method to devise a potentially generalizable strategy for overcoming resistance to targeted cancer therapies . We constructed cell-type specific network models of signaling from perturbation experiments in RAFi-resistant melanoma cells ( SkMel-133 cell line ) . The melanoma cells used for network modeling have a BRAFV600E mutation and homozygous deletions in PTEN and CDKN2A . The models quantitatively link 82 ( phospho ) proteomic nodes ( i . e . , molecular concentrations ) and 12 protein activity nodes with 5 cellular phenotype nodes ( e . g . , cell viability ) . As shown by cross-validation calculations , use of prior information significantly improved the predictive power of the models . Once the predictive power was established , we expanded the extent of the drug response information from a few thousand experimental data points to millions of predicted node values . Based on the predictions , which cannot be trivially deduced from experimental data , we nominated co-targeting of c-Myc and BRAF or MEK as a potential strategy to overcome RAFi-resistance . To test the prediction , we first experimentally showed that the BET bromodomain inhibitor , JQ1 , reduces c-Myc expression . Next , we showed that JQ1 blocks the growth of RAFi-resistant SkMel-133 cells in synergy with RAF/MEK signaling inhibition , and in this context , overcomes the drug resistance . Based on these results , we put forward the falsifiable hypothesis that co-targeting a specific upstream alteration and a general downstream point of vulnerability is a good strategy to prevent or overcome resistance to targeted drugs .
We predicted through quantitative simulations that melanoma cells were arrested in G1-phase of the cell cycle when c-Myc was targeted alone or in combination with other proteins , particularly BRAF , MEK , and cyclin D1 ( Figure 6A ) . We experimentally tested the key prediction from the network models . In order to target c-Myc expression , we treated melanoma cells with the BET bromodomain inhibitor , JQ1 , as a single agent and in combination with MEKi ( PD-0325901 ) or RAFi ( vemurafenib ) . JQ1 directly targets bromodomains , especially those of the bromodomain protein BRD4 , which marks select genes including MYC on mitotic chromatin . Inhibition of the BRD4 bromodomains with JQ1 downregulates MYC mRNA transcription and leads to G1 cell cycle arrest in diverse tumor types , such as multiple myeloma ( Delmore et al . , 2011; Loven et al . , 2013; Puissant et al . , 2013 ) . First , we asked whether we could affect c-Myc levels in SkMel-133 cells using JQ1 . As measured by Western blot experiments , c-Myc protein expression is reduced in response to JQ1 alone . c-Myc protein levels are further reduced when the cells are treated with a combination of JQ1 and MEKi or RAFi ( Figure 6B ) . To directly test the key prediction from the perturbation biology models , we measured the cell cycle progression response of melanoma cells to JQ1 in combination with the RAF and MEK inhibitors . We observed a strong synergistic interaction between JQ1 and RAFi ( Figure 6C , D ) . 51% and 46% of melanoma cells are in G1-stage 24 hr after treatment with JQ1 ( 500 nM ) and RAFi ( 200 nM ) , respectively , while 39% of cells are in G1-stage in the absence of any drug . On the other hand , when cells are treated with the combination of JQ1 and RAFi , a drastic increase in the fraction of cells arrested in G1-stage ( 84% ) is observed . The single agent MEKi ( 50 nM ) induces a strong G1-arrest phenotype in SkMel-133 cells ( 88% G1-stage in MEKi-treated cells vs 39% in nondrug treated cells ) . The combination of MEKi with JQ1 arrests an even higher fraction of the cells ( 92% ) in the G1-stage ( Figure 6—figure supplement 3 ) . Before assessing the effect of JQ1-MEKi/RAFi combination on viability of melanoma cells ( SkMel-133 ) , we tested the effect of single agent JQ1 and found that the melanoma cells were considerably sensitive to single agent JQ1 treatment ( cell viability IC50 = 200 nM ) . The sensitivity of SkMel-133 to JQ1 is similar to those of A375 and SkMel-5 lines ( RAFi/MEKi sensitive , carrying BRAFV600E mutation ) to another BRD4 inhibitor , MS417 ( Segura et al . , 2013 ) . The observed sensitivity is also comparable to those of multiple myeloma and MYCN-amplified neuroblastoma cell lines , reported to be potentially JQ1-sensitive tumor types ( Delmore et al . , 2011; Puissant et al . , 2013 ) , and substantially higher than those of lung adenocarcinoma and MYCN-WT neuroblastoma cell lines ( Lockwood et al . , 2012; Puissant et al . , 2013 ) . We tested the effect of combined targeting of c-Myc with MEK or BRAF on cell viability in SkMel-133 cells ( Figure 6E ) . Strikingly , when combined with JQ1 ( 120 nM ) , cell viability is reduced by 50% with 120 nM of RAFi ( PLX4032 ) , whereas the IC50 for single agent RAFi is >1 μM in RAFi-resistant SkMel-133 cells . Similarly , when combined with 5 nM MEKi ( PD901 ) , viability of SkMel-133 cells is reduced by 50% with 100 nM of JQ1 , an IC50 value , which is close to those of the most sensitive multiple myeloma cell lines ( Delmore et al . , 2011 ) . At higher doses ( IC80 ) , JQ1 is synergistic with both MEKi ( combination index , CI85 = 0 . 46 ) and RAFi ( CI85 = 0 . 47 ) in SkMel-133 cells . At intermediate doses , JQ1 synergizes with RAFi ( CI50 = 0 . 65 ) and has near additive interaction with the MEKi ( CI50 = 0 . 85 ) ( Figure 6F ) . Consistent with the observed synergy at high doses , both JQ1 combinations significantly improve the maximal effect level ( Amax , response to the drugs at highest doses ) , leading to lower cell viability beyond the levels reached by treatment with any of the agents alone . The observed improvement in Amax is particularly important since a subpopulation of cancer cells usually resist treatment even at highest possible drug doses . Treatments with drug combinations , such as those tested here will overcome or delay emergence of drug resistance if they can shrink the size of this resistant subpopulation ( i . e . , lead to improved Amax ) .
Website for the perturbation biology method: http://www . sanderlab . org/pertbio/ . The website stores the source code for BP-decimation algorithm , input scripts , and data files for running the BP-guided decimation code , perturbation response data , downloadable files for the executable model solutions , and the complete simulation results . A detailed description of the perturbation biology method is also provided in the webpage .
The improvements in perturbation biology have enabled us to establish our method as a widely applicable tool in genomically informed preclinical studies . Here , we obtained signaling models with increased scope ( increased coverage of diverse pathways using a data set 20-fold bigger than previously used data sets ) and nominated a particular drug combination to overcome the drug resistance in melanoma . The increased scope of the models is particularly critical for biologists to adopt our method as such models provide a nonreductionist insight into relations between the molecular changes and complex biological phenomena such as drug response , cell fate in diverse conditions , or any other cellular phenotype that can be measured quantitatively . Here , we improved the perturbation biology method and solved three key challenges in network inference to reach scopes and predictive power necessary for addressing complex problems in cancer biology . | Drugs that target the activity of specific genes could potentially form precise cancer therapies . Some cancers , including the aggressive skin cancer called melanoma , initially respond well to such treatments . However , resistance to drugs develops quickly , leading to the rapid regrowth of the tumors . Resistance can develop in a number of ways . For example , to prevent the drug from working or to compensate for the effects of a drug , cancer cells can adapt their signaling processes or acquire genetic mutations or other modifications that affect how genes are expressed . A well-designed combination of drugs that targets multiple molecular pathways can make it harder for cells to resist treatment , as this limits the number of available ‘escape’ pathways that bypass the drug targets . However , it is difficult to accurately predict how a cell will respond when treated with a particular drug , making it extremely challenging to design effective drug combinations . In 2013 , researchers developed a technique to build predictive models of cellular response pathways based on data collected from perturbation experiments followed by mathematical modeling . Now , Korkut et al . —including several of the researchers involved in the 2013 work—have refined this technology and applied it to the problem of preventing drug resistance in cancer cells . Computer simulations that used the mathematical models suggested a particular strategy of ‘upstream–downstream targeting’ in cells that were insensitive to the clinically successful drug vemurafenib ( an inhibitor of RAF proteins , which are often mutated in cancers ) . In the landscape of signaling pathways , the target of the upstream drug is on or near the mutated RAF protein . c-Myc , the indirect target of the downstream drug helps to express genes that trigger signals that cause the cells to grow . Inhibiting both targets in parallel may have the dual advantage of blocking the activation of the tumor-specific growth pathway while reducing the cancer cells' attempts to bypass the activation block . An initial test of the designed drug combination required moving from computer simulations to the laboratory using cell cultures originally derived from melanoma tumors . When Korkut et al . applied the drug combination , the combined treatment successfully blocked cell growth . The results suggest that the data-driven computer modeling strategy termed perturbation biology could be a useful tool for identifying effective cancer drug combinations for further preclinical research , possibly followed by clinical trials . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"computational",
"and",
"systems",
"biology"
] | 2015 | Perturbation biology nominates upstream–downstream drug combinations in RAF inhibitor resistant melanoma cells |
Atoh1-null mice die at birth from respiratory failure , but the precise cause has remained elusive . Loss of Atoh1 from various components of the respiratory circuitry ( e . g . the retrotrapezoid nucleus ( RTN ) ) has so far produced at most 50% neonatal lethality . To identify other Atoh1-lineage neurons that contribute to postnatal survival , we examined parabrachial complex neurons derived from the rostral rhombic lip ( rRL ) and found that they are activated during respiratory chemochallenges . Atoh1-deletion from the rRL does not affect survival , but causes apneas and respiratory depression during hypoxia , likely due to loss of projections to the preBötzinger Complex and RTN . Atoh1 thus promotes the development of the neural circuits governing hypoxic ( rRL ) and hypercapnic ( RTN ) chemoresponses , and combined loss of Atoh1 from these regions causes fully penetrant neonatal lethality . This work underscores the importance of modulating respiratory rhythms in response to chemosensory information during early postnatal life .
Hard-wired , transcriptionally defined neural circuit development is often complemented by synaptic plasticity that is driven by feedback from experience . Yet circuits giving rise to vital functions , such as respiration , have no time for such trial and error: the animal must be able to maintain its own O2/CO2 homeostasis from the moment it is born . This ability likely arises from a detailed genetic blueprint in the hindbrain respiratory circuit . Indeed , mapping the expression domains of key transcription factors in the developing hindbrain reveals a checkerboard pattern , with rostro-caudal and dorso-ventral stripes crisscrossing the entire region ( Gray , 2008 , 2013; Pagliardini et al . , 2008; Pasqualetti et al . , 2007 ) . The complexity of the circuit and the relative inaccessibility of some its individual components have made it difficult to tease out the specific contributions of various neuronal populations to neonatal survival . For example , mice lacking the transcription factor Atonal homolog 1 ( Atoh1 ) die at birth of respiratory failure even though they generate some respiratory movements ( Ben-Arie et al . , 1997; Rose et al . , 2009b; Tupal et al . , 2014 ) . The precise cause of this fully penetrant lethality has eluded a number of studies , which have nonetheless deepened our understanding of Atoh1’s contributions to the respiratory circuit . Atoh1 is expressed in the developing hindbrain along the entire rostro-caudal rhombic lip , where it promotes the development of several nuclei involved in respiratory control ( Gray , 2013; Rose et al . , 2009a , 2009b; Wang et al . , 2005 ) . Atoh1 is also expressed in postmitotic neurons of two paramotor nuclei: the intertrigeminal region ( ITR ) and the retrotrapezoid nucleus ( RTN ) ( Figure 1A and B ) ( Huang et al . , 2012; Rose et al . , 2009b ) . The chemosensitive RTN sends excitatory projections to the preBötzinger complex ( preBötC ) , which generates inspiratory rhythms ( Guyenet and Bayliss , 2015; Guyenet et al . , 2010; Kumar et al . , 2015; Onimaru and Homma , 2003 ) . Atoh1 loss from the RTN impairs respiratory responses to hypercapnia , but , rather remarkably , causes only partial neonatal lethality ( Huang et al . , 2012 ) . We therefore set out to find other Atoh1-lineage neurons contributing to neonatal survival and further delineate Atoh1’s function in respiratory development using intersectional genetics . We found that loss of Atoh1-lineage neurons developing from the rostral rhombic lip ( rRL ) impairs both respiratory rhythm and chemoresponsiveness . Some of these neurons are specifically activated during respiratory chemoresponses and project to the Atoh1-lineage paramotor nuclei ( ITR and RTN ) as well as the preBötC . Yet only the combined deletion of Atoh1 from the rRL and RTN recapitulated the fully penetrant lethality of Atoh1-null mice . This confirms that developmentally defined neural lineages have distinct roles in respiratory control and that , in neonatal mice , integration of chemosensory information is essential for survival .
Atoh1 is expressed along the entire rostro-caudal rhombic lip of the developing hindbrain ( Figure 1A , red ) , where it functions as a proneural transcription factor . Loss of Atoh1 results in loss of proliferating cells in the rhombic lip . Among the rhombic lip derived Atoh1-lineage are three populations of neurons that have been implicated in respiratory control: the parabrachial complex ( PBC ) , the rostral ventral respiratory group ( rVRG ) and the lateral reticular nucleus ( LRt ) ( Figure 1B , red ) ( Rose et al . , 2009b; Tupal et al . , 2014 ) . Two paramotor nuclei express Atoh1 during the postmitotic phase , and its expression is essential for their proper migration and connectivity from the RTN to the preBötC ( Figure 1A and B , blue ) ( Huang et al . , 2012; Rose et al . , 2009b ) . To date , the Atoh1-lineage PBC neurons are the only Atoh1-lineage neurons whose role in respiratory control and neonatal survival was not assessed . We first tested whether Atoh1-lineage PBC neurons might have a role in respiratory chemoresponses . To test this we labeled the Atoh1-lineage with tdTomato using a Cre-dependent reporter allele ( Atoh1Cre/+;Rosalsl-tdTomato/+ mice ) and exposed these mice to either room air , hypoxia , or hypercapnia prior to staining for the neural activity marker cFos . We found tdTomato+ , cFos+ double-positive cells in the medial and lateral parabrachial region as well as in the Kölliker Fuse after exposure to either hypoxia or hypercapnia , but not room air ( Figure 1C ) . This confirms that Atoh1-lineage PBC neurons are activated by changes in O2 and CO2 and might play a role in respiratory function . There are no previous reports that PBC neurons are intrinsically chemosensitive , so these neurons are likely activated by upstream neurons that are chemosensitive . Previous studies showed that Atoh1-null mice lose the substance P receptor NK1R in the PBC region ( Rose et al . , 2009b ) . We found that indeed all NK1R-expressing PBC neurons were Atoh1-lineage neurons ( Figure 1—figure supplement 1A ) . We also looked whether Atoh1-lineage PBC neurons expressed calcitonin gene-related peptide ( CGRP ) and pituitary adenylate cyclase‐activating polypeptide ( PACAP ) , because these peptides have been implicated in playing a role for hypercapnic and hypoxic responses respectively ( Arata et al . , 2013; Cummings et al . , 2004; Kaur et al . , 2017; Yokota et al . , 2015 ) . We found that all counted Atoh1-lineage neurons in the lateral PBC expressed CGRP and that some expressed PACAP ( Figure 1—figure supplement 1A ) . Together , these results suggest that Atoh1-lineage neurons might be important for respiratory responses through signaling with one or both of these neuro peptides . Next , we assessed whether these Atoh1-lineage neurons are required for neonatal survival . As PBC neurons develop from the rRL , we deleted Atoh1 only from this domain using an Engrailed-1-driven Cre-line ( En1Cre ) . We crossed Atoh1Flox/Flox females ( Shroyer et al . , 2007 ) to males heterozygous for En1Cre/+ and Atoh1LacZ/+ , which is a functional Atoh1-null allele that can be used to trace Atoh1-expressing neurons throughout development ( Ben-Arie et al . , 2000 ) . Atoh1Flox/LacZ;En1Cre/+ mice ( hereafter Atoh1::En1-CKO ) were born in Mendelian ratios ( 25/97 surviving pups , Figure 1D ) . We stained E14 . 5 embryos with X-gal and confirmed that Atoh1::En1-CKO lost all Atoh1-expressing neurons from the rRL , while leaving other Atoh1-lineage neurons intact ( Figure 1E and F ) . Anatomical analysis of postnatal ( P21 ) animals confirmed that this conditional deletion of Atoh1 resulted in loss of NK1R-expressing PBC neurons ( Figure 1G ) . Thus , despite losing these Atoh1-lineage PBC neurons , these animals survive , showing that Atoh1 expression in the rRL is not necessary for neonatal survival . Atoh1::En1-CKO mice developed severe ataxia , dystonia and tremor in the second to third week after birth and died shortly after weaning ( P22-25 ) , probably because the motor phenotypes impair their ability to get proper amounts of food and water . These phenotypes were likely the result of loss of Atoh1-lineage cerebellar neurons including glutamatergic deep cerebellar nuclei and cerebellar granule cells ( Figure 1—figure supplement 2A ) ( Ben-Arie et al . , 1997; Wurst et al . , 1994 ) . Unlike Atoh1-lineage PBC neurons , however , these Atoh1-lineage cerebellar neurons are not activated during respiratory chemochallenges and are thus less likely to be important for respiratory chemoresponses ( Figure 1—figure supplement 2B and C ) . Given that Atoh1::En1-CKO mice survive the early neonatal period , we were able to examine their respiration using unrestrained whole-body plethysmography ( UWBP ) at three weeks of age ( Figure 2A ) . In room air , Atoh1::En1-CKO mice had a greater number of sigh-induced and spontaneous apneas , sighs , and irregular respiratory rhythms than their control littermates ( Figure 2B ) . No other respiratory parameters were affected ( Figure 2—figure supplement 1 ) . Apneas , sighs , and rhythmic irregularity are hallmarks of immature respiration that can occur in some human infants ( Abu-Shaweesh and Martin , 2008; Martin et al . , 2004; Abu-Shaweesh , 2004 ) . When infants present with apnea of prematurity ( AOP ) in a clinical setting , they are treated with caffeine to stabilize their breathing rhythms ( Aranda et al . , 1977; Natarajan et al . , 2007 ) . We therefore tested whether caffeine treatment could rescue respiratory rhythms in Atoh1::En1-CKO mice by administering caffeine through the drinking water of lactating dams from P2 onward . This was sufficient to detect caffeine levels in the blood plasma of the pups ( treated: 6 . 15 ± 2 . 1 mg/L caffeine; untreated: 0 . 15 ± 0 . 04 mg/L caffeine; p=0 . 02 , two-tailed t-test ) . These levels are similar to those observed in infants treated with caffeine ( Natarajan et al . , 2007 ) . Caffeine treatment normalized apnea frequency and irregular breathing rhythms , but not sighs ( Figure 2B ) , showing that caffeine is sufficient to stabilize irregular breathing rhythms in our mice similar to human infants . Much to our surprise , this method of caffeine treatment also significantly decreased minute ventilation in control mice , as a result of both decreased tidal volume and breathing frequency ( Figure 2—figure supplement 1 ) . Although Atoh1::En1-CKO mice also showed a decrease in tidal volume , they seemingly compensated by increasing their rate of respiration , resulting in normal minute ventilation compared to control conditions . As Atoh1-lineage PBC neurons are specifically activated during hypoxia and hypercapnia , we hypothesized that these neurons also contribute to respiratory chemoresponses . We therefore assessed whether Atoh1::En1-CKO mice showed abnormal responses to hypoxia and whether the caffeine treatment rescued their irregular respiratory rhythms by restoring proper chemoresponses . Control mice had a short initial increase in tidal volume and breathing frequency that returned to baseline within several minutes of exposure to hypoxia ( Figure 3A and B ) . This resulted in a bimodal response in minute ventilation ( Figure 3C ) , similar to what others have reported in juvenile mice ( Haddad and Mellins , 1984; Waters and Gozal , 2003 ) . Atoh1::En1-CKO mice also showed a brief initial increase in tidal volume ( Figure 3A ) , but this was accompanied by a rapid , sustained decrease in respiratory rate that repressed minute ventilation during hypoxia ( Figure 3D ) . Interestingly , while caffeine treatment in control mice did prolong the respiratory response to hypoxia by limiting respiratory depression during exposure to hypoxic gas ( Figure 3C ) , it did not improve respiratory chemoresponses in Atoh1::En1-CKO mice ( Figure 3D ) . Respiratory depression in response to hypoxia resembles the suppression of fetal breathing movements during hypoxia in prenatal mammals ( Gluckman and Johnston , 1987; Haddad and Mellins , 1984; Abu-Shaweesh , 2004; Waters and Gozal , 2003 ) , underscoring how loss of rRL neurons recapitulates many aspects of immature breathing control . We next assessed how Atoh1::En1-CKO mice responded to hypercapnia . We found that control littermates showed a rapid increase in tidal volume and breathing frequency during hypercapnia , whereas respiratory chemoresponses of Atoh1::En1-CKO mice were severely attenuated ( Figure 3E and F ) . Caffeine treatment delayed the return to baseline minute ventilation in control mice , but did not improve the hypercapnic chemoresponses of Atoh1::En1-CKO littermates ( Figure 3G and H ) . Thus , the mechanism by which caffeine rescues apneas , sighs , and irregular respiratory rhythms in Atoh1::En1-CKO mice , cannot be through normalizing chemoresponses and these results confirm that Atoh1-lineage rRL neurons contribute to respiratory chemoresponses to low oxygen and high carbon dioxide . As neither Atoh1-lineage deep cerebellar nuclei nor cerebellar granule cells were activated during respiratory chemochallenges , we hypothesized that the abnormal chemoresponses , seen upon deletion of Atoh1 from the rRL , were not due to cerebellar dysfunction . To date , there is no Cre-line that can specifically delete Atoh1 from the pontine PBC without affecting the cerebellum , or vice versa . Therefore , we tested our hypothesis by silencing neurotransmission of Purkinje cells ( PCs ) , which are the sole output of the cerebellar cortex . PCs receive input from granule cells and directly project onto deep cerebellar nuclei . These Pcp2Cre/+;Slc32a1Flox/Flox mice have no defects in cell types other than PCs , and loss of PC signaling results in abnormal motor control including ataxia and poor performance on the rotarod ( White et al . , 2014 ) . Despite this abnormal cerebellar function , Pcp2Cre/+;Slc32a1Flox/Flox mice do not have more apneas or sighs than control littermates during room air breathing , although they have slightly more irregular respiratory rhythms than their control littermates ( Figure 3—figure supplement 1Aiv ) . No other respiratory parameters ( Figure 3—figure supplement 1A ) such as respiratory chemoresponses to either hypoxia or hypercapnia were altered upon silencing of PCs ( Figure 3—figure supplement 1B ) . Thus , abnormal cerebellar function or ataxia does not explain the abnormal respiratory chemoresponses observed in Atoh1::En1-CKO mice . Atoh1::En1-CKO mice display irregular breathing rhythms , sighs , and apneas , as well as respiratory depression in response to hypoxia and attenuated respiratory response to hypercapnia . As noted above , these features characterize immature respiration in some human infants , although they usually resolve on their own with time ( Abu-Shaweesh and Martin , 2008; Abu-Shaweesh , 2004 ) . Our results suggest that rRL neurons might play a prominent role in the postnatal maturation of respiratory control . To test this hypothesis , we evaluated respiratory control in one-week-old ( P7 ) Atoh1::En1-CKO mice and control littermates ( Figure 4A ) . At this age , Atoh1::En1-CKO mice display about twice as many apneas as their control littermates , but do not sigh more ( Figure 4B ) . We also observed that P7 Atoh1::En1-CKO mice have longer inhalation times and shallower breaths , resulting in a slower breathing rhythm and smaller minute ventilation ( Figure 4B ) . In response to hypoxia , control mice showed the bimodal respiratory response we observed in P21 Atoh1::En1-CKO mice with an initial increase in tidal volume and respiratory frequency , followed by respiratory depression ( Figure 4Ci to Ciii ) . Nevertheless , the decrease in tidal volume and respiratory frequency was initiated earlier during the hypoxic exposure and was more pronounced in Atoh1::En1-CKO mice . In contrast , there were no differences in respiratory response to hypercapnia between Atoh1::En1-CKO mice and control littermates at P7 ( Figure 4Civ to Cvi ) . Additionally , these results underscore our findings that cerebellar dysfunction is not the main driver of respiratory abnormalities in Atoh1::En1-CKO mice , the cerebellum is not yet developed at this age: cerebellar granule cells do not form their first functional synapses with Purkinje cells until P8 ( White and Sillitoe , 2013 ) . Thus , at P7 wild-type cerebelli are functionally more similar to those of Atoh1::En1-CKO mice , and if the cerebellum caused respiratory dysfunction in Atoh1::En1-CKO mice we would expect the differences between P7 control mice and Atoh1::En1-CKO mice to be smaller , not larger . Our results show that P7 Atoh1::En1-CKO mice have abnormal respiratory control in room air and in response to hypoxia . This suggests that although some respiratory phenotypes of P21 Atoh1::En1-CKO mice ( high number of sighs and abnormal hypercapnic chemoresponses ) might indeed result from abnormal postnatal maturation , others ( slow breathing rhythms and small tidal volume ) might be self-resolving with maturation . Yet at all ages tested rRL neurons are essential to prevent apneas and respiratory depression during hypoxia . Oxygen sensors in the carotid body communicate with the nucleus tractus solitarius ( NTS ) to integrate the sensory information into the central respiratory circuit , which is necessary for hypoxic chemoresponses ( Accorsi-Mendonça et al . , 2015; Dutschmann et al . , 2008; Ferreira et al . , 2015; Mayer et al . , 2015; Song et al . , 2011 ) . There is ample evidence that NTS neurons directly project to the PBC ( Bianchi et al . , 1995; Dutschmann et al . , 2008; Roman et al . , 2016; Song et al . , 2011 ) and our results suggest that the Atoh1-lineage PBC neurons are necessary to prevent respiratory depression in response to hypoxia . Yet it is unknown whether they modulate respiratory rhythms through activation of downstream rhythmogenic or chemosensitive nuclei , or act directly as premotor neurons . To trace the projections from the Atoh1-lineage rRL neurons , we made use of an intersectional reporter allele ( Ai65 ) that expresses tdTomato only after removal of both an FRT-flanked and a loxP-flanked stop-cassette ( Madisen et al . , 2015 ) . We generated an Atoh1FlpO knock-in mouse line that expresses FlpO recombinase in place of Atoh1 under the Atoh1 promoter ( Figure 5—figure supplement 1 ) . This mouse line can be used to remove the first stop-cassette in the reporter allele , exclusively in Atoh1-lineage neurons ( Figure 5A ) ; the second stop-cassette is removed using En1Cre , so that only neurons in the Atoh1;En1 intersectional domain will be labeled with tdTomato ( Figure 5A ) . We confirmed that the cell bodies of Atoh1-lineage parabrachial neurons were labeled red ( Figure 5B ) and tdTomato+ puncta overlapped with the synaptic marker synapsin , thus representing synapses on downstream neurons ( Figure 5C ) . We then assessed whether any tdTomato+ puncta were found in key respiratory nuclei or motor nuclei involved in respiratory motor rhythms ( Summarized in Figure 5D ) . We found tdTomato+ puncta in the Atoh1-lineage paramotor nuclei: the intertrigeminal region ( ITR ) and the chemosensitive retrotrapezoid nucleus ( RTN ) ( Figure 5E and F ) . We also found projections towards the rhythmogenic preBötC ( Figure 5G ) , but detected no feedback projections to the nucleus tractus solitarius ( NTS ) ( Figure 5H ) . We found no tdTomato+ puncta in any of the motor nuclei in the respiratory circuitry ( Figure 5I–L ) . This shows that the rRL neurons selectively innervate downstream neurons in the respiratory circuit that are important for chemoresponsive adaptations in the respiratory rhythm , but do not function directly as premotor neurons . Loss of activation from the PBC might subdue the respiratory responses , impairing adaptation to hypercapnia and hypoxia . We found that rRL neurons project to the Atoh1-lineage intertrigeminal region ( ITR ) ( Figure 5E ) , suggesting that part of the Atoh1::En1-CKO phenotype might be through loss of modulation of the ITR region . Loss of Atoh1 from the ITR was previously assessed only in Atoh1::Phox2b-CKO mice that also have impaired RTN development ( Huang et al . , 2012; Ruffault et al . , 2015 ) . To assess the function of Atoh1-lineage ITR neurons , we generated an ITR-specific conditional knockout mouse using the HoxA2::CreTG mouse line that expresses Cre only in r2 neurons ( Awatramani et al . , 2003 ) . We crossed homozygous Atoh1Flox/Flox females to males that were doubly heterozygous for HoxA2::CreTG and Atoh1LacZ to obtain Atoh1Flox/LacZ;HoxA2::CreTG/+ mice , hereafter called Atoh1::HoxA2-CKO mice . In situ hybridization for Atoh1 RNA confirmed that conditional deletion only from the ITR ( Figure 6—figure supplement 1A ) . ITR neurons were still present but mislocalized on the lateral side of the trigeminal motor nucleus instead of migrating medially at E14 . 5 ( Figure 6—figure supplement 1B and C ) . Because Atoh1::HoxA2-CKO mice survived the neonatal period ( Figure 6—figure supplement 2 ) , we could test their breathing behavior using unrestrained whole-body plethysmography ( UWBP ) at three weeks of age . We found that Atoh1::HoxA2-CKO mice had more sigh-induced and spontaneous apneas , without a change in the number of sighs or inter breath interval irregularity ( Figure 6Ai–Aiv ) . Interestingly , Atoh1::HoxA2-CKO mice did have a smaller tidal volume per breath , resulting in a smaller minute ventilation than control littermates ( Figure 6Avi and Aviii ) , phenotypes that were observed neither in our Atoh1::En1-CKO mice nor in Atoh1::Phox2b-CKO mice . Nevertheless , Atoh1::HoxA2-CKO mice did not have abnormal respiratory chemoresponses ( Figure 6B ) . Atoh1 is thus essential for normal ITR development , but loss of Atoh1 from the ITR does not impair neonatal survival or respiratory chemoresponses . The abnormal chemoresponses caused by loss of Atoh1 from the En1- or Phox2b-domain are thus caused by loss of rRL or abnormal development of RTN neurons , respectively . Through studying a series of conditional knockout mice , our lab has now mapped the role of all Atoh1-lineage neurons in neonatal survival ( Figure 7 ) . There is no single Atoh1-derived population that fully accounts for the perinatal death of Atoh1-null mice . In light of the fact that Atoh1 loss from either the En1 or Phox2b domains leads to abnormal chemoresponses , and that loss of Atoh1 from the Phox2b domain results in neonatal lethality in only about half of the mice , we hypothesized that combined loss of Atoh1-derived neurons involved in respiratory chemoresponsiveness might be responsible for the 100% neonatal lethality of Atoh1-nulls . To test this hypothesis , we crossed Atoh1Flox/+;En1Cre/+ males with Atoh1LacZ/+;Phox2b::CreTG/+ females to obtain mice that lack Atoh1 from both the En1- and Phox2b-intersectional domains ( Atoh1Flox/LacZ;En1Cre/+;Phox2b::CreTG/+ mice , hereafter Atoh1-dCKO mice ) . Out of nearly four hundred offspring observed , no Atoh1-dCKO mice survived past P1 ( Figure 7—figure supplement 1 ) . We did not find any alive Atoh1-dCKO mice , but several were found dead in the cage on the day of birth . These animals had turned blue and there was no visible milk spot , suggesting that they died of respiratory failure before drinking any milk . These results show that loss of Atoh1 from chemoresponsive nuclei accounts for the 100% neonatal death seen in Atoh1-null mice ( Ben-Arie et al . , 1997 ) .
Many of the transcription factors expressed in the developing hindbrain are necessary for respiratory circuit development and survival ( Gray , 2008 ) . In some cases , loss of a particular factor causes respiratory failure that is traceable to a single respiratory nucleus: for example , neonatal death in Dbx1-null mice is caused by loss of rhythmogenic preBötC neurons ( Bouvier et al . , 2010; Wu et al . , 2017 ) . The broad expression domains of factors such as Atoh1 , Tlx3 and Lbx1 , however , have made it difficult to pinpoint their role in a specific nucleus or functional impairment ( Gray , 2008; Huang et al . , 2012; Pagliardini et al . , 2008; Shirasawa et al . , 2000 ) . Here we used intersectional genetics to uncover a role for Atoh1 in the development of chemoresponsive neuronal populations that also express Engrailed1 and Phox2b ( the PBC and RTN , respectively ) . Concomitant loss of Atoh1 from these two domains causes fully penetrant neonatal lethality . This study thus answers the decades-old question about the cause of respiratory failure in Atoh1-null mice: they die of an inability to modulate respiratory rhythms in response to hypoxic and hypercapnic conditions . Several pieces of evidence suggest that the abnormal chemoresponses we observed in the Atoh1::En1-CKO mice are primarily due to loss of Atoh1-lineage parabrachial neurons . First , even though multiple nuclei develop from the intersectional domain of Atoh1 and En1 , the parabrachial neurons were the only Atoh1-lineage neurons that were activated specifically during hypercapnic and hypoxic chemochallenges ( Figure 1 ) . Second , silencing the output from the cerebellar cortex does not cause abnormal respiratory chemoresponses ( Figure 3—figure supplement 1 ) . Third , differences in abnormal chemoresponses between Atoh1::En1-CKO mice and control littermates are observed long before the cerebellum fully matures or Atoh1-lineage cerebellar granule cells even make their first functional synapses ( Figure 4 ) . Lastly , our results confirm and extend previous work showing that lesioning the parabrachial nucleus impairs respiratory chemoresponses ( Mizusawa et al . , 1995; Song and Poon , 2009a , 2009b ) and that the parabrachial nucleus projects to the ventrolateral medulla ( Fulwiler and Saper , 1984; Holstege , 1988; Yokota et al . , 2015 ) . We found rostral Atoh1-lineage neurons projecting to the retrotrapezoid nucleus and preBötC , but we did not observe any projections to motor nuclei or specifically the reported connections with the hypoglossal nucleus . This discrepancy can be because we used genetic tools ( Atoh1-driven ) that only target a subset of PBC neurons . Furthermore , these genetic tools are less prone to potential off target effects caused by needle tracks and local injections . Therefore , we propose that Atoh1-lineage PBC neurons are an essential component in the chemosensory pathways that relays information from peripheral , and perhaps central , chemosensors to enhance the firing of central chemosensitive and rhythmogenic neurons within the respiratory network . Nevertheless , we cannot exclude the possibility that Atoh1-lineage , deep cerebellar nuclei contribute to the observed phenotypes . Atoh1-lineage parabrachial and deep cerebellar neurons are born around the same embryonic day ( E9 . 5-E12 . 5 ) ( Rose et al . , 2009a ) and since we do not know what factors determine their differentiation , there are no developmental or genetic tools to target the deep cerebellar nuclei without affecting the parabrachial nucleus . Viral approaches do not work , either , as we discovered: deep cerebellar neurons send collateral projections throughout the cerebellum and the brainstem , so viral injections infect Atoh1-lineage deep cerebellar neurons regardless of the injection site . Nevertheless , the primary source of input to the deep cerebellar nuclei are the Purkinje cells , and silencing these cells did not recapitulate or alter respiratory chemoresponses . This latter finding , together with our data showing absence of chemoresponses in deep cerebellar neurons , make them unlikely to be essential for respiratory chemoresponses . Likewise , Atoh1 loss from the intertrigeminal region had little effect on respiratory chemoresponses . In agreement with earlier studies from our lab , we found that intertrigeminal neurons depend on Atoh1 expression for normal migration , which resulted in increased number of apneas and a shallower breaths . Previous studies suggested a role for intertrigeminal neurons in the attenuation of sleep and reflex apneas ( Radulovacki et al . , 2003 , 2004 ) . We indeed saw an increase in sigh-induced apneas but we saw an increase in the number of spontaneous apneas in only two out of thirteen Atoh1::HoxA2-CKO mice , suggesting that the previously reported apneas were not likely caused by a central mechanism ( Figure 6 ) . Despite the changes in respiratory control in room air , we did not observe any abnormal respiratory chemoresponses in Atoh1::HoxA2-CKO mice . These results support the hypothesis that the respiratory phenotype of mice that lack Atoh1 in both ITR and RTN ( the Atoh1::Phox2b-CKO mice ) results from abnormal RTN development . The current work provides evidence that there are two Atoh1-lineage nuclei that rely on Atoh1 expression for their function in respiratory chemoreflexes: the PBC and RTN . Removal of Atoh1 from both of these nuclei recapitulated the fully penetrant neonatal lethality of Atoh1-null mice , which we hypothesize is caused by combined loss of CO2-evoked glutamatergic signaling ( RTN ) and hypoxia-induced activation of PBC neurons . Previous studies in neonatal models have shown that arterial CO2 unloading removes respiratory drive and results in sustained apnea , in accordance with the notion that CO2 sensing is essential for neonatal breathing ( Nattie , 1999; Praud et al . , 1997 ) . We suspect that , when directly or indirectly activated during respiratory chemochallenges , Atoh1-lineage neurons increase glutamatergic input to the preBötzinger complex , which is thought to form the central respiratory pattern generator . Since neither Atoh1::En1-CKO mice nor Atoh1::Phox2b-CKO mice show complete lethality nor complete loss of respiratory chemoresponses , it is likely that these two pathways of chemosensation ( PBC and RTN-mediated ) are partially redundant , as would make sense for any circuit that controls a process as fundamental to survival as breathing . Indeed , en bloc recordings from Atoh1-null mice reveal that slow respiratory frequency can be fully restored by application of a glutamate reuptake inhibitor and partially restored by substance P application ( Rose et al . , 2009b ) . Our results are also in agreement with previous recordings in wild-type brainstems , where loss of input from the pons and RTN to the ventrolateral medulla results in a transition from a tri-phasic to a slow gasp-like , mono-phasic rhythm ( Rubin et al . , 2009; Smith et al . , 2009 ) . The present study provides further evidence that pontine PBC and RTN neurons are essential for the respiratory versatility observed in vivo . Despite the normal development of preBötzinger neurons that are sufficient for in situ respiratory rhythms , excitatory projections from Atoh1-lineage neurons are necessary for normal breathing in vivo . Caffeine treatment , along with glutamate and substance P , can stimulate endogenous , slow respiratory rhythms in rats ( Ruangkittisakul et al . , 2010 ) ; this is thought to be the mechanism by which caffeine reduces apneas of prematurity in newborn infants ( Aranda et al . , 1977; Natarajan et al . , 2007 ) . Caffeine drives this excitation mainly by acting as an antagonist to the A1 adenosine receptor that inhibits respiratory frequency ( Koos et al . , 2001 ) . Decreased inhibition might compensate for the loss of Atoh1-dependent glutamatergic drive in Atoh1::En1-CKO mice , thereby rescuing the apnea . Caffeine treatment did not , however , improve the respiratory chemoresponses in Atoh1::En1-CKO mice , suggesting that increased ventilation during chemochallenge relies on more precise signaling of distinct neurons in the circuitry . This is in agreement with the observation that some , but not all , Atoh1-dependent parabrachial neurons are activated during chemochallenges ( Figure 1 ) . To our surprise , caffeine treatment actually decreased respiratory output in control animals , which might be caused by inhibition of A2 adenosine receptors that increase respiratory frequency ( Koos , 2011; Koos and Chau , 1998; Koos et al . , 2001 ) . Future studies with specific adenosine agonists and antagonists are needed to elucidate the specific effects of long-term caffeine treatment on respiratory control in neonatal mice . The breathing abnormalities of Atoh1::En1-CKO mice resemble those seen in premature infants: increased apneas and sighs , abnormal rhythms , attenuated responses to hypercapnia , and , perhaps most surprisingly , respiratory suppression in hypoxia ( Martin et al . , 2004; Abu-Shaweesh , 2004 ) . These also mimic respiratory responses to hypoxia that occur in utero , when the oxygenation of fetal blood is regulated by the mother . During development , exposure to low O2 suppresses fetal movements , including breathing movements , likely to limit oxygen demand in relatively anaerobic environments . Several reports have suggested that pontine nuclei play a role in this respiratory suppression ( Gluckman and Johnston , 1987; Haddad and Mellins , 1984 ) and in the postnatal maturation of hypoxic responses ( Bissonnette and Knopp , 2001; Waters and Gozal , 2003 ) . Nevertheless , it was not known precisely which neurons were important for adaptation to hypoxia . Given that abnormal development of pontine nuclei and mutations in En1 have been observed in several sudden infant death syndrome ( SIDS ) cases ( Lavezzi , 2015 , 2016 , 2004; Weese-Mayer et al . , 2004 ) , we hypothesize that hypoxia-mediated respiratory suppression concomitant with immature responses to hypercapnia might be a major risk factor for SIDS .
All animals were housed in a Level 3 , AALAS-certified facility on a 14 hr light cycle . Husbandry , housing , euthanasia , and experimental guidelines were reviewed and approved by the Institutional Animal Care and Use Committee ( IACUC ) of Baylor College of Medicine . The following genetically engineered mouse lines were used: Rosalsl-tdTomato ( Gt ( ROSA ) 26Sortm9 ( CAG-tdTomato ) Hze , JAX:007914 ) , Atoh1Cre ( Yang et al . , 2001 ) , En1Cre ( En1tm2 ( cre ) Wrst/J , JAX:007916 ) : , Rosalsl-LacZ ( Gt ( ROSA ) 26Sortm1 ( CAG-lacZ , -EGFP ) Glh , JAX:012429 ) , Atoh1LacZ ( Atoh1tm2Hzo , JAX:005970 ) , Atoh1Flox ( Atoh1tm3Hzo , MGI:4420944 ) , Pcp2Cre ( B6 . 129-Tg ( Pcp2-cre ) 2Mpin/J , JAX: 004146 ) , Slc32a1Flox ( Slc32a1tm1Lowl/J , JAX: 012897 ) , HoxA2::CreTG ( Awatramani et al . , 2003 ) , Phox2b::CreTG ( Tg ( Phox2b-cre ) 3Jke , JAX:016223 ) , Atoh1FlpO , Ai65 ( Gt ( ROSA ) 26Sortm65 . 1 ( CAG-tdTomato ) Hze , JAX:021875 ) , and Sox2::CreTG ( Tg ( Sox2-cre ) 1Amc , MGI:3801167 ) . Ear tissue , collected from ear-marking , was used for PCR genotyping to minimalize stress on animals . For timed pregnancies , noon on the day of the vaginal plugging was set as embryonic day 0 . 5 ( E0 . 5 ) . The FlpO sequence from pQUAST-FLPo was ligated to a PGK-Neo cassette flanked by lox2722 sites in a pUC vector . This FlpO-PGK-Neo cassette was then cloned into a pre-existing pBlueScript II KS+ plasmid that contained the Atoh1 5’ and 3’ targeting arms without disrupting the Atoh1 transcriptional start site , identical to the approach previously described ( Rose et al . , 2009b ) . This construct was then electroporated into B57/6J ES cells with an agouti mutation . These ES cells were expanded under neomycin selection and screened for correct recombination by Southern blot using external 3’ probes , and internal 5’ PCR . Six clones were further expanded , tested correct recombination using PCR and sequence validated . Three clones were chosen to be injected into albino B57/6J blastocyst to generate chimeras . Chimeras were backcrossed to albino B57/6J females to generate heterozygote Atoh1FlpO-Neo mice . These were crossed to Sox2Cre to remove the PGK-Neo cassette . See Figure 5—figure supplement 1 for genomic targeting diagram , and example PCR genotyping ( forward primer: CTTCGTTGCACGCGAC , reverse primer: CACAATTTATCGTGTAGCCG . WT: 2 . 2 kb , FlpO-KI: 2 . 6 kb ) . IF and cryosectioning were performed using previously described protocols ( Huang et al . , 2012 ) . Mice older than one week were perfused with 4% PFA , after which brains were dissected and post-fixed overnight in 4% PFA at 4°C . Embryonic and P0 brains were directly dissected from pups and dropped fixed in 4% PFA at 4°C . After fixation brains were cryopreserved in 30% sucrose solution in PBS until sunk and frozen in OCT . Frozen sections were cut at 40 µm and kept at 4°C in the dark until used for immunostaining . For immunolabeling sections were first blocked in 5% normal donkey serum , 0 . 5% Triton-X in PBS for one hour at room temperature . Next , sections were incubated in primary antibody overnight at 4°C in blocking solution , followed by three washes and secondary labeling for at least two hours at room temperature . Nuclei were labeled using DAPI ( 1:10 , 000 ) and slides were mounted in DAPI-free mounting solution ( Vectashield ) . The following antibodies and their dilutions were used: rabbit anti-cFos ( 1:5000 , Santa Cruz ) , rabbit anti-NK1R ( 1:2000 , Advanced Targeting Systems ) , mouse anti-Synapsin ( 1:1000 , Synaptic Systems ) , goat anti-TPH2 ( 1:1000 , Santa Cruz ) , mouse anti-TH ( 1:200 , ImmunoStar ) , and goat anti-ChAT ( 1:200 , EMD Millipore ) . Secondary antibodies were conjugated with Alexa Fluor 488 ( 1:1000 , Molecular Probes ) . We used a Leica TCS SP5 confocal microscope system to image fluorescent signal . Image brightness and contrast were normalized using ImageJ . Embryos were collected at E14 . 5 , brains were dissected and fixed for 10 min in ice-cold formalin followed by brief washing in PBS on ice and in room-temperature wash buffer . β-galactosidase activity was assayed by embedding and sectioning tissue as previously described ( Ben-Arie et al . , 2000 ) . X-gal stained whole brains were stored at 70% ethanol . X-gal stained sections were first counterstained with nuclear fast red ( Vector laboratories ) and then imaged using a bright-field Axio Imager M2 microscope , equipped with an Axio Cam MRc5 color camera ( Carl Zeiss , Germany ) . Contrast and saturation were adjusted using ImageJ and Adobe Photoshop . In situ hybridization ( ISH ) was performed on 25µm-thick sagittal sections cut from unfixed , fresh frozen E14 . 5 embryonic control and CKO mice covering the entire embryo . We generated digoxigenin ( DIG ) -labeled mRNA antisense probes against Atoh1 using reverse-transcribed mouse cDNA as a template and a RNA DIG-labeling kit from Roche . Primer and probe sequences for the probe is available on the Allen Brain Atlas website ( www . brain-map . org ) and were validated previously ( Huang et al . , 2012 ) . ISH was performed by the RNA In Situ Hybridization Core at Baylor College of Medicine using an automated robotic platform as previously described ( Yaylaoglu et al . , 2005 ) . To analyze in vivo hypoxia or hypercapnia induced cFos expression , adult animals ( 6–8 weeks old ) were habituated to the plethysmography chambers five hours the day prior to the experiment and one hour on the day of the experiment . Freely moving mice were placed in whole-body plethysmography chambers ( Buxco ) through which fresh air was pumped at a basal flow rate of 0 . 5 L/min . Next , animals were exposed to either room air , hypoxia ( 10% O2 , balance N2 ) , or hypercapnia ( 5% CO2 , 21% O2 , balance N2 ) for 1 hr . Animals were sacrificed within 30 min of exposure and tissue was processed as described in ‘Immunofluorescence ( IF ) Assays . ’ Breathing analysis was performed using previously reported protocols with minor modifications ( Huang et al . , 2012; Orengo et al . , 2018; Yeh et al . , 2017 ) . To test respiratory parameters , weaning age mice ( P19-21 ) were placed in the plethysmography chambers and habituated for at least one hour prior to the experiment . After this habituation period , we recorded twenty minutes of room air breathing to determine the baseline . Next , we exposed animals to nine minutes in either hypoxia ( 10% O2 , balance N2 ) or hypercapnia ( 5% CO2 , 21% O2 , balance N2 ) . After this chemochallenge , we recorded breathing behavior for another fifteen-minute recovery period in room air . To determine changes in breathing behavior in response to the chemochallenge , we normalized all breathing parameters to baseline values using the following formula: normalized y ( t ) =y ( x ) / ( average ( y ( xbaseline ) ) , where y is the average value of a parameter during any given minute ( x ) and baseline is minutes one to twenty . We used Phonemah three software ( DSI ) to identify breath waveforms and used custom-written MATLAB ( Mathworks ) code to derive inspiratory time , expiratory time , and tidal volume ( Source code 1 ) . Breaths with an inspiration time shorter than 0 . 03 s or an expiration time longer than 10 s were excluded . Minute long segments in which the average breathing frequency ( breaths/min ) was higher than 500 were excluded from our analysis to prevent confounds from sniffing or exploring . Tidal volume was adjusted for body temperature as previously described ( Ray et al . , 2011 ) . Apneas were defined as breaths longer than 0 . 5 s and at least twice the length of the average of the six surrounding breaths ( three previous and three following ) . Apneas were divided in spontaneous apneas and apneas that were preceded by a sigh ( or a sigh in the previous breath ) . Sighs were defined as breaths with an inspiratory tidal volume 2 . 5 as big and an inspiratory time 1 . 25 as long as the previous five breaths . Inter Breath Interval ( IBI ) irregularity was defined as: IBI irregularity = abs ( breath length ( n + 1 ) -breath length ( n ) ) /breath length ( n ) . Minute ventilation was defined as the total amount of air breathed per minute ( breathing frequency * tidal volume ) . Source data from the plethysmography analysis can be found in the source data files: Figure 2—source data 1 , Figure 3—source data 1 , Figure 3—figure supplement 1—source data 1 , and Figure 6—source data 1 . Since we did not observe any difference in breathing parameters between male and female mice , or between Atoh1-heterozygote animals and any of the Cre-lines , we grouped these animals together in the analysis . Breathing analysis for neonatal mice was performed using a DSI pup-plethysmography set-up designed especially for these experiments . Specialized , small pup plethysmography chambers ( DSI ) were connected to a FinePointe Whole Body Plethysmography Unit with gas switch capability ( DSI ) . The unit provided the pup plethysmograph chambers with a constant 1 L/min airflow and amplified breathing waveforms . P7 ( day of birth is P0 ) animals were placed in the plethysmography chambers and could acclimate for 10 min , followed by 5 min of room air recording and a 5 min gas-challenge ( hypoxia ( 10% O2 , balance N2 ) or hypercapnia ( 5% CO2 , 21% O2 , balance N2 ) ) . FinePointe software was used to define breaths and calculate basic breathing parameters . Breaths were defined as followed: ( 1 ) peak inspiratory flow ( PIF ) must be before peak expiratory flow ( PEF ) ; ( 2 ) expiration or inspiration is less than a couple samples; ( 3 ) Rpef , EF50 , and Tr can be computed; ( 4 ) conditioning coefficient can be computed; ( 5 ) MinimumBoxFlowThreshold is larger than 0 . 005; ( 6 ) Start of expiration can be identified on the AC flow waveform . Breaths were rejected if TV was smaller than 0 . 005 ml or larger than 1 . 0 ml; if TI was larger than 0 . 02 s or twice as large as TE; or if inspiratory volume was not within 50% to 150% of the couple expiratory volume . For each minute of breathing recording a rejection index ( RINX ) was calculated as the percentage of the trace from which individual breaths were defined ( length breathing trace used for analysis/one minute * 100% ) . For our final analysis , we included only those minutes for which the RINX was less than 40% . All animals from which we recorded fewer than three reliable minutes of respiratory rhythms in both room air and during chemochallenge ( RINX <40% ) were excluded from our analysis . The percentage of animals of each genotype included in the final analysis were similar ( 23 control mice out of 33 ( 70% ) ; 12 Atoh1::En1-CKO mice out of 16 ( 75% ) ; Chi-Square = 0 . 133 , p=0 . 7149 ) . RINX of included animals was also similar between genotypes ( control mice: 18%; Atoh1::En1-CKO mice: 19 . 8%; t-test , p=0 . 614 ) . From the included minutes the following parameters were derived: inhalation time , exhalation time , breathing frequency , tidal volume , and minute ventilation . Apneas and sighs were determined by a trained observer who was blinded to the genotype of the mice . Source data from the pup plethysmography analysis can be found in the source data file: Figure 4—source data 1 . Caffeine was administered to pups through the drinking water of lactating dams ( 0 . 3 gram/L ) from postnatal day three until weaning . Caffeine was dissolved in water provided by the animal facility and stored in special dark bottles to prevent photodegradation . Plasma levels of caffeine in pups were determined using a commercially available Caffeine/Pentoxifylline ELISA kit ( Neogen ) . We obtained blood and isolated plasma from P19 pups raised in cages with normal water or caffeinated water . All quantification and statistical analyses were performed using MATLAB ( Mathworks ) . Data in text and figures represent the mean ± standard error of the mean ( SEM ) . Student’s t-test was used when comparing two independent groups . Chi-square test were performed to test Mendelian ratios of surviving offspring . We used two-way ANOVA to determine interaction effects between genotype ( control vs . conditional knockout ) and treatment ( no caffeine vs . caffeine ) , followed by a Tukey-Kramer post-hoc analysis in case significance was reached . To determine time-specific significant differences in respiratory chemoresponses , serial t-tests or ANOVA’s were performed and the p-value was Bonferroni adjusted for multiple comparisons . Statistical significance was accepted at p<0 . 05 for all other tests . | Breathing seems very simple: humans and other animals do it all the time without even thinking about it . Yet , many different cell types coordinate rhythmic breathing movements . Some cells set the breathing rhythm , motor neurons control the muscles , and other cells sense blood oxygen and carbon dioxide levels . Information about oxygen and carbon dioxide is necessary to trigger faster and deeper breaths when there is too little oxygen , for example , at high altitude . Or when there is too much carbon dioxide , for example , during exercise . At birth , most newborns can breathe as fast as needed because key genes oversee the development of all the cells involved in breathing . Learning more about these genes and what they do could lead to better understanding of why some newborns are at risk for sudden infant death or crib death . The Atoh1 gene , for example , helps carbon dioxide-sensing cells called retrotrapezoid neurons develop . Mice born without the Atoh1 gene are unable to breathe normally and die at birth . But when the gene is only deleted from these carbon dioxide-sensing cells in mice , just half of them die . This suggests that Atoh1 in other cells may also be important for breathing . Now , Van der Heijden and Zoghbi show that the Atoh1 gene also helps develop another set of cells that are essential for breathing called the parabrachial complex . These cells receive information from oxygen sensors and relay the information to cells that set breathing rhythms . Mice missing parabrachial complex cells do not breathe faster when oxygen levels in the air are low . Mice lacking Atoh1 from both the parabrachial complex cells and the retrotrapezoid cells have breathing problems and die at birth . Van der Heijden and Zoghbi show that the Atoh1 gene is essential for two cell types that make mice breathe faster when oxygen or carbon dioxide levels change . Together these two cell types are necessary for survival . The experiments also may provide insights into what goes wrong in babies who experience sudden infant death . Mutations in genes that are important to both cell types increase the risk of these infant deaths . Newborn babies with mutations in such key developmental genes will be at risk when in low oxygen or high carbon dioxide environments because their breathing systems are still maturing . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2018 | Loss of Atoh1 from neurons regulating hypoxic and hypercapnic chemoresponses causes neonatal respiratory failure in mice |
Paternal environmental conditions can influence phenotypes in future generations , but it is unclear whether offspring phenotypes represent specific responses to particular aspects of the paternal exposure history , or a generic response to paternal ‘quality of life’ . Here , we establish a paternal effect model based on nicotine exposure in mice , enabling pharmacological interrogation of the specificity of the offspring response . Paternal exposure to nicotine prior to reproduction induced a broad protective response to multiple xenobiotics in male offspring . This effect manifested as increased survival following injection of toxic levels of either nicotine or cocaine , accompanied by hepatic upregulation of xenobiotic processing genes , and enhanced drug clearance . Surprisingly , this protective effect could also be induced by a nicotinic receptor antagonist , suggesting that xenobiotic exposure , rather than nicotinic receptor signaling , is responsible for programming offspring drug resistance . Thus , paternal drug exposure induces a protective phenotype in offspring by enhancing metabolic tolerance to xenobiotics .
Environmental conditions experienced in one generation can affect phenotypes that manifest in future generations , a phenomenon sometimes referred to as the ‘inheritance of acquired characters . ’ In mammals , a substantial body of literature links various maternal exposures to offspring phenotypes ( Harris and Seckl , 2011; Rando and Simmons , 2015; Simmons , 2011 ) , and an increasing number of studies have shown that paternal environment can also alter offspring phenotype ( Rando , 2012 ) . Paternal effect paradigms are of particular mechanistic interest in mammals , given that it is challenging to disentangle maternal environment effects on the oocyte epigenome from effects on uterine provisioning during offspring development . In contrast , in many paternal effect paradigms , males contribute little more than sperm to the offspring , simplifying the search for the mechanistic underpinnings of paternal effects on children . A large number of paternal exposure paradigms have been used to show that a father’s diet can affect metabolic phenotypes in the next generation ( McPherson et al . , 2014; Rando , 2012 ) , while another large group of studies link paternal stress ( using paradigms such as social defeat stress , or early maternal separation ) to anxiety-related behaviors and cortisol release in offspring ( Bale , 2015 ) . Finally , a growing number of toxins and drugs have been shown to induce effects on various offspring phenotypes ( Skinner et al . , 2011; Vassoler et al . , 2013; Yohn et al . , 2015; Zeybel et al . , 2012 ) . A key challenge in such studies at present is to understand how the offspring phenotype is related to the stimulus presented in the paternal generation – in other words , how specific is the offspring response ? This challenge is compounded by the fact that many of the stimuli used for paternal effect paradigms – low protein and high fat diets , social stressors , and endocrine disruptors – have pleiotropic effects on organismal physiology . We therefore sought to develop a paternal effect paradigm based on a defined ligand-receptor interaction , to enable pharmacological interrogation of the specificity of the offspring phenotype . Nicotine is a commonly-used drug in humans , and acts by binding to and activating nicotinic acetylcholine receptors ( nAChRs ) , ligand-gated cation channels normally activated by the endogenous neurotransmitter acetylcholine . Maternal use of nicotine has been linked to multiple phenotypes in offspring ( Yohn et al . , 2015; Zhu et al . , 2014 ) , and although effects of paternal nicotine exposure have been less-studied , paternal smoking in humans has been suggested to affect metabolic phenotypes in children ( Pembrey et al . , 2006 ) . Here , we develop a rodent model for paternal nicotine effects , asking ( 1 ) whether exposure of male mice to nicotine could impact phenotypes in offspring , and ( 2 ) whether any affected phenotype would be specific for nicotine . We found that paternal exposure to nicotine induced a protective response in the next generation , as male offspring of nicotine-exposed fathers exhibited significant protection from nicotine toxicity . Importantly , this toxin resistance was not specific to nicotine , instead reflecting a more general xenobiotic response – offspring of nicotine-exposed fathers exhibited increased hepatic expression of a variety of genes involved in clearance of xenobiotics , and these animals were resistant to cocaine as well as to nicotine toxicity . Finally , we found that enhanced resistance to nicotine toxicity was also observed in offspring of males treated with the nicotine antagonist mecamylamine , strongly suggesting that drug resistance in offspring is a common outcome of paternal exposure to multiple xenobiotics rather than a specific response arising from nicotine signaling . Taken together , our results describe a novel paternal effect paradigm , and demonstrate that in the case of paternal nicotine exposure , the phenotype observed in offspring is a relatively generic response – enhanced xenobiotic resistance – rather than a selective downregulation of the specific molecular pathway subject to paternal perturbation .
We established a paternal exposure paradigm in which male mice were either provided with nicotine hydrogen tartrate ( nicotine 200 μg/ml free base , sweetened with saccharine ) in their drinking water , or a control solution of tartaric acid and saccharine . Mice consumed nicotine or control solutions ( NIC or TA , respectively ) from 3 weeks of age until 8 weeks of age . As previously described ( Zhao-Shea et al . , 2015 ) , this administration regimen maintains a high level of nicotine in the bloodstream ( Figure 1—figure supplement 1A–B ) , and results in nicotine dependence in exposed animals ( Zhao-Shea et al . , 2013 ) . Males were then withdrawn from nicotine for one week prior to mating in order to prevent any potential for seminal fluid transmission of nicotine ( the half-life of nicotine in mice is ~10 min , the half-life of its ‘long-lived’ metabolite cotinine is ~40 min [Siu and Tyndale , 2007] ) . Nicotine and control males were then mated with control females . Overall , we observed no difference in average size or sex ratio of litters arising from control or nicotine matings , or in offspring body weights ( Figure 1—figure supplement 1C–F ) . We first sought to determine whether the enforced nicotine withdrawal in our exposure paradigm might result in a paternal stress response that could affect the phenotype of progeny . As anxiety-related behaviors have been reported in offspring of males subject to several distinct stress paradigms ( Dietz et al . , 2011; Gapp et al . , 2014; Short et al . , 2016 ) ( albeit not all such paradigms – [Rodgers et al . , 2013] ) , we therefore assessed anxiety behaviors in TA and NIC offspring . Importantly , we observed no differences between TA and NIC offspring in time spent in the center during an open field anxiety test , or in time spent or number of entries into the open arms of an elevated plus maze ( Figure 1—figure supplement 2 ) . These results and results discussed below ( see Figure 6 ) indicate that our nicotine administration paradigm does not induce a stress response robustly enough , or for long enough prior to mating , to affect offspring phenotype . We next asked whether paternal nicotine administration could more specifically affect nicotine-related phenotypes in the next generation . We first focused on a physiological readout of offspring sensitivity to nicotine , using a well-established assay for suppression of locomotor activity by acute nicotine administration ( Tapper et al . , 2004 ) . Briefly , after acclimating animals to a saline injection protocol for three days , animals are injected with either nicotine ( 1 . 5 mg/kg ) or saline , and immediately introduced to a novel environment . Saline-injected animals actively explore the novel environment , and locomotor activity is quantified over a 40 min time course ( Figure 1 – Baseline ) . In this paradigm , injection of nicotine results in rapid suppression of locomotor activity , followed by a gradual recovery of exploratory behavior over the time course of the assay . Using this assay , we observed no significant difference in nicotine sensitivity between TA and NIC offspring , either for male or female offspring ( Figure 1 , Figure 1—figure supplement 3 ) . We therefore conclude that the acute locomotor suppression response to nicotine is not altered by our paternal nicotine exposure paradigm . 10 . 7554/eLife . 24771 . 003Figure 1 . Nicotine suppression of locomotor activity is unaffected by paternal nicotine history . Nicotine effects on locomotor activity were assayed in male offspring of control ( TA ) or nicotine-exposed ( NIC ) fathers . Data for females and alternative administration regimens are shown in Figure 1—figure supplement 3 . For each plot , males were injected with either saline or nicotine immediately prior to being placed in a novel environment for 40 min , during which locomotor activity was assessed by the number of times the animal interrupted a light beam during each minute . Each time point shows the number of beam crossings in that minute , shown as average plus/minus s . e . m . for all animals tested . Importantly , here and throughout the manuscript , the listed number of animals represent the number of litters analyzed , as we only assess one animal per litter in a given assay . Data are shown for saline injection ( ‘Baseline’ ) – exploratory behavior decreases over time in saline-injected animals as they habituate to the locomotor cage – and for 1 . 5 mg/kg nicotine injection in animals naïve to nicotine ( Day 1 ) or following five or eight prior days of the same nicotine injection and locomotor assessment protocol . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 00310 . 7554/eLife . 24771 . 004Figure 1—figure supplement 1 . Physiological effects of nicotine exposure on treated males . ( A ) Weight of males subject to 5 weeks of exposure to nicotine ( NIC ) or control ( TA ) solution . Data are shown for animals at the end of 5 weeks of nicotine exposure ( ** indicates p<0 . 01 ) , and following a week of withdrawal ( n . s . : not significant ) , as indicated . ( B ) Blood levels of cotinine , a relatively long-lived metabolite of nicotine , in males ( at 8 weeks of age , following 5 weeks of nicotine/control treatment ) consuming control or nicotine solutions . ( C ) Average litter size for offspring of control and nicotine-treated males . Data show average plus/minus s . e . m . ( D ) Average gender ratio for offspring of control and nicotine-treated males . Data show average plus/minus s . e . m . ( E–F ) Average weights for male ( E ) and female ( F ) TA and NIC offspring at 4 , 5 , and 6 weeks of age . Data are shown as average plus/minus S . D . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 00410 . 7554/eLife . 24771 . 005Figure 1—figure supplement 2 . Paternal nicotine exposure does not affect offspring anxiety-related behaviors . ( A–B ) Data are shown for elevated plus maze performance – time spent in open arms ( A ) , or total entries into the open arms ( B ) – for TA offspring ( n = 7 ) and NIC offspring ( n = 11 ) . ( C–J ) Open field test performance , shown for the first 10 min ( C–F ) or first 5 min ( G–J ) following introduction of the animal into the enclosure . Panels show total distance moved ( C , G ) , velocity ( D , H ) , fraction of time spent in the center of the open field enclosure ( E , I ) , and cumulative time spent in the central zone ( F , J ) . All data show average +/- s . e . m . None of the differences between TA and NIC offspring are statistically significant . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 00510 . 7554/eLife . 24771 . 006Figure 1—figure supplement 3 . No significant effects of paternal nicotine exposure on offspring locomotor response to nicotine . For each column , offspring of control and nicotine-treated males ( TA and NIC , respectively ) were subject to a locomotor activity assay as follows . Animals were first acclimated to intraperitoneal saline injections once a day for three days . On day three ( Baseline ) , offspring were injected with saline , then placed in a novel environment – a box equipped with infrared photodiodes to enable detection of locomotor activity . Saline-injected animals actively explore the novel environment , and locomotor activity is quantitated over a 40 or 90 min time course by the number of times the animal interrupts the light beam . Exploratory behavior decreases over time in saline-injected animals as they fully explore the enclosure . On nine subsequent days ( data for four representative days are shown in each column ) , animals are injected with nicotine ( 1 . 5 or 2 . 0 mg/kg , as indicated ) and immediately introduced to the measurement box . In this paradigm , injection of nicotine results in rapid suppression of locomotor activity , followed by a gradual recovery of exploratory behavior over the time course of the assay . Data here are shown as mean plus/minus s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 006 We next sought to identify any effects of paternal nicotine exposure on nicotine reinforcement in offspring using an operant self-administration assay ( Fowler et al . , 2011 ) . Here , after surgical implantation of a catheter into the superior vena cava , animals are subject to caloric restriction and trained to nose-poke an active portal to self-administer ( SA ) sucrose . TA and NIC offspring exhibited similar behavior during the training period , with the exception of a modest albeit significant difference in sucrose SA on the final day of dietary training ( Figure 2—figure supplement 1 ) . After seven days of food shaping , animals were placed in the operant chamber , a nicotine infusion pump was connected to the central catheter , and the dietary reward for nose-poking the active portal was replaced with a nicotine infusion . The amount of nicotine self-administered every day was then measured per session over the course of 35 days , with the nicotine infusion dose increasing every 4–8 days ( Materials and methods ) . Overall , there was no difference in daily nicotine SA between offspring of control males and offspring of nicotine-exposed males ( Figure 2A ) , indicating that nicotine reward behavior is not significantly reprogrammed by our paternal exposure paradigm . 10 . 7554/eLife . 24771 . 007Figure 2 . Paternal experience affects nicotine toxicity , but not self-administration , in offspring . ( A ) Paternal nicotine exposure does not affect nicotine self-administration in offspring . Each day , a mouse trained to self-administer nicotine ( Materials and methods ) was connected to the self-administration apparatus for one hour , with the dose of nicotine administered via cannula for every correct nose poke ramping up every 4–8 days , as indicated . Total nicotine self-administered is shown for each day of the protocol as average and s . e . m . Note that the numbers of animals participating in the trial decreased over time due to removal from the protocol ( clogged catheter ) or death – the listed n represents all animals that remained on the protocol until death . ( B ) Offspring of nicotine-exposed fathers exhibit significant protection from nicotine toxicity . Survival curve is shown for all animals on the self-administration protocol ( underlying data are provided in Figure 2—source data 1 ) . Nicotine offspring exhibited significantly increased survival during the time course of the assay relative to control offspring ( Kaplan-Meyer survival curve , p<0 . 0001 for both Log-rank test and Gehan-Breslow-Wilcoxon test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 00710 . 7554/eLife . 24771 . 008Figure 2—source data 1 . Offspring of nicotine-exposed fathers exhibit significant protection from nicotine toxicity . Survival curve for all animals on the nicotine self-administration protocol . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 00810 . 7554/eLife . 24771 . 009Figure 2—figure supplement 1 . Modest effect of paternal nicotine exposure on dietary training . Following surgical implantation of a central line , TA and NIC offspring were allowed to recover for 3 days . Animals were then subject to caloric restriction ( 80% of daily diet w/w compared to animals feeding ad libitum ) , placed in a self-administration box with two buttons , one of which was marked with a small light . Animals were then provided with sucrose pellets in response to a nose poke on the lit button – for 3 days a pellet was provided following each correct nose poke , then for one more day two nose pokes were required for a pellet , and finally five nose pokes were required for a food pellet for 3 days . Bars here show the number of food pellets earned in 1 hr for TA and NIC offspring – NIC offspring earned moderately more sucrose pellets in the final reward regime than TA offspring ( p=0 . 03 ) . This enhanced food training carried over to the first day of nicotine self-administration ( Figure 2A ) , when NIC animals self-administered slightly more nicotine than TA animals , but this difference only persisted for the first day . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 009 Nonetheless , a clear phenotype emerged serendipitously from the SA paradigm . We found that in our strain background , the escalating nicotine dosing schedule of SA resulted in death of nearly all animals tested at the highest doses used . Surprisingly , NIC offspring survived for many more days , on average , than TA offspring ( Figure 2B ) . This difference in survival was highly significant ( Gehan-Breslow-Wilcoxon p<0 . 0001 ) . As there was no difference in the daily levels of nicotine administered by either group ( Figure 2A ) , this result suggests that paternal nicotine exposure can protect offspring from nicotine toxicity . As TA and NIC offspring exhibit differences in their resistance to lethal doses of nicotine despite no difference in the daily level of nicotine consumed , we asked whether the effect of paternal nicotine exposure on offspring survival could be recapitulated using a single dose nicotine challenge , rather than the laborious self-administration protocol described above . This nicotine challenge was performed using two distinct paradigms . First , we simply challenged offspring of control or nicotine fathers with a single dose injection of nicotine – these ‘naïve’ animals had had no prior direct exposure to nicotine . In addition , we reasoned that since the animals in the self-administration paradigm were consuming nicotine for several weeks prior to eventual exposure to lethal levels of the drug ( Figure 2B ) , this would be expected to substantially alter nicotine-related biology in the tested animal . We therefore also subjected TA and NIC offspring to one week of chronic low-dose nicotine ( supplied in the drinking water ) – we refer to these animals as the ‘chronic’ cohort – then challenged these animals with an injection of a single LD50 dose of nicotine . As shown in Figure 3A , naïve TA and NIC offspring exhibited no significant difference in susceptibility to a toxic nicotine injection , indicating that paternal nicotine exposure does not program a constitutively nicotine-resistant state . In contrast , and consistent with the results of the self-administration test , male ( but not female ) offspring of nicotine-exposed fathers became significantly more tolerant to a lethal nicotine challenge than control offspring ( Figure 3B ) , but only once they had become acclimated to a week of chronic nicotine . Taken together , these data demonstrate that male offspring of nicotine-exposed fathers exhibit an enhanced ability to develop tolerance to toxic doses of nicotine , but that this tolerance is only revealed following prior exposure to sub-lethal levels of nicotine . 10 . 7554/eLife . 24771 . 010Figure 3 . Paternally-induced protection from nicotine toxicity is primed by nicotine exposure in offspring . ( A ) Survival of TA or NIC offspring following a single injection of nicotine at the indicated dose . Above each bar , fraction shows the number of surviving animals over number of animals injected . For all four doses tested , there was no significant difference in toxicity between TA and NIC offspring ( p>0 . 7 across all four doses for males , p>0 . 8 for females ) . ( B ) Survival of TA and NIC offspring following a single injection of nicotine at roughly the LD50 for naïve animals in ( A ) – 7 . 2 mg/kg for male offspring , shown in the top panel , 5 . 04 mg/kg for females , shown in the bottom panel . Here , offspring were acclimated to chronic nicotine in their drinking water for 6 days , with nicotine challenge being administered 24 hr following the last day of nicotine consumption . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 010 What is the physiological basis for the enhanced resistance to nicotine toxicity observed in NIC offspring relative to TA offspring ? Lethal doses of nicotine induce seizures originating in the hippocampus ( Fonck et al . , 2003 ) . Resistance to such seizures could result from highly specific resistance mechanisms such as downregulation of nicotinic acetylcholine receptors in the hippocampus , or from relatively nonspecific resistance mechanisms such as enhanced detoxification of xenobiotics in the liver . Although we cannot definitively rule out a neural basis for the enhanced nicotine resistance observed in NIC offspring , several lines of evidence – including extensive RNA-Seq analysis of isolated hippocampus – argue against this resistance resulting from altered neural physiology ( Figure 4—figure supplement 1 , Supplementary file 1 ) . In contrast to the lack of relevant molecular changes observed in the brains of NIC offspring , we discovered a significant effect of paternal nicotine exposure on hepatic detoxification of nicotine in offspring . As shown in Figure 4A , nicotine-acclimated NIC offspring exhibit significantly higher levels of the long-lived nicotine metabolite cotinine at earlier time points after nicotine injection than do TA offspring . This finding is consistent with enhanced nicotine clearance underlying the nicotine resistance phenotype displayed by these animals , suggesting that paternal nicotine exposure programs a state of enhanced metabolic tolerance in offspring . 10 . 7554/eLife . 24771 . 011Figure 4 . Paternal nicotine exposure induces an exaggerated protective response to xenobiotics . ( A ) Paternal nicotine exposure enhances nicotine metabolism in offspring . Male TA and NIC offspring were acclimated to nicotine for 6 days , then 24 hr later were injected with 1 . 5 mg/kg nicotine . Serum levels of the long-lived nicotine metabolite cotinine were measured at the indicated times after nicotine injection , with significantly ( p<0 . 0002 , t-test with Holm-Sidak correction ) elevated cotinine levels being observed at the earliest time point analyzed , indicating enhanced nicotine clearance in NIC offspring . ( B ) Schematic of hepatocyte RNA-Seq experiment . ( C ) Cluster of hepatocyte RNA-Seq dataset . For each paternal treatment group ( TA or NIC ) , data are shown for ten individual male offspring from ten separate litters , with hepatocytes from five animals also being cultured for varying times ( 0 to 21 hr ) following isolation . Data are z score normalized for each culture time point . The heatmap shows 60 genes ( filtered for average expression >25 ppm ) changing with a multiple hypothesis-corrected p value<0 . 1 . Underlying data are provided in Figure 4—source data 1 . ( D ) Genes upregulated in NIC offspring encode enzymes involved in all three phases of xenobiotic metabolism , as indicated . ( E ) Selected Gene Ontology categories enriched among genes upregulated ( adjusted p<0 . 1 ) in NIC hepatocytes . ( F ) ATAC-Seq coverage for TA and NIC hepatocytes , as indicated , across Nr1i3 . See also Figure 4—figure supplement 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 01110 . 7554/eLife . 24771 . 012Figure 4—source data 1 . Cluster of hepatocyte RNA-Seq dataset . RNA-Seq data for hepatocytes obtained from TA and NIC offspring , showing 60 genes ( filtered for average expression >25 ppm ) changing with a multiple hypothesis-corrected p value <0 . 1 . Data are z score normalized for each culture time point . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 01210 . 7554/eLife . 24771 . 013Figure 4—figure supplement 1 . Paternal nicotine has no significant effects on offspring hippocampal gene regulation or neural activity . ( A–B ) RNA-Seq of isolated hippocampus from TA and NIC offspring . Scatterplots show average mRNA abundance ( minimum tpm of 10 ) , with x axis showing average for TA offspring and y axis showing average for NIC offspring . ( A ) shows data for TA and NIC animals that had not experienced nicotine ( ‘naïve’ ) , ( B ) shows data for animals provided with chronic nicotine for 6 days . There are no significant effects of paternal nicotine exposure on any mRNAs in either condition . Similar results were obtained in preliminary studies of the ventral tegmental area , nucleus accumbens , and prefrontal cortex ( not shown ) . Importantly , while recent reports document an increase in Igf2 mRNA abundance in the hippocampus of 8 week old male offspring of stressed fathers ( Short et al . , 2016 ) , we observed no significant change in Igf2 levels in NIC offspring ( see Supplementary file 1 ) , providing another argument against the hypothesis that our paternal nicotine exposure paradigm affects offspring via a paternal stress response . ( C–D ) Paternal nicotine treatment does not affect offspring neural activity in the hippocampus . Representative images showing c-fos staining as a proxy for neural activity in hippocampus isolated from TA ( C ) and NIC ( D ) offspring . Here , animals were put on chronic nicotine ( 200 µg/ml nicotine free-base ) for six days . After 24 hr without nicotine , animals were injected with 1 . 5 mg/kg nicotine free-base . Tissue was collected 90 min after the injection . ( E ) Quantitation of c-fos staining data . Y axis shows number of c-fos-positive neurons in the gyrus dentatus for nicotine-injected TA ( n = 12 ) , and NIC ( n = 13 ) offspring . Several lines of evidence thus argue against the drug resistance of NIC offspring resulting from altered neural physiology . First , the fact that drug-acclimated animals exhibit enhanced resistance to both nicotine and cocaine toxicity ( Figures 3 and 5 ) rules out mechanisms involving downregulation or desensitization of either nicotinic acetylcholine receptors or the dopamine receptor . Second , RNA-Seq analysis of several brain regions – hippocampus , ventral tegmental area , nucleus accumbens , and prefrontal cortex -- revealed minimal effects of paternal nicotine exposure on the transcriptome . Finally , we found no significant differences in staining patterns of the activity marker c-fos in the hippocampus of TA and NIC offspring . Thus , while we cannot definitively rule out a neural basis for the phenotypes observed in NIC offspring , we found no evidence to support such a hypothesis . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 01310 . 7554/eLife . 24771 . 014Figure 4—figure supplement 2 . Paternal nicotine exposure affects multiple phenotypes in offspring . ( A–B ) Glucose tolerance ( A ) and insulin tolerance ( B ) are significantly altered in NIC offspring relative to TA offspring . Plasma glucose levels are shown for six male NIC or TA offspring at varying times after a 2 g/kg glucose bolus at 7 weeks of age ( A ) , or a 0 . 75 U/kg insulin bolus at 10 weeks of age ( B ) . * and ** represent p values of <0 . 05 and <0 . 01 , respectively ( t-test ) . ( C ) q-RT-PCR data for the indicated genes . In each case , expression level ( after normalizing to Actb and Gapdh ) is plotted relative to the average expression level for 4 TA livers ( n = 6 NIC livers ) , with bars showing average and s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 01410 . 7554/eLife . 24771 . 015Figure 4—figure supplement 3 . Global differences in hepatocyte chromatin architecture between TA and NIC offspring . Aggregated ATAC-Seq data for hepatocytes isolated from TA and NIC offspring ( n = 4 animals each , with dexamethasone-treated and -untreated samples for each animal ) . ( A ) Multimegabase-scale differences in the accessibility landscape of TA and NIC hepatocytes . Top two panels show ATAC-Seq data for NIC and TA offspring for chromosome 6 , along with an averaged NIC-TA score , followed by gene density . For chromosome 7 , only NIC-TA and gene density are shown . Red boxes highlight a subset of genomic regions of low gene density in which TA hepatocytes exhibit greater ATAC-Seq signal than NIC hepatocytes . Conversely , gene-dense regions generally exhibit higher ATAC signal in NIC hepatocytes ( not highlighted ) . The mechanistic basis for this global difference is unclear – it does not appear to reflect contamination of TA samples with dead cells , for example , as these samples ( from nicotine-naïve animals ) did not differ in viability , and plating of hepatocytes also effectively selects against dead cells . As fragment length distributions were consistent from library to library , it also seems unlikely that there were gross differences in the concentration or activity of the added Tn5 . Nonetheless , while this difference could reflect meaningful biology , such as a global difference in heterochromatin condensation , global differences in any genome-wide assay should of course be viewed with skepticism . ( B ) ATAC-Seq data for 500 bp surrounding all annotated transcription start sites ( TSSs ) , sorted from high to low average ATAC signal intensity . ( C ) Increased ATAC signal in NIC hepatocytes is shown for all TSSs , or RXRA or LXRB binding sites , as indicated . ( D–H ) Examples of loci exhibiting enhanced chromatin accessibility in NIC offspring , relative to TA offspring . ( D–E ) show ATAC-Seq tracks in which TA and NIC data are set to the same vertical range , as in Figure 4F . For panels F-H , y axes are set independently for TA and NIC datasets , visually correcting for the global differences between TA and NIC datasets . In these panels , a subset of significantly NIC-enriched peaks ( Supplementary file 3 ) are indicated with arrows . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 015 What is the molecular basis for the enhanced nicotine detoxification observed in NIC offspring ? As the liver is the primary site of nicotine and other xenobiotic clearance in mammals , we investigated changes in mRNA abundance in hepatocytes isolated from TA and NIC offspring ( Figure 4B–C , Supplementary file 2 ) . Paternal nicotine exposure significantly ( adjusted p<0 . 05 ) affected the expression levels of 51 genes , with upregulated genes being significantly enriched for those involved in lipid metabolism ( p=3 . 9e-14 ) , amino acid catabolism ( p=6 . 6e-8 ) , and various mitochondrial annotations including mitochondrial membrane ( p=1 . 9e-7 ) ( Figure 4D–E , Figure 4—figure supplement 2A–B ) . Most notably , given the nicotine resistance observed at the organismal level , NIC hepatocytes also exhibited increased expression of genes involved in drug metabolism ( p=4 . 3e-6 ) , with upregulated genes including ‘Phase I’ ( Cyp1a2 , Cyp2c68 ) and ‘Phase II’ ( Ugt2a3 , Ugt2b1 , Sult1d1 , and Sult1a1 ) detoxification enzymes , ‘Phase III’ membrane transporters ( Slco1a4 ) , as well as genes encoding the xenobiotic-responsive nuclear hormone receptors CAR and PXR ( Nr1h3 and Nr1i2 ) ( Figure 4C–D ) . In addition , the primary cytochrome involved in nicotine clearance in rodents , Cyp2a5 , was upregulated ~2 fold on average in NIC hepatocytes . Although this upregulation was not significant ( adjusted p=0 . 2 ) in the genome-wide dataset due to sample to sample variability in expression of this gene , we validated upregulation of Cyp2a5 in additional intact livers ( n = 6 NIC , n = 4 TA , p<0 . 01 ) by q-RT-PCR ( Figure 4—figure supplement 2C ) . These gene expression studies thus reveal that , relative to TA hepatocytes , NIC hepatocytes exhibit a general derepression of target genes for a broad range of nuclear hormone receptors . To investigate the mechanistic basis for this derepression , we characterized open chromatin genome-wide in TA and NIC hepatocytes ( n = 8 samples each ) using ATAC-Seq ( Buenrostro et al . , 2015 ) . Our ATAC-Seq dataset exhibited expected features such as strong peaks of accessibility over promoters and other regulatory elements ( Figure 4—figure supplement 3 ) . Comparing TA and NIC datasets , we observed a consistent global difference in overall chromatin accessibility – normalized ATAC peaks at regulatory elements were nearly 2-fold higher in NIC hepatocytes than in TA hepatocytes , while TA hepatocytes exhibited a consistently higher background of transposition throughout regions of the genome distant from regulatory elements ( Figure 4—figure supplement 3A–C ) . Whatever the basis for this global change in chromatin accessibility , we additionally identified 1861 peaks of chromatin accessibility ( Figure 4F , Figure 4—figure supplement 3D–H , Supplementary file 3 ) that differ significantly between TA and NIC hepatocytes after correcting for the global difference in peak height between these samples . Consistent with the changes in mRNA abundance observed in hepatocytes , these peaks were significantly enriched near genes involved in lipid metabolism ( p=2 . 8e-18 ) and xenobiotic metabolism ( p=1 . 3e-6 ) , along with many related GO categories . We conclude that a history of paternal drug exposure can influence the chromatin landscape of hepatocytes in offspring , resulting in a broad increase in accessibility at regulatory elements involved in metabolism and detoxification . Importantly , the gene expression program observed in isolated hepatocytes includes a broad variety of genes associated with drug metabolism , most of which are not specific for nicotine clearance . To test the hypothesis that the nicotine-resistant state of NIC offspring reflects a general xenobiotic response , rather than a nicotine-specific detoxification pathway , we asked whether NIC offspring also exhibit enhanced resistance to another toxic challenge , cocaine . As cocaine and nicotine operate through distinct molecular pathways – cocaine prevents dopamine reuptake at the synaptic cleft by binding to and blocking the dopamine transporter , while nicotine activates and desensitizes nicotinic acetylcholine receptors – a finding of enhanced tolerance to cocaine would strongly argue against NIC offspring exhibiting specific epigenetic effects on the direct molecular receptor for nicotine . We first assessed cocaine toxicity in ‘naïve’ animals that had not been previously directly exposed to nicotine or cocaine . Similar to our findings with nicotine toxicity ( Figure 3A ) , naïve NIC and TA animals did not exhibit significant differences in their resistance to cocaine toxicity ( Figure 5A ) . However , as the enhanced ability of NIC offspring to survive toxic nicotine levels was only revealed following pre-exposure of these animals to sub-lethal doses of nicotine ( Figure 3B ) , we next sought to determine whether acclimation of NIC offspring to cocaine could induce a cocaine-resistant state . To address this question , TA and NIC offspring were chronically treated with sub-lethal doses of cocaine – twice-daily injections of 15 mg/kg cocaine for five days – prior to challenge with a toxic dose of cocaine . Astonishingly , this acclimation protocol resulted in enhanced resistance to cocaine toxicity in NIC offspring , relative to TA controls ( Figure 5B ) , revealing that NIC offspring are hyper-responsive to multiple xenobiotics . 10 . 7554/eLife . 24771 . 016Figure 5 . NIC offspring are protected from multiple xenobiotics . ( A ) Paternal nicotine exposure does not affect susceptibility of drug-naïve offspring to cocaine toxicity . Male TA and NIC offspring were injected with a single 100 mg/kg dose of cocaine . Survival is shown as in Figure 3 . ( B ) Acclimation of TA and NIC offspring to either nicotine or to cocaine reveals protective effect of paternal nicotine exposure on offspring cocaine resistance . As in ( A ) , for male offspring acclimated to chronic nicotine ( 200 µg/mL nicotine free-base in drinking water for six days ) or cocaine ( twice-daily injections with 15 mg/kg cocaine for five days ) . Twenty-four hours following final drug exposure , animals were injected with a single 100 mg/kg dose of cocaine . ( C ) Cocaine acclimation induces nicotine resistance in NIC offspring . Here , male TA and NIC offspring were acclimated to cocaine injections ( twice-daily , 15 mg/kg ) over five days . Twenty-four hours after the final cocaine injection , animals were injected with 7 . 2 mg/kg nicotine . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 016 We next asked whether the process of acclimation to sub-lethal doses of nicotine or cocaine induces a drug-specific resistant state in NIC offspring . In other words , does pre-acclimation of NIC offspring to different molecules induce resistance specifically to the drug to which the animals were exposed , or do chronic exposures to multiple distinct drugs all induce a common state of general xenobiotic resistance ? To distinguish these possibilities , we pre-acclimated TA and NIC offspring to either nicotine or cocaine , then challenged acclimated animals with a lethal dose of the drug to which they had not yet been exposed . Consistent with the hypothesis that drug acclimation induces a general xenobiotic response , we found that pre-acclimation to nicotine induced a cocaine-resistant phenotype in NIC offspring , and , conversely , that chronic cocaine could induce nicotine resistance ( Figure 5B–C ) . Together , these data suggest that paternal nicotine exposure programs a hyper-responsive state in male offspring in which chronic xenobiotic exposure results in a generalized toxin resistance . The revelation that nicotine resistance in NIC offspring reflects a somewhat generic xenobiotic resistance program ( Figures 4C–D and 5 ) raises the question of what aspect of the paternal nicotine exposure paradigm is responsible for programming the offspring phenotype . The nicotine exposure paradigm utilized here induces nicotinic acetylcholine receptor ( nAChR ) signaling , with several physiological consequences: ( 1 ) nicotine dependence , ( 2 ) reduced caloric intake , and ( 3 ) physiological withdrawal resulting from the removal of nicotine for the final week prior to mating . To investigate the role of nAChR signaling in the paternal induction of offspring drug resistance , we made use of mecamylamine , a non-selective , non-competitive antagonist of nAChRs that readily crosses the blood-brain barrier . Male mice were provided with 2 . 0 mg/kg/day mecamylamine via a surgically-implanted infusion pump , and mecamylamine-treated mice were split to either nicotine or TA drinking water , as in our primary nicotine exposure paradigm . Studies have previously shown that mecamylamine administration prevents known physiological responses to nicotine such as nicotine-induced anorexia ( Mineur et al . , 2011 ) , hypothermia and locomotor effects ( Tapper et al . , 2004 ) , and nicotine reinforcement ( Corrigall and Coen , 1989 ) . Male offspring of these fathers were then acclimated to nicotine for 6 days , then subject to a toxic nicotine challenge , as in Figures 3 and 5 . Surprisingly , male mice concurrently treated with nicotine and its antagonist fathered offspring with the same enhanced nicotine resistance seen in NIC offspring ( Figure 6 ) . Importantly , this finding rigorously rules out the possibility that our nicotine exposure paradigm induces paternal effects on offspring as a consequence of the nicotine withdrawal stress imposed in the week before mating . 10 . 7554/eLife . 24771 . 017Figure 6 . Offspring drug resistance is induced by a nicotine antagonist . Here , we modified the paternal exposure paradigm by implanting pumps to deliver the nicotine antagonist mecamylamine to male mice . Mecamylamine-treated mice were provided with nicotine or control solution for four weeks , then mated to control females . Male offspring were acclimated to chronic nicotine for six days and then subject to a toxic nicotine challenge , and survival is shown as in Figures 3 and 5 . Data for no mecamylamine animals are reproduced from Figure 3B . Note that concurrent mecamylamine and nicotine exposure resulted in a protective effect on offspring , and even mecamylamine alone was able to modestly induce nicotine resistance in the next generation . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 017 Moreover , the drug resistance observed in nicotine+mecamylamine offspring strongly argues that this paternal effect does not even require nicotine signaling in treated fathers , instead suggesting that the paternal effect is perhaps induced simply by exposure to xenobiotics . Consistent with this hypothesis , mecamylamine exposure alone also induced drug resistance in the next generation , although this effect was not as robust as that induced by nicotine or nicotine+mecamylamine ( Figure 6 ) . Together , these data demonstrate that drug resistance in sons can be induced by paternal exposure to both nAChR agonists and nAChR antagonists , arguing that paternal xenobiotic exposure is likely to be the relevant feature of our nicotine exposure paradigm . Finally , we sought to understand the requirement for drug acclimation in revealing organismal drug resistance in NIC offspring . Curiously , the relative upregulation of xenobiotic processing genes ( XPGs ) in NIC offspring was observed in hepatocytes and livers isolated from ‘naïve’ animals that had not been exposed to nicotine or cocaine ( Figure 4 ) , yet enhanced resistance to toxins was only observed in animals that were first acclimated to one of these drugs ( Figures 3 and 5 ) . To test the hypothesis that XPG upregulation might be even stronger in NIC hepatocytes following drug exposure , we set out to characterize gene expression changes in nicotine- or cocaine-acclimated offspring . However , in attempting to isolate hepatocytes from drug-acclimated TA and NIC offspring for RNA-Seq analysis , we noticed much poorer recovery of hepatocytes from TA than from NIC offspring ( not shown ) , suggesting the possibility that NIC animals might be protected from drug-induced hepatotoxicity . Therefore , to quantify cell viability in vivo , we took a histochemical approach to assess apoptosis in livers from drug-acclimated TA and NIC offspring . Consistent with the relatively poor recovery of hepatocytes from TA animals , we observed substantial hepatocyte death in the livers of cocaine-exposed animals ( Figure 7A ) . Importantly , while hepatocyte apoptosis and necrosis were extremely common in livers from cocaine-exposed TA offspring , NIC offspring were significantly protected from such cocaine toxicity ( Figure 7 ) . We conclude that the upregulation of XPGs in naïve NIC offspring is not sufficient to significantly protect animals from a lethal nicotine or cocaine challenge , but that this upregulation can protect hepatocytes from sub-lethal doses of these drugs . Following a week of chronic toxin exposure , TA offspring are left with substantially reduced liver function , while NIC offspring maintain greater numbers of functional hepatocytes . We speculate that this greater hepatocyte functional capacity , as well as the upregulation of XPGs in hepatocytes ( Figure 4 ) , may both serve to protect the animal from a single toxic dose of xenobiotic . 10 . 7554/eLife . 24771 . 018Figure 7 . NIC offspring exhibit relative sparing of hepatocytes following chronic drug exposure . ( A–B ) Effects of chronic cocaine treatment on hepatocyte viability . Two representative sections are shown for TUNEL-stained livers from TA ( A ) and NIC ( B ) offspring following five days of cocaine injections ( twice-daily , 15 mg/kg ) . Prominent centrilobular apoptosis is seen in TA offspring , but is almost completely absent in NIC offspring . ( C–D ) Quantitation of TUNEL staining data . ( C ) shows the average ( plus/minus s . e . m . ) number of TUNEL+ centrilobular regions per slide ( staining of >25% of central vein circumference was counted as TUNEL+ , and was assessed at five different levels for each liver lobe I-IV ) for four individual TA ( blue ) and NIC ( red ) offspring , treated as in ( A–B ) . ( D ) shows data for all individual slides as dots , with boxplot showing median , one standard deviation , and 5th/95th percentile for the 80 data points . DOI: http://dx . doi . org/10 . 7554/eLife . 24771 . 018
The use of nicotine , a well-characterized small molecule that acts in vivo by binding to specific receptors , as the inciting paternal exposure enabled us to rigorously interrogate the specificity of the offspring response . Importantly , the enhanced toxin survival seen in offspring is not specific for the drug to which fathers were exposed – NIC offspring were hyper-resistant to both nicotine and to cocaine challenges – demonstrating that our paternal exposure paradigm does not result in transmission of a nicotine-specific phenotype to progeny ( at least for toxicity , locomotor effects , and reward behavior ) . Mechanistically , the drug resistance observed in NIC offspring presumably results from the enhanced hepatic drug clearance observed in these animals ( Figure 4A ) . Consistent with this increased nicotine clearance , isolated hepatocytes exhibited upregulation of a variety of xenobiotic processing genes ( XPGs ) accompanied by greater chromatin accessibility at relevant regulatory regions . A variety of XPGs are induced in NIC hepatocytes in addition to those known to play a role in nicotine clearance ( Figure 4C ) , suggesting that NIC offspring may prove resistant to many toxins beyond the two tested in this study . In addition to the significant derepression of xenobiotic response genes observed in NIC offspring , we note that the most significant effects of paternal nicotine on offspring hepatocyte gene expression occurred at metabolic genes ( Figure 4C , E ) . This finding suggested that NIC offspring might also exhibit metabolic alterations , in addition to the documented changes in xenobiotic resistance . Alterations in glucose control and lipid metabolism are commonly observed in paternal effect studies , being observed not only in dietary paradigms but also in some stress and toxin-related paternal effect studies ( Rando and Simmons , 2015 ) , suggesting that multiple distinct stimuli experienced by males might in some way convergently influence metabolic traits in offspring . As a detailed metabolic phenotyping of NIC offspring is beyond the scope of this study , we chose here to simply focus on the most common phenotype observed in other paternal effect experiments , assaying glucose and insulin tolerance in TA and NIC offspring ( Figure 4—figure supplement 2A–B ) . Consistent with the ability of multiple paternal environments to alter glucose control in offspring , we observed that NIC offspring exhibited significantly diminished clearance of a glucose bolus , as well as a moderately diminished response to insulin . Taken together , our data reveal ( 1 ) that paternal nicotine exposure induces a pleiotropic set of phenotypes in male offspring , and ( 2 ) that the induced phenotypes in offspring are not specific for nicotine . It will be of great interest in future studies to interrogate a wide variety of phenotypes in offspring of males subject to a broad range of exposure paradigms – including stress , nicotine treatment , and various diets – to identify common and divergent phenotypes induced by distinct paternal exposure paradigms . A curious feature of many , but not all , paternal effect paradigms reported in mammals is that phenotypic effects often manifest preferentially in offspring of one gender . For example , while paternal social defeat was reported to affect anxiety-related behavior in both male and female offspring , locomotor activity and sucrose preference were only altered in male offspring ( Dietz et al . , 2011 ) . Here , we find that paternal nicotine exposure only affects drug resistance in male offspring , raising once again the unsolved question of why paternal environments induce gender-specific outcomes in progeny . Here , we consider three potential explanations for this phenomenon . First , a subset of epigenetic information carriers – cytosine methylation and chromatin packaging – are associated in cis with a specific genomic locus , meaning that epigenetic changes occurring on the sex chromosomes will only affect progeny inheriting that chromosome . Thus , it is plausible that nicotine exposure affects epigenetic modification of the Y chromosome to program drug resistance in male offspring ( or , less simply , that epigenetic marks on the X chromosome suppress an autosomal or small RNA-directed phenotype that would otherwise affect both male and female progeny ) . Second , X chromosome dosage compensation in mammals occurs via silencing of one of the two X chromosomes in females . The inactive X chromosome could thus act as a ‘sink’ for epigenetic silencing machinery in females ( Blewitt et al . , 2005 ) , such that the effective levels of this machinery available for autosomal gene regulation could differ between males and females . In this scenario , paternal transmission of an epigenetically-marked autosomal locus , or RNA , could cause differential effects in developing male vs . female offspring based on differences in the available levels of epigenetic effector machinery . Finally , we note that an emerging theme in many paternal effect paradigms is that the phenotypic changes observed in offspring are known to be regulated by various nuclear hormone receptors ( NHRs ) . For example , the phenotypes described in paternal stress paradigms are related to glucocorticoid receptor signaling , while the metabolic gene expression changes resulting from paternal dietary interventions exhibit significant overlap with genes regulated by NHRs such as PPARα ( Carone et al . , 2010 ) . Here , we find that paternal nicotine exposure affects hepatic expression of many targets of metabolic NHRs , as well as the xenobiotic-responsive NHRs CAR and PXR ( Figure 4 ) . As sex hormones also act through NHR signaling – androgen receptor and estrogen receptor – we speculate that levels or activity of NHR coactivators or corepressors could differ in male vs . female progeny , resulting in altered penetrance or magnitude of paternal effects on NHR-mediated gene regulation . A crucial feature of the drug resistance exhibited by NIC offspring is that the toxin-resistant state is only revealed by pre-exposure of these animals to xenobiotics . This requirement for drug pre-exposure/acclimation emphasizes the key role of the offspring’s environment in the manifestation of an epigenetically ‘reprogrammed’ phenotype . In other words , the development of an animal’s phenotype here involves an interaction between environmental conditions in two consecutive generations ( see ( Rodgers et al . , 2013; Zeybel et al . , 2012 ) for similar examples ) – as with gene X environment effects , epigenetic marks also have context-dependent effects on organismal phenotype . What is the mechanism by which low level drug exposure enhances the survival of NIC offspring ? NIC hepatocytes exhibit derepression of xenobiotic response genes even before exposure to any drugs , yet these drug-naïve animals are no more resistant to nicotine or cocaine toxicity than control animals ( Figures 3A and 5A ) . Instead , the enhanced xenobiotic metabolism in NIC livers appears to protect susceptible hepatocytes from toxicity during a course of sublethal drug exposure ( Figure 7 ) . The loss of hepatocytes in drug-exposed TA animals presumably explains why fewer than 50% of these animals survive an LD50 dose – calculated using drug-naïve animals – of nicotine or cocaine ( Figures 3 and 5 ) , with the preservation of hepatic capacity in NIC offspring preventing this degradation in survivability . That said , not only do drug-acclimated NIC offspring simply preserve their survival in the face of an LD50 dose of these drugs , but they exhibit dramatically improved survival , as far more than half of these animals survive this challenge . We have yet to uncover the mechanistic basis for this enhanced survival , as RNA-Seq analysis of the hepatocytes isolated from drug-acclimated animals does not reveal further upregulation of XPGs than that documented for naïve hepatocytes ( not shown ) . Future studies will investigate whether drug acclimation might ( 1 ) affect mRNA abundance in a limited subset of hepatocytes ( which would be diluted out in whole liver or hepatocyte culture experiments ) , ( 2 ) affect mRNA abundance only transiently during drug exposure ( and not in cultured hepatocytes ) , leaving behind higher levels of the encoded proteins without an mRNA-Seq signature , or ( 3 ) affect xenobiotic metabolism not at the level of mRNA abundance , but post-transcriptionally . The pleiotropic response observed in nicotine-exposed offspring raises the question of how nicotine is sensed in the paternal generation in this system . A key question in this regard is whether stress experienced by the nicotine-exposed males might be responsible for inducing the offspring phenotype , as it is known that a variety of paternal stress exposure paradigms – including early maternal separation , social defeat stress , and chronic variable low level stress – affect multiple phenotypes in offspring , from glucose control to anxiety-related behaviors ( Bale , 2015 ) . While we have not formally ruled out a role for paternal stress in our system – it will of course be of interest to assay offspring nicotine resistance in well-studied paternal stress paradigms – two findings strongly argue against this paternal effect arising from a general stress response . First , chronic exposure to the nicotinic receptor antagonist mecamylamine , which blocks nicotine dependence in nicotine-treated fathers , does not interfere with induction of xenobiotic resistance in offspring ( Figure 6 ) , thus definitively ruling out a role for paternal withdrawal stress in induction of this phenotype . This first point is further supported by the finding that mecamylamine alone – which on its own has little effect on anxiety , locomotor behavior , or physical withdrawal symptoms in nicotine-naïve mice ( Zhao-Shea et al . , 2013 ) – is sufficient to induce xenobiotic resistance in offspring . Second , in contrast to multiple reported paternal stress paradigms , we do not find any evidence that paternal nicotine exposure affects anxiety-related behavior in offspring ( Figure 1—figure supplement 2 ) . What , then , is the relevant feature of nicotine in inducing xenobiotic resistance in offspring ? Paternal effects on toxin resistance in offspring did not require nicotinic receptor signaling , as both nicotine itself as well as a nicotine antagonist were able to induce the protective response in offspring . As both nicotine and mecamylamine exposure can result in reduction of nAChR signaling via desensitization or antagonism , respectively , it is formally possible that nAChR deactivation is the inciting stimulus in the paternal generation ( or , less likely , that the surgical stress of mecamylamine infusion pump implantation , and nicotine consumption , both convergently induce the same effect in offspring ) . However , we favor the simpler hypothesis that both of these molecules serve to program offspring drug resistance via effects on paternal xenobiotic sensing . This model naturally raises the question of how xenobiotic exposure is sensed . As a diverse variety of xenobiotics can affect gene regulation via activation of the NHRs CAR and PXR , these NHRs represent appealing candidates for the relevant xenobiotic sensor in fathers . Whatever the nature of the relevant xenobiotic sensor , a key challenge to address is why experimental exposure to nicotine or mecamylamine ( or , presumably , many other xenobiotics ) reprograms offspring drug resistance relative to control animals , given that control animals are also exposed to a multitude of small molecules even in controlled laboratory conditions . Do nicotine and mecamylamine somehow induce a switch-like ‘all or none’ change in some epigenetic mark that is not present in control sperm , or is the overall activity level of a xenobiotic sensor translated into quantitative changes in the levels of some continuous signal present in sperm ? In the former case , what aspects of a given exposure paradigm are required to induce alterations to the sperm epigenome ? We offer that one appealing mechanism for sensing increased levels of environmental xenobiotics would rely on comparing changes in sensor activity over an animal’s lifetime . For instance , if CAR/PXR signaling early in life – in utero perhaps , or early in postnatal life – were to result in a long-lasting ‘setpoint’ for the levels of CAR/PXR activity expected later in life , then the organism could detect increased xenobiotic exposure later in life via changes in overall CAR/PXR activity compared to this setpoint . Future studies will explore the nature of the ‘nicotine’ sensor in the paternal generation , and how information about exposure history is transmitted to offspring . Taken together , our studies define a novel paternal exposure paradigm based on a specific ligand-receptor interaction , and show that paternal nicotine exposure programs offspring for enhanced resistance to multiple distinct toxins . Our data also reveal broad metabolic gene expression changes in NIC offspring , with potential implications for metabolic and cardiovascular health of offspring . Future studies will determine whether paternal nicotine exposure affects offspring via epigenetic marks in the sperm ( vs . seminal fluid , etc . ) , and how paternally-transmitted information alters the course of development to result in xenobiotic-resistant hepatocytes . It will also be of interest to extend these studies to human populations , where the longer half-life of nicotine could potentially result in self-administration phenotypes not observed in the mouse model .
C57BL/6J mice ( RRID:IMSR_JAX:000664 ) , three weeks old , were obtained from Jackson labs on a weekly basis and group-housed ( four mice/cage ) on a 12 hr light-dark cycle ( 7:00 A . M . to 7:00 P . M ) . After arrival , males were immediately put on either tartaric acid ( TA , 375 µg/ml ) or nicotine ( 200 µg/ml nicotine free-base ) in drinking water for five consecutive weeks , followed by an additional week on tap water prior to mating . Nicotine-exposed and control males were then allowed to mate ( for six days ) with control females from the same shipment date . F1 offspring from nicotine-exposed and control fathers were used for all experiments reported , generally at eight weeks of age unless otherwise noted . Animals were maintained on-site in accordance with an approved IACUC protocol ( A-1788 ) . F1 males from nicotine-exposed and control fathers were pre-conditioned to handling and injections with 0 . 9% saline ( 100 µl , i . p . ) for three days prior to start of the study . For the nicotine test sessions , animals were injected with nicotine and transferred to individual cages placed within an infrared photobeam frame ( San Diego Instruments ) . Test sessions lasted 40 or 90 min per day for nine consecutive days . Locomotor activity was defined as the number of beam breaks during a session , whereupon the animal had to cross at least two photobeams from the original location to count as ambulation . Results were statistically quantified using unpaired t-tests with multiple comparison adjustment ( Holm-Sidak correction ) . Microsurgical catheter implant was performed on 7-week old F1 males from nicotine-exposed and control fathers . Animals were anaesthetized with ketamine ( 100 mg/kg BW ) and xylazine ( 10 mg/kg BW ) followed by a intrascapular and right midclavicular incision at the level of the carotid sheath . Blunt preparation was used to create a subcutaneous canal between the two incisions . Subsequently , the vena jugularis dextra was located and a catheter ( 2Fr , PV 10 cm , Instech Labs ) was inserted and gently pushed forward into the vena cava superior , where it remained for the length of the study . The catheter was ligated to the vein using Ethibond Excel 4 . 0 . The distal end of the catheter was connected to a button ( 25 G , VAB , Instech . Labs ) , which was placed subcutaneously in an intrascapular position for easy access . After verifying that there was no leakage , the incision sites were closed with Ethibond Excel 4 . 0 . Through the catheter , the mouse was treated with heparin ( 15 I . U . , Sigma-Aldrich , St . Louis , MO ) and an antibiotic mix of Ticarcillin ( 20 mg booster , Sigma-Aldrich ) and Amikacin ( 10 mg/kg BW , Sigma-Aldrich ) . Animals received Ketoprofen ( 5 mg/kg BW , Sigma-Aldrich ) once daily during a 3-day recovery phase . Afterwards , mice were put on a caloric restriction diet ( 85% w/w of regular 24 hr consumption ) three days prior to start of the experiment . We preconditioned animals on sucrose pellets in a 60 min session once a day for seven consecutive days , whereby animals learned to nose-poke the active portal in a self-administration chamber in order to receive food reward . The number of nose-pokes required to get a sucrose pellet escalated starting with a fixed ratio of 1:1 ( FR1 ) up to a fixed ratio of 5:1 ( FR5 ) . Only animals that had successfully been conditioned on sucrose pellets advanced to the testing phase , during which they administered nicotine to themselves through the implanted catheter . Catheter patency was verified daily by aspiration of blood and subsequent heparin infusion . Animals with blocked or dislocated catheters were excluded from the study . The self-administered nicotine doses started with 0 . 03 mg/kg/injection for 4 days , then 0 . 1 mg/kg/injection for 8 days , 0 . 25 mg/kg/injection for 8 days , and 0 . 4 mg/kg/injection for 8 days . The number of nose-pokes of the active versus the inactive portal , as well as the number of injections administered , were recorded and analyzed using GraphPad Prism 7 . 0 and multiple t-tests with Holm-Sidak correction . Survival was plotted as a Kaplan-Meier curve with significance levels calculated using modified Chi-square tests ( Log-rank and Gehan-Breslow-Wilcoxon ) . Blood of F1 males from nicotine-exposed or control fathers was collected in EDTA-coated tubes after injection of 1 . 5 mg/kg nicotine free-base i . p . at 15 min , 30 min , and 45 min post-injection . Cellular components were separated from serum by centrifugation at 12 , 000 ×g for 10 min . Cotinine levels in serum of chronic F1s were measured using a Direct ELISA kit ( CalBiotech Inc . ) . Samples were run as two technical replicates together with a cotinine standard curve for each 96-well plate . Analysis was performed using GraphPad Prism 7 . The elevated plus maze consisted of four arms connected by a central axis ( 5 × 5 cm ) and was elevated 45 cm above the floor . Two of the arms contained plastic black walls ( 5 × 30 × 15 cm ) while the other two remained open ( 5 × 30 × 0 . 25 cm ) . Mice were individually placed on the center of the maze with their heads facing one of the open arms and allowed 5 min of free exploration . The number of entries into the open and closed arms , and the total time spent in the open and closed arms was measured by MED-PC IV software ( MED associates , Inc . ) . The apparatus was thoroughly cleaned between animals . For activity in the open field , mice were placed in a rectangular arena made of Plexiglas ( 40 × 40 × 30 cm ) and mouse activity was video recorded for 10 min . Total activity , velocity , and time spent in the peripheral and central area of the open field was analyzed using video tracking software ( Noldus Ethovision ) . F1 males from TA- and nicotine-exposed fathers were treated as for transcriptome analysis and phenotype studies . Briefly , animals received nicotine in their drinking water ( 200 µg/ml nicotine free-base ) for six consecutive days starting at seven weeks post-natum . Afterwards , mice were put on filtered tap water from 12:00 P . M . until 7:00 A . M . the next day followed by immediate tissue collection . Brains of additional eight-week old control animals are dissected 90 min after i . p . injection of 1 . 5 mg/kg BW nicotine free-base . Animals were anesthetized with sodium pentobarbital i . p . ( 200 mg/kg BW ) followed by intracardial infusion of 10 ml ice-cold PBS and 10 ml paraformaldehyde ( PFA; 4% w/v in PBS ) . Brains were kept at 4°C in 4% PFA for 2 hr and then transferred into 30% sucrose ( w/v in PBS ) until slice preparation . Brains were sectioned using a microtome ( Leica ) into 25 µm slices and immersed in a 50% glycerol , 50% ethylene glycol solution ( Sigma ) to preserve the tissue . Brain slices were stored in −20°C until further processing . Using the free-floating immunostaining method , slices were washed with PBS for 5 min , permeabilized with 0 . 5% ( v/v ) Triton X-100 ( Sigma ) for 10 min , and blocked with 3% donkey serum for 30 min . The slices were incubated overnight at 4°C with antibodies against c-Fos ( 1:1000 , catalog number: sc-52 , lot number: D2315 , Santa Cruz Biotechnology , Santa Cruz , CA ) . After washes with PBS , slices were incubated with Alexa Fluor 594 secondary antibodies ( 1:1000 , ref number: A21207 , lot number: 1602780 , Life Technologies , Carlsbad , CA ) . Counterstaining was carried out with DAPI through mounting media ( Cat number: H-1200 , lot #: ZB0730 , Vector , Burlington , CA ) . Fluorescent images were captured using an AxioCam MRm camera ( Carl Zeiss , Peabody , MA ) attached to a Zeiss Axiovert inverted fluorescent microscope equipped with Zeiss filter sets 38HE , 49 , and 20 . Zeiss objectives A-p were subsequently processed using Axiovision version 4 . 8 . 2 . Quantification of c-Fos-positive cells was performed using ImageJ , with a minimum of 6 hippocampal brain slices analyzed per animal . Seven week-old male F1 animals from control ( TA ) and nicotine-exposed fathers were divided into three treatment groups: naïve , chronic , and chronic + stimulation . Naïve mice were not exposed to nicotine before tissue collection at 8 weeks of age . Chronic animals received nicotine in their drinking water ( 200 µg/ml ) for six consecutive days . Afterwards , chronic mice were put on filtered tap water from 12:00 P . M . until 7:00 A . M . the next day followed by tissue collection as for naïve animals . Chronic + stimulation animals were treated as chronic animals , but received an additional nicotine injection ( 1 . 5 mg/kg BW nicotine free base i . p . ) 30 min before organ harvest . For all three sets of animals , following sacrifice brains were explanted and put on ice . A midline incision was executed and midbrain , hypothalamus , and hippocampus of either side were dissected . Tissues were immediately immersed in liquid nitrogen , then stored at −80°C until further processing . Eight week-old male F1 animals from control ( TA ) and nicotine-exposed fathers were anaesthetized using ketamine ( 100 mg/kg BW ) and xylazine ( 10 mg/kg BW ) . The abdominal cavity was opened with a transverse incision below the rib cage . The portal vein was dissected with blunt forceps and a 26 G catheter needle was inserted . After cutting the vena cava inferior cranial of the liver , the organ was perfused firstly with 1X HBBS +200 mM EDTA ( 10 ml at 7 ml/min ) and secondly with 50 ml DMEM containing collagenase type I ( 0 . 4 mg/ml ) at 7 ml/min . The liver was then removed from the abdominal cavity , put in a petri dish containing culture medium ( DMEM , 20% FBS , 1X ITS , 1X Penicillin/Streptomycin , 0 . 1 µM Dexamethasone , Sigma-Aldrich ) , and gently dissected to allow release of hepatocytes and supporting cells from connective tissue . Note that due to the disaggregation of the entire liver , mRNA abundance changes observed in a subset of hepatocytes ( such as , for example , dying cells in drug-acclimated animals – Figure 7 ) will be diluted out by the majority of unaffected hepatocytes . After filtration through a 70 µm nylon cell strainer , cells were washed twice with PBS 1X and once with culture media ( centrifugation at 500 rpm for 5 min ) , and plated on a 0 . 1% gelatin-coated well . Hepatocytes were allowed to adhere to the bottom of the well for three hours . Nonadherent cells were then removed , and fresh culture medium was then added , initiating our time course ( T0 , T1 , T3 , T21 hours ) . Cells were collected after a PBS 1X wash by adding TriZol to the well for RNA experiments . Strand-specific libraries were prepared as previously described ( Zhang et al . , 2012 ) . Briefly , brain and liver were collected from nicotine-exposed and control F1 males . Hepatocytes were isolated as described above . For the hippocampus , after sectioning of brain into 1 mm slices , areas of interest were identified according to the Mouse Brain Atlas by Paxinos and Franklin and dissected using 0 . 5 mm punches . RNA from brain and liver was isolated using standard TriZol protocols , followed by rRNA depletion ( RiboZero kit , Illumina , Inc . ) . After first- and second-strand synthesis , adapters were ligated to fragments and amplified using multiplexed PCR primers . Libraries were sequenced on a NextSeq 500 platform from Illumina , Inc . Quality-controlled reads were aligned to the reference genome ( Mus musculus/mm10 ) with Bowtie2 and differential expression was calculated using DESeq2 . For multiple comparison adjustments , we used Holm-Bonferroni correction as a more conservative approach . RNA-Seq data are available at GEO , accession # GSE94059 . ATAC-seq libraries were prepared for 16 hepatocyte samples ( 4 NIC and 4 TA animals , with each sample split into untreated and dexamethasone-treated aliquots ) as previously described ( Buenrostro et al . , 2015 ) using the Nextera DNA Library Preparation Kit ( Illumina ) . Libraries were paired-end sequenced on a NextSeq 500 , and reads were aligned to mm10 using Bowtie2 , v2–2 . 1 . 0 with the parameters -D 15 R 2 N 1 L 20 -i S , 1 , 0 . 50 --maxins 2000 --no-discordant --no-mixed . Mitochondrial DNA and random chromosome mapped reads were removed , and PCR duplicates were removed . Genome browser images were generated from merged datasets with reads extended to 150 bp , and normalized by total mapped reads per sample . For differential peak analysis , HOMER was used to identify NIC-specific peaks using TA peak files as background . ATAC-Seq data are available at GEO , accession # GSE92240 . Livers were harvested from F1 males from nicotine-exposed and control fathers under various conditions ( pre-treatment with nicotine 1 . 5 mg/kg BW intraperitoneal b . i . d . for five days or cocaine 15 mg/kg BW intraperitoneal b . i . d . or acetaminophen 400 mg/kg BW q . d . for one day ) and washed with PBS . A 4 mm slice was taken from each lobe and put in ice-cold 4% formaldehyde overnight . The next day , samples were dehydrated in a series of escalating ethanol solutions starting with 70% and ending with 100% , embedded in paraffin , and sectioned ( 4 µm slices ) , each section containing all four lobes , which were then mounted onto a glass slide . For H/E staining , slices were de-parafinized , incubated with xylene and a series of descending ethanol solutions . Incubation times for Mayer’s hematoxylin ( Sigma-Aldrich ) and 1% Eosin Y ( Sigma-Aldrich ) were 30 s and 20 s , respectively . After dewaxing of tissue , TUNEL staining was performed following the manufacturer’s recommendations ( in Situ Cell Death Detection Kit , POD , Roche ) . Apoptotic areas per lobe were counted under a light microscope with 20X magnification at five different levels through the sample and analyzed with Image J . | Until recently , it seemed impossible that the conditions a person or animal experiences during their lifetime might affect the health of their offspring and future generations . Research over the past decade , however , has shown that a parent’s environment can cause changes that can be passed to future generations . For example , studies in rodents have shown that a father’s diet influences the way their offspring metabolize food . Moreover , a male mouse exposed to stress or toxins fathers pups that often respond differently in stressful situations relative to other mice . So , how do these traits get transferred to offspring via sperm and how specific is the next generation’s response to the environmental pressures faced by their fathers ? Many studies so far have looked at environmental influences that may have broad biological effects , for example a high fat diet . Now , some scientists are trying to understand whether exposure to nicotine , which has a more targeted effect , causes drug-specific effects in offspring . Vallaster et al . now show that mice whose fathers had been exposed to nicotine before mating are more able to withstand toxic levels of the chemical than mice whose fathers were never exposed to the drug . In the experiments , some male mice were given water with nicotine in it over the course of five weeks . Later , the offspring of these mice were exposed to nicotine to see whether they were more or less sensitive to it than offspring of unexposed males . It turns out the mice with nicotine-exposed fathers have a higher resistance to the toxic effects of nicotine and , unexpectedly , to toxic levels of cocaine as well . This suggests that the pups of nicotine-exposed fathers are not specifically programmed to respond to nicotine , but instead are more resistant to toxins in general . Vallaster et al . found that the livers of the offspring of nicotine-exposed fathers appear to be better able to metabolize both drugs . Exposing the fathers to another drug called mecamylamine ( which can prevent many of nicotine’s effects on the body ) also made their offspring more resistant to nicotine , showing that multiple drugs may make offspring more toxin-resistant . Studies in humans will be needed to confirm whether a father’s nicotine use affects children the same way it does mice . Similar mice studies also may help scientists to study how other types of environmental exposure might affect a man’s future children . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2017 | Paternal nicotine exposure alters hepatic xenobiotic metabolism in offspring |
Transcription activator-like effectors ( TALEs ) bind DNA through an array of tandem 34-residue repeats . How TALE repeat domains wrap around DNA , often extending more than 1 . 5 helical turns , without using external energy is not well understood . Here , we examine the kinetics of DNA binding of TALE arrays with varying numbers of identical repeats . Single molecule fluorescence analysis and deterministic modeling reveal conformational heterogeneity in both the free- and DNA-bound TALE arrays . Our findings , combined with previously identified partly folded states , indicate a TALE instability that is functionally important for DNA binding . For TALEs forming less than one superhelical turn around DNA , partly folded states inhibit DNA binding . In contrast , for TALEs forming more than one turn , partly folded states facilitate DNA binding , demonstrating a mode of ‘functional instability’ that facilitates macromolecular assembly . Increasing repeat number slows down interconversion between the various DNA-free and DNA-bound states .
Transcription activator-like effectors ( TALEs ) are bacterial proteins containing a domain of tandem DNA-binding repeats as well as a eukaryotic transcriptional activation domain ( Kay et al . , 2007; Römer et al . , 2007 ) . The repeat domain binds double stranded DNA with a register of one repeat per base pair . Specificity is determined by the sequence identity at positions twelve and thirteen in each TALE repeat , which are referred to as repeat variable diresidues ( RVDs ) ( Boch et al . , 2009; Miller et al . , 2015; Moscou and Bogdanove , 2009 ) . This specificity code has enabled design of TALE-based tools for transcriptional control ( Cong et al . , 2012; Geissler et al . , 2011; Li et al . , 2012; Mahfouz et al . , 2012; Morbitzer et al . , 2010; Zhang et al . , 2011 ) , DNA modifications ( Maeder et al . , 2013 ) , in-cell microscopy ( Ma et al . , 2013; Miyanari et al . , 2013 ) , and genome editing ( TALENs ) ( Christian et al . , 2010; Li et al . , 2011 ) . TALE repeat domains wrap around DNA in a continuous superhelix of 11 . 5 TALE repeats per turn ( Deng et al . , 2012; Mak et al . , 2012 ) . Because TALEs contain on average 17 . 5 repeats ( Boch and Bonas , 2010 ) , most form over 1 . 5 full turns around DNA . Many multisubunit proteins that form rings around DNA require energy in the form of ATP to open or close around DNA ( reviewed in O'Donnell and Kuriyan , 2006 ) , yet TALEs are capable of wrapping around DNA without energy from nucleotide triphosphate hydrolysis . One possibility is that TALEs bind DNA through an energetically accessible open conformation . Consistent with this possibility , we previously demonstrated that TALE arrays can populate partly folded and broken states ( Geiger-Schuller and Barrick , 2016 ) . By measuring the length-dependence of protein stability and employing a statistical mechanics Ising model , we previously described several different TALE partly folded states termed ‘end-frayed’ , ‘internally unfolded’ , and ‘interfacially fractured’ states . Although the calculated populations of partly folded states in TALE repeat arrays are small , they are many orders of magnitude larger than populations of partly folded states in other previously studied repeat arrays ( consensus ankyrin [Aksel et al . , 2011] and DHR proteins [Geiger-Schuller et al . , 2018] ) suggesting a potential functional role for the high populations of partially folded states in TALE repeat arrays . Consensus TALEs ( cTALEs ) are designed homopolymeric arrays composed of the most commonly observed residue at each of the 34 positions of the repeat ( Geiger-Schuller and Barrick , 2016 ) . In addition to simplifying analysis of folding and conformational heterogeneity in this study , the consensus approach simplifies analysis of DNA binding , eliminating contributions from sequence heterogeneity and providing an easy means of site-specific labeling . Here we characterize DNA binding kinetics of cTALEs using total internal reflection fluorescence single-molecule microscopy . We find that consensus TALE arrays bind to DNA reversibly , with high affinity . Analysis of the dwell-times of the on- and off-states reveals multiphasic binding and unbinding kinetics , suggesting conformational heterogeneity in both the free and DNA bound state . We develop a deterministic optimization analysis that supports such a model , and provides rate constants for conformational changes in the unbound and bound states , and rate constants for binding and dissociation . Comparing the dynamics observed here to previously characterized local unfolding suggests that locally unfolded states inhibit binding of short cTALE arrays ( less than one full superhelical turn around DNA ) , whereas they promote binding of long arrays ( more than one full superhelical turn ) . Whereas local folding of transcription factors upon DNA binding is well documented ( Spolar and Record , 1994 ) , local unfolding in the binding process is not . Our results present a new mode of transcription factor binding where the major conformer in the unbound state is fully folded , requiring partial unfolding prior to binding . The critical role of such high energy partly folded states is an exciting example of ‘functional instability’ , in which formation of a functional complex is impeded by the fully folded native state , and is instead facilitated by partial disruption of native structure .
Consensus TALE ( cTALE ) repeat sequence design was described previously ( Geiger-Schuller and Barrick , 2016 ) . To avoid self- association of cTALE arrays , we fused arrays to a conserved N-terminal extension of the PthXo1 gene . Although the sequence of this domain shows little similarity to TALE repeat sequences , the structure of this domain closely mimics four TALE repeats ( Gao et al . , 2012; Mak et al . , 2012 ) and is required for binding and full transcriptional activation ( Gao et al . , 2012 ) . In this study , all repeat arrays contain this solubilizing N-terminal domain . In a previous study of the folding of a series of cTALE arrays , we used a nearest-neighbor Ising model to determine energies of intrinsic repeat folding and interfacial interaction between repeats . This analysis allows us to quantify the energies of different partly folded states . Figure 1A depicts three types of partly folded states of a generic repeat protein . In the fully folded state , all repeats are folded , and all interfaces are intact . In the end-frayed states , one or more terminal repeats are unfolded but all interfaces ( except the interface between the unfolded and adjacent folded repeat ) , are intact . In the internally unfolded states , a central repeat is unfolded but all interfaces ( except the interfaces involving the unfolded repeats ) are intact . In the interfacially ruptured state , all repeats are folded but one interface is disrupted due to local structural distortion . Figure 1B shows calculated free energy differences between various partly folded states and the fully folded repeat array for two different RVDs ( NS and HD ) in an otherwise identical consensus sequence background , using the intrinsic and interfacial engeries we determined previously ( Geiger-Schuller and Barrick , 2016 ) . The distribution of partly folded states is calculated for different 20-repeat arrays containing two types of TALE arrays ( with the NS RVD in red and with the HD RVD in blue ) as well as consensus ankyrin arrays ( cAnk in black; Materials and methods ) . For cTALE arrays , end frayed states are within a few kBT of the folded state , internally unfolded states are highest in energy , and interfacially ruptured states fall energetically between end frayed and internally unfolded states . Changing the RVD sequence affects the distribution of these partly folded states: arrays containing HD repeats are more likely to internally unfold or interfacially rupture than arrays containing NS repeats . However , both types of cTALEs are more likely to populate many of these partly folded states than cAnk is to populate even the lowest energy partly folded state , the end frayed state . Thus , compared to ankyrin repeats , cTALEs are locally unstable , meaning they are likely to form partly folded states . As these states disrupt the superhelix , they may facilitate DNA binding . To ask if cTALE local instability is relevant for DNA binding kinetics , DNA binding trajectories were measured using single molecule total internal reflection fluorescence ( smTIRF ) . Figure 2A shows a schematic of the smTIRF experiments performed to measure DNA binding . For site-specific cTALE labeling , R30 is mutated to cysteine in a single repeat . Position 30 is frequently a cysteine in naturally occurring TALEs ( in earlier folding studies , arginine was chosen in the consensus sequence to avoid disulfide formation; Geiger-Schuller and Barrick , 2016 ) . This cysteine was Cy3 ( FRET donor ) -labelled using maleimide chemistry , and the Cy3-lablelled TALE array was attached to biotinylated slides via the C-terminal His6 tag and α-Penta•His antibodies . At salt concentrations below 300 mM NaCl , cTALEs aggregate . Because DNA binding is weak at high salt concentrations , measuring binding kinetics in bulk at high salt is not possible . However , tethering cTALEs to the quartz slide at high salt prevents self-association , even at the low salt concentrations required to study DNA binding kinetics . A histogram of NcTALE8 ( 8 NS-type repeats and the N-terminal domain ) labeled via a cysteine in the first repeat shows a single peak at zero FRET efficiency , as expected for donor-only constructs ( Figure 2B ) . To test for DNA binding to tethered cTALE constructs , we added 5'-Cy5 ( FRET acceptor ) -labeled 15 bp-long DNA ( Cy5 . A15/T15 ) to tethered NcTALE8 . This results in a new peak at a FRET efficiency of 0 . 45 , indicating that DNA binds directly to cTALE arrays . As DNA concentration in solution is increased , the peak at 0 . 45 FRET efficiency increases in population ( Figure 2C–D ) , suggesting a measurable equilibrium between free and bound DNA rather than saturation or irreversible binding . In support of this , single molecule time trajectories show interconversion between bound and unbound states , providing access to rates of binding and dissociation . As expected for reversible complex formation , the peak at 0 . 45 FRET efficiency can be competed away by adding mixtures of labeled and unlabeled DNA to pre-formed cTALE-labelled DNA complexes ( schematic shown in Figure 2E; pre-formed complex shown in Figure 2F , competition data shown in Figure 2G–H ) . Challenging pre-formed complexes with a mixture of 5 nM unlabeled DNA and 15 nM labeled DNA results in a slight decrease in the population of the peak at 0 . 45 FRET ( compare Figure 2F and G ) . Challenging with a mixture of 50 nM unlabeled DNA and 15 nM labeled DNA further decreases the peak at 0 . 45 FRET ( Figure 2H ) . To test whether tethering of cTALE arrays impacts DNA binding , we also performed experiments with tethered dsDNA and free Cy3-labeled NcTALE8 . Here , we tethered biotinylated , Cy5 ( FRET acceptor ) -labeled 15 bp-long DNA ( Cy5 . A15/biotin . T15 ) . Addition of Cy3-labeled NcTALE8 at high salt concentration results in a new peak at a FRET efficiency of 0 . 5 ( Figure 2—figure supplement 1 ) . These FRET distributions are similar to those obtained from tethering cTALEs to the surface and adding free dsDNA , suggesting that the interaction of cTALE arrays with dsDNA is not significantly impacted by surface immobilization . Since cTALEs tend to associate at physiological salt concentration , we used the format where cTALEs were tethered to the slide and incubated with freely diffusing dsDNA for all subsequent experiments . In addition to the short smTIRF movies used to generate smFRET histograms from many molecules , long movies were also collected to examine the extended transitions of individual molecules between the low- and high-FRET ( 0 and 0 . 45 ) ( Figure 3A–B ) . A transition from low to high FRET ( 0 to 0 . 45 ) indicates that the acceptor fluorophore on DNA moved close enough to the donor on the protein for FRET and is likely a binding event . A transition from high to low FRET ( 0 . 45 to 0 . 0 ) indicates the acceptor fluorophore on DNA moved too far away from the donor on the protein for FRET and is likely an unbinding event . Low-FRET states show low colocalization with signal upon direct excitation of the acceptor , confirming that high-FRET states are DNA-bound states and low-FRET states are DNA-free states ( Figure 3—figure supplement 1 ) . These long single molecule traces show both long- and short-lived low- and high-FRET states , indicating that kinetics are multi-phasic ( Figure 3A–B ) . Binding events ( transitions from low to high FRET ) become more frequent as bulk DNA concentration increases ( compare representative traces at 1 nM dsDNA to 15 nM dsDNA; Figure 3A and Figure 3B ) . Cumulative distributions generated from dwell times in the low FRET state at a given DNA concentration are best-fit by a double-exponential decay , indicating a minimum of two kinetic phases associated with binding events ( Figure 3C ) . Cumulative distributions generated from dwell times in the high FRET state are also best-fit by a double-exponential decay , indicating that there are a minimum of two kinetic phases for unbinding as well ( Figure 3D ) . The rate constant for the fast phase in DNA binding shows a linear increase with DNA concentration ( Figure 3E ) , indicating that this step involves an associative binding mechanism . The slope of the rate constant for the fast phase as a function of DNA concentration gives a bimolecular rate constant of 5 . 9 × 108 nM−1s−1 , close to the diffusion limit . The rate constant for the slower phase ( 0 . 14 s−1 ) is independent of DNA concentration indicating a unimolecular isomerization mechanism ( Figure 3E ) . In contrast , neither of the two fitted rate constants for transitions from high to low FRET ( 0 . 45 to 0 . 0; unbinding events ) depends on DNA concentration , suggesting that unbinding involves two ( or more ) unimolecular processes ( Figure 3F ) . The rate constants of these two phases are 1 . 2 s−1 and 0 . 13 s−1 respectively . To rule out kinetic contributions of TALEs threading axially onto the ends of short DNAs , binding kinetics were measured with capped double-helical DNA sites . Capped DNA was generated by forming 5’digoxygenin-A5-Cy5-A15 duplexed with 5’-digoxygenin-T26 and adding a three-fold molar excess of anti-Digoxygenin . Low and high FRET dwell time cumulative distributions generated from capped DNA-binding kinetics are bi-phasic , similar to distributions from uncapped DNA ( Figure 3—figure supplement 2 ) . The DNA concentration-independent rate constant for binding is roughly the same for capped DNA as for uncapped DNA ( compare FRETL→H red and blue triangles in Figure 3—figure supplement 2 ) , as are the dissociation rate constants ( compare FRETH→L red and blue triangles in as well as FRETH→L red and blue circles in Figure 3—figure supplement 2 ) . The rate constant for bimolecular binding of capped DNA decreases compared to that for uncapped DNA ( compare FRETL→H red and blue circles in Figure 3—figure supplement 2 ) , which is consistent with the expected decrease in the rate of diffusion of the larger capped DNA . To assess the effect of molecular weight increase on diffusion of capped versus uncapped DNA , Sednterp ( Laue et al . , 1992 ) , a program commonly used to estimate sedimentation and diffusion properties of biomolecules , was used to estimate maximum diffusion coefficients . Including the two antibodies bound on the ends of capped DNA ( 320 kDa total ) gives an estimated diffusion coefficient of 4 . 7 × 10−7 cm2s−1 , which is much lower than the estimated diffusion coefficient of the uncapped DNA ( 1 . 5 × 10−6 cm2s−1 ) . This ~3 . 6 fold decrease in the diffusion constant for capped DNA is similar to the 6 . 7-fold decrease in the bimolecular rate constant for binding of capped DNA ( Figure 3—figure supplement 2 ) . The heterogeneity in the DNA-bound state may either result from conformational heterogeneity of the bound cTALE array , or from heterogeneity in the registry between the cTALE array and the DNA . Because there are three more base pairs than TALE repeats , there are several available binding registers where all TALE repeats are bound to DNA; ignoring end-effects , these registers are expected to have similar energetics . Although variation in registry would not be expected for natural TALE arrays that bind to high-complexity DNA sequences , it is more likely for the simple poly-A sequence here . To test whether bound-state heterogeneity results from a variation in cTALE-DNA registry , we altered the dsDNA sequence to promote a specific bound state . Previous studies indicate that the addition of a 5’ T to the binding sequence , referred to as a T-anchored binding site , greatly enhances TALE binding activity ( Boch et al . , 2009 ) depending also on the presence of certain RVDs and degree of mismatch relative to cognate DNA ( Schreiber and Bonas , 2014 ) . To monitor kinetics with a T-anchored DNA , we added Cy5-labeled 15 bp-long DNA ( Cy5 . TA14/T14A ) to tethered NcTALE8 . Similar to the homopolymeric DNA ( Cy5 . A15/T15 ) , this results in a peak at a FRET efficiency of 0 . 55 , which increases with increasing DNA concentration ( Figure 4A , B ) . To examine binding and unbinding kinetics of cTALEs interacting with T-anchored DNA , long movies were recorded to visualize multiple transitions of individual molecules between the low- and high-FRET states ( Figure 4A–B ) . As with A15/T15 DNA , these long single molecule traces show both long- and short-lived low-FRET states , consistent with multiphasic binding kinetics . In contrast , dissociation of T-anchored DNA only shows long-lived high-FRET states , suggesting a single kinetic phase for unbinding ( Figure 4A–B ) and cumulative distributions generated from dwell times in the high FRET state are well-fitted by a single-exponential decay ( Figure 4D ) . As with the binding kinetics measured for the homopolymeric DNA , the rate constant for the fast phase in T-anchored DNA binding shows a linear increase with DNA concentration ( Figure 4E ) . The bimolecular rate constant for this phase ( 3 . 9 × 108 nM−1s−1 ) is close to the diffusion limit , as was seen for homopolymeric DNA ( 5 . 9 × 108 nM−1s−1 ) . The rate constant for the slower phase ( 0 . 34 s−1 ) is also similar to the homopolymeric DNA rate constant for the slower phase ( 0 . 14 s−1 , Figure 4E ) . As with the unbinding kinetics measured for the homopolymeric DNA , the fitted rate constant for transition from high to low FRET ( 0 . 55 to 0 . 0; unbinding events ) does not depend on DNA concentration ( Figure 4F ) . Compared to the unbinding kinetics measured for the homopolymeric DNA , the anchored unbinding cumulative distributions are well-fitted by a single exponential ( although the model with bound-state heterogeneity still fits slightly better , as shown below ) . There are two possible interpretations of this result . Either the T-anchored DNA impacts the binding mechanism such that unbinding involves one simple unimolecular processes , or the T-anchored DNA shifts the microscopic rate constants such that , although unbinding involves two ( or more ) unimolecular processes , the amplitudes are very different , or apparent rates are too close to resolve them . Either way , the large effect of T-anchor on unbinding kinetics supports the idea that bound-state heterogeneity results from variation in the the registry between the cTALE array and the DNA . To examine how increasing the length of the cTALE array influences DNA binding , we generated Cy3-labelled constructs with 16 and 12 cTALE repeats , and measured binding to a longer Cy5-labelled DNA ( A23/T23 ) . We observed a low FRET value near 0 . 2 for the bound cTALE12 state , indicating that the first cTALE12 repeat is farther from the 5'-DNA-bound acceptor fluorophore than in the A15/T15 DNA complex . Attempts with the 16-repeat cTALE to increase FRET efficiency by moving the position of the mutated cysteine to the fourteenth repeat were unsuccessful . Thus , we used a fluorescence colocalization microscopy protocol to monitor binding of longer cTALE arrays to A23/T23 DNA ( Figure 5—figure supplement 1 ) . In this protocol , Cy3 was first imaged for ten camera frames ( 1017 . 5 msec total ) to identify positions of single TALE molecules . Then a long time series of fluorescence images of Cy5 signal were collected through directly exciting Cy5 on the DNA , and time trajectories of Cy5 signal were generated from the initially identified single TALE positions . Increasing the number of cTALE repeats from 8 to 12 and 16 dramatically affects DNA binding kinetics . Long movies collected over a range of DNA concentrations show short- and long-lived Cy5 signal on and off states , indicating a level of kinetic heterogeneity similar to NcTALE8 ( Figure 5—figure supplement 1 ) . Single molecule traces were analyzed using a thresholding filter ( see Materials and methods and Figure 5—figure supplement 1 ) to identify states and dwell times . Cumulative distributions were generated from dwell times at low Cy5 signal ( unbound states , with lifetimes representing binding kinetics ) , and at high Cy5 signal ( bound states , with dwell times representing unbinding kinetics ) . As with the eight repeat constructs , unbound cumulative distributions for these longer TALE arrays are best-fit by double exponential decays , particularly at high DNA concentrations ( compare the cumulative distribution at low DNA concentration , Figure 5—figure supplement 2A , to cumulative distribution at 5 nM DNA , Figure 5—figure supplement 2B ) . Bound cumulative distributions for longer TALE arrays are best-fit by double exponential decays ( Figure 5—figure supplement 2C–D ) . All apparent rate constants are much smaller for NcTALE16 and NcTALE12 ( green/black circles and triangles , Figure 5A–B ) compared to NcTALE8 , indicating that binding and unbinding is impeded by increasing the length of the binding surface between cTALEs and their cognate DNA ( Figure 5C ) . To address whether differences in binding kinetics are related to experimental differences between colocalization and FRET assays , alternating laser experiments were performed by switching between FRET and colocalization detection ( every five frames ) within single molecule trajectories ( Figure 3—figure supplement 1 ) . Changes in FRET and colocalization signals occurred simultaneously according to single molecule time traces , showing that differences in binding and unbinding kinetics of short and longer cTALEs are not due to differences in colocalization and FRET assays ( Figure 3—figure supplement 1 ) . To determine how the kinetic changes above are partitioned into underlying kinetic steps in binding , we fitted various kinetic models to the cumulative distributions for binding and unbinding . In addition to providing information about the mechanism of binding , this approach allows us to estimate the underlying microscopic rate constants and compare them for different constructs . This approach is generally applicable to studies of complex single molecule kinetics . Numerical integration was used to calculate the relative population of cTALE states as a function of time ( Figure 6A–C and G–I ) , given a binding mechanism , an associated set of rate laws , and a set of initial conditions . Cumulative distributions of unbound dwell times represent the distribution of times single molecules spent in the unbound state before transitioning into the bound state , allowing us to split the kinetic scheme when fitting to single-molecule dwell times . Among the various models tested , the model that is most consistent with the data has two unbound DNA-free states and two DNA-bound states . This is consistent with alternating laser experiments showing that DNA is only colocalized when cTALEs are in the high FRET state ( Figure 3—figure supplement 1 ) . This four-state model includes a TALE isomerization step in the absence of DNA from a DNA-binding incompetent conformation ( which we refer to as TALE ) to DNA-binding competent conformation ( which we refer to as TALE* ) . The DNA-binding competent TALE* conformer binds and unbinds DNA ( called TALE* when DNA free and TALE*~DNA when DNA-bound ) . Before unbinding , a fraction of TALE*~DNA isomerizes to a longer-lived DNA-bound state called TALE‡~DNA . Based on this mechanism , the rate laws for binding are given in Equations 1a - 1d . ( 1a ) d[TALE]dt=−k1[TALE]+k−1[TALE*] ( 1b ) d[TALE*]dt=k1[TALE]−k−1[TALE*]−k2[TALE*][DNA] ( 1c ) d[TALE*~DNA]dt=k2[TALE*][DNA] ( 1d ) Keq , DNA−free=k1k−1 Since the single-molecule dwell-time histograms of the unbound states are insensitive to the isomerization after DNA binding , the equation describing the time evolution of the long-lived bound state ( TALE‡~DNA ) is not relevant to our analysis of unbound-state lifetimes . To determine microscopic rate constants k1 , k-1 , and k2 , Equations 1a-1c were numerically integrated in Matlab , and the fraction of TALE*~DNA as a function of time was fitted to the low-FRET cumulative distributions ( NcTALE8; Figure 6D–E ) or to the no colocalization cumulative distributions ( NcTALE16; Figure 6J–K ) . Microscopic rate constants were adjusted to reduce sum of the squared residuals between the concentration of TALE*~DNA ( the direct product of binding ) as a function of time and single-molecule cumulative distributions . In both cases , cumulative distributions at different bulk DNA concentrations were fitted globally . Initial fractions of TALE and TALE*~DNA were set to zero , and the initial fraction of TALE* was set to one . Confidence intervals ( CI ) were estimated by bootstrapping ( Table 1; mean and 68% CI from 2000 or 8000 bootstrap iterations ) . Rate laws for dissociation are given in Equations 2a - 2d ( 2a ) d[TALE*~DNA]dt=−k−2[TALE*~DNA]−k3[TALE*~DNA]+k−3[TALE‡~DNA] ( 2b ) d[TALE‡~DNA]dt=k3[TALE*~DNA]−k−3[TALE‡~DNA] ( 2c ) d[TALE*]dt=k−2[TALE*~DNA] ( 2d ) Keq , DNA−bound=k3k−3 As with the system of equations above ( 1a-d ) , the equation describing the time evolution of the binding-incompetent free state ( TALE ) is not relevant to our analysis of bound-state lifetimes . To determine microscopic rate constants k-2 , k-3 , and k3 , Equations 2a-2c were numerically integrated in Matlab , and the fraction of TALE* as a function of time was fitted to the high-FRET cumulative distributions ( NcTALE8; Figure 6F ) or to the low colocalization cumulative distributions ( NcTALE16; Figure 6L ) . Microscopic rate constants were adjusted to reduce sum of the squared residuals between the concentration of TALE* ( the direct product of dissociation ) as a function of time and single-molecule cumulative distributions . In both cases , cumulative distributions at different bulk DNA concentrations were fitted globally . The initial fraction of TALE*~DNA conformer was set at one; all other initial fractions were set to zero . Confidence intervals were estimated by bootstrapping ( Table 1; mean and 68% CI from 2000 iterations ) . Fitted curves reproduce the experimental cumulative distributions for binding and unbinding ( Figure 6 ) , both for the short and long cTALE arrays , with reasonably small residuals , over a range of DNA concentrations . Generally , fitted rate constants have confidence intervals of 10% or smaller ( Table 1 ) . Comparison of microscopic rate constants for 8 , 12 , and 16 repeats show some significant differences . The bimolecular microscopic binding rate constant , k2 , is slightly larger for eight repeats than for 12 and 16 repeats ( 1 . 1 , 0 . 31 , and 0 . 39 nM−1s−1 for 8 , 12 , and 16 repeats respectively ) . However , microscopic unbinding rate constant , k-2 , is higher for eight repeat cTALEs ( 0 . 66 s−1 for NcTALE8 versus 0 . 13 s−1 for NcTALE12 and 0 . 299 s−1 for NcTALE16 ) . Clauß et al . ( 2017 ) have also observed TALE dissociation rates that are non-monotonic with repeat number in live cells . In addition , bound state isomerization ( interconversion between TALE*~DNA and TALE‡~DNA ) is 5–10 times slower for 16 and 12 repeat cTALEs than eight repeat cTALEs . The value of Keq , DNA-free , which is a measure of the equilibrium proportion of the unbound TALE that is DNA-binding competent ( TALE* ) to that which is binding-incompetent ( TALE ) , is larger for cTALEs with eight repeats ( Keq , DNA-free = 1 . 32 ) than for cTALEs with 12 and 16 repeats ( Keq , DNA-free = 0 . 11 and Keq , DNA-free = 0 . 61 , respectively ) .
NS is an uncommon RVD in natural TALEs . Previous reports suggest that NS is fairly nonspecific , but may bind with higher affinity than other common RVDs ( NG , NI , NN , and HD ) ( Miller et al . , 2015 ) . Our fitted rate constants can be used to calculate the apparent Kd ( Kapp ) calculated as follows: ( 3 ) Kapp=[TALE−DNA+TALE∗−DNA][DNA][TALE+TALE∗]=Keq , DNA−freeK2+Keq , DNA−freeK2Keq , DNA−bound1+Keq , DNA−free=k1k−1×k2k−2 ( 1+k3k−3 ) 1+k1k−1where K2=k2/k−2 . Using fitted rate constants from Table 1 in the final equality in Equation 3 gives values for Kapp of 2 . 5 nM for the eight repeat cTALE array , 0 . 5 nM for the 12 repeat cTALE array , and 1 . 0 nM for the 16 repeat cTALE array . Increasing the number of repeats from 8 to 12 repeats decreases the apparent Kd modestly , but further increasing from 12 to 16 repeats leaves the Kd unchanged . This affinity increase is small compared to that reported in a previous report studying length dependence on the affinity of designed TALEs ( dTALEs ) ( Rinaldi et al . , 2017 ) , although in that study affinity also became insensitive to repeat number for large arrays when KD , app was in the low nM range . Because the KD , app of cTALE8 is already in the low nM regime ( 2 . 5 nM ) , we speculate that this may represent a similar maximum binding affinity observed for the longer arrays in the previous Rinaldi et al . report . Thus , because cTALE8 is near a maximum affinity , the addition of four and eight cTALE repeats has only a modest impact on the apparent binding affinity . TALEs are believed to read out sequence information from one strand ( Boch et al . , 2009 ) . Due to the asymmetry of our DNA sequences ( poly-dA base-paired with poly-dT ) , in principle , the FRET efficiency contains information on the binding orientation ( and thus strand preference ) . However , based on the crystal structure of the DNA-bound state of TAL-effector PthXo1 ( Mak et al . , 2012 ) , we estimate that the distance between the donor site of NcTALE8 ( repeat 1 ) to the 5’ acceptor site on the DNA ( Cy5-A15/T15 ) should be similar for both the dA-sense or dT-sense orientations ( Figure 5—figure supplement 3A ) . Thus , the FRET data does not discriminate between the two modes of binding for the eight-repeat construct . However , for the 16 repeat NS RVD cTALE arrays , the PthXo1 model suggests very different distances ( 25 Å versus 73 Å for the dT-sense or dA-sense respectively , Figure 5—figure supplement 3B ) between the donor site ( TALE repeat 14 ) and the acceptor site ( 5’ Cy5-A23/T23 ) . To restrict the number of binding positions available to longer cTALE arrays , the 23 base pair DNA used for NcTALE16 measurements ( as well as DNA depicted in Figure 5—figure supplement 3B ) has the same number of additional base pairs as repeats ( eight additional repeats and eight additional base pairs ) compared to the 15 base pair DNA used for NcTALE8 measurements ( as well as DNA depicted in Figure 5—figure supplement 3A ) . While we limited the number of available binding positions , it may be possible for cTALEs to slide along DNA . However , taking into account the four repeat N-terminal capping domain , there are only three available base pairs in the bound complex . Thus we don’t expect the distance measurements to change by more than 10 Å ( ~3 base pairs ) if sliding occurs . The observation that there is colocalization but no measurable FRET when NcTALE16 is bound to DNA suggests that cTALEs containing the NS RVD prefer adenine ( the dA-sense mode ) compared with thymine bases , consistent with previous reports ( Boch et al . , 2009 ) . The cumulative distributions of dwell-times in Figure 3 provide clear evidence for conformational heterogeneity in both the free and DNA-bound cTALEs . Although the deterministic modeling supports such heterogeneity , puts it in the framework of a molecular model , and provides a means to determine the microscopic rate and equilibrium constants , such analysis provides little information about the structural nature of TALE conformational heterogeneity . Figure 7 shows a model of cTALE conformational change consistent with DNA binding kinetics . In this model there are four TALE states . DNA-free cTALEs comprise both incompetent and binding competent states . DNA-bound cTALES comprise at least two states that are likely to differ in their registry relative to the DNA . For eight repeat cTALE arrays , the DNA-binding competent state is more highly populated than the DNA-binding incompetent state . In this reaction scheme , the DNA-binding incompetent state can be regarded as an off-pathway conformation that inhibits DNA binding ( Figure 7A ) . Because the eight repeat cTALE array does not form multiple turns of a superhelix , unfolding to bind DNA is not required . In the model in Figure 7 , the binding competent state is the fully folded conformation , whereas the binding incompetent state includes partly folded conformations . Consistent with this interpretation , increasing populations of partly folded states through addition of 1M urea and through entropy enhancing mutations decreases apparent binding rates of 8 repeat cTALEs ( Figure 7—figure supplement 1 ) . This is also consistent with a partly folded DNA-binding incompetent state in shorter cTALE arrays . For 12 and 16 repeat cTALE arrays , the DNA-binding incompetent state is more highly populated than the DNA-binding competent state . In the model in Figure 7 , the DNA-binding competent state is a high-energy conformation required for DNA binding ( Figure 7B–C ) . Because 12 and 16 repeat cTALEs are expected to form 1 and 1 . 4 turns ( excluding the N-terminal domain ) , we hypothesize that the binding competent state includes some partly folded states that allow access to DNA . Not all partly folded states open the array to access DNA; therefore , the binding incompetent state includes some nonproductive partly folded states in addition to the fully folded state . In arrays containing 12 or more repeats , the binding competent and binding incompetent states likely include mixtures of many specific partly folded states . Because the types of partly folded states are unknown , connecting equilibria between binding competent and binding incompetent states to calculated partly folded equilibria ( using folding free energies similar to Figure 1 ) is challenging . Future work towards understanding the structural characteristics of the binding competent state in TALE arrays of one or more turns would inform which partly folded states to include in the calculation , making this comparison meaningful . A better structural understanding of the DNA binding competent state may also allow an opportunity for precise placement of destabilized repeats in designed TALEN arrays which may enable more efficient gene editing methodologies in both clinical and basic research applications . Here we demonstrate kinetic heterogeneity in DNA-bound and unbound TALE arrays , and we subsequently link the observed heterogeneity to partial unfolding of TALE arrays . We propose a model where binding requires partial unfolding of TALE arrays longer than one superhelical turn providing a functional role for previously observed moderate stability of TALE arrays . The functional instability described is particularly surprising given the small population of partly folded states which we expect to be DNA binding competent ( partly folded states similar to internally unfolded and interfacially fractured states depicted in Figure 1A ) . Discovery of a functional role for the observed conformational heterogeneity is even more surprising , given the sequence identity of each of our repeats . Sequence heterogeneity in naturally occurring TALE arrays may further enable access to partly folded binding-competent states . While it is well understood that many transcription factors sometimes undergo local folding transition upon DNA binding ( Spolar and Record , 1994; Tsafou et al . , 2018 ) , the findings here indicate that for TALE arrays , the major conformer is fully folded , and must undergo a local unfolding transition in order to bind DNA . Taken together , these findings suggest a new mode of transcription factor binding and provide compelling evidence for functional instability in TALE arrays . Previous reports show that TALEs have multiple diffusional modes when searching nonspecific DNA ( Cuculis et al . , 2015 ) . Our work with the homopolymeric DNA sequences suggests that cTALEs have multiple bound states ( Figure 7 ) . To gain more insight into conformational heterogeneity in the bound state , we performed binding experiments with T-anchored binding sequences and with divalent magnesium cation ( Figure 4 and Figure 7—figure supplementa 2–4 ) . The significant changes in unbinding kinetics suggest that the two kinetically distinct bound-states ( TALE*-DNA and TALE-DNA in Figure 6 ) differ in the registry of the TALE-DNA complex . Although we have no structural information on how these two registers differ ( for NcTALE8 binding to A15/T15 DNA , the two registers appear to have the same FRET efficiency ) , our deterministic modeling suggests that the two registers differ in their ability to dissociate . The TALE*-DNA state , which we refer to as ‘register 1’ in the mechanistic model in Figure 7 , can directly dissociate to the unbound state; likewise , it appears to be the direct product of association . In contrast , the TALE‡-DNA state , which we refer to as ‘register 2’ in Figure 7 , does not directly dissociate; rather , dissociation from register two involves conversion back to register 1 . To determine how the observed kinetic changes are partitioned into underlying kinetic steps in unbinding , we fitted various kinetic models to the cumulative distributions for unbinding ( binding to T-anchored DNA is shown in Figure 7—figure supplement 2; binding in the presence of 40 mM MgCl2 is shown in Figure 7—figure supplement 4 ) . All unbinding distributions were best-fit by the three-state unbinding model shown in Figure 5 , yielding lower chi-squared values than fits to a single-exponential model ( Table 2 ) . Distributions of the best-fit parameters obtained after 2000 bootstrap iterations are normally distributed with small confidence intervals ( Table 2 ) . Comparison of the microscopic rate constants for the homopolymeric DNA and T-anchored DNA show some significant differences . The addition of a T-anchor to the binding DNA sequence substantially decreases the rate constant for conversion from register 1 to register 2 ( k3 , from 0 . 36 s−1 to 0 . 06 s−1 , Figure 7—figure supplement 2 and Table 2 ) , and modestly decreases the rate constant for conversion from register 2 to register 1 ( k-3 , from 0 . 22 s−1 to 0 . 31 s−1 ) . The T-anchor also modestly decreases the unbinding rate constant ( k-2 , from 0 . 66 s−1 to 0 . 48 s−1 ) . Taken together , the rate constants from deterministic fits indicate that the addition of T to the binding DNA sequence stabilizes the bound register one state relative to the unbound and register two states . Comparisons of the microscopic rate constants for cTALE binding to A15/T15 DNA in the presence of monovalent K+ and Mg2+ also show some significant differences . With Mg2+ , the unbinding rate constant is larger ( k-2 = 1 . 28 s−1 vs . 0 . 66 s−1 with K+ ) , as is the rate constant for conversion from register 2 to register 1 ( k-3 = 1 . 40 s−1 vs . 0 . 222 s−1 with K+ ) ; Figure 7—figure supplement 4 and Table 2 ) . Table 1 shows that microscopic rate constants for transition between the bound register 1 and register two states become much slower in 12 and 16 repeat cTALEs compared with eight repeat cTALEs ( k-3 and k3 ) . These rate constants decrease much more than the microscopic unbinding rate constant ( the k-2 values are 0 . 66 s−1 , 0 . 13 s−1 , and 0 . 30 s−1 for NcTALE8 , NcTALE12 , and NcTALE16 respectively ) indicating that the rates of register shifting depend on the number of repeats . Although the model does not provide information on the structure of this conformational change , it is likely that this conformational change involves cTALEs shifting register by 1–3 base pairs on the homopolymeric DNA . Overall , the dissociation results demonstrate that TALE-DNA complexes are heterogeneous , and their rates of interconversion and dissociation depend on sequence , repeat number , and solution conditions .
Consensus TALE repeat constructs were cloned with C-terminal His6 tags via an in-house version of Golden Gate cloning ( Cermak et al . , 2011 ) . TALE constructs were grown in BL21 ( T1R ) cells at 37°C to an OD of 0 . 6–0 . 8 and induced with 1 mM IPTG . Following cell pelleting and lysis , proteins were purified by resuspending the insoluble material in 6M urea , 300 mM NaCl , 0 . 5 mM TCEP , and 10 mM NaPO4 pH 7 . 4 . Constructs were loaded onto a Ni-NTA column . Protein was eluted using 250 mM imidazole and refolded during buffer exchange into 300 mM NaCl , 30% glycerol , 0 . 5 mM TCEP , and 10 mM NaPO4 pH 7 . 4 . Labelling of cTALE arrays followed a previously reported protocol ( Rasnik et al . , 2004 ) . NcTALE8 and NcTALE12 were labeled at residue R30C in the first repeat , while NcTALE16 was labeled at residue R30C in the fourteenth repeat . 1 mg protein was loaded onto 500 uL NiNTA spin column . The column as washed with 10 column volumes of 300 mM NaCl , 0 . 5 mM TCEP , and 10 mM NaPO4 pH 7 . 4 . Tenfold molar excess Cy3 maleimide dye was resuspended in 10 μL DMSO and added to column . The column was rocked at room temperature for 30 min , then at 4°C overnight . Cy3-labeled protein was eluted with 250 mM imidazole , 300 mM NaCl , 30% glycerol , 0 . 5 mM TCEP , and 10 mM NaPO4 pH 7 . 4 . Protein was stored in 300 mM NaCl , 30% glycerol , 0 . 5 mM TCEP , and 10 mM NaPO4 pH 7 . 4 at −80°C . Free energies of partly folded cTALE conformations were determined using previously reported TALE intrinsic and interfacial free energies ( Geiger-Schuller and Barrick , 2016 ) . Free energies in Figure 1 are difference between the partly and fully folded states ( where all 20 repeats are folded with coupled interfaces ) . Previously determined intrinsic and interfacial free energies were used to calculate probabilities of the fully folded state ( all repeats folded ) , the end-frayed state ( the first of twenty repeats unfolded ) , the internally unfolded state ( the tenth of twenty repeats unfolded ) , and the interfacially fractured state ( the interface between repeat ten and eleven disrupted ) . As an example , to calculate the free energy of end-fraying , the end-frayed state probability is divided by the fully folded state probability to generate the equilibrium constant for end-fraying ( Kend-frayed ) , which is used to calculate the free energy of end-fraying: ( 4 ) ΔGend−frayed=−RTlnKend−frayed The conceptual framework and mathematical description of the Ising model and folding free energies are described in Aksel and Barrick ( 2009 ) . Sequences used for binding studies were 5’-Cy5-A15-3’ and 5’ T15-3’ duplex ( Cy5-A15/T15 ) as well as 5’-Cy5-A15-3’ and 5’-biotin-T15-3’ duplex ( Cy5-A15/biotin-T15 ) for eight repeat binding studies , and 5’-Cy5-A23-3’ and 5’ T23-3’ duplex ( Cy5-A23/T23 ) for 12 and 16 repeat binding studies . Sequences used for T-anchored binding studies were 5’-Cy5-TA14-3’ and 5’ T14A-3’ duplex ( Cy5-TA14/T14A ) for eight repeat binding studies . DNA was annealed at 5 μM concentration with 1 . 2-fold molar excess unlabeled strand in 10 mM Tris pH 7 . 0 , 30 mM NaCl . Biotinylated quartz slides and glass coverslips were prepared as previously described ( Rasnik et al . , 2004 ) . Cy3-labeled cTALEs were immobilized on biotinylated slides taking advantage of neutravidin interaction with biotinylated α-penta•His antibody which binds the His6 cTALE tag . Slides were pretreated with blocking buffer ( 5 μL yeast tRNA , 5 μL BSA , 40 μL T50 ) before addition of 250 pM labeled cTALE . Cy5-labeled duplex DNA was mixed with imaging buffer ( 20 mM Tris pH 8 . 0 , 200 mM KCl , 0 . 5 mg mL−1 BSA , 1 mg mL−1 glucose oxidase , 0 . 004 mg mL−1 catalase , 0 . 8% dextrose and saturated Trolox ~1 mg mL−1 ) and molecules were imagined using total internal reflection fluorescence microscopy . The time resolution was 50 msec for NcTALE8 and 100 msec for NcTALE16 and NcTALE12 . Collection and analysis was performed as previously described ( Roy et al . , 2008 ) . A minimum of 20 short movies were collected , and the first five frames ( 50 msec exposure time ) were used to generate smFRET histograms . FRET was calculated as IA/ ( IA +ID ) where IA and ID are donor-leakage and background corrected fluorescence emission of acceptor ( Cy5 ) and donor ( Cy3 ) fluorophores . In competition experiments , unlabeled DNA with the same sequence as labeled DNA was mixed at indicated concentrations with labeled DNA prior to imaging . Long movies were collected with 50 msec exposure time for NcTALE8 and 100 msec exposure time for NcTALE16 and NcTALE16 . At least 20 representative traces at each DNA concentration were selected and dwell times were determined by fitting as previously described using HaMMy ( McKinney et al . , 2006 ) for FRET in NcTALE8 . Dwell times in NcTALE12 and NcTALE16 colocalization experiments are determined by using a thresholding procedure for Cy5 excitation ( Figure 5—figure supplement 1 ) . The algorithm used to identify low and high emission states here is slightly different than previously described thresholding algorithms ( Blanco and Walter , 2010 ) . To reduce the number of incorrectly identified transitions arising from increased background and noise at higher Cy5-labeled DNA concentrations , a thresholding algorithm with two limits was implemented ( see Figure 5—figure supplement 1 ) . All FRET and colocalization data are well described by models with two distinct states ( 0 . 0 FRET and ~0 . 45 FRET as well as low colocalization and high colocalization ) . Dwell times of the same state ( low versus high FRET or low versus high colocalization ) for all traces at a given DNA concentration are compiled , and cumulative distribution is generated with spacing equal to imaging exposure time . To determine apparent rate constants using model-independent analysis , cumulative distributions were fitted with single and double exponential decays ( Figures 3 and 4 ) . Observed rates from exponential decay fits were plotted as a function of DNA concentration . Apparent rate constants were calculated as slope of DNA concentration-dependent observed rates or average of DNA concentration-independent observed rates . Equations 1a-1c and 2a-2c were numerically integrated using the ODE15s and ODE45 solver in MATLAB . Microscopic rate constants were adjusted to minimize the sum of squared residuals between ODE-determined concentration of bound or free TALE and single molecule cumulative distributions using lsqnonlin in MATLAB . 68% confidence intervals were estimated by performing 2000 or 8000 bootstrap iterations in which residuals from the best fit of the model to the data were randomly re-sampled ( with replacement ) and re-fitted . All scripts and source data required to run this MATLAB program called Determinstic Modeling for Analysis of complex Single molecule Kinetics ( DeMASK ) are publicly available on GitHub at https://github . com/kgeigers/DeMASK ( Geiger-Schuller , 2019; copy archived at https://github . com/elifesciences-publications/DeMASK ) . | The DNA contains all the information needed to build an organism . It is made up of two strands that wind around each other like a twisted ladder to form the double helix . The strands consist of sugar and phosphate molecules , which attach to one of for bases . Genes are built from DNA , and contain specific sequences of these bases . Being able to modify DNA by deleting , inserting or changing certain sequences allows researchers to engineer tissues or even organisms for therapeutical and practical applications . One of these gene editing tools is the so-called transcription activator-like effector protein ( or TALE for short ) . TALE proteins are derived from bacteria and are built from simple repeating units that can be linked to form a string-like structure . They have been found to be unstable proteins . To bind to DNA , TALES need to follow the shape of the double helix , adopting a spiral structure , but how exactly TALE proteins thread their way around the DNA is not clear . To investigate this , Geiger-Schuller et al . monitored single TALE units using fluorescent microscopy . This way , they could exactly measure the time it takes for single TALE proteins to bind and release DNA . The results showed that some TALE proteins bind DNA quickly , whereas others do this slowly . Using a computer model to analyze the different speeds of binding suggested that the fast binding comes from partly unfolded proteins that quickly associate with DNA , and that the slow binding comes from rigid , folded TALE proteins , which have a harder time wrapping around DNA . This suggest that the unstable nature of TALEs , helps these proteins to bind to DNA and turn on genes . These findings will help to design future TALE-based gene editing tools and also provide more insight into how large molecules can assemble into complex structures . A next step will be to identify TALE repeats with unstable states and to test TALE gene editing tools that have intentionally placed unstable units . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2019 | Functional instability allows access to DNA in longer transcription Activator-Like effector (TALE) arrays |
Distributed neural activity patterns are widely proposed to underlie object identification and categorization in the brain . In the olfactory domain , pattern-based representations of odor objects are encoded in piriform cortex . This region receives both afferent and associative inputs , though their relative contributions to odor perception are poorly understood . Here , we combined a placebo-controlled pharmacological fMRI paradigm with multivariate pattern analyses to test the role of associative connections in sustaining olfactory categorical representations . Administration of baclofen , a GABA ( B ) agonist known to attenuate piriform associative inputs , interfered with within-category pattern separation in piriform cortex , and the magnitude of this drug-induced change predicted perceptual alterations in fine-odor discrimination performance . Comparatively , baclofen reduced pattern separation between odor categories in orbitofrontal cortex , and impeded within-category generalization in hippocampus . Our findings suggest that odor categorization is a dynamic process concurrently engaging stimulus discrimination and generalization at different stages of olfactory information processing , and highlight the importance of associative networks in maintaining categorical boundaries .
Object categorization is an adaptive function of the brain , allowing organisms to sort information from the external world into behaviorally relevant classes . Importantly , sensory systems must generalize across different objects sharing similar features , but at the same time maintain the specificity of individual objects and categories ( Roach , 1978; Riesenhuber and Poggio , 2000 ) . Mechanisms of pattern recognition have been proposed to underlie the neural basis of object categorization , which requires a balance between generalizing inputs across a certain range of variations ( known as pattern completion ) and discriminating between distinct inputs ( known as pattern separation ) ( Riesenhuber and Poggio , 2000; Haberly , 2001; Wilson and Sullivan , 2011; Chapuis and Wilson , 2012 ) . Such computations can be achieved by associating sensory inputs with internal templates that are established through a lifetime of experience and encoded into memory ( Bar , 2007 ) . Most neuroscientific research on pattern recognition has concentrated on the visual system , where associative areas in the visual ventral stream and the CA3 region of the hippocampus have been shown to support processes of object categorization ( Riesenhuber and Poggio , 2000; Yassa and Stark , 2011; Haxby et al . , 2001 ) . In the olfactory system , information in a whiff of scented air is transformed into distributed patterns of neural activity in the piriform cortex , with both animal and human studies demonstrating that different odor objects evoke distinguishable ensemble activity patterns without spatial topography ( Wilson and Sullivan , 2011; Gottfried , 2010; Bekkers and Suzuki , 2013; Stettler and Axel , 2009; Howard et al . , 2009 ) . Recent work has revealed that fMRI multivariate patterns in posterior piriform cortex ( PPC ) encode not only odor identity , but also category information ( e . g . , minty or woody ) , whereby odor patterns belonging to the same category are more similar ( more overlapping ) than those across different categories ( Howard et al . , 2009 ) . Despite these insights , the mechanisms by which olfactory inputs are organized into categorical percepts through their associations with olfactory cortical areas are poorly understood . The neural architecture of the piriform cortex makes it an attractive model for investigating mechanisms of odor object recognition . As the largest subregion of primary olfactory cortex , the piriform cortex receives afferent ( bottom-up ) inputs from the olfactory bulb through the lateral olfactory tract , and extensive associative ( top-down ) inputs from higher-order association areas such as orbitofrontal cortex ( OFC ) , amygdala , and entorhinal cortex ( Carmichael et al . , 1994; Johnson et al . , 2000; Haberly and Price , 1978; Insausti et al . , 1987; Insausti et al . , 2002 ) . This convergence of bottom-up and top-down projections , along with the presence of dense recurrent collaterals , is thought to support olfactory pattern recognition and associative learning ( Haberly , 2001; Haberly and Bower , 1989; Wilson , 2009 ) . For example , when confronted with highly overlapping odor mixtures , rats can learn to discriminate or ignore detectable differences between these mixtures , with piriform activity patterns exhibiting either separation ( enhanced discrimination ) or completion ( enhanced generalization ) , respectively ( Chapuis and Wilson , 2012 ) . Evidence from humans has also pointed to PPC as a substrate for odor discrimination ( Li et al . , 2008 ) and categorization ( Howard et al . , 2009 ) . Together these findings suggest that piriform cortex is capable of modulating pattern representations along a discrimination-generalization spectrum in order to encode behaviorally adaptive meaning through perceptual experience . While theoretical modelling and empirical evidence propose that piriform associative connections are essential for odor recognition ( Haberly , 2001 ) , few studies have explicitly investigated the relative contributions of afferent inputs versus associative networks in supporting odor categorization . In a previous fMRI study , human subjects were deprived of afferent sensory input for one week , resulting in a reduction of odor-evoked mean activity in PPC , without alteration of pattern-based piriform representations of odor categories ( Wu et al . , 2012 ) . Here we address the inverse question , namely , how attenuation of piriform associative connections influences odor category coding in primary sensory regions and higher-order cortical areas . To this end , we took advantage of the GABA ( B ) receptor agonist , baclofen , to modify the relative balance between afferent and associative inputs within piriform cortex . Baclofen selectively suppresses synaptic transmission of association fibers into piriform cortex , but leaves afferent inputs from the olfactory bulb unaffected ( Tang and Hasselmo , 1994 ) . In vivo local application of baclofen in the piriform cortex of anesthetized rats modified the strength of odor-evoked responses of pyramidal neurons , by blocking broadly-tuned neurons and increasing odor-selective responses ( Poo and Isaacson , 2011 ) . In behaving animals , injection of baclofen into the piriform cortex following an olfactory fear conditioning session resulted in fear memory generalization , indicating that piriform associative connections are essential for consolidation of stimulus-specific memories ( Barnes and Wilson , 2014 ) . Inspired by these animal studies , we conducted a double-blind , placebo-controlled drug study in human subjects to examine fMRI ensemble representations of familiar odor categories before and after treatment with baclofen . Given that odor object codes take the form of distributed ensemble patterns , we used multivariate fMRI analyses to characterize baclofen effects in olfactory areas found to represent categorical information . The placebo group served as a control to account for session-effect confounds between pre- and post-drug phases of the study . As such , we examined the effects of baclofen by comparing pre-to-post changes relative to those observed in placebo subjects ( i . e . , group-by-session interaction ) . We predicted that baclofen would disrupt associative connections , leading to perceptual and neural reorganization of odor categories in piriform cortex and in olfactory downstream areas including OFC , amygdala , entorhinal cortex , and hippocampus .
We first established that baclofen did not generally compromise cognitive or perceptual performance . Specifically , we found no significant differences between baclofen and placebo groups on neuropsychological assessments of basic cognition , short-term memory , visual attention , or task switching ( Table 1 ) . We also collected subjective reports of sleepiness using the Stanford Sleepiness Scale ( SSS ) during test sessions , given that the most common adverse reaction to baclofen medication is transient drowsiness ( RxList The internet Drug Index , 2007 ) . Baclofen subjects reported feeling sleepier after taking the drug ( Figure 2 ) , though reaction times during the fMRI categorization task did not differ from placebo subjects ( Table 1 ) . Finally , we examined whether baclofen altered general odor perception . Placebo and baclofen groups did not differ on olfactory measures of detection threshold , identification , fine odor discrimination ( Figure 3d ) , or intensity and pleasantness ratings ( for stimuli used in the main fMRI experiment ) ( Table 1 ) , thereby reducing the possibility that baclofen-induced changes in odor perception could have influenced the imaging results . 10 . 7554/eLife . 13732 . 004Table 1 . Behavioral performance . DOI: http://dx . doi . org/10 . 7554/eLife . 13732 . 004TaskPlacebo ( n = 18 ) Baclofen ( n = 14 ) P value of group × session interactionPrePostPrePostMMSE29 . 89 ± 0 . 1129 . 94 ± 0 . 0629 . 93 ± 0 . 0730 . 00 ± 00 . 86Digit span ( forward ) 7 . 22 ± 0 . 227 . 72 ± 0 . 147 . 14 ± 0 . 317 . 43 ± 0 . 200 . 55Digit span ( backward ) 6 . 00 ± 0 . 206 . 00 ± 0 . 295 . 57 ± 0 . 315 . 71 ± 0 . 270 . 69Trail making test B ( s ) 48 . 32 ± 2 . 7039 . 05 ± 2 . 0157 . 62 ± 5 . 4247 . 11 ± 5 . 670 . 85Stanford sleepiness scale2 . 44 ± 0 . 172 . 22 ± 0 . 211 . 93 ± 0 . 202 . 50 ± 0 . 330 . 041*Sniffin’ Sticks ( odor detection threshold ) 7 . 08 ± 0 . 849 . 65 ± 1 . 067 . 82 ± 0 . 958 . 57 ± 1 . 010 . 22UPSIT ( odor identification ) 36 . 28 ± 0 . 6136 . 00 ± 0 . 5634 . 57 ± 0 . 4933 . 79 ± 0 . 630 . 55α- vs . β-pinene triangle test ( fine odor discrimination ) 0 . 66 ± 0 . 050 . 72 ± 0 . 050 . 72 ± 0 . 050 . 73 ± 0 . 070 . 45Odor intensity ratings4 . 00 ± 0 . 324 . 13 ± 0 . 313 . 05 ± 0 . 182 . 91 ± 0 . 280 . 39Odor pleasantness ratings5 . 43 ± 0 . 165 . 63 ± 0 . 165 . 64 ± 0 . 135 . 63 ± 0 . 160 . 12Odor category descriptor ratings ( within – across ) 7 . 47 ± 0 . 447 . 49 ± 0 . 367 . 44 ± 0 . 467 . 78 ± 0 . 340 . 59Odor pairwise similarity ratings ( within – across ) 4 . 16 ± 0 . 605 . 14 ± 0 . 553 . 93 ± 0 . 324 . 53 ± 0 . 440 . 59Odor categorization catch trial accuracy0 . 87 ± 0 . 040 . 89 ± 0 . 030 . 81 ± 0 . 040 . 81 ± 0 . 040 . 80Odor categorization catch trial RT ( s ) 3 . 29 ± 0 . 232 . 85 ± 0 . 153 . 89 ± 0 . 383 . 48 ± 0 . 340 . 93Visual categorization catch trial accuracy0 . 97 ± 0 . 01 ( n = 14 ) 0 . 99 ± 0 . 0040 . 97 ± 0 . 01 ( n = 11 ) 0 . 96 ± 0 . 010 . 21Visual categorization catch trial RT ( s ) 0 . 42 ± 0 . 020 . 40 ± 0 . 030 . 44 ± 0 . 040 . 52 ± 0 . 060 . 25Data are shown for cognitive and olfactory tests , as well as for behavioral performance in fMRI experiments from placebo and baclofen groups in pre- and post-drug sessions . Scores are presented as mean ± s . e . m . P values reported are for the interaction effects between group and session , based on a 2-way ANOVA , with one between-group ‘drug’ factor ( placebo/baclofen ) and one within-subject ‘session’ factor ( pre/post ) . *P < 0 . 05 . 10 . 7554/eLife . 13732 . 005Figure 2 . Effect of baclofen on subjective sleepiness . Ratings from the Stanford Sleepiness Scale ( 1 = ‘wide awake’ , 7 = ‘sleep onset soon’ , mean ± within-subject s . e . m . , placebo n = 18 , baclofen n = 14 ) indicate that there was a significant interaction between drug groups ( placebo vs . baclofen ) and session ( pre vs . post ) ( F1 , 30 = 4 . 57 , P = 0 . 041; *P < 0 . 05 ) . Post-hoc within-group comparisons showed no effect of session in placebo subjects ( F1 , 17 = 0 . 88 , P = 0 . 36 ) , and a marginal effect of session in baclofen subjects ( F1 , 13 = 3 . 85 , ‡P = 0 . 072 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13732 . 00510 . 7554/eLife . 13732 . 006Figure 3 . Subjects successfully classified odors into their relevant categories . ( a ) Category descriptor ratings of the six odors ( two citrus: C1 , C2; two minty: M1 , M2; two woody: W1 , W2 ) from all subjects during the pre-drug session ( mean ± within-subject s . e . m . , n = 32 ) . Repeated-measures ANOVA was conducted separately on each odor ( ** = P < 0 . 001 ) . Subjects robustly classified the odors into the appropriate perceptual categories ( C1: F1 . 86 , 57 . 77 = 31 . 62; C2: F1 . 72 , 53 . 40 = 74 . 58; M1: F1 . 92 , 59 . 61 = 144 . 04; M2: F1 . 60 , 49 . 46 = 373 . 79; W1: F1 . 82 , 56 . 33 = 140 . 96; W2: F1 . 49 , 46 . 10 = 166 . 84; all P’s < 0 . 001 ) . ( b ) Average of category descriptor ratings across odors , sorted by within-category condition and across-category condition in pre- and post-drug sessions for placebo ( n = 18 ) and baclofen ( n = 14 , mean ± within-subject s . e . m . ) groups . ( c ) Pair-wise similarity ratings of within- and across-category odor pairs in pre- and post-drug sessions for placebo and baclofen groups ( mean ± within-subject s . e . m . ) . ( d ) Fine odor discrimination between α- and β-pinene in pre- and post-drug sessions for placebo and baclofen groups ( mean ± within-subject s . e . m . ) . ( e ) Dendrogram plots obtained from a cluster analysis of the average pair-wise similarity ratings for placebo and baclofen subjects during pre- and post-drug sessions showed that both groups sorted the six odors into three categories in both sessions . Shorter distance indicates greater similarity . DOI: http://dx . doi . org/10 . 7554/eLife . 13732 . 006 Before and after drug administration , subjects participated in an fMRI odor categorization task . On each trial , subjects smelled one of six odors belonging to three categories: citrus ( C1 and C2 ) , mint ( M1 and M2 ) , and wood ( W1 and W2 ) ( Figure 1b ) . Prior to each scanning session , subjects first provided category descriptor ratings ( i . e . , “how citrusy/minty/woody is odor X ? ” ) , as well as pair-wise similarity ratings , for each of the six odors . During the pre-drug session , within-category descriptor ratings were significantly higher than across-category descriptor ratings ( simple main effect of category at pre: F1 , 31 = 572 . 66 , P < 0 . 001; mixed-model ANOVA ) ( Figure 3a , b ) , in the absence of an interaction between placebo and baclofen groups ( F1 , 30 = 0 . 0019 , P = 0 . 97; Figure 3b ) . Moreover , during the pre-drug session , the within-category odor pairs were rated as significantly more similar than across-category odor pairs ( simple main effect of category at pre: F1 , 31 = 125 . 80 , P < 0 . 001; Figure 3c ) , again without an interaction between groups ( F1 , 30 = 0 . 097 , P = 0 . 76; Figure 3c ) . Finally , based on group averages of similarity ratings , we performed a cluster analysis and found that both placebo and baclofen subjects successfully group the six odors into three classes appropriately in both pre- and post-drug sessions ( Figure 3e ) . These data confirm that subjects were highly familiar with the odor categories prior to initiating the experiment . To quantify categorization performance , we calculated the difference between within-category and across-category descriptor ratings , as well as the difference between within-category and across-category similarity ratings ( Table 1 ) . There was a main effect of session on categorization showing a general improvement based on similarity ratings ( F1 , 30 = 5 . 65 , P = 0 . 024 ) , probably reflecting a practice effect . However , the session-related changes did not differ between groups ( F1 , 30 = 0 . 30 , P = 0 . 59; Figure 3c , Table 1 ) . For categorization based on descriptor ratings , there was no main effect of session ( F1 , 30 = 0 . 33 , P = 0 . 57 ) or interaction between groups ( F1 , 30 = 0 . 30 , P = 0 . 59; Figure 3b , Table 1 ) . Collectively , these results indicated that baclofen did not affect behavioral measures of odor categorization at the group level . During fMRI scanning , subjects received occasional ‘catch trials’ ( every 4–8 trials ) , in which they were prompted to indicate the category of the previously delivered odor . In the pre-drug session , subjects categorized odors with high accuracy ( 84 . 4% ± 2 . 7% , chance level at 33% , t31 = 19 . 37 , P < 0 . 0001 ) . Of note , neither the catch trial accuracies nor reaction times ( RT ) differed significantly as a function of treatment group from pre- to post-drug session ( Table 1 ) . During the fMRI odor categorization task , the six odors were delivered in a pseudorandom order , and subjects were cued to sniff upon odor delivery . They were asked to pay attention to the quality of the odors throughout the task , and make category judgments during catch trials . As olfactory information takes the form of distributed patterns of fMRI activity in the human brain ( Howard et al . , 2009; Wu et al . , 2012 ) , multivariate pattern analyses are well-suited for examining the impact of baclofen on odor pattern recognition . We first used a support vector machine ( SVM ) classifier to identify brain areas where odor category information is represented , among several anatomically defined regions of interest ( ROIs ) including piriform cortex , higher-order areas that directly project to piriform ( olfactory subregion of OFC , amygdala , entorhinal cortex ) , and hippocampus ( Figure 4a ) . This analysis was conducted for all subjects in the pre-drug session , in order to constrain our investigation of baclofen-induced drug effects to those ROIs that had robust odor category coding at baseline ( thus independent of drug administration ) . We trained the SVM classifier on patterns evoked by one pair of odors belonging to different categories ( e . g . , C1 vs . M1 ) , and then tested the classifier on patterns evoked by the complementary pair of odors from the same categories ( e . g . , C2 vs . M2; Figure 4b ) . Importantly , because training and test sets were based on data evoked by different odor identities , significant above-chance decoding is only possible if fMRI patterns encode category information independent of the specific odor identities . Across all subjects in the pre-drug session , we found significant above-chance decoding accuracy in PPC ( t31 = 2 . 05 , P = 0 . 024 ) , OFC ( t31 = 1 . 96 , P = 0 . 029 ) , amygdala ( t31 = 3 . 17 , P = 0 . 0017 ) , and posterior hippocampus ( pHIP , t31 = 1 . 90 , P = 0 . 034; Figure 4c ) . All subsequent analyses were constrained to these four regions where fMRI ensemble patterns encode odor category information . 10 . 7554/eLife . 13732 . 007Figure 4 . Ensemble pattern coding of odor category information at baseline ( pre-baclofen session ) . ( a ) Axial and coronal slices of the averaged , normalized T1-weighted structural scan from all subjects showing anatomically defined regions of interest . Odor-evoked ensemble patterns across all voxels within a given ROI were used in a two-step multivariate classification analysis . First , we trained a linear SVM on a training data-set ( b , left panel ) to separate two odors belonging to different categories . Second , odor category coding was assessed in an independent test data-set ( b , right panel ) , specifically by testing how well the SVM classified the other pair of odors from the corresponding categories; here , cross-decoding is only successful if similar patterns code different odors of the same category . ( c ) Category decoding from all subjects during the pre-drug session showed that classification accuracy in PPC , OFC , amygdala , and pHIP significantly exceeded chance ( mean ± between-subject s . e . m . , n = 32 , *P < 0 . 05 , one-tailed ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13732 . 007 In order to characterize the continuous degree of pattern similarity between stimuli ( Nili et al . , 2014 ) , we next used a linear correlation analysis ( Haxby et al . , 2001; Howard et al . , 2009; Kriegeskorte et al . , 2008 ) , which provides a more direct assessment of pattern overlap . Specifically , to examine how baclofen alters the categorical organization of odors , we assembled vectors of ensemble pattern activity from all voxels within PPC , and measured the dissimilarity ( correlation distances ) of pattern vectors evoked by across-category odors ( e . g . , C1/M1 ) and within-category odors ( e . g . , C1/C2 , Figure 5a ) . In order to control for within-session and between-session variations that could arise from training , familiarity , increasing boredom , drug effect , scanner drift or other potential artifact , we computed pattern distances between the same odor as baseline patterns , and then subtracted these from the within-category and across-category pattern distances . A two-way mixed-model ANOVA on same-odor pattern distances showed that there was no main effects of session or group , or group × session interaction ( F1 , 30 = 1 . 88 , P = 0 . 18 ) , indicating that the baseline patterns were consistent across time and group , and were not affected by the drug . We then tested a three-way analysis of variance ( ANOVA ) , with two within-subject factors of session ( pre/post ) and category type ( within-/across-category ) , and one between-subject factor of drug ( placebo/baclofen ) . This yielded a significant session × category type × drug interaction effect ( F1 , 30 = 5 . 49 , P = 0 . 026 ) in the absence of other main effects or two-way interactions ( all P’s >0 . 15 ) , and suggests that baclofen significantly affected the categorical structure of odor pattern representations in PPC . 10 . 7554/eLife . 13732 . 008Figure 5 . Baclofen effect on odor pattern changes in PPC . ( a ) Schematic illustrating within-category and across-category relationships among categorically organized odors , and how changes of each distance parameter alter the categorical structure . Worse categorization emerges when within-category distances increase or when across-category distances decrease . Better categorization emerges when within-category distances decrease or when across-category distances increase . ( b ) Odor pattern distance in PPC in pre- and post-drug sessions , sorted by within-category and across-category distances , from placebo ( n = 18 ) and baclofen ( n = 14 , mean ± within-subject s . e . m . ) subjects . Placebo subjects showed increased within-category distances without across-category changes . There was no significant odor distance change in baclofen subjects . ( c ) A scatterplot showing the correlation between the magnitude of within-category odor pattern separation in PPC and behavioral changes in a fine odor-discrimination task , from pre- to post-drug session ( ρ = 0 . 51 , P = 0 . 031 , n = 14 , one-tailed ) . Each diamond represents one baclofen subject . *P<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 13732 . 008 Based on inspection of the odor-evoked pattern changes in PPC ( Figure 5b ) , it is evident that these changes were actually more prominent in the placebo group , and for the within-category condition . To assess these hypotheses , we examined drug-related categorization effects separately in each group . In placebo subjects the interaction of session × category type was significant in PPC ( F1 , 17 = 9 . 35 , P = 0 . 0071; repeated-measures ANOVA ) , whereas no such interaction was identified in the baclofen group ( F1 , 13 = 0 . 62 , P = 0 . 45 ) . This effect was driven by a significant increase of the within-category odor distance in the placebo group ( F1 , 17 = 5 . 23 , P = 0 . 035 ) , but not in the baclofen group ( F1 , 13 = 2 . 61 , P = 0 . 13 ) , with a significant difference between groups ( F1 , 30 = 7 . 36 , P = 0 . 011; mixed-model ANOVA , session × group interaction ) . On the other hand , across-category odor distances did not differ for either group ( placebo , F1 , 17 = 0 . 75 , P = 0 . 40; baclofen , F1 , 13 = 0 . 27 , P = 0 . 61 ) or between groups ( F1 , 30 = 0 . 00014 , P = 0 . 99 , Figure 5b ) . These results highlight a divergence in PPC pattern representations for odors belonging to the same category , but only in the placebo group . One implication is that repeated exposure to the odors ( in absence of drug ) induced pattern separation or differentiation , a process that appears to be blocked in the presence of baclofen . Interestingly , this conceptualization – greater pattern separation over time in the control subjects – is in close accordance with an earlier olfactory perceptual learning study from our lab , where prolonged passive exposure to one target odor increased its discriminability from categorically related odors ( Li et al . , 2006 ) . Viewed in this context , it is reasonable to speculate that baclofen interferes with the natural emergence of olfactory pattern separation in PPC , possibly reflecting a disruption in consolidation mechanisms that normally underlie perceptual learning . If pattern separation in PPC is critical for differentiating categorically related odors , it follows that subjects with greater disruption of PPC pattern separation ( as a result of baclofen treatment ) should exhibit greater olfactory perceptual deficits . This hypothesis was tested by regressing subject-wise measures of fine odor discrimination ( Figure 3d ) against the magnitude of baclofen-induced pattern changes in PPC . We found a significant correlation between perceptual performance change and the degree of odor-evoked pattern separation in PPC ( ρ = 0 . 51 , P = 0 . 031 , one-tailed; Figure 5d ) . Thus , subjects with less within-category odor separation in PPC showed greater difficulty in discriminating between odors sharing semantic features . It is worth considering that because baclofen produced significant sleepiness , the associated tiredness and sedation may have also been associated with a lack of attention on the hardest discriminations . To investigate this possibility , we tested the correlation between pre-post changes in sleep scale ratings and changes in fine odor discrimination performance across baclofen subjects . This relationship was not significant ( Spearman ρ = -0 . 33 , P = 0 . 25 , n = 14 ) . Likewise , there was no significant correlation between sleep scale changes and within-category pattern distance changes in PPC across baclofen subjects ( Spearman ρ = 0 . 11 , P = 0 . 70 , n = 14 ) . These data suggest that there was no direct evidence of a systematic link between sleepiness and odor discrimination at the behavioral or neural level . Because olfactory categorical codes were also identified in OFC , amygdala , and pHIP in the pre-treatment session ( Figure 4 ) , we also investigated the effects of baclofen on categorical organization of odor ensemble patterns in these regions . Significant three-way interactions of session × category type × drug were found in OFC ( F1 , 30 = 4 . 48 , P = 0 . 043 ) and pHIP ( F1 , 30 = 5 . 90 , P = 0 . 021 ) without other main effects or two-way interactions . No significant interaction was observed in amygdala ( F1 , 30 = 0 . 047 , P = 0 . 83; Figure 6c ) . 10 . 7554/eLife . 13732 . 009Figure 6 . Baclofen effect on odor pattern changes in OFC and pHIP . Odor pattern distances in ( a ) OFC , ( b ) pHIP , and ( c ) amygdala in pre- and post-drug sessions , sorted by within-category and across-category distances , for placebo ( n = 18 ) and baclofen ( n = 14 , mean ± within-subject s . e . m . ) subjects . ( a ) In the baclofen group , across-category distances in OFC decreased significantly without change in within-category distances , leading to disrupted categorical structure . There was no change in the placebo group . ( b ) In pHIP , the placebo group showed a trend decrease in within-category odor distances without change in across-category distances . There was no significant odor distance change in the baclofen group . ( c ) In amygdala there was no baclofen effect on the categorical representation of odors . ‡ P<0 . 1 , *P<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 13732 . 009 Following the same approach used for PPC , we next asked how odor category organization in OFC changes in each group . We found that in OFC , the session × category interaction approached significance in the baclofen group ( F1 , 13 = 4 . 14 , P = 0 . 063 ) , driven by a significant decrease in across-category distances ( F1 , 13 = 5 . 93 , P = 0 . 030; Figure 6a ) , and leading to a net effect of category disruption ( that is , greater convergence of across-category patterns; e . g . , C1 and M1 becoming more alike , Figure 5a ) . There was no categorical change in the placebo group ( session × category interaction , F1 , 17 = 0 . 17 , P = 0 . 69 ) . By contrast , in pHIP , there was a significant session × category interaction in the placebo group ( F1 , 17 = 7 . 68 , P = 0 . 013 ) , here driven by a trend decrease in within-category distances ( F1 , 17 = 3 . 09 , P = 0 . 097 , Figure 6b ) , and giving rise to an enhanced categorical structure among odors ( that is , greater convergence of within-category patterns; e . g . , C1 and C2 becoming more alike , Figure 5a ) . Of note , this profile is opposite to that seen in PPC ( Figure 5b ) . On the contrary , there was no categorical change in the baclofen group ( session × category interaction , F1 , 13 = 0 . 77 , P = 0 . 40 ) . The above findings indicate that baclofen had selective effects on odor category coding in PPC , OFC , and pHIP . However , because baclofen was administered systemically , it remains unclear whether the effects were specific to odor categorization , or merely altered semantic or conceptual processing independently of sensory modality . Therefore , in a parallel fMRI experiment , the same subjects performed a visual categorization task ( Figure 7a ) , viewing six images belonging to three categories ( chairs , teapots , and houses ) and identifying the category on catch trials . There was no effect of baclofen on response accuracies and reaction times ( Table 1 ) . 10 . 7554/eLife . 13732 . 010Figure 7 . Visual control experiment . ( a ) Paradigm of the fMRI visual categorization experiment . Subjects viewed six images belonging to three categories . On catch trials that occasionally followed image presentations , names of the three categories appeared on screen , and subjects indicated the category of the image with a mouse click . ( b ) Visual category decoding from all subjects during the pre-drug session showed that classification accuracy in LOC significantly exceeded chance ( *P = 0 . 013 , one-tailed ) . ( c ) The effect of baclofen on visual categorical representations in LOC was not significant ( P = 0 . 50 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13732 . 010 We then utilized the same multivariate analysis pipeline to explore the effects of baclofen on visual pattern recognition . First we used the SVM classifier to decode visual category information in the same ROIs as in the olfactory task . We also included two additional visual ROIs located in lateral occipital complex ( LOC ) and fusiform gyrus as defined by an independent functional localizer scan , and which are known to be involved in visual object recognition ( Haxby et al . , 2001; Cox and Savoy , 2003; Grill-Spector , 2003; Kriegeskorte et al . , 2008 ) . Across all subjects in the pre-drug session , category decoding accuracy was significantly above chance in LOC ( t30 = 2 . 33 , P = 0 . 013 ) , but not in fusiform cortex ( t30 = -1 . 20 , P = 0 . 88 ) or in any of the olfactory ROIs ( PPC: t30 = 0 . 36 , P = 0 . 72; OFC: t30 = -1 . 36 , P = 0 . 18; amygdala: t30 = -2 . 47 , P = 0 . 99; pHIP: t30 = 0 . 078 , P = 0 . 47; Figure 7b ) . Next , we performed an fMRI pattern correlation analysis to test the drug effect on visual category representations . A three-way session × category type × drug interaction was not significant in LOC ( F1 , 29 = 0 . 46 , P = 0 . 50; Figure 7c ) , suggesting that baclofen did not alter coding of categories in the visual domain . Finally , we compared the effect of baclofen on categorization between olfactory and visual tasks , and found that the impact of baclofen on category coding in PPC was specific to olfaction . A mixed four-way ANOVA ( three within-subject factors of modality , category type , and session; one between-subject factor of group ) revealed a significant interaction of modality × category × session × drug ( F1 , 29 = 4 . 41 , P = 0 . 044 ) . Thus , while baclofen blocked within-category separation in PPC , it did not alter the visual categorization compared to placebo ( category × session × drug interaction: F1 , 29 = 0 . 056 , P = 0 . 81 ) . These findings imply that the observed effect of baclofen in PPC was not due to generic changes in semantic processing , nor to non-specific changes in hemodynamic parameters , but instead was due to alterations in information coding in the presence of olfactory inputs .
In this study we investigated the role of piriform associative connections in the neural coding of odor categories . We used the GABA ( B ) receptor agonist , baclofen , to reduce associative input in the olfactory network while sparing afferent input from the periphery . This pharmacological manipulation , combined with multivariate pattern analysis , enabled us to examine how baclofen treatment alters fMRI pattern representations of odors within and across categories relative to placebo . We found that in PPC , baclofen ( compared to placebo ) interfered with the emergent pattern separation of odors belonging to the same categorical class . The magnitude of this effect correlated with difficulties in fine-odor discrimination at the perceptual level . In contrast , baclofen disrupted across-category separation in olfactory downstream region of OFC , and impeded within-category generalization in pHIP . Interestingly , the baclofen effect observed in PPC was opposite to our original prediction that baclofen would simply weaken the boundaries between categories , leading to reduced pattern separation between citrus , mint , and wood odors . Instead , there was increased pattern separation for within-category odors over time , but specifically in the placebo group , likely reflecting perceptual training or stimulus-specific consolidation from the pre- to post-session . The significant group × session interaction implies that the natural process of within-category pattern separation in controls was disrupted in the presence of baclofen . For example , PPC pattern representations of the two citrus odors became more distinct under the control condition , but failed to diverge under baclofen . We speculate that piriform associative input normally supports the separation of patterns corresponding to unique identities of individual odors , especially those sharing perceptual features and associated with the same semantic labels . This mechanism would be compatible with prior work showing that perceptual learning enhances discriminability of within-category odor pairs , with concomitant fMRI changes in PPC as well as OFC ( Li et al . , 2006 ) . It is worth considering why baclofen had no effect on across-category odor separation in PPC . One plausible explanation is that piriform cortex has the capacity to enhance either pattern separation or completion , as a function of task demands ( Chapuis and Wilson , 2012; Li et al . , 2008; Shakhawat et al . , 2014 ) . Various rodent paradigms of olfactory associative learning have shown that the direction of piriform pattern changes can flexibly match the behavioral requirements for either odor discrimination ( i . e . , pattern separation ) or odor generalization ( i . e . , pattern completion ) . In the current experiment , subjects were asked to perform an odor categorization task , in which differences across categories , but not within category , were emphasized . As such , our experimental design might have helped stabilize category-specific differences in PPC , even in the presence of baclofen , though at the expense of within-category odor separation . The fact that categorical representations of citrus , mint , and wood odors were already highly familiar to the subjects also could have created further stability against across-category pattern changes . Thus , it is fair to say that while we have no evidence to support baclofen-induced disruption of across-category codes , the nature of our task leaves open the possibility that in a different task where categorical distinctions were less relevant , baclofen might have a modulatory influence on across-category differences . Another potential factor related to the experimental design that might complicate interpretation of the PPC results is a short-term order effect between trials . That is to say , pattern representations of the same odor could differ depending on whether the preceding odor belonged to the same category ( category repetition , e . g . , C1 preceded by C2 ) or a different category ( non-repetition , e . g . , C1 preceded by M1 ) . This repetition factor could induce short-term ‘learning’ or adaptation effects that involve local synaptic interactions mediated by GABA ( B ) receptors ( Brenowitz et al . , 1998; Ziakopoulos et al . , 2000 ) , and thus could be susceptible to baclofen manipulation . Examination of our data indicates that across subjects and pre/post sessions , there were 39 . 3 ± 0 . 6 ( mean ± SE ) category repetition trials as opposed to 121 . 5 ± 0 . 9 category non-repetition trials . That the majority ( ~75% ) of trials in our experiment belonged to non-repetition trials suggests that category repetitions would not have had a pronounced effect on the findings . In additional analyses ( see Materials and methods ) , the proportion of category repetition versus non-repetition trials ( per imaging run ) was not associated with the strength of pattern categorization effects in PPC , and PPC patterns evoked by the same odor under repetition and non-repetition conditions were not significantly different . Thus , it is reasonable to conclude that the sequence order of the odors did not have an impact on categorical pattern coding in PPC , either at baseline or in the context of drug . In contrast to PPC , fMRI patterns in olfactory downstream areas , including OFC and pHIP , showed deficient category coding in the baclofen group . Thus in OFC , the discrete categorical patterns for citrus , mint , and wood became less separated , in the presence of baclofen . In spite of these changes , there was no parallel impact on behavior . Indeed , baclofen had no perceptual effect on categorical discrimination , and we would argue that such a finding would have been unlikely , presumably due to high familiarity and discriminability of odor categories . However , to the extent that the existence of an odor category necessitates an association between an olfactory stimulus and semantic conceptual knowledge , these results are consistent with the recognized integrative role of OFC in guiding olfactory-based behavior . Both animal and human studies have demonstrated that OFC patterns can differentiate between odor objects and categories ( Howard et al . , 2009; Wu et al . , 2012; Schoenbaum and Eichenbaum , 1995; Critchley and Rolls , 1996 ) . Moreover , the OFC has been proposed to integrate taste and visual information associated with odor stimuli ( Critchley and Rolls , 1996; Gottfried and Dolan , 2003 ) , encode the reward value of odors ( Howard and Gottfried , 2014 ) , disambiguate mixtures of categorically dissimilar odors ( Bowman et al . , 2012 ) , and represent olfactory lexical-semantic content ( Olofsson et al . , 2014 ) . Viewed in this context , our results highlight the role of OFC in preserving the perceptual distinctions between different odor categories , likely through its associative access to multimodal and semantic information streams . The demonstration of olfactory category coding in pHIP , and its vulnerability to baclofen , echoes hippocampal findings in the visual modality ( Seger and Miller , 2010; Seger and Peterson , 2013; Kumaran and McClelland , 2012 ) . For example , single-unit recordings from the hippocampus have identified neurons in both humans and monkeys that are able to categorize visual information ( Kreiman et al . , 2000; Hampson et al . , 2004 ) , and fMRI activity in human hippocampus is selectively increased when memory performance relies on perceptual generalization across stimuli ( Preston et al . , 2004; Shohamy and Wagner , 2008 ) . Considered in this framework , the trend effect of decreased within-category separation in pHIP in placebo subjects may reflect the role of hippocampus to generalize , or to make inferences , across shared odor features , essentially bringing odors of the same category closer together , and creating more separation between different categories ( Figure 5a ) . It is interesting to note that both piriform cortex and hippocampus have long been regarded as canonical models of autoassociative networks where pattern separation and pattern completion computations can be flexibly achieved ( Yassa and Stark , 2011; Bekkers and Suzuki , 2013; Wilson , 2009; Leutgeb and Leutgeb , 2007; Hunsaker and Kesner , 2013; LaRocque et al . , 2013; Eichenbaum et al . , 2007 ) . That the effect of baclofen was to impede within-category discrimination in PPC , while simultaneously impeding within-category generalization in pHIP , highlights a unique functional difference between these two anatomically homologous regions , and may help bring new mechanistic understanding of the contributions of piriform cortex , hippocampus , and piriform-hippocampal interactions to human olfactory processing and perception . Our behavioral data indicate that the 50-mg baclofen dose did not impair general cognition or olfactory perceptual performance , suggesting that off-target effects of the drug were minimal , other than a modest effect on subjective sleepiness that did not interfere with online task accuracy or response times . While it is possible that the 50-mg dose may not have been potent enough to exert a physiological effect , the study medication schedule was similar to those used in other human studies that administered baclofen to induce reliable changes in brain activity or behavior ( Terrier et al . , 2011; Franklin et al . , 2012; Young et al . , 2014; Franklin et al . , 2011 ) . In our study , we did not find evidence of drug effect on odor categorization behavior , in spite of significant changes in fMRI pattern representations . There are at least three possibilities to account for this discrepancy . One possibility is that our behavioral tests were simply not sensitive enough to detect changes in perceptual performance . Across both placebo and baclofen groups , there was a general trend towards improved performance , likely reflecting effects of training and exposure . As such , any further subtle effect of baclofen on perception may not have emerged beyond these training effects per se . Related to this , even if the three-way forced-choice pinene triangle test was arguably the most ‘sensitive’ or difficult test of odor discrimination , this test might not have revealed a significant change if the perceptual learning effects had been confined to those odors presented repeatedly during the fMRI . A second possibility is that because all of the subjects were explicitly informed of the categorical features of the odor stimuli , there may have been an implicit tendency to anchor their perceptual responses to semantic categorical attributes . Moreover , even at baseline , all of the odors were easy to discriminate and highly familiar , and subjects were regularly called upon to make category judgments of the odors throughout the experiment . Thus , even though there was reorganization of categorical representations in PPC , these factors could have obscured our ability to observe perceptual plasticity across the set of odors . Finally , while changes in piriform odor representations might have induced parallel changes in perception , it is possible that other brain areas would be able to compensate for these perturbations , helping to stabilize olfactory perceptual performance . For example , in the placebo group , increased within-category pattern separation in PPC ( Figure 5b ) would be counteracted by decreased within-category pattern separation in pHIP ( Figure 6b ) , resulting in no detectable change at the behavioral level . One potential issue is that baclofen can also target GABA ( B ) receptors that have been identified in area CA1 of the hippocampus , influencing visual object recognition and memory ( Lanthorn and Cotman , 1981; Ault and Nadler , 1982 ) . Therefore , to establish that our findings were specific to the olfactory system , and to ensure that baclofen did not disrupt general semantic processing and object categorization , subjects also performed a visual categorization fMRI task in which they viewed pictures rather than smelled odors . This control study confirmed that our pharmacological manipulation induced both regional and modality specificity , thus ruling out possible confounds such as altered global attention , arousal , or hemodynamic reactivity . As an added way to minimize mere drug effects , we explicitly focused our imaging analyses on the interactions between group ( baclofen/placebo ) , session ( pre/post ) , and category level ( within/across ) , effectively cancelling out any other session-related confounds . An unavoidable limitation of this study was that baclofen was administered systemically . While our findings demonstrate regionally selective treatment effects in PPC , it is not possible to confirm that these changes were due to the direct action of baclofen solely at piriform cortex . There are at least three mechanisms by which baclofen could affect categorization in the olfactory network , none of which are mutually exclusive . First , baclofen might directly target the layer 1b synapses in piriform cortex where associative intracortical and extracortical inputs predominate . This would most closely mirror what has been tested using focal baclofen injections in animal models ( Poo and Isaacson , 2011; Barnes and Wilson , 2014 ) , and would underscore the idea that categorical odor representations rely on associative information processing within this layer of piriform cortex . Second , baclofen might target neurons in OFC , entorhinal cortex , and other associative brain areas that project onto piriform cortex . Given that the fMRI BOLD response is thought to reflect local dendritic processing and population activity ( Logothetis and Wandell , 2004; Hipp and Siegel , 2015 ) , our findings could reflect a distant action of baclofen on OFC ( or other areas ) , which in turn alters distributed fMRI patterns measured in piriform cortex . Third , the changes seen in PPC could theoretically have arisen in the olfactory bulb , where GABA ( B ) receptors have also been described ( Nickell et al . , 1994; Palouzier-Paulignan et al . , 2002; Okutani et al . , 2003; Wachowiak et al . , 2005; Aroniadou-Anderjaska et al . , 2000; Isaacson and Vitten , 2003; Karpuk and Hayar , 2008 ) . In this instance , one might have predicted a more profound olfactory perceptual deficit , including impairments of odor threshold , identification , and perceived intensity , though such a profile was not found in our study . Irrespective of the specific mechanism or mechanisms , these findings establish a critical role of the GABA ( B ) receptor in modulating categorical representations in PPC and OFC , with specificity for the olfactory modality . In summary , our study provides a foundation for understanding the contribution of afferent and associative inputs to odor categorical perception in the human brain . Of note , this work forms a counterpoint to an earlier study from our lab in which subjects underwent a 7-day period of odor deprivation ( Wu et al . , 2012 ) : by reducing olfactory afferent input , we were able to show that multivariate pattern representations of odor category were selectively altered in OFC , without any pattern-based changes observed in PPC . By comparison , in the current study , we were able to test the inverse manipulation , using baclofen to reduce olfactory associative input . In this instance , we again observed a disruption of odor categorization in OFC , but also an interference of session-related pattern changes in PPC and pHIP . The fact that within-category pattern changes in PPC were complementary to those in pHIP , in conjunction with the different course of across-category changes in OFC , underscores the idea that odor categorization is a dynamic process involving multiple stages of an extended olfactory network . We surmise that under normal conditions , the ability to refine discriminability of within-category odors in PPC through experience helps to improve perceptual acuity and decision making , and to prevent perceptual generalization from becoming maladaptive . With the interruption of associative input , in the setting of experimental baclofen or even perhaps as the consequence of a neurological disorder , within-category boundaries can become obscured , leading to perceptual over-generalization that can result in detrimental choices . As such , our findings may point toward an important mechanism by which associative networks regulate perceptual processing . Whether such mechanisms are restricted to the olfactory modality , or apply more widely across different sensory systems , remains to be determined .
We obtained informed consent from 36 subjects ( mean age , 25 years; 18 baclofen and 18 placebo , with equal numbers of men and women in each group ) to participate in this study , which was approved by the Northwestern University Institutional Review Board . Subjects were right-handed nonsmokers with no history of significant medical illness , psychiatric disorder , or olfactory dysfunction . Four female baclofen subjects were excluded from the results due to either excessive movement or falling asleep in the scanner , leaving a total of 14 baclofen subjects . Prior to the main experiment , we conducted a screening session to ensure that subjects had normal olfactory abilities and were able to categorize odors reliably . Subjects rated intensity , pleasantness , and familiarity for six odors belonging to three categories . Each odor was presented three times . Familiarity was rated on a visual analog scale ( VAS ) with end-points of ‘extremely unfamiliar’ and ‘extremely familiar’ , and the cursor was reset at the mid-point on every trial . For analysis , the VAS ratings were scaled to a range from 0 ( extremely unfamiliar ) to 10 ( extremely familiar ) . The mean familiarity rating of the six odors across all subjects was 7 . 68 ± 0 . 20 , suggesting that subjects found the odor set to be relatively familiar . There was also no difference in odor familiarity ratings between subjects assigned to the placebo and baclofen groups ( two-sample t-test , t30 = -1 . 18 , P = 0 . 25 ) . Additionally , during the screening session , and following the odor ratings , subjects were asked to smell the six odors from glass bottles and to sort them into three categories . All subjects enrolled in the study were able to sort the odors appropriately into the three categories . In this manner , subjects were pre-exposed to the odors , found them to be familiar , and could associate them with categorical knowledge . The total length of the experiment spanned 5 consecutive days . Following enrollment , subjects were randomly assigned to the baclofen ( n = 14 ) or placebo ( n = 18 ) group by the research pharmacy at Northwestern Memorial Hospital . Experimenters and subjects were both blinded to these assignments . Subjects took 10 mg of baclofen or placebo on the first day and progressively increased the dosage by 10 mg per day to reach 50 mg at day 5 . On day 1 before drug administration , subjects underwent pre-drug baseline tests including cognition , olfactory psychophysics , and fMRI imaging measures . On day 5 after medication , subjects completed post-drug tests which were the same as the pre-drug session . Six odorants were used in the fMRI odor categorization experiment and included two ‘citrus’ smells ( R- ( + ) -limonene and Citral ) , two ‘mint’ smells ( L-Menthol and Methyl Salicylate ) , and two ‘wood’ smells ( Cedrol and Vetiver Acetate ) . For the fine odor discrimination task outside the scanner , two perceptually similar isomers , α- and β-pinene ( 5% diluted in mineral oil ) , were used in an olfactory three-way forced choice triangular task . Odors were delivered using a custom-built olfactometer . In this system , clean air or odorized air was directed towards subjects ( wearing a nasal mask ) via Teflon tubing at a rate of 3 L/min . On days 1 and 5 , subjects were tested on four cognitive measures before olfactory testing and fMRI scanning: ( 1 ) Mini-mental state examination ( MMSE ) , a short questionnaire used to measure cognition impairment ( Folstein et al . , 1975 ) ; ( 2 ) an auditory digit span test ( in forward and backward order ) to assess short-term memory; ( 3 ) Trail Making Test B as a measure of visual attention and cognitive flexibility ( Bowie and Harvey , 2006 ) ; and ( 4 ) subjective report of degree of alertness using the Stanford Sleepiness Scale ( SSS ) ( Hoddes et al . , 1973 ) , which ranges from “Feeling active , vital , alert , or wide awake” ( 1 point ) to “No longer fighting sleep , sleep onset soon; having dream-like thoughts” ( 7 points ) . Four behavioral measures were tested outside of the scanner . ( 1 ) Odor detection thresholds and ( 2 ) odor identification ability were assessed using Sniffin’ Sticks ( Burghart ) and the University of Pennsylvania Smell Identification Test ( UPSIT , Sensonics ) , respectively ( Doty et al . , 1984; Hummel et al . , 1997 ) . ( 3 ) A triangular odor discrimination task was performed to assess the ability to discriminate α- and β-pinene ( Li et al . , 2008 ) . ( 4 ) For the six odorants used in the fMRI odor categorization experiment , visual analog ratings of odor intensity ( anchors , ‘undetectable’ and ‘extremely intense’ ) , pleasantness ( anchors , ‘dislike” , ‘neutral’ , and ‘like’ ) , pair-wise similarity of odor quality ( anchors , ‘not alike at all’ and ‘identical’ ) ( Howard et al . , 2009 ) were collected . Subjects also rated the applicability of descriptors of the three categories ( citrus , mint and wood ) with anchors ( ‘not at all’ and ‘extremely citrusy/minty/woody’ ) . Subjects underwent an odor categorization task designed to assess the multivoxel pattern specificity of odor-evoked fMRI activity across pre- and post-drug sessions . The task was divided into six 8-min runs of 28 trials each , during which the six odors were presented for 4 or 5 trials ( depending on the run ) . On each trial , subjects were presented a visual sniff cue prompting them to sniff . Odor stimuli were presented for 1 . 5 s , with a 13-s stimulus-onset asynchrony ( SOA ) . Each odor was presented 28 times in pseudorandom order . Four out of the 28 trials in each run were randomly chosen as ‘catch trials’ , where subjects were asked to indicate the category of the received odor with a mouse click . The catch trials were not included in the fMRI pattern analysis . The total task lasted for 48 min . Subjects also performed a visual categorization task which was parallel to the olfactory version with the equivalent number of trials and runs , and visual and olfactory runs were interleaved . On each trial , an image ( from a total of six possible images , Figure 7a ) was presented for 0 . 5 s , with a jittered interval of 3–4 s between trials . The visual fMRI data were absent from 1 male placebo subject due to technical problems during the experiment . A separate functional localizer scan was performed to identify regions of image-evoked activity to be used in the visual pattern analysis . This scan was done in the pre-drug session , in which subjects were shown seven 20-s blocks of images ( 0 . 3 s presentation and 0 . 7 s inter-stimulus interval ) with 20-s resting gaps between blocks . Each block contained one of six object categories ( chairs , houses , teapots , cars , keys , and scissors ) or scrambled version of the same images . The scrambled images were created by dividing the images into 20 × 20 unit grids and shuffling the units . During the image presentation blocks , subjects performed a one-back detection task by pressing a button to maintain their focus and attention . Breathing behavior was monitored during olfactory scanning with a spirometer ( affixed to the nasal mask ) measuring the flow of air during inhalation and exhalation . Respiration signals from each run were first smoothed and then scaled to have a mean of 0 and standard deviation of 1 . The cued sniff waveforms were extracted from each trial , and inhalation peak flow , duration , and volume were computed . In the pre-drug session , there were no systematic differences in peak flow ( F3 . 4 , 105 . 52 = 1 . 44 , P = 0 . 23 , repeated measures ANOVA ) or duration ( F3 . 97 , 123 . 18 = 0 . 89 , P = 0 . 47 ) across odors , but the inhalation volumes were different ( F3 . 93 , 121 . 88 = 3 . 27 , P = 0 . 014 ) . Therefore the inhalation volume was included in the fMRI analysis as a nuisance regressor ( see below ) . Gradient-echo T2*-weighted echoplanar images were acquired with a Siemens Trio 3T scanner using parallel imaging and a 12-channel head-coil ( repetition time , 2 . 3 s; echo time , 20 ms; matrix size , 128 × 120 voxels; field-of-view , 220 × 206 mm; in-plane resolution , 1 . 72 × 1 . 72 mm; slice thickness , 2 mm; gap , 1 mm ) . A 1 mm3 T1-weighted MRI scan was also obtained for defining anatomical regions of interest ( ROIs ) . fMRI data were pre-processed with SPM8 software ( http://www . fil . ion . ucl . ac . uk/spm/ ) . All functional images across pre- and post-drug sessions were spatially realigned to the first scan of the first run to correct for head movement . The T1 structural image was also co-registered to the mean aligned functional image . Realigned functional images were then normalized into a standard space using the transformation parameter from each individual’s T1-weighted scan to the standard T1 template . For multivariate fMRI analysis of olfactory and visual categorization scans , we did not perform subsequent spatial smoothing in order to preserve the voxel-wise fidelity of the signal . Images from visual localizer scans were smoothed for generating functional visual object recognition ROIs . Results are shown as mean ± s . e . m . for subjects and sessions . For determining category encoding regions , we used one-tailed t tests to compare decoding accuracy to chance . To test for drug effects on behavior and fMRI patterns , we used a mixed-model three-way ANOVA , with one between-group ‘drug’ factor ( placebo/baclofen ) and two repeated-measures within-subject factors of ‘session’ ( pre/post ) and ‘category type’ ( within/across ) . Here the critical contrast was the group × session × category interaction , with post-hoc analysis of the simple effects where appropriate . Pearson’s linear correlation coefficient was calculated for the correlation analysis of behavioral and fMRI pattern data across subjects . Significance threshold was set at p<0 . 05 , two-tailed , unless otherwise stated . | Imagine bringing your groceries home and tucking them into the refrigerator . You’ll probably organize the items by categories: lemons and oranges into the fruit drawer , carrots and cauliflower into the vegetable drawer . Categorization is essential , allowing us to interact with the world in the most efficient way possible . If the differences between objects are not relevant to the task at hand , the brain will group objects together based on their shared properties and develop mental representations of the “categories” . Importantly , we are still aware of the distinctions between objects within the same category . Categories of odor ( for example , minty or fruity ) are represented in a part of the brain called the olfactory ( or piriform ) cortex , which receives information from odor cues as well as “top-down” information from other areas of the brain . But how do these top-down pathways influence odor categorization ? Bao et al . asked how the brain solves the problem of categorizing odors . For the experiments , human volunteers smelled six familiar odors belonging to three different categories while their brain activity was monitored using a magnetic resonance imaging ( fMRI ) scanner . Then , half of the participants were given a drug called baclofen that prevents top-down inputs , but not odor cues , from reaching the piriform cortex , while the rest received a placebo . After five days of taking the medication , all of the volunteers had another session of fMRI where they had to categorize the same odors as before . The experiments show that when comparing the fMRI scans before and after the drug treatment , the representations of odors belonging to the same category became more distinct in the piriform cortex in the placebo group . Put differently , as the volunteers were repeatedly exposed to odors of well-known categories , they became better at discriminating individual odors within the same category . However , these changes were disrupted in the group of volunteers that took baclofen . Bao et al . ’s findings indicate that this “practice makes perfect” approach to recognizing odors relies on top-down inputs into the piriform cortex . In future work it will be important to study the roles of these inputs in learning new categories of odors , and to investigate whether the mechanisms identified here apply to other sensory information and to more abstract knowledge . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | The role of piriform associative connections in odor categorization |
Mammalian visual behaviors , as well as responses in the neural systems underlying these behaviors , are driven by luminance and color contrast . With constantly improving tools for measuring activity in cell-type-specific populations in the mouse during visual behavior , it is important to define the extent of luminance and color information that is behaviorally accessible to the mouse . A non-uniform distribution of cone opsins in the mouse retina potentially complicates both luminance and color sensitivity; opposing gradients of short ( UV-shifted ) and middle ( blue/green ) cone opsins suggest that color discrimination and wavelength-specific luminance contrast sensitivity may differ with retinotopic location . Here we ask how well mice can discriminate color and wavelength-specific luminance changes across visuotopic space . We found that mice were able to discriminate color and were able to do so more broadly across visuotopic space than expected from the cone-opsin distribution . We also found wavelength-band-specific differences in luminance sensitivity .
The mouse visual system is increasingly ( Baker , 2013; Priebe and McGee , 2014 ) being used as a model system for studying both cortical sensory processing ( Glickfeld et al . , 2013 , 2014; Niell and Stryker , 2008 , 2010; Wang et al . , 2011a ) and behavior ( Busse et al . , 2011; Harvey et al . , 2012; Histed et al . , 2012; Hoy et al . , 2016; Montijn et al . , 2015 ) . While most physiological work has used achromatic stimuli ( Durand et al . , 2016; Niell and Stryker , 2008 ) , mice , like most other mammals , display physiological color-opponent signals in the retina ( Baden et al . , 2013; Baden et al . , 2016; Breuninger et al . , 2011; Chang et al . , 2013; Joesch and Meister , 2016 ) , through LGN ( Denman et al . , 2017 ) and possibly V1 ( Tan et al . , 2015 ) . The mouse retina displays asymmetric and mixed expression of its two opsins along the dorsal-ventral axis of the retina , creating opposing gradients of short and middle opsins ( Applebury et al . , 2000; Wang et al . , 2011b ) and resulting in gradients of wavelength-band-specific responses ( Chang et al . , 2013; Denman et al . , 2017; Rhim et al . , 2017 ) . Therefore , the substrate for cone-driven color-opponent signals , and any color sensitivity , exists only in the overlapping ‘opsin transition zone’ ( Baden et al . , 2013; Denman et al . , 2017 ) . However , short and middle opsin responses broadly overlap in V1 and higher visual areas ( Rhim et al . , 2017 ) and rod-cone antagonism can also create color opponency in some mouse retinal ganglion cells ( Joesch and Meister , 2016 ) , presenting the possibility that behaviorally relevant color information could be extracted more broadly across retinotopic space . Whether mice can use color information to guide visual behavior is an open question . There is some evidence for color discrimination ( Jacobs et al . , 2004 ) , but it remains unclear how this depends on overall luminance , luminance contrast , or retinotopic position . Further , it is not known if the gradients in opsin distribution lead to variations in behavioral luminance sensitivity across space . Such non-uniformity would impact studies of visuotopically extended V1 populations , such as studies of population sparsity ( Froudarakis et al . , 2014 ) , population correlations ( Montijn et al . , 2016a ) and other notions of population coding ( Montijn et al . , 2015; Okun et al . , 2015 ) . Here , we use a simple behavior , change detection , to determine where in visual space mice can discriminate changes in chromaticity and luminance at ethologically-relevant mesopic ( 10−3 - 3 cd/m2 , [Wyszecki and Stiles , 1982] ) luminance levels . By measuring detectability of luminance and color changes separately across elevation ( spanning ~75° ) , we are able to generate an estimate of wavelength-specific contrast sensitivity across visual space . Mice were able to discriminate color , but only at elevations above the horizon . We find both wavelength-specific luminance and color contrast sensitivity to be dependent on retinotopic location , but that these differences in sensitivity were less dramatic than expected from the cone opsin distribution , suggesting behavioral access to differential activation of rods and cones .
To examine the psychophysical and physiological basis of mouse color vision , we first trained mice in a go/no-go change detection task ( Histed et al . , 2012 ) in an immersive visual stimulation environment customized for delivering stimuli in the spectral bands of the mouse short and middle wavelength opsins ( Figure 1A; Materials and methods ) . We use the system here to deliver a video stimulus driven by a green and ultraviolet LED projector; for each point on the stimulus , the green and ultraviolet intensity could be independently modulated . Total luminance was in the mesopic range ( <3 cd/m2 ) , over which mice are both behaviorally active ( Daan et al . , 2011 ) and color opponent signals have been demonstrated in the retina ( Baden et al . , 2013; Joesch and Meister , 2016 ) . Briefly , under this paradigm , mice indicate that they have perceived a change in the stimulus by licking a reward spout within 1 s of the change ( Figure 1B ) ; subsequent licks allow reward consumption ( Figure 1C ) . Following pre-training on a luminance change detection task ( see Materials and methods ) , we switched to change detection sessions in which the ultraviolet and green intensity , centered on the mouse short and middle wavelength bands , respectively , were varied independently on each trial . Each trial contained a change in intensity within a 15° test circle on a mean luminance background at one of four elevations: −10° , 10° , 30° , and in some cases 50° ( relative to both the horizon and the placement of the rotating mouse platform ) . We varied position only along elevation because both rods and cones are relatively uniform across the azimuthal axis of the retina ( Sterratt et al . , 2013 ) . Eye position did not change with stimulus location ( Figure 1—figure supplement 1 ) . Stimuli at change times were modulations in ultraviolet and green intensities within the stimulus spot , resulting in changes in short and middle band stimulation . Changes in ultraviolet and green intensity were independent of each other , and of the previous intensities , resulting in mainly trials that contained luminance changes , but also some in which short and middle opsin activation oppose each to create an effective hue exchange without a luminance change . Changes covered all of the color space available using our display . There is no explicit level of chance performance in this task design , as false-positive rate could vary depending on the strategy used to guess; we include catch trials in which the stimulus did not change in order to estimate false positive rate at each elevation . To achieve sufficient trials to cover this space , we presented a total of 127 , 659 trials ( n = 4/5 total mice trained , 284 sessions ) . To control for motivation , we calculated a running average of the reward rate and selected trials where this reward rate remained above four rewards per minute; only these engaged trials ( 44%; 56 , 112/127 , 659 ) were used for analysis . We first examined our results to estimate the relative luminance contrast sensitivities to short and middle-wavelength band stimulation across the visual field . Although the green and ultraviolet projector LEDs nearly isolate responses of the middle and short wavelength-sensitive opsins ( Estévez and Spekreijse , 1982 ) , they do not necessarily isolate responses of individual cones , most of which express a combination of the two opsins . Nor do they necessarily measure the relative weight of the cone opsins themselves , as rods also may contribute to light sensitivity at these luminance levels . Rather , we present a measure of the relative perceptual weight to stimuli of the middle and short wavelength bands covered by our stimulus LEDs ( Figure 2—figure supplement 1 ) , as combined through both cone opsins and rods . Total luminance change detection saturated by ~30% at all elevations ( Figure 2A ) and the half-saturation threshold ( hereafter referred to as ‘threshold’ ) was less than 12% for each elevation ( Figure 2B ) . The highest sensitivity was in the upper visual field , at 5 . 5% threshold , a threshold that is consistent with previous reports for a 15° ( ~0 . 07 cyc/° ) stimulus ( Histed et al . , 2012; Prusky et al . , 2000; Sinex et al . , 1979; Umino et al . , 2008 ) . Sensitivity to increments in contrast was similar for all elevations ( 10 . 5–12 . 1% ) . Consistent with previous physiological measurements in V1 of mouse ( Tan et al . , 2015 ) and other species ( Kremkow et al . , 2016; Wang et al . , 2015 ) , sensitivity to decrements in contrast was higher than sensitivity to increments ( 5 . 4%–10 . 4% ) . Notably , this was most pronounced in the upper visual field ( difference: 5% ) than the lower visual field ( difference: 1 . 3% ) . To determine the independent contributions of short and middle wavelength bands , we examined change trials that contained increments or decrements of only one of the two LEDs . For the short wavelength band ( i . e . UV ) , contrast sensitivity was non-uniform , with the highest sensitivity in the upper visual field ( Figure 2C , D; 8% threshold ) . As with total luminance , mice were more sensitive to decrements than increments in contrast . The non-uniformity across elevation was more pronounced for short-wavelength-specific sensitivity than total luminance , but was restricted to decrements . The middle-wavelength ( i . e . green ) luminance contrast sensitivity was also non-uniform , across elevation , but with the opposite relationship as the short wavelength and total luminance: higher sensitivity at the two lower elevations and the highest tested elevation ( Figure 2E–F ) . Middle-wavelength sensitivity was very similar for increments and decrements , again different than total and short-wavelength luminance contrast sensitivity . In summary , we found that luminance contrast sensitivity was non-uniform , with significant opposing wavelength-band-specific non-uniformities , although less than what would be predicted from the opsin expression or photoreceptor response alone , as shown below . We next determined the relative strength of short- and middle-wavelength stimulation across the visual field . Despite existing measurements of the cone gradients and non-uniformity in V1 response ( Rhim et al . , 2017 ) , we were uncertain about the relative contributions of rods and cones at the tested light levels ( i . e . relative contributions of the rod and middle opsin to middle band stimulation ) , so we first determined which combinations of wavelength band activation effectively opposed each other at each elevation . For both this determination of relative wavelength band weight , and our study of color sensitivity ( see below ) , our approach is schematized in Figure 3A: for each trial , we plot the change ( at the change time in the task , see Figure 1B ) of each LED intensity against the other . Equal but opposite changes in the activation of the short and middle bands should oppose each other and lead to some change in chromaticity ( colorfulness relative to luminance ) , but only a subset of changes will have sufficient change in chromaticity to be distinguishable ( MacAdam , 1942 ) , if the mouse can distinguish chromaticity at all . Unbalanced changes in intensity create a luminance change , and often a chromaticity and hue change as well , with only those that have equal changes to each wavelength band produce a pure luminance change . Because sensitivity is higher to any luminance change than pure chromaticity changes , the resulting performance forms an elongated ellipse when plotted this way . If our stimulation of each band were equal , the major axis of this ellipse would fall along the unity line , indicating equiluminant stimuli that contain chromaticity change without a luminance change . However , the relative expression of each opsin are not equal across the retina ( Applebury et al . , 2000; Baden et al . , 2013; Ortín-Martínez et al . , 2014 ) , so the slope of the major axis of the observed ellipse , hereafter called the ‘equiluminant line’ , indicates the relative weight of each luminance band at that location . As such , a slope of 1 lies on the unity and equal short and middle-band weight ( Figure 3A , left ) ; slopes >1 indicate middle band domination ( Figure 3A , right ) and <1 short band domination . Note that the axes of these plots are changes in band activation , not absolute intensity levels . The trials in any given bin in the plot have different absolute values of color , both before and after the change point , but the movement within any color space at the change point is always the same ( in both direction and magnitude ) . We measured the axis along which opposite changes in short and middle wavelength stimulation effectively canceled each other at several elevations . The performance across pairwise combinations of changes in short and middle band luminance ( Figure 3B ) was fit with an ellipse ( Figure 3C , top ) , and the major axis of this ellipse was taken to be the equiluminance line ( Figure 3C , bottom ) . We found the mouse to be more sensitive to middle band stimulation than expected at all elevations tested , including at 30° where , given the eye positioning ( Figure 1—figure supplement 1 ) , short-opsin expression dominates . In fact , surprisingly , the mice were more sensitive to middle than short-band changes at all elevations , with the following middle/short ratios: 3 . 4 , 3 . 6 , and 2 . 25 at –10° , 10° , and 30° , respectively . We compared our measure of the wavelength-band-specific perceptual contributions to predictions from the cone expression distribution , the cone functional response , and intrinsic imaging of the mouse visual system ( Baden et al . , 2013; Rhim et al . , 2017 ) . By projecting ( Sterratt et al . , 2013 ) the spatial profile of cones into visuotopic coordinates based on estimations of the mean eye position during our experiments ( Figure 1—figure supplement 1 ) , we computed expected middle/short ratios for each of the elevations we tested ( Figure 3—figure supplement 1 ) . Similar to the estimated relative opsin weights across V1 under photopic conditions ( Aihara et al . , 2017; Rhim et al . , 2017 ) , 2 . 3 , 0 . 81 , 0 . 81 , the cone functional distribution predicts 2 . 3 , 1 . 0 , and 0 . 44 at –10° , 10° , and 30° , respectively . Both predict far less middle-band sensitivity that we observed . This result suggests that rod opsins , centered near the middle-band , contribute significantly to mouse perceptual sensitivity at these light levels , at least as much as 60% ( Figure 3—figure supplement 1 ) at higher elevations . We continued to ask if mice could report a change in chromaticity independent of any luminance change . Instead of explicitly creating a device that normalized total luminance during chromaticity changes ( Gagin et al . , 2014; MacAdam , 1942; Wyszecki and Stiles , 1982 ) , we presented sufficient combinations of wavelength-band-specific luminance changes to experimentally determine when chromaticity changes occurred independent of luminance changes . We began with the approach ( and data ) described above for Figure 3A–C , but now specifically examining which combinations indicate color discrimination independent of luminance change . Again note that the axes of these plots are changes in band activation , not absolute intensity levels , so the major axis of the ellipse is equivalent to chromaticity changes and the minor axis lightness change , collapsed across all hues . We start with the assumption that a lack of change in luminance or chromaticity is not discriminable to the observer . All luminance and color contrast changes that are behaviorally indistinguishable from this ‘no change’ condition ( the 0 , 0 point in Figure 3A , B ) form an ellipse of non-discriminability analogous to a MacAdam ellipse of human non-discriminability on the human chromatcity diagram ( MacAdam , 1942 ) ; in our case , this ellipse is in a chromaticity-luminance space ( Figure 3A , B ) . As noted above , the major axis of this ellipse specifies this equiluminance line for the wavelength band sensitivities at that elevation ) . This line is equivalent to a slice through the equiluminant plane in a DKL color space ( Derrington et al . , 1984 ) , and by examining change detection performance along this experimentally defined axis of chromaticity change , we can ask if mice can discriminate color independent of luminance ( Figure 3A , left , orange ellipses ) . Indeed , qualitative inspection of the extremes of the equiluminance lines in Figure 3B indicated color discrimination ability at some elevations , and differences across visuotopic space . However , because of the shift of short versus middle contributions across space ( slopes of the lines in Figure 3C ) , we were concerned that the stimuli along the measured equiluminance line at each elevation ( Figure 3B ) could contain a within-stimulus luminance gradient that allowed the mice to exploit small within-stimulus luminance changes to detect changes we interpreted as color changes . Further , because the range of chromaticities presented at each elevation was unequal ( lengths of the lines in Figure 3C ) , we were concerned that the mouse could be sensitive to chromaticity changes at −10° outside of the range of the stimuli presented . To overcome these issues , we adjusted the intensity of the middle band stimulation by fitting an exponential to the slopes of the equiluminance lines across space , normalizing this to the maximum possible intensity of the green LED ( Figure 3D , top ) . This should create conditions of uniform luminance contrast across all elevations , shifting the major axis of the ellipse at all elevations to the unity line ( Figure 3D , bottom ) . As such , we limited our testing conditions to the changes around the unity line in order to reduce the number of sessions needed to achieve sufficient trials in each condition; this results in spear-like plots focused on luminance-independent color change . While our adjustment failed to perfectly compensate for variable short-middle weights , we were still able to capture more chromaticities at each elevation and minimize potential luminance gradients . We found color discrimination to depend on elevation . Examining the adjusted data ( Figure 3F–H ) , we found color discrimination was negligible at −10° , but mice were capable of varying levels of color sensitivity at all other elevations tested ( Figure 3F , G ) . The performance along the equiluminant line at −10° was not well fit by a hyperbolic ratio function , and the performance in catch trials ( 0% contrast change , false-positive rate ) was not significantly different from any point along the line ( p>0 . 05 , student’s t-test ) . We were able to fit the performance at each of the other elevations tested , up to 50° above the horizon . Color contrast sensitivity was highest for decrements in short-middle opponency at 10° elevation ( 13 . 4% ) ; color sensitivity was nearly identical for decrements in short-middle opponent contrast at 30 and 50° and for increments at all elevations above the horizon ( 29 . 3–32 . 1% ) . Notably , the false-positive rate ( 0% contrast change , catch trials ) was markedly different between the lowest and other elevations . We interpret this is a shift in strategy in order to maximize reward volume , as the total volume of reward did not differ between −10° and the other elevations ( 0 . 8−0 . 92 mL/session , p>0 . 2 , Welch’s t-test ) , despite the lack of color discrimination at −10° .
The finding that both wavelength-specific luminance ( Figure 2 ) and color contrast ( Figure 3E–F ) sensitivity is not uniform across the visual field is in accordance with the distributions of both retinal and primary visual cortical ( Rhim et al . , 2017 ) responses . However , we found that middle-band sensitivity was both higher and more uniform than expected ( Figure 2 ) . This suggests that rod sensitivity contributes significantly to perceptual sensitivity at these light levels ( Figure 3—figure supplement 1 ) . This finding may be important for studies of the mouse visual system that use visuotopically extended stimulation ( Froudarakis et al . , 2014; Garrett et al . , 2014; Montijn et al . , 2015; Montijn et al . , 2016b; Pouille et al . , 2009; Zhuang et al . , 2017 ) , especially those that measure the underlying population representation of the stimuli . Because the spatial scale of luminance and contrast adaptation can be large ( Smirnakis et al . , 1997 ) , the adaptation to large single-band stimuli ( such as those produced by LCD or other sRGB displays ) in these studies may underestimate the contrast sensitivity for cells in upper visual field . This spatial scale is especially relevant because of the scale of mouse vision – 50% differences can be seen across a small number ( ~5 ) receptive field diameters . Our results also demonstrate that color sensitivity depends on retinotopy , and that some retinotopic locations appear to not support color discrimination . A goal of many large-scale data collection efforts , both completed ( Baden et al . , 2016 ) and underway ( brain-map . org/visualcoding ) as well as smaller-scale surveys ( Durand et al . , 2016; Gao et al . , 2010; Niell and Stryker , 2008; Piscopo et al . , 2013 ) from retina to V1 is the classification or clustering of response properties in order to define functional channels . Because color-opponent cells , both single and double ( Shapley and Hawken , 2011 ) , are thought to underlie such behavior , our findings indicate that mice may have at least one , likely at least two , color-opponent cell types; the presence of such functional cell types may depend strongly on retinotopy . Notably , our animals are housed in an environment with fluorescent lighting that does not provide UV-B for reflection ( Figure 2—figure supplement 1 ) , suggesting that the behavior we observed is developmentally specified , not learned , and further not lost through lack of use . Any color-opponency depends on the underlying distribution of opsins in the retinal photoreceptor layer . The distributions of cone opsins in the mouse retina has been well described , with only a subset of pure cones expressing only a single opsin and 40% of cones expressing a mix of both , and a non-uniform distribution of both pure and mixed cones: middle-wavelength opsin dominates the dorsal retina , although it also extends ventrally , while the short-wavelength opsin is more constrained to the ventral retina . ( Applebury et al . , 2000; Baden et al . , 2013; Ortín-Martínez et al . , 2014; Wang et al . , 2011b ) . In total , cones are relatively uniform across the dorsal-ventral axis of the retina , although others report some increased density in the ventral retina . Rods , on the other hand , are uniform across the dorsal-ventral axis of the retina and are much more numerous than cones ( Jeon et al . , 1998 ) . If color responses , both physiological and behavioral , rely only on cone signals they should follow the cone opsin distribution exactly: more in the transition zone , some in the upper visual field , and none in the lower visual field . Our results are consistent with cone-cone opponency , as we observed a lack of color discrimination only in the lower visual field . However , our result also indicates some rod-cone opponency , as color change detection behavior thresholds relatively unchanged across 10° to 50° , despite the steep decline in middle opsin expression across this range of the retina ( Baden et al . , 2013 ) . Mice , while often considered nocturnal ( Febinger et al . , 2014 ) , can be behaviorally active across a range of luminance conditions; C57BL/6 mice in particular can shift between diurnal , crepuscular , and nocturnal behavioral patterns over the course of the year ( Daan et al . , 2011 ) . Previous studies on color signaling in the mouse have offered several hypotheses for the ethological uses of color signals . Our results are consistent with the hypothesis ( Baden et al . , 2013 ) that non-uniform wavelength-specific sensitivity is matched to the luminance statistics of natural scenes ( UV in the upper visual field , green in the lower ) , and high sensitivity to decrements in the short wavelengths may be particularly helpful during the shift toward UV in the spectral radiance distribution the during twilight hours ( Spitschan et al . , 2016 ) . Another hypothesis , that short-middle opponency is useful for identifying mouse urine posts ( Joesch and Meister , 2016 ) , is inconsistent with our demonstration of a lack of color discrimination at −10° , at least in absence of significant excursions in eye position or head movements . Our results suggest that , under the mesopic condition during which mice are often active , color signals may be mediated by both cone-cone and some rod-cone opponency , and this may facilitate the specialization of cone opsin distributions for sampling natural luminance statistics ( Baden et al . , 2013; Chiao et al . , 2000 ) .
All animals used in this study ( n = 5 ) were C57Bl/6J male mice aged 30–300 days obtained from The Jackson Laboratories ( IMSR_JAX:000664 ) . To fix the animal's head within the behavioral apparatus , a single surgery to permanently attach a headpost was performed . During this surgery , the animal was deeply anesthetized with 5% isoflurane and anesthesia maintained throughout the surgery with 1 . 5–2% continuous inhaled isoflurane . The mouse was secured in a strereotax with ear bars; hair was removed and the exposed skin sterilized with three rounds of betadine . An anterior-posterior incision was made in the skin from anterior of the eyes to posterior of the ears . The skin was removed in a tear drop shape exposing the skull . The skull was leveled and the headpost was placed using a custom stereotaxic headpost placement jig . A custom 11 mm diameter metal headpost with mounting wings was affixed to the skull using dental cement . The exposed skull inside the headpost was covered with a thin protective layer of clear dental cement and further covered with Kwikcast . The animal was allowed to recover for at least 5 days prior to the initiation of behavioral training . After headpost implantation , animals were kept on a reverse light cycle ( lights OFF from 9AM to 9PM ) and behavioral testing was done between 9AM and 1PM . Mice were habituated to handling gradually , through sessions of increasing duration . Mice were also habituated to the behavioral apparatus , first by allowing periods of free exploration and subsequently with head fixation sessions increasing from 10 min to 1 hr over the course of 1 week . Water restriction began with habituation; all mice were maintained at 85% of the original body weight for the duration of training and testing . Ultraviolet and human-visible stimuli were provided across a range of retinotopic locations using a custom spherical stimulus enclosure ( Denman et al . , 2017 ) ( Figure 1A , Figure 2—figure supplement 1 ) . A custom DLP-projector designed for the mouse visual system provided independent spatiotemporal modulation of ultraviolet ( peak 380 nm , Figure 1B ) and green ( peak 532 nm , Figure 1B ) light . The projection system operated at 1024 × 768 pixel resolution and a refresh rate of 60 Hz , achieving a maximum intensity of ~3 cd/m2 . Planar stimuli were spatially warped according to a custom fisheye warp for presentation on a curved screen; the fisheye warp was created through an iterative mapping protocol using the meshmapper utility ( http://paulbourke . net/dome/meshmapper/ ) calibrated on the behavioral environment to achieve maximal accuracy . Stimuli were presented in the right visual hemifield and consisted of 15° diameter circles of varying color on a mean intensity background using custom written software extensions of the PsychoPy package ( http://www . psychopy . org , RRID:SCR_006571 ) . The background intensity was 1 . 52 cd/m2 . For some testing sessions , the color of the display was adjusted to match the mouse's spectral sensitivity in order to create uniform and balanced sensitivity to the projector's LED sources across the visual stimulus enclosure . To do so , a second custom warp was applied that included a spatially dependent adjustment of the intensity of each LED ( near-UV and ‘green’ ) , according to the results shown in Figure 3C . Animals were head-fixed on a freely rotating disc in the center of the spherical enclosure and allowed to run freely during the course of training and testing . A lick spout was positioned approximately 0 . 5 cm in front of the mouse within range of tongue extension . In some experiments , infrared short-pass dichroic mirrors ( 750 nm short-pass filter , Edmund Optics ) were placed in front of each eye to allow for video tracking of the pupil . Cameras ( Mako and Manta , AVT technologies ) placed behind the animal were aligned to record a reflected image of the pupil; infrared illumination and a reference corneal reflection was provided via an LED positioned near the camera . Movies of the eye position during presentation of the stimuli used in the task was acquired at >=60 Hz , with the eye occupying >60% of the image at 300 × 300 pixels . Data from these sessions were not included in the performance analysis to avoid any potential artifact caused by the infrared dichroic . Animals were first shaped to associate changes in luminance with a reward . After each change in luminance , a water reward was automatically delivered , regardless of mouse licking behavior . During these sessions , the reward was constant at 10 µL . Incorrect licks were punished by resetting the trial , such that the mouse had to wait longer for the next change . This ‘shaping’ phase lasted a minimum of 2 days , but for most mice extended to several weeks . For some animals ( 2/5 ) , subsequent epochs of this automatic reward shaping served as task reinforcement when performance in testing blocks dropped . During each testing session , a circle was presented at a single visuotopic location and remained at this location for the duration of the session . At non-regular intervals , again selected from an exponential distribution , the color and/or luminance of the stimulus was changed , and the mouse had to report detection of change by licking the reward spout within 1 s of the change in order to receive reward ( Figure 1C ) . Licks were detected through a capacitive sensor connected to the reward spout . No water was present on the reward spout before the first lick; if the animal correctly detected a change , a water reward ( 3–10 µL , depending on animal and stage of training ) was delivered through this spout ( Figure 1D ) . Sessions were 50–60 min and typically included ~300 trials . The percent correct reported throughout this work is the total number of trials of that condition in which the mouse licked within 1 s of the change , divided by the total number of trials of that condition . The percent correct on catch trials , in which no stimulus change occurred but rewards were delivered for licks within 1 s of fictive change , is interpreted as the false positive or ‘chance’ performance level . Mice were first trained to associate changes of a 15° stimulus at 100% luminance contrast with a reward . In these sessions ( total of 3 to 25 sessions ) , the contrast of a stimulus ( 10° elevation , relative to the horizon ) changed at exponentially distributed intervals from 50% positive relative to the background to 50% negative ( from white to black ) , or vice versa ( black to white ) . If a lick occurred within 1 s of an actual stimulus change ( Figure 1B ) , a reward was delivered to the spout and liquid reward was consumed subsequent licks ( Figure 1C ) . If a lick occurred outside of this window the trial was aborted , extending the time the mouse must wait and effectively creating a ‘time-out’ period . Mice advanced from this protocol after performance exceeded 75% for consecutive sessions . In subsequent testing sessions , the intensity of the ultraviolet and green intensities were varied independently on each trial . Each trial contained a change in LED intensity for a 15° test circle on a mean luminance background at one of four elevations: −10° , 10° , 30° , and in some cases 50° . The first 8–20 trials of each session were 100% contrast changes , as described for the training blocks , with rewards automatically delivered . The number of these daily ‘free’ rewards was reduced to eight for as long as the mouse received >1 . 0 mL of reward during training or performed well enough to reach satiety and disengage from the task . We attempted to correct for sessions with poor performance by increasing these ‘free’ rewards on subsequent days before gradually reducing them again . To control for motivation in the results , we calculated a running average of the reward rate and selected trials where this reward rate remained above four rewards per minute; only these engaged trials ( 44%; 56 , 112/127 , 659 ) were used for analysis . All analyses were done using Python ( RRID:SCR_008394 ) and common scientific packages ( numpy RRID:SCR_008633 , scipy RRID:SCR_008058 , matplotllib RRID:SCR_008624 , and pandas ) . Code is publicly available ( Denman , 2017 ) and includes a Jupyter notebook that contains code for generation of our figures from the data . Data from each training session was saved and combined into a common data structure , also available from Denman , 2017 , that was used for all analysis . Individual sessions were analyzed to drive adjustments in the training parameters such as the number of automatic ‘free’ rewards . Following data collection for all animals , all sessions were loaded into a single object for analysis . This data structure can be recreated from the files made available from https://github . com/danieljdenman/mouse_chromatic ( copy archived at https://github . com/elifesciences-publications/mouse_chromatic ) . To quantify performance , from each trial the following parameters were extracted: change times ( the time of stimulus change ) , lick times ( the time of each lick , as detected through the capacitive sensor connected to the reward spout ) , and the stimulus conditions . A trial was scored ‘correct’ if the first lick after a change time occurred with one second , and if there was actually change in intensity of either green or ultraviolet at that change time . For each mouse , the percent correct was computed for each pair of LED state transitions , that is , each pairwise combination of change in short-band luminance and change in middle-band luminance ( e . g . Figure 3C ) . For each mouse , performance was ignored if three trials were not presented for those conditions . For fitting , missing data were replaced via a nearest neighbors approach , with the mean of the surrounding data . Our sampling strategies focused on the areas of changing performance , ensuring that cases of missing data were limited to the areas where performance had saturated at or near the lapse rate . Psychophysical curves for wavelength band-specific and color sensitivity were taken from the appropriate slices of this color space . Sensitivity was taken from the c50 parameter of fit a hyperbolic ratio fit ( Contreras and Palmer , 2003 ) with an additional offset term , R0 , to account for shift in false-positive rate across conditions . A total of five mice entered training on the task; one mouse failed to reach consecutive sessions of 75% performance during the initial high luminance contrast change detection phase , and so did not continue to testing in the color contrast discrimination phase . We did not use any statistical methods to determine mouse or trial sample size prior to the study , determining based on stability and consistency of results when sufficient samples had been collected . Statistical tests were student’s t-test unless otherwise specified . Eyetracking analysis was done via a semi-automated starburst algorithm; full details are available from the Allen Brain Observatory Visual Coding Overview v . 4 , June 2017 <http://help . brain-map . org/display/observatory/Documentation> . Briefly , the algorithm fits an ellipse to the pupil or corneal reflection ( CR ) area within user specified windows outlining the location of the pupil and corneal reflection . . A seed point is identified by convolution with a black square ( for the pupil ) and white square ( for the corneal reflection ) . An ellipse was fit to candidate boundary points identified using ray tracing using a random sample consensus algorithm . The fit parameters were first reported in coordinate centered on the mouse eye . To convert the tracked pupil positions from pixel coordinates in the eye tracking images to visuotopic coordinates , we first converted the pixel coordinates to spherical coordinates using an orthographic projection of the image onto a sphere with a radius of 1 . 16 mm , the reported radius if the mouse eye . Because the mouse was positioned in the center of the stimulus enclosure ( see the diagrams in Figure 1—figure supplement 1 , panel A ) , which is also a sphere , the position of the pupil in these ‘eye’ spherical coordinates can be converted to visuotopic coordinates by projecting the center of the ‘eye’ sphere on to the visuotopic sphere . The reflection of an infrared LED off of the cornea indicated the center of each eye , and we use the placement of each mirror relative to the mouse ( right eye: 51 . 2° in azimuth , 0° in elevation; left eye: 60° in azimuth and −10° in azimuth ) to determine the direction of gaze . As such , the corneal reflection in each eye tracking movie indicates a known reference point in visuotopic space , determined by the relationship of the imaging plan to the eye sphere , and the difference between the corneal reflection and the pupil allows for estimation of gaze position in visual coordinates . Coordinates for eye position were extracted independently for each frame of the eye position movie . This calculation assumes that the center of both eyes are at the center of the visual stimulation dome , that the each eye is a sphere , and that the movements of the eye are rotations about the center , none of which is strictly true . While left and right eye movies had different contrast and noise levels ( see Figure 1—figure supplement 1 , panels C-D ) , estimations of eye position in visual degrees brought disparate pixel measurement from each eye into good agreement with each other ( compare y pixel measures to elevation in Figure 1—figure supplement 1 , panel C ) . We believe this calculation to be accurate enough for comparison of relative eye positions and , given the small displacement of the eyes from the center ( <1 cm ) of the dome ( 30 cm radius ) , reasonable for using published retinotopic distributions to estimate relative opsin distributions in visuotopic coordinates . | Color is a key part of our visual experience . Humans can distinguish between colors thanks to light-sensitive cells at the back of the eye called cones . Our eyes contain three types of cones , most simply named red , green and blue . When light enters the eye , it activates each cone type to a different degree . The combined activity of the three types of cone determines which color we see . However , in about 8% of men , one of the cone types is missing or faulty . This leads to color blindness , usually in the form of an inability to distinguish between reds and greens . Most other mammals can also see colors . This includes mice , which are used increasingly to study the mechanisms underlying vision . But it was not clear if mice also use color to guide their behavior . Mice have only two types of cones , some of which respond to green light and others to ultraviolet light . To complicate matters , the two types of cones are distributed unevenly across the back of the mouse eye . This suggests that mice may see colors differently in different parts of a visual scene . Denman et al . trained mice to lick a spout whenever they noticed a change in the color or brightness of a dot appearing at various locations on a screen . The results revealed that mice could detect changes in both color and brightness . But the mice's ability to do so depends on where the change occurs . In the upper part of the visual field , corresponding to the area above the horizon , mice could distinguish between different colors . In the lower part of the visual field , below the horizon , they could not . By contrast , mice were able to detect changes in brightness at many different locations . This information will make it easier to design and interpret experiments that use mice to try to understand how the brain generates vision . This should help scientists develop new treatments for disorders that affect vision , and possibly for many forms of cognitive impairment too . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Mouse color and wavelength-specific luminance contrast sensitivity are non-uniform across visual space |
HIV replication requires nuclear export of unspliced and singly spliced viral transcripts . Although a unique RNA structure has been proposed for the Rev-response element ( RRE ) responsible for viral mRNA export , how it recruits multiple HIV Rev proteins to form an export complex has been unclear . We show here that initial binding of Rev to the RRE triggers RNA tertiary structural changes , enabling further Rev binding and the rapid formation of a viral export complex . Analysis of the Rev-RRE assembly pathway using SHAPE-Seq and small-angle X-ray scattering ( SAXS ) reveals two major steps of Rev-RRE complex formation , beginning with rapid Rev binding to a pre-organized region presenting multiple Rev binding sites . This step induces long-range remodeling of the RNA to expose a cryptic Rev binding site , enabling rapid assembly of additional Rev proteins into the RNA export complex . This kinetic pathway may help maintain the balance between viral replication and maturation .
Intron-containing RNA transcripts are usually retained in the nucleus until they are either spliced or degraded ( Luo and Reed , 1999; Bousquet-Antonelli et al . , 2000; Zhou et al . , 2000 ) . In contrast , during the late stage of HIV infection , unspliced viral transcripts must be exported to the cytoplasm either to be translated into structural proteins or to serve as genomic RNA packaged into new virions ( Cullen , 2003 ) . To ensure such nuclear export , HIV transcripts assemble into ribonucleoprotein ( RNP ) particles in which multiple copies of the viral Rev protein , a translation product of fully spliced HIV transcripts , bind to the Rev-response element ( RRE ) located within a viral RNA intron ( Malim et al . , 1989; Pollard and Malim , 1998; Cullen , 2003 ) . Upon forming a nuclear–export complex with Crm1 and RanGTP , the Rev-RRE RNP translocates across the nuclear pore into the cytoplasm ( Fornerod et al . , 1997; Pollard and Malim , 1998; Cullen , 2003; Yedavalli et al . , 2004 ) . Formation of the Rev-RRE RNP involves binding of multiple copies of Rev to viral transcripts . Rev can bind RNA as well as self-associate to form dimers and higher-order oligomers . It also contains a nuclear export sequence ( NES ) that can be recognized by Crm1 . Both the RNA binding and dimerization/multimerization properties of Rev play key roles in Rev-RRE complex assembly ( Mann et al . , 1994; Pond et al . , 2009; Daugherty et al . , 2010a , 2010b ) . Initial association of Rev with Stem IIB of the RRE ( Heaphy et al . , 1990 , 1991; Cook et al . , 1991; Huang et al . , 1991; Kjems et al . , 1991; Malim and Cullen , 1991 ) leads to additional Rev protein association with a secondary binding site near Stem IA ( Daugherty et al . , 2008 ) ( Figure 1A ) . A total of at least six copies of the Rev protein are thought to be necessary to form a functional RNP , although the exact number of Rev proteins in the mature complex remains a matter of debate ( Mann et al . , 1994; Daugherty et al . , 2008; Robertson-Anderson et al . , 2011 ) . 10 . 7554/eLife . 03656 . 003Figure 1 . HIV RRE RNA adopts a pre-organized compact fold . ( A ) SHAPE-based secondary structure of the RRE RNA . Red , orange , and blue dots highlight nucleotides with high , medium , and low SHAPE reactivity , respectively . Nucleotides with no SHAPE reactivity are in black . Nucleotides with SHAPE reactivity unidentified are in gray . The box shows the SHAPE handles in the RNA construct . The region of secondary structure rearrangement in our prediction is highlighted by gray shadow . The commonly used secondary structure is shown in the upper left with the rearranged region shadowed . The green line shows the ends of the 233-nt RRE construct ( Fang et al . , 2013 ) . The yellow curve indicates the position of nucleotides 54-58 . ( B ) SHAPE profiles of the RRE RNA alone ( black ) and the RNA-oligo complex ( red ) with AS 54-84 . The anti-sense oligo region is shown in cyan , and the regions of SHAPE reactivity change are shown in magenta . Those regions are labeled both under the data chart and on the secondary structure model . ( C ) SHAPE profiles of the RRE RNA alone ( black ) and the RNA-oligo complex ( blue ) with AS 100-113 . The anti-sense oligo region is shown in cyan , and the regions of SHAPE reactivity change are shown in magenta . Those regions are labeled both under the data chart and on the secondary structure model . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 00310 . 7554/eLife . 03656 . 004Figure 1—figure supplement 1 . Designed oligonucleotides can invade and hybridize to the RRE RNA at specific sites . ( A ) EMSAs showing AS 100-113 and AS 54-84 can form complexes with the RRE RNA , while shorter oligonucleotides targeting the Stem I region cannot bind to the RRE . ( B ) Toe-printing assays indicate that oligonucleotides only bind at designed locations . Peaks here indicate positions of reverse transcription dropoff . Oligonucleotides are color coded and the regions they are complementary to are indicated by lines of the same colors under the graph . Experiments with oligo 54-68 , 100-113 , and 318-330 are done using the same batch of RNA . The peak region in the black box at ∼300-nt ( Insert shows the expanded region within the box ) reflects a small amount of RNA degradation product . This is not a result of oligonucleotide binding at a secondary site as the same signals are observed regardless of input oligonucleotides and this pattern disappeared when a different batch of RNA is used as shown in blue . While by EMSAs complex between antisense oligo 54-68/318-330 and RRE were not detectable , small fractions of complex can be detected from toe-printing assays , indicating that the toe-printing experiments are highly sensitive for small fraction of oligonucleotide binding . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 00410 . 7554/eLife . 03656 . 005Figure 1—figure supplement 2 . SHAPE changes induced by oligonucleotides interactions . To make this plot , any SHAPE value >1 . 0 was set to 1 . 0 . This is because changes for any SHAPE reactivities beyond that is not relevant as those positions will be considered highly reactive in both cases . Change in a local area is considered relevant only if two or more consecutive nucleotides show significant SHAPE changes . ( A ) SHAPE changes induced by binding of AS 54-84 . ( B ) SHAPE changes induced by binding of AS 100-113 . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 005 Several lines of evidence suggest that Rev assembles into the Rev-RRE complex in a sequential manner ( Lam et al . , 1998; Van Ryk and Venkatesan , 1999; Daugherty et al . , 2008 , 2010a; Pond et al . , 2009; Robertson-Anderson et al . , 2011 ) . Some evidence suggests that Rev is recruited to the RRE one molecule at a time ( Pond et al . , 2009 ) , while other data are consistent with recruitment in the form of a dimeric complex ( Daugherty et al . , 2008 , 2010b ) . Studies of the Rev-RRE assembly pathway have focused on the state of the RNA and protein components either alone or in the fully assembled complex ( Cook et al . , 1991; Mann et al . , 1994; Battiste et al . , 1996; Charpentier et al . , 1997; Watts et al . , 2009; Daugherty et al . , 2010b ) , or on the dynamics of Rev protein within the complex ( Van Ryk and Venkatesan , 1999; Pond et al . , 2009 ) . In the absence of information about structural changes in the RRE that might occur during association with Rev , contradicting Rev-RRE assembly models were proposed ( Daugherty et al . , 2010b; Fang et al . , 2013 ) . To address this problem , we interrogated the folding pathway of the RRE RNA during the course of Rev-RRE complex formation , both as a function of time and Rev association state . We find that three regions within the RRE , bridged by previously unidentified RNA tertiary interactions , exhibit sequential structural changes during RNA-protein assembly . Comparison of the kinetic and thermodynamic pathways of Rev-RRE complex formation suggests a two-step assembly mechanism beginning with Rev binding to a pre-organized RRE structure critical for enhanced binding rate . This interaction triggers rearrangement of long-range RNA contacts to facilitate higher-order Rev-RRE assembly . This complex promotes a switch to late HIV protein expression and packaging .
To determine the structural features of an extended RRE transcript including sequences beyond the primary and secondary Rev binding sites , we performed selective 2′-hydroxyl acylation analyzed by primer extension ( SHAPE ) analysis ( Merino et al . , 2005 ) of the 354-nt RRE RNA of HIV-1 isolate ARV-2/SF2 ( Figure 1A ) . This RNA segment was chosen because it forms an independently folded region flanked by unstructured sequences within the HIV genome ( Watts et al . , 2009 ) . We determined its SHAPE profile and used this information for secondary structure prediction on the RNAstructure Web Server ( Low and Weeks , 2010; Reuter and Mathews , 2010 ) . The resulting SHAPE-based secondary structure model is largely consistent with current models for the RRE , including that obtained from structural probing of the entire HIV-1 genome ( Mann et al . , 1994; Charpentier et al . , 1997; Daugherty et al . , 2008 , 2010a; Pond et al . , 2009; Watts et al . , 2009; Fang et al . , 2013 ) . We note that the SHAPE data predict a rearrangement in the Stem III/IV and Stem V regions relative to previous models , yielding a modified secondary structure more similar to that of the RRE in SIV ( Pollom et al . , 2013 ) , which is used for mapping additional structural features and protein interactions detected in this study ( Figure 1A ) . Surprisingly , although predicted to be in a loop region , nucleotides 54-58 exhibit either low or no SHAPE reactivity , indicating that they could be constrained by tertiary contacts . To test whether this loop forms long-range interactions with another part of the RRE , we designed a 31-nt oligonucleotide complementary to nucleotides 54-84 ( AS 54-84 ) , which can efficiently invade and hybridize with the pre-folded RRE RNA to disrupt its local structure , and determined the SHAPE profile for the resulting RNA-oligo complex ( Sztuba-Solinska and Le Grice , 2012 ) . The length of this oligonucleotide was determined experimentally to be the minimum required for stable association with the RNA , presumably due to competing stability of the RRE secondary structure in this region ( Figure 1—figure supplement 1A ) . In addition to the Stem I region ( around nucleotides 50-80 and 300-340 ) , which was directly affected by the oligonucleotide binding , another segment covering nucleotides 100-113 showed significant changes in SHAPE reactivity , even though it is distant from the target region based solely on predicted secondary structure ( Figure 1B , Figure 1—figure supplement 2A ) . Results from toe-printing assays showed that AS 54-84 did not bind to a secondary site on the RRE ( Figure 1—figure supplement 1B ) , indicating the SHAPE change could be a result of long-range crosstalk . To confirm this , a 14-nt oligonucleotide complementary to nucleotides 100-113 ( AS 100-113 ) was used in a reciprocal experiment to perturb its potential long-range contacts . A shorter oligonucleotide was used here because this segment is predicted to be more accessible for oligonucleotide hybridization based on thermodynamic predictions . Toe-printing assays again showed no secondary oligonucleotide binding on the RRE ( Figure 1—figure supplement 1A , B ) . This resulting RNA-oligo complex showed substantially altered SHAPE reactivity in several patches along the Stem I region , including patches spanning around nucleotides 50 , 310 , and 340 ( Figure 1C , Figure 1—figure supplement 2B ) . Therefore , disruption of the tertiary structure in the Stem I region specifically affects the 100-113 segment and the converse is also observed . These results suggest the existence of a long-range structural contact between Stem I and nucleotides 100-113 . No sequence complementarity exists between these two regions , indicating that any tertiary interaction may be more complex than direct base-pairing . Based on observations above , we hypothesized that tertiary interactions bridging Stem I and the 100-113 region could make the overall fold of the 354-nt RRE relatively compact . This is consistent with results of a recent SAXS study , which revealed that a 233-nt RRE construct adopts an A shape with a maximum diameter ( Dmax ) of ∼195 Å ( Fang et al . , 2013 ) . Similarly , we found that the 354-nt RRE construct used here has a Dmax of 198 Å as detected by SAXS ( Figure 2A–D ) . This result suggests that the long Stem I in the RRE folds back towards the core of the multi-way junction instead of forming an extended tail ( Figure 2E ) . 10 . 7554/eLife . 03656 . 006Figure 2 . Model of the compact RRE RNA conformation . ( A ) RRE-oligo complexes show scattering patterns different from that of the RRE RNA alone . ( B ) Porod-Debye plot of RRE RNA and RNA-oligo complexes indicates the native RNA is more folded and the RNA-oligo complexes are more open and extended . ( C ) Distance distribution function ( P ( r ) ) of RRE RNA and RNA-oligo complexes . ( D ) Comparison of particle maximum diameter ( Dmax ) of RRE RNA and RNA-oligo complex . Data for 233-nt RNA is as published ( Fang et al . , 2013 ) . ( E ) Model for the compact fold of the RRE mediated by tertiary interactions between the 100-113 region and Stem I . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 00610 . 7554/eLife . 03656 . 007Figure 2—figure supplement 1 . Guinier plots of the SAXS data . Left , Guinier plot of the full-length RRE; middle , Guinier plot of the RRE with antisense oligonucleotide 54-84; Right , Guinier plot of the RRE with antisense oligonucleotide 100-113 . Guinier analysis were performed on the data within the q-range limited by q * Rg < 1 . 3 ( 11–12 data points for each sample ) , a few more data points are plotted to demonstrate a longer linear range . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 007 If the relatively compact conformation of the RRE results from the tertiary interactions identified earlier , disruption of those interactions is predicted to result in a more extended conformation . To test this , we used the antisense oligonucleotides described above ( AS 54-84 or AS 100-113 ) to interfere with long-range interactions within the RRE and performed SAXS measurements on the resulting complexes . As expected , the RNA-oligo complexes showed markedly altered scattering patterns ( Figure 2A , Figure 2—figure supplement 1; Supplementary file 1 ) and became more open and extended compared to the native RRE based on the Porod-Debye plot ( Rambo and Tainer , 2011; Figure 2B ) . The Dmax values for the resulting complexes also increased dramatically to 302 Å and 292 Å , respectively . It has been reported that a 30-nt single-stranded DNA oligonucleotide shows a Dmax value of ∼90 Å under similar Mg2+ concentration ( Meisburger et al . , 2013 ) , indicating that even if free oligonucleotide remained in the system after extensive washes , it would not bias the distance distribution functions ( P ( r ) ) to a longer range . Therefore , the large variations in the P ( r ) indicate a substantial conformational change in the RRE ( Figure 2C , D ) . Taken together , these results demonstrate that the Stem I and 100-113 regions are linked through long-range interactions , which enables the RRE to fold into a compact structure ( Figure 2E ) . We speculate that the sharp bend in the RNA ( Figure 2E ) results from a series of bends that occur throughout the noncanonical parts of Stem I . To determine how the pre-formed RRE structure affects RNP assembly , we used SHAPE to locate all Rev binding-induced structural changes on the RRE RNA and associate them with individual Rev binding events . Experiments were conducted on equilibrated complexes formed using different Rev:RRE ratios ( Figure 3A ) . In parallel , the percentage of different sub-complexes within each binding reaction was quantified by electrophoretic mobility shift assays ( EMSAs ) using aliquots from the same samples ( Figure 3B , Figure 3—figure supplement 1A ) . 10 . 7554/eLife . 03656 . 008Figure 3 . Thermodynamic studies on the Rev-RRE assembly pathway . ( A ) SHAPE profiles from samples at different Rev:RRE ratio . The positions showing increase/decrease of SHAPE reactivity are indicated with arrows at the bottom of the plots . ( B ) Trend of emergence of each Rev-RRE sub-complex at increasing ratios of Rev . ( C ) k-means clustering result of nucleotides following distinct SHAPE signatures with the SHAPE signatures shown in cyan boxes . 5 of the 10 SHAPE signatures ( Signatures* ) represent increased SHAPE reactivity as a function of Rev concentration , while the other five ( Signatures ) reflect decreased SHAPE reactivity . They fall into seven clusters because some signatures are not distinct enough from others . Red indicates higher SHAPE reactivity while black indicates lower SHAPE reactivity . ( D ) Nucleotide positions affected by Rev binding . In red , orange and yellow shows positions with increased SHAPE reactivity upon Rev binding . In dark , medium and light blue shows positions with decreased SHAPE reactivity upon Rev binding . Changes emerging at lower Rev:RRE stoichiometry are shown in darker colors , while changes emerging at higher Rev:RRE stoichiometry are shown in lighter colors . Font size reflects the amplitude of SHAPE change . Region1 , Region2 and Region3 are highlighted by gray shadows . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 00810 . 7554/eLife . 03656 . 009Figure 3—figure supplement 1 . SHAPE signatures generated by EMSA . ( A ) EMSAs for quantification of different Rev-RRE intermediate states . ( B ) SHAPE signatures calculated based on EMSAs results . Left , SHAPE signatures representing nucleotides with increased SHAPE reactivity upon Rev binding . Right , SHAPE signatures representing nucleotides with decreased SHAPE reactivity upon Rev binding . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 009 From the EMSA data , we detected the formation of different Rev bound sub-complexes as a function of Rev:RRE stoichiometry ( Figure 3B ) , and used that information to deduce distinct SHAPE modification signatures that represent progressive Rev binding states ( Figure 3—figure supplement 1B; ‘Materials and methods’ ) . We then performed k-means clustering ( Eisen et al . , 1998 ) on nucleotides with significant SHAPE reactivity changes ( ΔSHAPE > 0 . 15 ) ( ‘Materials and methods’ ) to group them together according to SHAPE signatures . All SHAPE signatures fall into seven clusters , each containing a group of nucleotides sharing a common pattern of SHAPE reactivity changes as a function of Rev concentration ( Figure 3C ) . Based on the SHAPE signatures and the Rev bound sub-complexes they represent , nucleotides within the seven clusters are associated with RNA structural changes triggered by Rev binding at low , intermediate and high Rev:RRE stoichiometries and those regions are named Region1 , Region2 and Region3 , respectively ( Figure 3D ) . Region1 covers the primary , high-affinity Rev binding site reported previously ( Heaphy et al . , 1990 , 1991; Cook et al . , 1991; Huang et al . , 1991; Kjems et al . , 1991; Malim and Cullen , 1991 ) . It also includes the three-way junction of Stems IIA , IIB and IIC . According to earlier reports , Region1 is likely recognized by a Rev dimer ( Zemmel et al . , 1996 ) , with one molecule binding in the widened RNA major groove at the primary binding site ( Battiste et al . , 1996 ) and the other binding at the three-way junction ( Zemmel et al . , 1996; Van Ryk and Venkatesan , 1999; Daugherty et al . , 2008 ) ( Jayaraman et al . , unpublished manuscript ) . Region2 covers the previously identified secondary Rev binding site ( Daugherty et al . , 2008 ) . By comparing the footprint of this region with that of Region1 together with a previous truncation study ( Van Ryk and Venkatesan , 1999 ) , we infer that Region2 could accommodate at least a Rev dimer . Region3 , a previously undefined Rev binding site , is located in the center of Stem I ( Figure 3D ) . It comprises an array of purine-rich bulges , which resemble the preferred RNA site for Rev binding ( Heaphy et al . , 1991; Tan et al . , 1993; Battiste et al . , 1996 ) . In fact , Region3 assignment is consistent with previous reports showing that positions on Stem I can be protected by Rev oligomers , and truncations in this area affect oligomeric Rev binding ( Mann et al . , 1994; Robertson-Anderson et al . , 2011 ) . Interestingly , Region3 also overlaps with the area on Stem I that could form long-range interactions with the nucleotides 100-113 within Region2 ( Figure 3D , Figure 1B , C ) . Although Region2 and Region3 are distant from each other based on secondary structure , our observations indicate that tertiary interactions could bring them into close proximity , which might facilitate Rev binding at those regions while forming a continuous oligomer at the same time ( Figure 2E ) . To follow the dynamics of Rev-RRE RNP formation , we performed time-resolved SHAPE to examine how the RRE RNA changes during the course of RNP assembly ( Figure 4A ) . To obtain snapshots of the RNA on a timescale of seconds , a fast-reacting SHAPE reagent , benzoyl cyanide , was used in these experiments ( Mortimer and Weeks , 2008 , 2009 ) . Based on a previous single-molecule study of the Rev-RRE system , the timescale for each Rev binding step is ∼2–5 s ( Pond et al . , 2009 ) . This indicates that under our experimental conditions , intermediate stages of RNP assembly should be detectable . To increase the precision of the measurements , we adapted the SHAPE-Seq method ( Lucks et al . , 2011 ) to determine SHAPE reactivities and rates of SHAPE-monitored structural changes at different positions on the RRE . The accuracy of this method depends on the number of reads used for calculating each SHAPE profile . In this study , the amount of data for one replicate at each time point range from 256 , 424 to 1 , 145 , 824 reads . With this amount of data , the reactivity values ( theta , the modification probability [Lucks et al . , 2011] ) are highly reproducible with Pearson correlation coefficients ranging from 0 . 967 to 0 . 974 between any two of the three replicates ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 03656 . 010Figure 4 . Dynamic assembly of the Rev-RRE RNP . ( A ) SHAPE profiles for second-resolution snapshots of the RRE at different time points over the course of Rev-RRE assembly . The positions showing increase or decrease of SHAPE reactivity are indicated with arrows at the bottom of the plots . For ease of comparison , the theta values are converted to SHAPE reactivity values by renormalization following the 2%/8% rule for each SHAPE profile , with the top 2% of theta values being excluded and the next 8% of theta values being averaged to get the normalization factor , against which all the theta values are normalized ( Low and Weeks , 2010; Lucks et al . , 2011 ) . ( B ) Nucleotide positions affected by Rev binding . Different nucleotides are labeled across the spectrum based on the rate of SHAPE change , with the fastest rate in red and the slowest in purple . Font size reflects the amplitude of SHAPE change . Region1 , Region2 and Region3 are highlighted by gray shadows . ( C ) Heat map showing the SHAPE pattern as a function of time for the nucleotides with complex SHAPE-changing features . For each nucleotide in this panel , the SHAPE values are normalized to 0–1 in order to emphasize the trend of SHAPE change . ( D ) Scatterplot ( left ) and histogram ( right ) showing the distribution of the SHAPE change half-life for nucleotides showing significant SHAPE change . Colors used in the scatter plot correspond to those used in panel B . Both plots show a fast-SHAPE-changing group and a slow-SHAPE-changing group . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 01010 . 7554/eLife . 03656 . 011Figure 4—figure supplement 1 . Quality and reproducibility test for the SHAPE-Seq data . ( A ) Quality statistics for SHAPE-Seq reads used for analysis . Index1 and index2 are two barcodes introduced during the PCR amplification step . R1 and R2 represent the paired-end reads from both directions . ( B ) Correlation between the reactivity ( theta ) in any two of the three time-series datasets used for SHAPE change rate determination . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 01110 . 7554/eLife . 03656 . 012Figure 4—figure supplement 2 . Rev-RRE assembly process exhibits two-step features . ( A ) Dendrogram of the rates for SHAPE-reactivity changes at different nucleotides . All the fast-reacting and slow-reacting nucleotides are colored in the same way as in Figure 4 , while intermediate reacting nucleotides are colored in black . ( B ) Second-phase SHAPE changes for positions with complex SHAPE-changing patterns . All these positions are located in Region3 . Data from the first 10 s were excluded from the fitting . As a reference to show the slow decrease of SHAPE reactivity , all data is fitted to one-phase decay with automatic outlier elimination using GraphPad Prism 5 . 0b for Mac OS X . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 012 Using this method of analysis , distinct dynamic features were observed at different regions on the RRE during RNP assembly as a function of time ( Figure 4A , B ) . At most nucleotide positions , the SHAPE-reactivity change patterns follow single exponential decay . However , more complex SHAPE kinetic patterns are observed for several other nucleotides , most of which show an increase of SHAPE reactivity at earlier time points followed by a gradual decrease ( Figure 4C ) . Next , we calculated the rate for all positions showing a significant SHAPE change ( ΔSHAPE > 0 . 15 ) by fitting the time-resolved data using exponential decay/association kinetics . Only the earlier time points for nucleotides with complex kinetic behavior were used for fitting in our initial comparison ( Supplementary file 2 ) . During Rev-RRE assembly , structural changes of the RRE originate from the primary Rev binding site and subsequently propagate along the RNA . The pattern is similar to that observed based on the thermodynamic experiments described above ( Figure 3D , Figure 4B ) . Even though the accuracy of the fastest SHAPE reactivity change rates is limited by the time resolution of the experiment , the overall trend of sequential SHAPE changes is evident ( Figure 4B ) . Upon encountering Rev , Region1 exhibits SHAPE reactivity changes consistent with previous structural data for complexes of Rev and RNA fragments covering this region ( Battiste et al . , 1996 ) ( Jayaraman et al . , unpublished manuscript ) . At this time resolution , the two Rev binding events that occur in this region , one at Stem IIB and one at the three-way junction at Stem IIA , IIB and IIC , are indistinguishable . And the majority of SHAPE reactivity changes at Region1 occur within 1 s ( Figure 4D; Supplementary file 2 ) . The overall rates of SHAPE reactivity changes at Region2 are only marginally slower than those observed at Region1 , and the half-life for the majority of changes in Region1 and Region2 are clustered together below 1 . 5 s ( Figure 4D; Supplementary file 2 ) . Since the RRE RNA at this local region is pre-organized into a compact fold ( Fang et al . , 2013; Figure 2E ) , its conformation can facilitate Rev multimerization from Region1 to Region2 with little RNA rearrangement necessary , leading to rapid and highly cooperative binding of multiple Rev proteins . Based on previous reports , binding of the third Rev molecule ( the second Rev dimer ) and beyond requires higher oligomerization capability of the protein ( Daugherty et al . , 2008 ) . Therefore , Rev self-association should also be a fast process that can be completed within the same time period . In contrast , Region3 shows by far the slowest rate of folding , with the initial stage of most SHAPE changes in this region showing half-lives of 2–12 s ( Figure 4D; Supplementary file 2 ) . All nucleotides with complex kinetic behaviors are located in this region , and the second phase of their SHAPE reactivity change continues beyond 50 s at many positions ( Figure 4A , Figure 4—figure supplement 2B ) . Combined with data revealing the tertiary fold of the RRE , for the nucleotides with complex kinetic behaviors , the increase in SHAPE reactivities at earlier time points is consistent with rearrangement of the RRE tertiary structure at Region3 . This could occur after Rev disrupts the long-range contact within the RRE by binding to Region2; the subsequent decrease in SHAPE reactivity could represent additional Rev binding in this region ( Figure 4C ) . The slower rates of structural change in Region3 could be due to RNA conformational changes , or adjustment of Rev conformation as influenced by surrounding RNA and proteins to allow additional Rev oligomerization , or a combination of both effects . Nonetheless , characteristics exhibited by this stage of assembly best resemble an induced-fit model of RNA-protein recognition . Due to the slow assembly of the higher-order Rev-RRE complex , the full RNP formation could take minutes ( Figure 4—figure supplement 2B ) . The latter phase of SHAPE reactivity changes at those positions is much noisier ( Figure 4—figure supplement 2B ) . This could reflect a combination of ‘conformational selection’ and ‘induced-fit’ events , leading to a series of specific and non-specific contacts that facilitate finding the optimal binding configuration , similar to observations for other RNP assembly pathways ( Bokinsky et al . , 2006; Boehr et al . , 2009; Rau et al . , 2012; Kim et al . , 2014 ) . Previously , it has been reported that in the absence of Region3 , a truncated ∼240-nt RRE can mediate HIV RNA nuclear export but with lower efficiency ( Malim et al . , 1989; Huang et al . , 1991 ) , indicating a facilitating role of the sequences outside of the ∼240-nt RRE . These observations are consistent with our results showing that Region3 on the extended Stem I ( Figure 1A ) is a preferred Rev binding location in the context of the full-length RRE . Moreover , sequences on the extended Stem I also mediate tertiary interactions within the RRE . We then asked besides providing additional Rev binding site , whether the extended Stem I region contribute to Rev-RRE complex assembly , especially at the early stage , due to the tertiary interactions it mediates . SHAPE snapshots were taken at 0 , 1 , 6 , 11 , and 900 s of the complex assembly on three mutants of the extended Stem I . One of them is a 242-nt truncation mutant and the other two are site mutants in Region3 alone ( Region3 mut1 ) and Region3/Stem I ( mut2 ) , which reduce the flexibility of Stem I by introducing additional base-pairing ( Figure 5A ) . SHAPE reactivity changes at key nucleotides in Region1 and Region2 were compared between different constructs and the 354-nt RRE . For the ease of comparison , the SHAPE reactivity values were normalized so that zero represents the SHAPE state for unbound RRE at 0 s and 1 represents the SHAPE state of the RRE in the fully assembled complex at 900 s . At 6 and 11 s , all RNA constructs behave similarly , and the normalized SHAPE changes indicate that Rev binding at Region1 and Region2 are mostly completed at those time points . In contrast , at 1 s the 354-nt RRE shows overall higher normalized SHAPE changes than all three mutants , indicating that Rev binding rate is enhanced in the context of the full-length RRE with intact Stem I region ( Figure 5B , C ) . This enhancement is observed at early Rev binding sites , underlining the importance of the RNA tertiary folding . These results reveal that the extended Stem I of the RRE , which covers a cryptic Rev binding site , can enhance the rate of the Rev-RRE complex assembly , likely by pre-organizing the RRE into a favorable conformation for Rev binding . Assembly of this RNP governs the balance for nuclear export of different types of HIV transcripts , which is essential for promoting HIV infectious cycle . Offset in the assembly rate of the Rev-RRE complex could break the optimal balance of cytoplasmic HIV transcripts , which not only affect the distribution of HIV transcripts but could also make a greater difference in the encapsidation of genomic RNA into infectious viral particles ( Brandt et al . , 2007 ) . 10 . 7554/eLife . 03656 . 013Figure 5 . Functional importance of the Region3 . ( A ) Table showing the mutations used in this figure . ( B ) Normalized SHAPE change at 1 , 6 and 11 s after Rev binding . The ∼354-nt RRE is shown in black , Region3 mut1 is shown in blue , mut2 is shown in red and the 242-nt RRE is shown in green . ( C ) Table indicating the significance of the differences in normalized SHAPE changes between different constructs at 1 s . All mutants are compared to the 354-nt WT RRE . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 013 Together , these results fill in a long missing piece of this RNP assembly puzzle: how the Rev-response element structure responses to the addition of Rev in a dynamic manner . Our data suggest a concerted Rev-RRE complex assembly mechanism and indicate how specificity can be achieved here with limited components . We propose that Rev-RRE assembly features two distinct stages ( Figure 4D , Figure 4—figure supplement 2A ) . The first stage largely utilizes pre-organized RNA structure for protein recruitment , while the second stage involves more global RNA conformational changes and induced-fit RNA-protein recognition ( Figure 4D , Figure 6 ) . The similarity in the thermodynamic and kinetic RNP assembly pathways indicates that this process is hierarchical , with the RRE RNA driving RNP assembly by organizing sequential Rev binding ( Figure 6 ) . The highest affinity Rev-binding site , Stem IIB , ensures that complex assembly nucleates from a single origin on the RNA . Binding of up to four Rev proteins at Region1 and Region2 is tightly coupled to obtain a stable intermediate state ( Zemmel et al . , 1996; Van Ryk and Venkatesan , 1999; Daugherty et al . , 2008; Pond et al . , 2009 ) . This step also releases a cryptic Rev binding site at Region3 , which can make further contacts with additional Rev molecules . The tertiary folding of the RRE facilitates the efficiency of Rev-RRE complex formation . In addition , the hierarchical nature of the assembly could ensure its selectivity and accuracy of the final complex with limited number of components . These features together provide a fine control over cytoplasmic distribution of various HIV transcripts as well as viral packaging . Intriguingly , these steps in Rev-RRE assembly resemble those that occur during bacterial ribosome assembly in which stably formed rRNA structures recruit initial ribosomal protein binding partners . The resulting local structures trigger long-range RNA conformational rearrangements that enable binding of additional ribosomal proteins ( Adilakshmi et al . , 2008; Shajani et al . , 2011; Kim et al . , 2014 ) . Such similarities suggest that common mechanisms for RNP assembly could be shared among diverse biological processes . Properties of the HIV RNP assembly pathway elucidated here present opportunities for antiviral strategies that could block the nuclear export step of HIV replication by targeting important intermediates of the Rev-RRE complex ( Fenster et al . , 1994; Chapman et al . , 2002; Shuck-Lee et al . , 2008; Ward et al . , 2009 ) . 10 . 7554/eLife . 03656 . 014Figure 6 . Model for pre-organized RRE RNA guides sequential binding of Rev to form the Rev-RRE RNP . RRE RNA forms a compact fold in the absence of Rev . Rev assembly on the RRE starts from a single nucleation point . Region1 and Region2 binding are coupled and the four-Rev complex state can serve as a checkpoint to ensure specificity . Region2 Rev binding leads to conformational change of the RRE to allow additional Rev binding through induced-fit . Both Rev and RRE could sample a number of interaction conformations at the same time until an optimal binding state is reached . The high-oligomer complex is then ready for Crm1 binding and nuclear export . DOI: http://dx . doi . org/10 . 7554/eLife . 03656 . 014
The RRE RNA construct used here contains the 354-nt full-length RRE from isolate ARV-2/SF2 with SHAPE handles at both ends as previously reported ( Berry et al . , 2011 ) . RNA samples from all constructs were prepared by in vitro transcription using T7 polymerase , with either linearized plasmid or PCR product DNA as templates . For SHAPE analysis RNA samples were column-purified using RNA Clean & Concentrator-25 ( Zymo Research , Irvine , CA , USA ) , and for SAXS experiments RNA samples were gel-purified and washed multiple times through filtration . Purified RNA samples were annealed in a buffer containing 50 mM HEPES-KOH pH 7 . 5 , 200 mM KOAc , and 3 mM MgCl2 by heating at 75°C for 2 min and snap cooling on ice . For SHAPE-Seq experiments , barcodes on the RRE molecules were introduced by PCR and placed within the 3′ SHAPE handle as previously described ( Lucks et al . , 2011; Mortimer et al . , 2012 ) . Both the His-GB1-Rev fusion protein construct and its purification procedure were reported previously ( Daugherty et al . , 2008 , 2010a ) . SHAPE probing was performed as previously reported ( Bai et al . , 2013 ) with 40 mM of benzoyl cyanide ( BzCN ) ( Sigma-Aldrich , St . Louis , MO , USA ) used as the 2′ hydroxyl-selective electrophile . To make RRE-oligonucleotide complexes , RRE RNA was first annealed at 0 . 1 mg/ml followed by a 5 min incubation at room temperature . A large excess of oligonucleotides complementary to either nucleotides 54-84 ( AS 54-84 ) or nucleotides 100-113 ( AS 100-113 ) was added to pre-folded RRE and the mixture was incubated for 20 min at room temperature to allow the oligonucleotides to form a complex with the RNA . For SAXS measurements , the RRE-oligonucleotide complexes were washed extensively by ultrafiltration . To make Rev-RRE complex at different stoichiometries , GB1-Rev fusion protein was first diluted to various concentrations ( 0 . 25–4 mg/ml ) in Rev buffer containing 40 mM Tris pH 8 . 0 , 200 mM NaCl , 100 mM Na2SO4 , 400 mM ( NH4 ) 2SO4 , 2 mM β-ME , and 10% glycerol . 30 μl of annealed RRE RNA at 0 . 1 mg/ml was then mixed with 3 μl of either Rev protein solution or Rev buffer alone . The resulting mixtures were incubated for 20 min at room temperature . 10 μl of each resulting solution was analyzed by EMSAs , while the proceeding SHAPE protocol was used on the remaining sample . SHAPE-Seq experiments were performed as described ( Mortimer et al . , 2012 ) with slight modification of the primer design , as listed below . 10 parts of annealed RRE RNA at 0 . 1 mg/ml was mixed with one part of 4 mg/ml Rev protein and SHAPE reaction was performed at different time points . For the ∼1 s time point , the SHAPE reagent was introduced at the same time with the Rev protein by centrifugation and the ∼1 s time is estimated based on the half-life of BzCN ( Mortimer and Weeks , 2008 ) . The RRE RNA samples used in this assay are barcoded within the SHAPE handle as described ( Lucks et al . , 2011 ) and additional barcodes were introduced during the PCR amplification step of library preparation using the NEBNext Multiplex Oligos for Illumina kit ( NEB , Ipswich , MA , USA ) to increase the multiplex capacity . RT_index1:AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTNNNNNNRRRYGAACCGGACCGAAGCCCGRT_index2:AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTNNNNNNYYYRGAACCGGACCGAAGCCCGA_adapter_b_short:5′ Phos-ATGCNNNNNNNNAGATCGGAAGAGCACACGTCTGAACTCCAGTCAC-C3Paired_end_reverse:CAAGCAGAAGACGGCATACGAGATGTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT For capillary electrophoresis based experiments , raw traces from fragment analysis were analyzed using ShapeFinder ( Vasa et al . , 2008 ) . For the RRE-oligonucleotide complexes , samples treated with only DMSO but not SHAPE reagent were also analyzed as toe-printing assays to determine the binding sites of the oligonucleotides . For sequencing based experiments , raw data was processed using FASTX-Toolkit ( http://hannonlab . cshl . edu/fastx_toolkit/ ) . A quality filter was applied so that only reads with a minimum of 95% bases having a quality score of over 30 were retained for data analysis . Paired-end reads were aligned to the RRE RNA using Bowtie 0 . 12 . 7 ( Langmead et al . , 2009 ) and SHAPE reactivity was determined using Spats ( Aviran et al . , 2011b; Lucks et al . , 2011; Aviran et al . , 2011a ) . Reads with identical sequences were not collapsed during our data process due to file format compatibility consideration . However , primer ID ( Jabara et al . , 2011 ) was included in the primers to control for PCR amplification bias . Only <3% of the reads are from PCR duplication and most of them are aligned to adapter dimers . Therefore , there is no significant PCR bias in our experiment . SHAPE-based secondary structure of the RRE RNA was calculated using RNAstructure Web Servers ( Reuter and Mathews , 2010 ) with the SHAPE reactivity file obtained for the free RRE RNA ( Low and Weeks , 2010 ) . The SHAPE Intercept and SHAPE Slope used for this prediction were −0 . 6 and 1 . 8 , respectively . SHAPE signatures for each Rev binding event were generated based on EMSA results . First , the trend of emergence for each of the RRE or Rev-RRE complex was derived directly from the quantification of the EMSA data . With the assumption that a SHAPE change associated with an earlier Rev binding event remains in all subsequent complexes , we next calculated a series of SHAPE signatures , which reflect the emergence of RNP with one Rev and above , two Rev and above , etc . This series of curves were used to represent positions showing increased SHAPE reactivity upon Rev binding . Subsequently , another series of curves were generated reflecting the disappearance of species with less than one Rev , less than two Rev , etc . These series of curves were used to represent positions showing decreased SHAPE reactivity . k-means clustering of SHAPE profile at different nucleotide positions was performed using Cluster 3 . 0 ( Eisen et al . , 1998; de Hoon et al . , 2004 ) as previously reported ( Grohman et al . , 2013 ) with slight modification to also include that SHAPE signatures as guidance points . Results were visualized using Java TreeView ( Saldanha , 2004 ) . Here ΔSHAPE > 0 . 15 is considered as significant SHAPE change . However , cutoffs of ΔSHAPE > 0 . 2 and ΔSHAPE > 0 . 1 gave qualitative similar result after data analysis . SHAPE-reactivity changing rates for nucleotides showing significant SHAPE change ( ΔSHAPE > 0 . 15 ) from time-resolved SHAPE-Seq experiments were determined by fitting the data to either one-phase association or one-phase decay function with automatic outlier elimination using GraphPad Prism 5 . 0b for Mac OS X . SAXS data were collected at the Advanced Light Source ( Lawrence Berkeley National Laboratory ) beamline 12 . 3 . 1 ( Hura et al . , 2009; Classen et al . , 2013 ) . Two-dimensional scattering curves obtained at different exposure time were merged and processed using PRIMUS ( Konarev et al . , 2003 ) and distance distribution functions P ( r ) , radius of gyration ( Rg ) and Porod volume were generated using GNOM ( Svergun , 1992 ) . Porod-Debye plot was calculated based on raw data using GraphPad Prism 5 . 0b for Mac OS X . | HIV is a virus that causes the immune system of an infected person to gradually fail , which can eventually result in AIDS . The virus consists of an RNA molecule—which encodes its genetic information—surrounded by coats of proteins . Once HIV enters a host cell , its RNA genome is converted into a DNA molecule , which travels to the nucleus and becomes part of the host's genome . The integrated viral genome can remain dormant for an extended period before the virus starts to replicate . HIV replication begins with the production of RNA copies of the viral genome . For certain types of viral RNA molecules to be translated and packaged into new virus particles they need to be exported from the nucleus as part of the ‘nuclear–export complex’ . This is made up of: a HIV RNA molecule , a HIV protein called Rev , and two host proteins . Formation of the nuclear–export complex begins with multiple copies of the Rev protein attaching to specific stretches of the viral RNA , but how the Rev proteins assemble on the RNA molecule was previously unclear . Bai et al . have now used both structural and biochemical techniques to dissect the individual steps in this process . First , Rev proteins rapidly bind to a pre-formed region of the RNA molecule where multiple binding sites are compactly organized . This causes the overall shape of the RNA to change , and exposes a previously hidden extra binding site for Rev proteins . More Rev proteins then quickly bind to the newly exposed site , before finally the two host proteins bind and the whole complex is exported from the nucleus . Bai et al . propose that checkpoints during this two-step assembly process are required to ensure that Rev proteins specifically bind to viral RNAs , and that such checkpoints may be important for controlling viral replication . The findings of Bai et al . may , in future , help to develop new drugs that treat HIV infection by blocking the export of the virus from the nucleus and thus inhibiting HIV replication . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | RNA-guided assembly of Rev-RRE nuclear export complexes |
The mammalian circadian clock drives daily oscillations in physiology and behavior through an autoregulatory transcription feedback loop present in central and peripheral cells . Ablation of the core clock within the endocrine pancreas of adult animals impairs the transcription and splicing of genes involved in hormone exocytosis and causes hypoinsulinemic diabetes . Here , we developed a genetically sensitized small-molecule screen to identify druggable proteins and mechanistic pathways involved in circadian β-cell failure . Our approach was to generate β-cells expressing a nanoluciferase reporter within the proinsulin polypeptide to screen 2640 pharmacologically active compounds and identify insulinotropic molecules that bypass the secretory defect in CRISPR-Cas9-targeted clock mutant β-cells . We validated hit compounds in primary mouse islets and identified known modulators of ligand-gated ion channels and G-protein-coupled receptors , including the antihelmintic ivermectin . Single-cell electrophysiology in circadian mutant mouse and human cadaveric islets revealed ivermectin as a glucose-dependent secretagogue . Genetic , genomic , and pharmacological analyses established the P2Y1 receptor as a clock-controlled mediator of the insulinotropic activity of ivermectin . These findings identify the P2Y1 purinergic receptor as a diabetes target based upon a genetically sensitized phenotypic screen .
Type 2 diabetes is an escalating epidemic involving gene-environment interactions that culminate in β-cell failure and insulin resistance . Recent epidemiological evidence has shown that shift work and sleep disturbance are environmental risk factors for diabetes ( Perelis et al . , 2016 ) , while experimental genetic studies have shown that clock gene disruption within the endocrine pancreas causes hypoinsulinemic diabetes ( Marcheva et al . , 2010; Sadacca et al . , 2011 ) . At the molecular level , the circadian clock is composed of an autoregulatory transcriptional loop in which CLOCK/BMAL1 activate the repressors PER1/2/3 and CRY1/2 , which feedback to inhibit CLOCK/BMAL1 in a cycle that repeats itself every 24 hr . An additional stabilizing loop involving ROR/REV-ERB regulates BMAL1 expression ( Kim and Lazar , 2020 ) . Recent chemical screens have identified new factors that modulate the core clock , including casein kinase 1 inhibitors that lengthen the circadian period through stabilizing PER proteins ( Hirota et al . , 2010; Chen et al . , 2012 ) , and a separate series of cryptochrome stabilizer compounds have been discovered that control glucose homeostasis in vivo ( Hirota et al . , 2012 ) . Modulators of clock transcription factors may also control whole animal metabolism ( He et al . , 2016 ) , though such compounds lack specificity ( Dierickx et al . , 2019 ) . Here , we developed a high-throughput small-molecule screen to identify insulinotropic compounds that act downstream of the circadian clock rather than through modulation of the core clock itself . We reasoned that compounds that enhance insulin secretion in the setting of β-cell clock disruption might in turn uncover therapeutic targets for more common forms of diabetes mellitus ( Marcheva et al . , 2010; Perelis et al . , 2015; Marcheva et al . , 2020; Moffat et al . , 2017 ) . To do so , we generated β-cells harboring a circadian gene mutation by CRISPR-Cas9 and co-expressing a luminescent insulin reporter that has previously been used to identify factors that either activated or repressed glucose-stimulated insulin secretion ( GSIS ) in wild-type β-cell lines ( Burns et al . , 2015 ) . In our screen of 2640 drug or drug-like compounds in circadian mutant β-cells , we identified the macrolide ivermectin ( IVM ) as an insulinotropic compound that activates the P2Y1 purinergic receptor . We further identified the P2Y1 receptor as a direct transcriptional target of the molecular clock factor BMAL1 and a potent regulator of glucose-dependent calcium signaling . Our findings establish a chemical genetic strategy to identify novel endocrine cell therapeutics .
Based upon our finding that circadian genes regulate β-cell function , we developed a chemical genetic screen to identify pathways that enhance glucose-coupled insulin secretion in a cell-based model of circadian β-cell failure ( Figure 1A ) . We previously showed that clonal Bmal1-/- Beta-TC-6 β-cell lines recapitulate the secretory defects observed in primary clock-deficient islets ( Marcheva et al . , 2010; Perelis et al . , 2015; Marcheva et al . , 2020 ) . We next generated stable WT and Bmal1-/- β-cell lines with a luciferase readout for insulin secretion using an insulin-NanoLuciferase ( NanoLuc ) -expressing lentivirus ( Figure 1B ) . We validated the direct correspondence between insulin-NanoLuc bioluminescence and levels of peptide secretion under increasing physiological concentrations of glucose ( 2–20 mM; R2 = 0 . 8937; Figure 1C ) . We further confirmed impaired insulin secretion by reduced bioluminescence in Bmal1-/- compared to WT β-cell lines expressing insulin-NanoLuc in response to stimulatory concentrations of glucose ( 20 mM ) , potassium chloride , forskolin , and the phosphodiesterase inhibitor 3-isobutyl-1-methylxanthine ( IBMX ) ( Figure 1D ) . We also validated the use of the DAG mimetic phorbol 12-myristate 13-acetate ( PMA ) as a positive control for the screen ( Figure 1D–F; Perelis et al . , 2015 ) . A feasibility test with a Z'-factor score of 0 . 69 indicated a significant separation between the distribution of bioluminescent signal from the positive ( 10 μM PMA + 20 mM glucose ) and negative ( 20 mM glucose ) controls , suggesting that the assay provides a suitable platform for a high-throughput screen ( Figure 1F; Zhang et al . , 1999 ) . We next used insulin-NanoLuc-expressing Bmal1-/- β-cell lines to screen 2640 drugs and drug-like molecules from the Spectrum Collection ( MicroSource Discovery Systems , Inc , New Milford , CT ) to identify compounds that enhance insulin secretion ( Figure 1E ) . Insulin-NanoLuc-expressing Bmal1-/- Beta-TC-6 cells were plated at 40 , 000 cells/well in a total of nine 384-well plates , incubated for 3 days , and then treated for 1 hr with either ( i ) 20 mM glucose alone ( negative control that elicits reduced insulin secretion in Bmal1-/- cells ) , ( ii ) 20 mM glucose plus 10 μM of one of the 2640 compounds , or ( iii ) 20 mM glucose plus 10 μM PMA ( positive control known to enhance insulin secretion in both Bmal1-/- mouse islets and Beta-TC-6 cells ) ( Perelis et al . , 2015 ) . Luciferase intensity from the supernatant was measured following exposure to NanoGlo Luciferase Assay Substrate ( Figure 1E ) . We initially identified 19 hit compounds that both significantly enhanced insulin secretion and elicited a response of greater than 3 standard deviations from the mean ( Z-score > 3 ) with more than a 1 . 25-fold increase , exceeding the upper 99% confidence interval of the negative control ( Figure 2A , Figure 2—figure supplement 1A , Supplementary file 1 ) . Of these , seven were excluded from further analysis because of reported toxic effects or lack of availability of the compound ( Figure 2—figure supplement 1A ) . The remaining 12 hit compounds mediate activity of ligand-gated cell surface receptors and ion channels that stimulate second messenger signaling cascades ( Figure 2B and C; Gaulton et al . , 2010; Carrano et al . , 2017 ) . Of these , four target ion channels ( tacrine hydrochloride , suloctidil , dyclonine hydrochloride , and IVM ) ( Figure 2B and C; Karlsson and Ahrén , 1992; Chatelain et al . , 1984; Khanna et al . , 2011; Chen and Kubo , 2018; Freeman et al . , 1988; de Gaetano et al . , 1976; Kornhuber et al . , 2008; Sahdeo et al . , 2014; Roghani et al . , 1999; Ikeda , 2003 ) . Five target seven-transmembrane G-protein coupled receptors ( GPCRs ) that signal through phospholipase C ( PLC ) and diacylglycerol ( DAG ) to activate insulin secretion and β-cell gene transcription ( benzalkonium chloride , carbachol , isoetharine mesylate , pipamperone , and IVM ) ( Figure 2B and C; Chen and Kubo , 2018; Higashijima et al . , 1990; Rinne et al . , 2015; Bierman , 1983; Van Craenenbroeck et al . , 2006; Nagata et al . , 2019; Ratajewski et al . , 2015; Ohtani et al . , 2011 ) . Similar to the hit compounds of our screen , our previous results showed that carbachol , a muscarinic Gq-coupled receptor agonist , and the DAG mimetic PMA rescue insulin secretion in Bmal1-/- islets ( Perelis et al . , 2015 ) . Four additional hit compounds act as acetylcholinesterase inhibitors , promoting enhanced glucose-dependent insulin secretion in response to acetylcholine through the muscarinic GPCRs , as well as the ionotropic nicotinic acetylcholine receptors ( tyrothricin , tomatine , carbachol , and tacrine hydrochloride ) ( Figure 2B and C; Changeux et al . , 1969; Milner et al . , 2011; Rosenberry et al . , 2008; Marco and Carreiras , 2003; Lang and Staiger , 2016; Shih et al . , 2009 ) . One compound has been shown to promote insulin secretion by inhibition of the mitochondrial protein tyrosine phosphatase PTPM1 ( alexidine hydrochloride ) ( Figure 2B and C; Doughty-Shenton et al . , 2010; Nath et al . , 2015 ) , and another likely affects β-cell function by signaling through the mineralocorticoid receptor ( deoxycorticosterone ) ( Figure 2B and C; Lu et al . , 2006 ) . Finally , in addition to ion channels and GPCRs , the macrolide IVM has also been shown to signal in micromolar concentrations though several ionotropic receptors , including purinergic , GABAergic , and glycine receptors , as well as through the farnesoid X nuclear receptor ( Chen and Kubo , 2018; Dawson et al . , 2000; Soltani et al . , 2011 ) . 10 of these 12 hit compounds were not considered for further analysis because of either the high dose required to achieve insulin secretion ( Figure 2—figure supplement 1B ) or because they augmented insulin release in low basal glucose ( 2 mM ) in intact WT mouse primary islets ( Figure 2D ) . One of the remaining compounds induces hepatotoxicity after prolonged use ( tacrine hydrochloride ) ( Galisteo et al . , 2000 ) . We therefore focused our attention on IVM due to its dose-dependent enhancement of GSIS in insulin-NanoLuc-expressing Beta-TC-6 cells , as well as its robust rescue of insulin secretion in Bmal1-/- islets ( Figure 2D and E ) . To test whether IVM drives GSIS in β-cell lines and primary mouse islets , we first assessed the impact of both acute treatment ( 1 hr ) and overnight exposure ( 24 hr ) with 10 μM IVM on the ability of WT β-cells and mouse islets to secrete insulin ( Figure 3A , Figure 3—figure supplement 1A ) . Consistent with our initial bioluminescence assay , we observed that IVM enhanced insulin secretion in a glucose-dependent manner following both 1 hr IVM exposure and 24 hr pretreatment with IVM in β-cell lines and WT mouse islets , suggesting that both acute and longer-term exposure to IVM enhance β-cell function ( Figure 3A , Figure 3—figure supplement 1A ) . Since there was not a significant increase in insulin secretion with overnight ( approximately twofold ) compared to acute ( ~1 . 5–1 . 6-fold ) IVM exposure , further analysis of IVM as a potentiator of insulin secretion was performed only with acute treatment . Chemical energy from ATP generated by glucose metabolism within the β-cell triggers closure of the sulfonylurea-linked potassium channel , depolarization of the plasma membrane , and opening of voltage-gated calcium channels , leading to stimulus-secretion coupling . To assess the mechanism of IVM-induced insulin secretion , we next monitored real-time calcium influx using ratiometric fluorescence imaging in WT β-cells in the presence of both glucose and IVM . We observed an immediate and robust glucose-stimulated intracellular calcium response within 2 min of IVM stimulation ( p<0 . 05 ) ( Figure 3—figure supplement 1B ) . Importantly , this effect was only observed in the presence of high glucose , consistent with results of our initial NanoLuc 384-well plate screening and subsequent ELISA-based analyses of GSIS . In contrast , the Ca2+ channel inhibitor isradipine completely suppressed Ca2+ influx and insulin secretion ( Figure 3—figure supplement 1C and D; Berjukow et al . , 2000 ) . To determine whether increased calcium influx corresponded with productive insulin release following IVM treatment , we used a dynamic perifusion system to directly measure NanoLuc activity in eluates harvested from IVM-treated β-cells every 2 min over the course of 30 min following stimulation with either 20 mM glucose or 20 mM glucose plus 10 µM IVM ( Figure 3—figure supplement 1E ) . IVM significantly increased insulin release during the initial burst of secretion within the first 12 min post-stimulation ( p<0 . 05 ) and continued to enhance insulin secretion during the remainder of the stimulation period ( 12–30 min ) , consistent with continuous release of reserve insulin granules ( Rorsman and Renström , 2003 ) . Since our cell-based studies indicated that IVM stimulates GSIS within immortalized β-cell lines , we next sought to determine whether IVM restores insulin secretion in the context of circadian disruption within primary islets , which are composed of multiple hormone-releasing cell types ( Arrojo E Drigo et al . , 2020 ) . To test this idea , we administered IVM to mouse islets isolated from pancreas-specific Bmal1-/- mice , revealing a 3 . 3-fold elevation of GSIS following exposure to the drug in the Bmal1 mutant islets ( Figure 3B ) . Furthermore , perifusion experiments in islets from Bmal1 mutant mice revealed that IVM significantly increased insulin release during both the initial burst of secretion ( first 12 min post-stimulation ) and during the sustained release ( 12–30 min ) in both WT and Bmal1 mutant islets ( Figure 3C ) . Additionally , we observed a similar 2 . 9-fold increase in GSIS following administration of IVM to islets isolated from an independent mouse model of circadian disruption ( Cry1-/-;Cry2-/- mice ) ( Figure 3D ) , suggesting that IVM ameliorates secretory defects caused by disruption of the circadian clock network . To determine if IVM can improve glucose homeostasis in diabetic animals , we next tested the effects of chronic IVM administration in the well-characterized C57BL/6-Ins2Akita/J Akita model of β-cell failure ( Yoshioka et al . , 1997 ) . Daily intraperitoneal IVM ( 1 . 3 mg/kg body weight ) was administered to Akita mice over a 14-day period ( Jin et al . , 2013 ) , terminating in assessment of glucose tolerance and ex vivo GSIS . Treatment with IVM significantly improved glucose tolerance and augmented glucose-stimulated insulin release from islets isolated from these mice ( Figure 3—figure supplement 1F and G ) . Given that our prior genomic and cell physiological studies have localized the β-cell defect in circadian mutant mice to impaired insulin exocytosis ( Marcheva et al . , 2020 ) , and as IVM augmented insulin secretion in Bmal1 mutant islets , we next sought to determine whether IVM might enhance depolarization-induced exocytosis using electrophysiological analyses ( Fu et al . , 2019 ) . We assessed cumulative capacitance , a measure of increased cell surface area as insulin granules fuse to the plasma membrane , in β-cells from islets of control and pancreas-specific Bmal1 mutant mice , as well as from human cadaveric islets . While Bmal1 mutant cells displayed reduced rates of exocytosis following direct depolarization ( as indicated by reduced capacitance ) , 10 µM IVM treatment rescued the defect in Bmal1 mutant cells , increasing cumulative capacitance from 11 . 0 to 20 . 7 fF/pF after 10 consecutive depolarization steps ( Figure 3E ) . IVM treatment also enhanced cumulative capacitance in human β-cells from 17 . 9 to 39 . 7 fF/pF ( Figure 3F ) . Together , these data show that IVM augments β-cell early calcium influx in a glucose-dependent manner to promote increased vesicle fusion and release . Several of the predicted targets of the insulinotropic compounds from our screen , including IVM , involve second-messenger signaling , raising the possibility that circadian disruption may be overcome by augmenting hormonal or metabolic factors that promote peptide exocytosis . IVM is a readily absorbable and potent derivative of avermectin B1 that allosterically regulates several different types of cell surface receptors , including purinergic and GABA receptors , as well as nuclear transcription factors such as the farnesoid X receptor ( FXR ) ( Jin et al . , 2013; Khakh et al . , 1999; González Canga et al . , 2008; Estrada-Mondragon and Lynch , 2015 ) . Since IVM augments insulin secretion in Bmal1-/- cells , we hypothesized that the expression of putative IVM targets may be reduced during circadian disruption . We first identified the purinergic receptor P2Y1 ( P2ry1 ) as the most highly expressed putative IVM target in wild-type β-cells ( Figure 4A ) . We then observed that P2ry1 was one of the most highly downregulated targets in Bmal1-/- cells , with mRNA expression levels reduced by ~3 . 1-fold ( adjusted p=10–55; Figure 4A , Figure 4—figure supplement 1A; GSE146916 ) . We found decreased levels and loss in rhythmicity of P2ry1 in synchronized Bmal1-/- pseudoislets ( Figure 4—figure supplement 1B ) . BMAL1 chromatin immunoprecipitation-sequencing in Beta-TC-6 cells further revealed enrichment of BMAL1 chromatin binding within enhancer regions 266–41 kb upstream of the P2ry1 gene transcription start site ( GSE69889; Figure 4A , Figure 4—figure supplement 1A ) . Finally , analysis of RNA-sequencing data from human islets ( SRA accession ERP017126 ) indicates that P2RY1 expression is enriched within β-cells among hormone-secreting cell types , with little to no detectable expression in the glucagon-secreting α cells ( Figure 4—figure supplement 1C; Segerstolpe et al . , 2016 ) . Together , these data reveal direct rhythmic control of the P2ry1 gene by the β-cell circadian clock . Based upon evidence that IVM targets purinergic receptors ( Weng et al . , 2008; Priel and Silberberg , 2004; Bowler et al . , 2003; Hansen et al . , 2008 ) , that the predominant purinergic receptor on β-cells is P2Y1 , and that BMAL1 specifically controls P2ry1 amongst the purinergic receptor family in the β-cell ( Figure 4A , Figure 4—figure supplement 1A and B ) , we sought to test the functional role of the P2Y1 receptor in the insulinotropic action of IVM . Pharmacological inhibition of P2Y1 using a subtype-specific inhibitor , the nucleotide analog MRS2179 , in the presence of both high glucose and 10 μM IVM resulted in a 52% reduction in insulin secretion by bioluminescence and a reduction in calcium influx to levels similar to those observed during high glucose alone , as assessed by Fura2-AM ratiometric determination of intracellular calcium ( Figure 4B and C ) . In addition to evidence that pharmacological blockade of P2Y1 receptor signaling attenuates IVM activity , we also tested the requirement of P2Y1 receptor signaling following CRISPR-Cas9-mediated knockout of the P2Y1 receptor in both WT and Bmal1-/-β-cells ( Figure 4—figure supplement 2A ) . While IVM enhanced GSIS in WT and Bmal1-/-β-cells 1 . 6- and 1 . 8-fold , respectively , IVM did not significantly enhance GSIS in cells lacking the P2Y1 receptor ( Figure 4D ) . Similar to the pharmacological findings with the P2Y1 antagonist MRS2179 , these results demonstrate a requirement for P2Y1 in IVM-induced GSIS . P2Y1 receptor signaling involves activation of Ca2+ entry and intracellular release , which results in both acute stimulation of insulin granule trafficking and activation of transcription factors that may be involved in β-cell function ( Léon et al . , 2005; Khan et al . , 2014; Balasubramanian et al . , 2010 ) . To analyze gene expression changes induced by P2Y1 activation , we performed RNA-sequencing to compare the IVM response within both WT and P2ry1-/- β-cells following stimulation with glucose or glucose plus IVM . Principal component analysis ( PCA ) was performed using log-transformed count data from the top 500 most variable genes across all samples ( Love et al . , 2014 ) . This revealed distinct patterns in mRNA expression between IVM- and control-treated WT cells along PC2 , while there was no separation between IVM- and control-treated P2ry1-/- β-cells , suggesting that P2Y1 is required for IVM-mediated transcriptional changes in β-cells ( Figure 4E ) . In WT cells , IVM induced differential expression of 65 transcripts ( 1 . 5-fold change , adjusted p-value<0 . 05 ) , including upregulation of the immediate early gene Fos ( Murphy et al . , 1991 ) and downregulation of Aldolase B , whose expression has been linked to reduced insulin secretion in human islets ( Gerst et al . , 2018; Figure 4F , Figure 4—figure supplement 2B , Supplementary file 2 ) . Strikingly , none of these transcripts were significantly altered by IVM in the P2ry1-/- β-cells ( all adjusted p-value>0 . 05 ) ( Figure 4F , Figure 4—figure supplement 2B , Supplementary file 2 ) . Taken together , these data suggest that the circadian clock program controls P2Y1 expression to modulate GSIS and highlight the utility of a genetic-sensitized drug screen for identification of therapeutic targets in circadian dysregulation and diabetes .
We have identified an unexpected role for the P2Y1 receptor as a BMAL1-controlled insulinotropic factor required for enhanced β-cell glucose-stimulated Ca2+ influx and insulin secretion in response to IVM . While P2Y receptors have been previously implicated in calcium and insulin secretory dynamics in β-cells , modulation has been primarily demonstrated using agonists that mimic ATP/ADP derivatives that have deleterious effects on thrombosis ( Léon et al . , 2005; Khan et al . , 2014; Balasubramanian et al . , 2010; Gąsecka et al . , 2020 ) . Little is known about P2Y1 targeting in disease states , such as circadian disruption and/or type 2 diabetes , or whether P2Y1 is controlled at a transcriptional level . Our evidence that P2Y1 is expressed under control of the circadian clock derives from analyses at the level of both chromatin binding by the core clock factor BMAL1 and genome-wide differential RNA expression analysis in circadian mutants . Intriguingly , P2X and P2Y receptors are required for Ca2+ signaling in the suprachiasmatic nucleus ( Lommen et al . , 2017; Svobodova et al . , 2018 ) , yet their role in circadian regulation of peripheral tissues has not been well studied . Our data suggests that IVM action requires the presence of P2Y1 receptors in β-cells since functional ablation of the P2Y1 receptor attenuates the effect of IVM on insulin secretion in both wild-type and circadian mutant β-cells ( Figure 4D ) . Our analyses reveal that pharmacological enhancement of P2Y1 receptor activity may therefore bypass pathological and circadian alterations in expression of the P2Y1 receptor in β-cells to restore insulin secretion . Recently , the P2Y1 receptor was implicated in nutrient- and ATP/ADP-dependent regulation of insulin release through an adipocyte-islet axis , further suggesting that P2RY1 may play a role in physiological regulation of islet hormone release ( Prentice et al . , 2021 ) . Future studies will be required to determine whether IVM affects paracrine ATP/ADP release to affect P2RY1 or whether IVM directly binds purinergic receptors in the β-cell . One possibility is that IVM may augment P2X-P2Y1 crosstalk to drive insulin secretion , which has been shown to drive Ca2+ and P2Y1-dependent activation of other cell types ( Weng et al . , 2008; Woehrle et al . , 2019 ) . Previous physiological and transcriptomic studies have shown that circadian regulation of insulin exocytosis involves control of the expression and activity of cell-surface receptors and second messenger systems ( Perelis et al . , 2015; Gil-Lozano et al . , 2014 ) . We based our drug screen on the idea that modulators of insulin secretion in cells that lack a functional clock would complement prior genomic analyses revealing circadian control of peptidergic hormone exocytosis and also to provide proof of principle that the clock can be leveraged to sensitize screening for new chemical modulators of β-cell function . This approach identified Ca2+-dependent pathways as a potential route to ameliorate circadian disruption and enhance GSIS . Previous small-molecule screens have identified glucose-dependent insulinotropic compounds in wild-type cells ( Burns et al . , 2015 ) . However , several of these compounds , including the anti-inflammatory bufexamac and anti-giardiasis drug lobendazole , were found to be ineffective or even inhibitory in our circadian mutant screen ( Supplementary file 1 ) . In the future , high-throughput screens may lead to more personalized therapeutics through comparison of insulinotropic compounds identified using cells without known mutations versus those discovered in cells harboring monogenic or polygenic diabetes variants . Several of the compounds identified in our screen have been used in disease treatment and have known mechanisms of action , including the cholinergic activators carbachol and tacrine ( Linn et al . , 1989; Crismon , 1994 ) . The identification of these compounds in our screen raises the intriguing possibility of using drug derivatives related to these molecules for type 2 diabetes treatment , particularly in the context of circadian/sleep disruption . The study of transcriptional rhythms across the 24 hr circadian cycle has previously revealed a diverse landscape of clock-controlled genes and pathways ( Zhang et al . , 2014 ) . Despite the identification of thousands of tissue-specific and clock-controlled transcripts , limited advances have been made in utilizing this information to treat diseases associated with circadian disruption , including type 2 diabetes . One approach to this challenge has been to intervene and restore the molecular clock program using pharmacology ( Nobiletin ) ( He et al . , 2016 ) , micronutrient supplementation ( NAD+ precursors ) ( Levine et al . , 2020; Sato et al . , 2017 ) , or enforced behavioral rhythms ( such as time-restricted feeding ) ( Sutton et al . , 2018 ) . However , it remains unclear how altering the whole-body clock will affect nutritional and hormonal dynamics at a cellular level . Another approach has been to directly target clock-controlled genes with known function in health and disease ( Lamia et al . , 2008 ) or to look at gain/loss of circadian control in health versus disease ( Petrenko et al . , 2020a ) . This approach requires an understanding of gene function within a given tissue , and thus limits the identification of novel therapeutic targets . In the studies performed here , we sought to address the challenge of connecting clock control of transcription with druggable targets by using an unbiased small-molecule drug screen , in tandem with functional genomics , to elucidate mechanisms of insulin secretory dynamics . Since the circadian timing system has been shown to not only regulate the function of mature β-cells , but also the regenerative capacity of islets in both the context of the mouse ( Petrenko et al . , 2020b ) and in human embryonic stem cell differentiation ( Alvarez-Dominguez et al . , 2020 ) , molecules identified in cell-based genetic screens may provide broad applicability as therapeutics .
IVM , ( + ) -bicuculline , and MRS2179 tetrasodium salt were obtained from Tocris ( R&D Systems , Inc , Minneapolis , MN ) . Isradipine and alexidine hydrochloride were purchased from Cayman Chemical Company ( Ann Arbor , MI ) . PMA , carbamoylcholine chloride ( carbachol ) , forskolin , tyrothricin , and benzalkonium chloride were obtained from Sigma-Aldrich ( St . Louis , MO ) . Suloctidil , tomatine , isoetharine mesylate , tacrine hydrochloride , pipamperone , dyclonine hydrochloride , and desoxycorticosterone acetate were purchased from MicroSource Discovery Systems , Inc . Male WT C57BL6J mice and C57BL/6-Ins2Akita/J mice were purchased from the Jackson Laboratory ( Bar Harbor , ME ) . PdxCre;Bmal1flx/flx and Cry1-/-;Cry2-/- mice were produced and maintained on C57BL6J background at the Northwestern University Center for Comparative Medicine ( Protocols IS00000466 , IS00003253 , IS00008732 , IS0005838 ) ( Peek et al . , 2013; Vitaterna et al . , 1999 ) . Unless otherwise stated , animals were maintained on a 12:12 light:dark cycle and allowed free access to water and regular chow . All animal care and use procedures were conducted in accordance with regulation of the Institutional Animal Care and Use Committee at Northwestern University . Beta-TC-6 cells were obtained from ATCC ( Manassas , VA ) ( CRL-11506 ) , and Bmal1-/- Beta-TC-6 β-cell lines were previously derived as described ( Marcheva et al . , 2020 ) . Cells were cultured in Dulbecco’s modified Eagle’s medium ( DMEM; Gibco , Aramillo , TX ) supplemented with 15% fetal bovine serum ( Bio-Techne , Minneapolis , MN ) , 1% penicillin-streptomycin ( Gibco ) , and 1% L-glutamine ( Gibco ) at 37°C with 5% CO2 . Culture medium was exchanged every 2–3 days . All cells used in experiments were at <15 passages . Cells were routinely checked for mycoplasma contamination . We used the proinsulin-NanoLuc plasmid ( David Altshuler , Addgene plasmid #62057 , proinsulin-NanoLuc in pLX304 ) to provide a low-cost , scalable , and rapid method to detect insulin secretion . The gene encoding NanoLuc was cloned into the C-peptide portion of mouse proinsulin such that cleavage within insulin vesicles by pH-sensitive prohormone convertase results in the co-secretion of NanoLuc with endogenous insulin in a stimulus-dependent manner ( Burns et al . , 2015 ) . The pLX304 lentivirus packaging plasmid containing the proinsulin-NanoLuc construct was transfected into HEK293T ( ATCC CRL-11268 ) cells with pCMV-VSVG ( envelope vector , Addgene plasmid #8454 ) and pCMV delta R8 . 2 ( packaging vector , Addgene plasmid #12263 ) . Supernatant containing lentivirus particles was harvested 48 hr after transfection . Beta-TC-6 and Bmal1-/- Beta-TC-6 cells were infected with insulin-NanoLuc lentivirus , and stably expressing cells were selected by treating with puromycin ( 2 µg/ml , 2 days ) . Exon 1 of the mouse P2yr1 gene was deleted in WT and Bmal1-/- Beta-TC-6 cells by CRISPR-Cas9 and homology-directed repair ( HDR ) . Cells were co-transfected with guide RNA , P2Y1 CRISPR/Cas9 KO , and P2Y1 HDR plasmids ( Santa Cruz Biotechnology , Dallas , TX , sc-422095 and sc-422095-HDR ) by Lipofectamine 3000 ( Thermo Fisher Scientific , Amarillo , TX ) . The locations of the three sites targeted for ablation by the P2Y1 CRISPR/Cas9 KO plasmids are indicated in Figure 4A . After 48 hr of transfection , stably integrated clones were selected for puromycin resistance ( puromycin dihydrochloride , Sigma-Aldrich ) . RNA and protein were extracted from these colonies , and P2ry1 expression was assessed by qPCR and Western blot . The Spectrum Collection small-molecule compound library ( MicroSource Discovery Systems , Inc ) , which consists of 2640 known drugs and drug-like molecules , was screened for compounds that augment insulin secretion in Bmal1-/- Beta-TC-6 cells . Insulin-NanoLuc-expressing Bmal1-/- Beta-TC-6 cells ( 40 , 000 cells/well ) were placed into 384-well plates and cultured for 3 days at 37°C and 5% CO2 . The cells were washed once and incubated in KRB buffer containing 0 mM glucose for 1 hr . Then , KRB buffer containing 20 mM glucose in addition to the small molecules ( 10 µM ) were added , and the cells were incubated for 1 hr . As a negative control , 16 wells received KRB buffer with only 20 mM glucose , which fails to elicit appropriate insulin secretion in Bmal1-/- cells , and as a positive control , 16 wells received KRB buffer containing 20 mM glucose and 10 μM PMA , which is known to induce insulin secretion in both Bmal1-/- mouse islets and Beta-TC-6 cells ( Perelis et al . , 2015 ) . After 1 hr , the supernatant was collected and centrifuged at 500 × g for 30 min . The supernatant was transferred into a fresh 384-well assay plate containing NanoGlo Luciferase Assay Substrate ( Promega , Madison , WI ) , and luciferase intensity was measured by EnSpire Plate Reader ( PerkinElmer , Waltham , MA ) within 30 min . All liquids for the high-throughput screen were dispensed using Tecan Fluent Automated Liquid Handling Platform ( Tecan , Mannedorf , Switzerland ) at the High-Throughput Analysis Laboratory at Northwestern University . Screen feasibility was determined by calculating Z'-factor using the following formula: Z'-factor = 1–3 ( σp + σn ) / ( μp - μn ) ( where σp is the standard deviation of positive control [20 mM glucose + PMA] , σn is the standard deviation of negative control [20 mM glucose only] , μp is the mean intensity of positive control , and μn is the mean intensity of the negative control ) ( Zhang et al . , 1999 ) . Z-scores are a measure of how many standard deviations above or below the population mean a raw score is . Z-scores for luciferase intensities produced by screened compounds were calculated from the following formula: z = ( X – μ ) /σ ( where z is the Z-score , X is the luciferase intensity of the compounds , μ is the intensity of negative control [20 mM glucose] , and σ is the standard deviation of negative control ) . A row-based correction factor was applied to all luciferase readings to adjust for logarithmic signal decay . Hit compounds were defined as those that elicited a response of greater than 3 standard deviations from the mean ( Z-score > 3 ) and more than 1 . 25-fold increase compared to negative control , which is the cutoff for ~10% chance of the observation occurring by random chance . Validated hit compounds that augmented insulin secretion at low drug dose were considered lead compounds . Mouse pancreatic islets were isolated via bile duct collagenase digestion ( Collagenase P , Sigma ) and Biocoll ( Millipore ) gradient separation and left to recover overnight at 37°C in RPMI 1640 with 10% FBS , 1% L-glutamine , and 1% penicillin/streptomycin . For insulin release assays , duplicates of five equally sized islets per mouse were statically incubated in Krebs-Ringer Buffer ( KRB ) at 2 mM glucose for 1 hr and then stimulated for 1 hr at 37°C with 2 mM or 20 mM glucose in the presence or absence of 10 μM of each compound . Supernatant was collected and assayed for insulin content by ELISA ( Crystal Chem Inc , Elk Grove Village , IL ) . Islets were then sonicated in acid-ethanol solution and solubilized overnight at 4°C before assaying total insulin content by ELISA . For insulin release assays from pseudoislets , 3 × 106 cells were plated for 3 days in 60 mm suspension dishes and allowed to form pseudoislets for 2–3 days . Glucose-responsive insulin secretion was performed as described above using 10 pseudoislets per sample and a basal glucose level of 0 mM glucose instead of 2 mM ( Marcheva et al . , 2020 ) . For secretion from insulin-NanoLuc cell lines , 1 × 105 cells were cultured on poly-L-lysine-coated 96 well plates for 2–3 days , starved for 1 hr in 0 mM glucose KRB , then stimulated with indicated compounds and/or receptor antagonists for 1 hr in conjunction with 0 mM basal glucose or 20 mM stimulatory glucose conditions . Luciferase intensity after addition of NanoGlo to supernatant was measured by Cytation3 Plate Reader ( BioTek , Winooski , VT ) . Primary islets from PdxCre;Bmal1flx/flx and Bmal1flx/flx mice were isolated as described above and left to recover overnight . Perifusion of 100 islets per mouse per treatment was performed using a Biorep Technologies Perifusion System Model PERI-4 . 2 with a rate of 100 μl/min KRB ( 0 . 1% BSA ) . After 1 hr of preincubation and equilibration at a rate of 100 μl/min with 2 mM glucose KRB , islets were perifused for 10 min with 2 mM glucose KRB , followed by perifusion for 30 min with 20 mM glucose or 20 mM glucose plus IVM . Perifusate was collected in 96-well plates , and insulin secreted was analyzed via ELISA . Perifusion of insulin-NanoLuc pseudoislets was performed in an identical manner using 0 mM glucose KRB instead of 2 mM glucose KRB . Pseudoislet perifusate was analyzed for NanoLuc activity using NanoGlo Luciferase Assay Substrate ( Promega ) as per the manual instructions . Mice were injected intraperitoneally for 14 days with 1 . 3 mg/kg body weight of IVM , which was dissolved in 40% w/v 2-hydroxypropyl-β-cyclodextrin ( Sigma-Aldrich ) ( Jin et al . , 2013 ) . At the end of IVM treatment , mice were fasted for 14 hr and glucose tolerance tests were performed at ZT2 following intraperitoneal glucose injection at 2 g/kg body weight . Plasma glucose levels were measured by enzymatic assay ( Autokit Glucose , Wako-Fujifilm , Cincinnati , OH ) . Where indicated , circadian synchronization was performed using 200 WT pseudoislets by first exposing cells to 10 μM forskolin for 1 hr , followed by transfer to normal media and RNA collection every 4 hr 24–44 hr following forskolin synchronization pulse . RNA was extracted from Beta-TC-6 cells and pseudoislets using Tri Reagent ( Molecular Research Center , Inc , Cincinnati , OH ) and frozen at −80°C . RNA was purified according to the manufacturer’s protocol using the Direct-zol RNA Microprep kit ( Zymo Research , Irvine , CA ) with DNase digestion . cDNAs were then synthesized using the High Capacity cDNA Reverse Transcription Kit ( Applied Biosystems , Amarillo , TX ) . Quantitative real-time PCR analysis was performed with SYBR Green Master Mix ( Applied Biosystems ) and analyzed using a Touch CFX384 Real-Time PCR Detection System ( Bio-Rad , Hercules , CA ) . Target gene expression levels were normalized to β-actin and set relative to control conditions using the comparative CT method . Primer sequences for qPCR are as follows: β-actin forward: 5'-TGCTCTGGCTCCTAGCACCATGAAGATCAA-3' , reverse: 5'-AAACGCAGCTCAGTAACAGTCCGCCTAGAA-3'; P2ry1 forward: 5'- TTATGTCAGCGTGCTGGTGT-3' , reverse: 5'-ACGTGGTGTCATAGCAGGTG-3' . Following RNA isolation ( described above ) , RNA quality was assessed using a Bioanalyzer ( Agilent , Santa Clara , CA ) , and sequencing libraries were constructed using a NEBNext Ultra Directional RNA Library Prep Kit for Illumina ( New England Biolabs , Ipswich , MA , E7420L ) according to the manufacturer’s instructions . Libraries were quantified using a NEBNext Library Quant Kit for Illumina ( New England Biolabs , E7630L ) and sequenced on an Illumina NextSeq 500 instrument using 42 bp paired-end reads . For differential expression analysis , RNA raw sequence reads were aligned to the reference genome ( mm10 ) using STAR version 2 . 7 . 2b , and raw and transcripts per million ( TPM ) count values determined using RSEM version 1 . 3 . 3 . Differentially expressed RNAs were identified by a false discovery rate ( FDR ) -adjusted p-value<0 . 05 and a fold change > 1 . 5 using DESeq2 version 1 . 32 . 0 in R 4 . 1 . 0 . Heatmaps were generated using the pheatmap package in R . Raw mRNA sequencing data and gene abundance measurements have been deposited in the Gene Expression Omnibus under accession GSE186469 . Beta-TC-6 cells were plated at a density of 100 , 000 cells per well in black 96-well plates with clear bottoms and cultured overnight at 37°C and 5% CO2 . Cells were then washed with BSA-free KRB buffer with no glucose and loaded with 5 μM Fura-2 ( Invitrogen , Amarillo , TX ) and 0 . 04% Pluronic F-127 ( Invitrogen ) for 30 min at 37°C . Following a wash with BSA-free KRB , Fura-2 intensity was measured after stimulation with either glucose alone or glucose plus the indicated compounds . Cells were alternately excited with 340 nm and 380 nm wavelength light , and the emitted light was detected at 510 nm using a Cytation 3 Cell Imaging Multi-Mode Reader ( BioTek ) at sequential 30 s intervals . Raw fluorescence data were exported to Microsoft Excel and expressed as the 340/380 ratio for each well . Human islet isolations and human islet cell biology experiments approved by the University of Alberta Human Research Ethics Board ( approval identifiers: Pro00013094; Pro00001754 ) were performed at the Alberta Diabetes Institute Islet-Core according to the methods deposited in the protocols . io repository ( Isolation of Human Pancreatic Islets of Langerhans for Research V . 3 , 2021 ) . Organ donation was coordinated by the appropriate regional organ procurement organization , including obtaining written next-of-kin consent for use of donor organs in this study . Donor organs were deidentified by the organ procurement organization prior to shipment to the Alberta Diabetes Institute Islet-Core , and no identifying donor information was made available to the research team . A total of three nondiabetic ( ND ) donors were examined in this study . Full details of donor information , organ processing , and quality control information can be assessed with donor number ( donors R224 , R225 , and R226 in this study ) at https://www . isletcore . ca . Patch-clamp measurement of exocytic responses in mouse β-cells was performed as previously described ( Marcheva et al . , 2020 ) . Dispersed human islets were cultured in low glucose ( 5 . 5 mM ) DMEM media ( supplemented with L-glutamine , 110 mg/l sodium pyruvate , 10% FBS , and 100 U/ml penicillin/streptomycin ) in 35 mm culture dishes overnight . On the day of patch-clamp measurements , human or mouse islet cells were preincubated in extracellular solution at 1 mM glucose for 1 hr and capacitance was measured at 10 mM glucose with DMSO or 10 µM IVM as previously described ( Marcheva et al . , 2020 ) . Mouse β-cells were identified by cell size and by half-maximal inactivation of Na+ currents near –90 mV , and human β-cells were identified by immunostaining for positive insulin , following the experiment as described ( Fu et al . , 2019 ) . Data analysis was performed using GraphPad Prism ( v8 . 0c ) . Comparison of multiple groups was done by one- or two-way ANOVA , followed by Bonferroni or Tukey’s post test . Data are expressed as means ± SEM , where p<0 . 05 is considered significant . Sequencing data from the study under SRA accession ERP017126 ( Segerstolpe et al . , 2016 ) were downloaded and converted to fastq files using the commands ‘prefetch’ followed by ‘fastq-dump’ through the sra-toolkit ( v2 . 10 . 5 ) . Each individual cell was aligned and transcript abundance quantified using RSEM with Hg38 ( GRCh38 . p12 ) as a reference . Raw single-cell expression count values were imported into RStudio for analysis using Seurat ( Hao et al . , 2021 ) . Following low-quality cell removal , normalized expression values were used in uniform manifold approximation and projection ( UMAP ) dimensional reduction analyses to cluster distinct cell types . The R script , raw count tables , and parameters of these analyses are made publicly available under the Gene Expression Omnibus accession GSE186469 . Beta-TC-6 cells lysates were isolated by treating cell pellets with RIPA buffer ( Sigma-Aldrich ) supplemented with 1× protease and 1× phosphatase inhibitors ( Roche , Basel , Switzerland ) . Protein levels were quantified using Quick Start Bradford Protein Assay , and protein extracts were subject to SDS-PAGE gel electrophoresis and transferred to nitrocellulose membranes ( Bio-Rad ) . Primary antibodies used were anti-P2Y1 ( Santa Cruz , sc-377324 ) and anti-β-actin ( Cell Signaling , Danvers , MA , CST 4970 ) . Results were expressed as mean ± SEM unless otherwise noted . Information on sample size , genotype , and p values is provided within each figure and figure legend . Statistical significance of capacitance , Fura2 , and perifusion data was performed using a two-way ANOVA or mixed effects model ( for datasets with missing values ) with repeated measures followed by multiple comparison tests using a Bonferroni p-value adjustment via Prism ( v9 . 2 . 0 ) . Statistical analysis was performed by unpaired two-tailed Student’s t-test unless otherwise indicated . p<0 . 05 was considered statistically significant . JTK_Cycle ( v3 ) was used to determine rhythmicity in qPCR data using a period length of 24 hr and considering a Benjamini–Hochberg ( BH ) -adjusted p-value<0 . 05 as statistically rhythmic ( Hughes et al . , 2010 ) . | Circadian rhythms – ‘inbuilt’ 24-hour cycles – control many aspects of behaviour and physiology . In mammals , they operate in nearly all tissues , including those involved in glucose metabolism . Recent studies have shown that mice with faulty genes involved in circadian rhythms , the core clock genes , can develop diabetes . Diabetes arises when the body struggles to regulate blood sugar levels . In healthy individuals , the hormone insulin produced by beta cells in the pancreas regulates the amount of sugar in the blood . But when beta cells are faulty and do not generate sufficient insulin levels , or when insulin lacks the ability to stimulate cells to take up glucose , diabetes can develop . Marcheva , Weidemann , Taguchi et al . wanted to find out if diabetes caused by impaired clock genes could be treated by targeting pathways regulating the secretion of insulin . To do so , they tested over 2 , 500 potential drugs on genetically modified beta cells with faulty core clock genes . They further screened the drugs on mice with the same defect in their beta cells . Marcheva et al . identified one potential compound , the anti-parasite drug ivermectin , which was able to restore the secretion of insulin . When ivermectin was applied to both healthy mice and mice with faulty beta cells , the drug improved the control over glucose levels by activating a specific protein receptor that senses molecules important for storing and utilizing energy . The findings reveal new drug targets for treating forms of diabetes associated with deregulation of the pancreatic circadian clock . The drug screening strategy used in the study may also be applied to reveal mechanisms underlying other conditions associated with disrupted circadian clocks , including sleep loss and jetlag . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2022 | P2Y1 purinergic receptor identified as a diabetes target in a small-molecule screen to reverse circadian β-cell failure |
Over the last 50 years , anatomical models of memory have repeatedly highlighted the hippocampal inputs to the mammillary bodies via the postcommissural fornix . Such models downplay other projections to the mammillary bodies , leaving them largely ignored . The present study challenged this dominant view by removing , in rats , the two principal inputs reaching the mammillary bodies: the postcommissural fornix from the hippocampal formation and Gudden's ventral tegmental nucleus . The principal mammillary body output pathway , the mammillothalamic tract , was disconnected in a third group . Only mammillothalamic tract and Gudden's ventral tegmental nucleus lesions impaired behavioral tests of spatial working memory and , in particular , disrupted the use of extramaze spatial landmarks . The same lesions also produced widespread reductions in immediate-early gene ( c-fos ) expression in a network of memory-related regions , not seen after postcommissural fornix lesions . These findings are inconsistent with previous models of mammillary body function ( those dominated by hippocampal inputs ) and herald a new understanding of why specific diencephalic structures are vital for memory .
Since the first neuroanatomical circuits of emotion ( e . g . , Papez , 1937 ) and memory ( e . g . , Barbizet , 1963; Delay and Brion , 1969 ) , the perceived contributions of the mammillary bodies have been dominated by their direct inputs from the hippocampal formation ( via the fornix ) . Consequently , the mammillary bodies are principally seen as a relay of hippocampal projections to the anterior thalamic nuclei and , from there , to the cingulate ( Barbizet , 1963; Delay and Brion , 1969 ) and prefrontal ( Warrington and Weiskrantz , 1982 ) cortices . The notion that the mammillary bodies form part of such an ‘extended hippocampal system’ ( Aggleton and Brown , 1999 ) is seemingly supported by the unidirectional nature of the hippocampal projections to the mammillary bodies ( Aggleton et al . , 2005 ) , along with clinical evidence of the importance for memory of the mammillary body projections to the thalamus , via the mammillothalamic tract ( MTT; Carlesimo et al . , 2007; Van der Werf et al . , 2000; Vann and Aggleton , 2003 ) . However , these standard models suffer two major shortcomings . First , they provide no real function for the mammillary bodies and second , they ignore other mammillary body afferents . The mammillary bodies receive dense inputs from Gudden's tegmental nuclei ( e . g . , Takeuchi et al . , 1985 ) , making it possible that these midbrain inputs are functionally important for memory ( Vann , 2010 ) . Three groups of rats contrasted the standard model with an alternative model of mammillary body function based on its tegmental inputs . Rats with lesions blocking either of the two main mammillary body afferents ( hippocampus formation or ventral tegmental nucleus of Gudden ) were tested alongside rats with lesions of the main mammillary body efferent pathway , the mammillothalamic tract . Only by comparing these separate lesions within the same study is it possible to measure their relative contributions . In one group the descending postcommissural fornix ( PCF ) was lesioned , a procedure that disconnects the hippocampal formation projections to the mammillary bodies whilst leaving other hippocampal connections intact , including those with the septum , striatum and anterior thalamic nuclei ( Vann et al . , 2011 ) . A second group received neurotoxic lesions of Gudden's ventral tegmental nucleus ( VTNg ) . Although the lateral mammillary nucleus receives dense inputs from Gudden's dorsal tegmental nucleus , the VTNg was targeted because of its inputs to the medial mammillary nucleus . Previous clinical and behavioral evidence highlights the particular importance of the medial mammillary nucleus for memory ( Vann , 2005 , 2010 , 2011; Vann and Albasser , 2009 ) . A third group received mammillothalamic tract lesions while a fourth group underwent control surgery . Rats were tested on three behavioral tasks , the first two of which are sensitive to hippocampal damage: T-maze alternation , a working memory task in the radial-arm maze , and a geometric discrimination task ( Aggleton et al . , 2009; Vann , 2011 ) . In addition to impairing spatial memory , MTT lesions also result in widespread neural hypoactivity , as measured by expression of the immediate-early gene c-fos ( Vann and Albasser , 2009 ) . To determine whether these changes reflect the indirect loss of hippocampal afferents or the disconnection of tegmental pathways , tissue from all four groups of animals was processed immunohistochemically for the expression of c-fos .
A stringent lesion criterion was adopted for final inclusion in the study and , as a consequence , the final groups comprised seven Gudden's ventral tegmental nuclei lesion animals ( VTNx ) , eight mammillothalamic tract lesion animals ( MTTx ) , seven postcommissural fornix lesions ( PCFx ) and eight surgical controls ( Sham ) .
Although the mammillary bodies ( MBs ) have a very longstanding association with memory ( Gudden , 1896 ) , remarkably little is known about their specific contributions . Models of MB function have inevitably focused on their dense hippocampal inputs ( Aggleton and Brown , 1999 ) so that the MBs are principally thought to relay hippocampal information to the anterior thalamic nuclei ( via the mammillothalamic tract ) , and then beyond ( e . g . , Barbizet , 1963; Delay and Brion , 1969 ) . However , the MBs have heavy , reciprocal connections with Gudden's tegmental nuclei whose functional significance has been largely overlooked . The present study contrasted the relative contributions of these two , principal MB inputs for ( i ) spatial memory and ( ii ) control of the expression of the immediate early gene c-fos in distal limbic sites . The second goal arose because mammillothalamic tract lesions cause c-Fos hypoactivity in numerous sites including the hippocampus , retrosplenial cortex , and prelimbic cortex ( Vann and Albasser , 2009 ) . The present study determined whether it is the loss of hippocampal or tegmental inputs that is the main cause of this neural hypoactivity . For these joint reasons , rats with either lesions of the descending postcommissural fornix ( PCFx ) , which carries the hippocampal formation projections , or lesions of Gudden's ventral tegmental nucleus ( VTNx ) were tested alongside mammillothalamic tract lesion rats ( MTTx ) , that is , the major efferent tract by which the MBs have their principal effects ( Vann and Aggleton , 2003 ) . The present study focused on the ventral tegmental nucleus of Gudden as it is reciprocally connected with the medial mammillary nucleus while the dorsal tegmental nucleus connects with the lateral mammillary nucleus . Although the lateral mammillary nucleus forms part of the head-direction network ( e . g . , Vann and Aggleton , 2004; Hopkins , 2005 ) , it is the loss of the medial mammillary system that appears principally responsible for the spatial memory impairments associated with complete lesions of this structure ( Vann , 2005 , 2010 , 2011; Vann and Albasser , 2009 ) . Furthermore , it is the medial mammillary nucleus that is always atrophied in the amnesic Korsakoff's syndrome ( Victor et al . , 1989; Kopelman , 1995 ) . In contrast to the predictions of traditional models of MB function ( e . g . , Delay and Brion , 1969; Gaffan , 1992; Aggleton and Brown , 1999 ) , only the MTTx and VTNx groups were impaired on the behavioral tasks relative to the Sham groups . At no stage did these two lesion groups significantly differ from one another . The only comparison where just one of these two groups was impaired was for Stage 1 of T-maze alternation where the VTNx , but not MTTx , group made significantly more errors than the Sham group . This apparent exception makes sense when the results from Stage 2 are considered . In Stage 1 of the T-maze alternation task both the sample and test runs were in the same maze; in Stage 2 the sample and test runs were in adjacent mazes to prevent the use of intra-maze cues . The performances of the VTNx and MTTx groups in Stage 2 were equivalent , with both groups impaired relative to the Sham group , suggesting the MTTx group had been using intra-maze cues to solve Stage 1 , so masking their spatial memory impairments . The MTTx and VTNx groups were also equivalently impaired on the radial-arm maze task . Although all groups showed an improvement across training , it appears that in the VTNx and MTTx groups this reflected a greater reliance on the use of intra-maze cues given the disruptive effect of rotating the maze midway through the trial . Thus both behavioral tasks indicated that the MTTx and VTNx rats were less able to use allocentric cues and more reliant on intra-maze cues . In contrast , the PCFx animals were not impaired relative to the Shams on any of the behavioral tasks . Consequently , the PCFx animals performed significantly better than both the VTNx and MTTx groups on both the radial-arm maze and T-maze tasks despite producing MB shrinkage ( resulting from loss of fiber inputs ) equivalent to that found after complete fornix lesions . The present PCFx lesions were indistinguishable from those that had previously been demonstrated to completely disconnect the hippocampal formation-MB projections ( Vann et al . , 2011 ) . The pattern of behavioral impairments was paralleled by the c-Fos findings . Both the VTNx and MTTx groups had significantly fewer c-Fos positive cells in the hippocampus , retrosplenial cortex , and prelimbic cortex . Reduced levels of c-Fos have also been found in this same network of brain regions following anterior thalamic nuclei lesions ( Jenkins et al . , 2002 ) . However , unlike the c-Fos reduction following anterior thalamic lesions , the current changes reflect ‘indirect’ effects , as the MTT and VTNg lesions do not result in direct de-afferentation of any of these sites . Although reduced activity in this memory-related network seems to be a consistent marker of damage to the medial diencephalon , in animal models and humans ( Joyce et al . , 1994; Kapur , 1994; Reed et al . , 2003; Savage et al . , 2003; Caulo et al . , 2005; Anzalone et al . , 2010 ) , it is not known whether this reduced activity contributes to mnemonic deficits or whether it simply a symptom of the system no longer working effectively . Evidence for the former account comes learning and memory deficits found after infusions of antisense that block c-Fos activity ( He et al . , 2002; Seoane et al . , 2012; Katche et al . , 2013 ) . The present PCFx lesion findings were remarkably similar to those found previously ( Vann et al . , 2011 ) , for example , they resulted in a 5% reduction in accuracy of T-maze alternation . The lack of effect of PCFx lesions is quite surprising and does raise the question of what does this pathway do . It is possible that the tasks used in the present study did not sufficiently tax this pathway and that greater task demands could result in lesion-induced impairments . In addition , it may be that these projections contribute to memory under normal conditions but the general lack of effects following PCFx lesions are due to redundancies in this pathway , given the direct projections from the hippocampal formation to the anterior thalamic nuclei . While complete fornix lesions disconnect the hippocampal formation from a number of sites , including their projections to both the MBs and anterior thalamic nuclei , selective lesions of the descending PCF will spare all but the MB projections . The hippocampal projections to the MBs and anterior thalamic nuclei arise from different cell populations ( Wright et al . , 2010 ) so it is not clear whether the same information is being conveyed ( Tsanov et al . , 2011 ) and , therefore , how much replication there is in the system . Those examples where fornix lesions appear more disruptive than MB lesions ( Aggleton et al . , 1995; Gaffan et al . , 2001 ) are consistent with the idea that the direct hippocampal-thalamic inputs are supportive , but there are some situations where anterior thalamic or MB lesion effects can be greater than those of fornix lesions ( Tonkiss and Rawlins , 1992; Vann et al . , 2000; Jenkins et al . , 2004; Aggleton et al . , 2009 ) , so highlighting the potential functional importance of additional information streams , for example , from the limbic midbrain . The present findings clearly suggest that the more crucial inputs for MB spatial function are from VTNg rather than the hippocampal formation . Questions immediately arise about the information provided by these tegmental nuclei and how might this relate to spatial memory . VTNg cells fire rhythmically and coherently with hippocampal theta ( Kocsis et al . , 2001 ) , and theta-related cells in the medial mammillary nucleus have been linked to memory function ( Alonso and Llinas , 1992; Kocsis and Vertes , 1994; Bland et al . , 1995; Kirk et al . , 1996 ) . It is also the case that the VTNg have been linked to vigilance states ( Bassant and Poindessous-Jazat , 2001 ) , and so might additionally modulate hippocampal activity . The presumption , based on their dense interconnectivity , is that the VTNg lesion effects are primarily via the MBs . The VTNg does , however , innervate a number of other sites , including the lateral hypothalamus , preoptic area , medial septum , parts of the reticular pontine nucleus , median raphe nuclei , supramammillary nucleus and ventral tegmental area ( Petrovicky , 1973; Leichnetz et al . , 1989; Hayakawa et al . , 1993 ) . These other VTNg efferents appear light ( see Hayakawa and Zyo , 1984 ) , and in some instances not been confirmed ( Barone et al . , 1981; Vertes , 1988 ) , a contribution from their disconnection must be considered , particularly because some of these other sites have hippocampal connections . However , lesions of these alternative sites rarely result in the pattern of spatial memory impairments seen following VTNg lesions ( e . g . , Sarihi et al . , 2000; Pan and McNaughton , 2002 ) . In addition , the lack of Fos changes in the supramammillary and septal nuclei following VTNg lesions provides some evidence that they are not having their effects via these routes , while the striking similarity of the VTNg and MTT lesion effects in the present study provides further support for the VTNg-MB pathway . In rats , the MTT and VTNg lesion effects appear to be selective for spatial working memory and not due to non-specific effects such as loss of arousal or motivation . Both MTTx and VTNx groups were able to acquire a geometric discrimination task at the same rate , and to the same level , as the Sham group . This task may be particularly sensitive to lesions within the head-direction system ( Aggleton et al . , 2009; Vann , 2011 ) . As the MTT lesions in the present study are most likely to disconnect the medial MB efferents , while leaving many lateral MB efferents largely intact ( Vann and Albasser , 2009 ) this sparing provides further support for the dissociation of two memory systems within the MBs ( Allen and Hopkins , 1989; Vann and Aggleton , 2004 ) . Clinical studies show that both the MBs and the MTT are important for human memory , and recollective memory in particular ( Dusoir et al . , 1990; Van der Werf et al . , 2000; Carlesimo et al . , 2007; Tsivilis et al . , 2008; Vann et al . , 2009 ) . The situation regarding the human VTNg is far less clear . Indeed , for a number of years the very existence of this nucleus in humans had been in doubt ( Petrovicky , 1971; Hayakawa and Zyo , 1983 ) . However , it is now apparent that the structure and connections of the VTNg are remarkably similar across species ( Petrovicky , 1973; Irle and Markowitsch , 1982; Hayakawa and Zyo , 1984; Huang et al . , 1992; Saunders et al . , 2012 ) . To date , there is only one reported amnesic patient with restricted damage in the VTN area , but this study is limited in that only CT images were available ( Goldberg et al . , 1981 ) . The current findings support a novel model of limbic brain interactions . They show that the MBs have important roles in memory that can be independent of their hippocampal inputs; instead , the results highlight the VTNg . Not only do MTT and VTNg lesions result in strikingly similar spatial memory impairments but they also both result in widespread hypoactivity across a memory-related network . These findings highlight the importance of looking beyond the hippocampus to larger networks of brain regions that contribute to memory ( Vann and Albasser , 2011 ) .
Subjects were 40 male , pigmented rats ( Dark Agouti strain; Harlan , Bicester , United Kingdom ) weighing between 214 g and 245 g at the time of surgery . Animals were housed in pairs under diurnal light conditions ( 14 hr light/10 hr dark ) and testing was carried out during the light phase . Animals were given free access to water and a large cardboard tube and wooden chew-stick were available in the home-cage throughout . For all behavioral experiments other than the water-maze , where food was available ad libitum , the animals were placed on a food restricted diet where they were still able to gain weight; their weights did not fall below 85% of their equivalent free feeding weight . All experiments were carried out in accordance with UK Animals ( Scientific Procedures ) Act , 1986 and associated guidelines . Prior to surgery , all animals were deeply anesthetized by intraperitoneal injection of sodium pentobarbital ( 60 mg/kg pentobarbital sodium salt; Sigma-Aldrich , United Kingdom ) . The 12 rats receiving Gudden's ventral tegmental nuclei lesions ( VTNx ) were then placed in a stereotaxic headholder ( David Kopf Instruments , Tujunga , CA ) , with the nose bar at 0 . 0; the scalp was cut and retracted to expose the skull . The lesions were made by injecting 0 . 09M N-methyl-D-aspartate ( NMDA; Sigma Chemical Company Ltd , United Kingdom ) , dissolved in phosphate buffer ( pH 7 . 2 ) . Injections ( 0 . 18 μl ) were made in one site per hemisphere using a 1 μl syringe ( Hamilton Bonaduz AG , Switzerland ) . The injection was made over 10 min and the needle was then left in situ for a further 10 min . The stereotaxic ‘arm’ was set at 20° from vertical so that the needle entered the same hemisphere for both injections . The stereotaxic co-ordinates of the lesion placements relative to bregma were anteroposterior ( AP ) −8 . 8 , lateral ( L ) +2 . 6/+3 . 0 and the depth , from top of cortex , was −7 . 2 mm . For the mammillothalamic tract lesions ( MTTx; n = 11 ) and descending postcommissural fornix lesions ( PCFx; n = 12 ) , the nose bar was set at +5 . 0 . These lesions were made by radiofrequency using a thermocouple radiofrequency electrode ( 0 . 7 mm active tip length , 0 . 25 mm diameter; Diros Technology Inc . , Toronto , Canada ) . For both lesion types , the electrode was lowered vertically and the tip temperature was raised to 60°C for 15 s using an OWL Universal RF System URF-3AP lesion maker ( Diros Technology Inc . , Toronto , Canada ) . The stereotaxic coordinates for the MTT lesions were: AP , −1 . 2; L , ±0 . 9 ( both relative to bregma ) ; and the depth , from top of cortex , was −6 . 9 mm . For the PCF lesions the coordinates were: AP , −0 . 2 , L , ± 1 . 2 , and the depth , from top of cortex , was −8 . 2 mm . There were three surgical control rats for each lesion type ( Sham; n = 9 ) and for these surgeries the same procedures were used except the probe/needle was lowered to +1 . 0 mm above the lesion site; the temperature of the probe was not raised nor was any injection made . During surgery , rats were maintained on oxygen and given an analgesic ( Meloxicam; Boehringer Ingelheim , Rhein , Germany ) . At the completion of surgery , the skin was sutured and an antibiotic powder ( Acramide; Dales Pharmaceuticals , Skipton , UK ) was applied topically . Animals also received subcutaneous injections of 5 ml glucose saline . All animals recovered well following surgery . Behavioral testing began four weeks following the completion of surgery . 90 min after completing the final radial-arm maze session in the novel room , rats were deeply anesthetized with sodium pentobarbital ( 60 mg/kg; Euthatal , Rhone Merieux , United Kingdom ) and transcardially perfused with 0 . 1 M phosphate buffer saline ( PBS ) followed by 4% paraformaldehyde in 0 . 1 M PBS ( PFA ) . The brains were removed and postfixed in PFA for 4 hr and then transferred to 25% sucrose overnight at room temperature with rotation . Sections were cut at 40 μm on a freezing microtome in the coronal plane . One series ( one-in-four sections ) was collected in PBS . Sections were processed for c-Fos immunostaining using c-Fos rabbit polyclonal antibody ( 1:3000; SC-52; Santa Cruz Biotechnology , Santa Cruz , CA ) . The methods have been described previously ( Vann and Albasser , 2009 ) . A second , one-in-four series of sections was mounted onto gelatin-coated slides and stained with cresyl violet , a Nissl stain , for histological assessment . A further series ( also one-in-four sections ) was collected in PBS to be processed for either serotonin or acetylcholinesterase staining . Tissue from the VTNx rats and three Sham rats underwent acetylcholinesterase staining using the modified Koelle method as previously described ( Vann , 2009 ) . | The hippocampus is a seahorse-shaped structure in the brain and its role in memory has been recognized since the 1950s . However , much less is known about two small structures called the mammillary bodies that are found near the hippocampus . These bodies are part of the limbic system—a network of brain regions that also includes the hippocampus and the amygdala—and this system is known to be involved in the regulation of emotion and the formation of long-term memories . In 1937 , James Papez injected rabies virus into the hippocampus and , by tracing its movement through the brain , identified a distinct circuit within the limbic system . This circuit , which is today known as Papez’ circuit , consists of projections from the hippocampal formation to the mammillary bodies , and from the mammillary bodies on to another region called the anterior thalamus . From here , projections form a loop via several other regions back to the hippocampus . It is widely thought that the mammillary bodies are required for memory formation due to their role in relaying projections from the hippocampus . However , the mammillary bodies also receive projections from other regions , including Gudden's ventral tegmental nucleus , and it is possible that these could contribute to the role of the mammillary bodies in memory . To distinguish between these possibilities , Seralynne Vann compared the performance of three groups of lesioned rats in tests of spatial short-term memory . The first group had lesions of the hippocampal inputs to the mammillary bodies; the second had lesions of the ventral tegmental inputs to the mammillary bodies; and the third group had lesions of the mammillary body outputs to the thalamus . Vann found that the third group was impaired in the memory tasks , consistent with the idea that outputs sent from the mammillary bodies to the thalamus are required for memory formation . Surprisingly , however , blocking signals sent from the hippocampal formation to the mammillary bodies had little impact on the formation of memories , whereas blocking inputs from Gudden's ventral tegmental nucleus led to significant impairments in memory . By revealing that limbic midbrain inputs to the mammillary bodies have an essential role in the formation of memories , these new results challenge dogma in the field , and highlight the importance of looking beyond the hippocampus when considering memory . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2013 | Dismantling the Papez circuit for memory in rats |
A substantial fraction of the metazoan transcriptome undergoes circadian oscillations in many cells and tissues . Based on the transcription feedback loops important for circadian timekeeping , it is commonly assumed that this mRNA cycling reflects widespread transcriptional regulation . To address this issue , we directly measured the circadian dynamics of mouse liver transcription using Nascent-Seq ( genome-wide sequencing of nascent RNA ) . Although many genes are rhythmically transcribed , many rhythmic mRNAs manifest poor transcriptional rhythms , indicating a prominent contribution of post-transcriptional regulation to circadian mRNA expression . This analysis of rhythmic transcription also showed that the rhythmic DNA binding profile of the transcription factors CLOCK and BMAL1 does not determine the transcriptional phase of most target genes . This likely reflects gene-specific collaborations of CLK:BMAL1 with other transcription factors . These insights from Nascent-Seq indicate that it should have broad applicability to many other gene expression regulatory issues .
Most organisms from bacteria to humans possess circadian rhythms , which generate oscillations in biochemistry , physiology and behavior . The circadian system in eukaryotes is based on cell-autonomous molecular oscillators , which rely on transcriptional feedback loops . In mammals , the transcription factor BMAL1 acts as a dimer with either CLOCK ( CLK ) or Neuronal PAS domain protein 2 ( NPAS2 ) to activate the transcription of many genes , including the transcriptional repressors Period ( Per1 , Per2 and Per3 ) and Cryptochrome ( Cry1 and Cry2 ) . The PERs and CRYs are expressed , post-translationally modified , feedback to inhibit their own transcription and are then rhythmically degraded to lead to a new round of BMAL1:CLK or BMAL1:NPAS2 -mediated transcription ( reviewed in Ko and Takahashi , 2006; Dardente and Cermakian , 2007 ) . This temporal regulation of clock gene transcription cycles with a period of about 24 hr and probably underlies much of circadian biology . Over the past decade , clock gene transcriptional regulation has been described in many species and tissues , where it drives the rhythmic expression of a large fraction of the mRNA population ( up to 10–15% of all mRNAs in a single mammalian tissue; Lowrey and Takahashi , 2004; Vollmers et al . , 2009 ) . Rhythmic mRNA expression has mostly been characterized by analyzing temporal changes of steady-state mRNA levels , using techniques such as microarrays ( e . g . , McDonald and Rosbash , 2001; Panda et al . , 2002; Storch et al . , 2002 ) and more recently high-throughput sequencing ( Hughes et al . , 2012 ) . It is generally assumed that these rhythms in mRNA expression directly result from temporal changes in transcription . There are , however , a few reports indicating that post-transcriptional regulation contributes to rhythmic mRNA expression of several genes , including core clock genes ( reviewed in Kojima et al . , 2011; Staiger and Green , 2011; Staiger and Koster , 2011; Zhang et al . , 2011 ) , but this has never been studied in detail at the genome-wide level . Circadian post-transcriptional regulation may impact rhythmic mRNA expression at many different levels , such as mRNA splicing , stability and translation . For example , post-transcriptional events rhythmically regulate the mRNA half-life of the mammalian clock genes Per2 and Cry1 and the Drosophila clock gene per ( So and Rosbash , 1997; Woo et al . , 2009; Woo et al . , 2010 ) . Moreover , several RNA-binding proteins such as LARK , hnRNP I , hnRNP P , hnRNP Q or the circadian deadenylase NOCTURNIN have been shown to regulate circadian gene expression and/or circadian behavior ( reviewed in Kojima et al . , 2011 ) . These different modes of post-transcriptional regulation are not restricted to circadian biology ( Keene , 2007 ) and have been shown in other systems to regulate cellular mRNA levels independent of transcriptional regulation ( Giege et al . , 2000; Cheadle et al . , 2005 ) . To address the genome-wide contribution of transcriptional and post-transcriptional regulation to mammalian mRNA rhythms , we used Nascent-Seq ( high-throughput sequencing of nascent RNA; Carrillo Oesterreich et al . , 2010; Khodor et al . , 2011 ) to assay global rhythmic transcription in mouse liver . We performed a parallel analysis of rhythmic mRNA expression with RNA-Seq and compared the two sequencing datasets . Although many genes are rhythmically transcribed in the mouse liver ( ∼15% of all detected genes ) , only 42% of these rhythmically transcribed genes show mRNA oscillations . More importantly , about 70% of the genes that exhibit rhythmic mRNA expression do not show transcriptional rhythms , suggesting that post-transcriptional regulation plays a major role in defining the rhythmic mRNA landscape . To assess the contribution of the core molecular clock to genome-wide transcriptional rhythms , we also examined how rhythmic CLK:BMAL1 DNA binding directly affects the transcription of its target genes . Although maximal binding occurs at an apparently uniform phase , the peak transcriptional phases of CLK:BMAL1 target genes are heterogeneous , which indicates a disconnect between CLK:BMAL1 DNA binding and its transcriptional output . The data taken together reveal novel regulatory features of rhythmic gene expression and highlight Nascent-Seq as an important genome-wide assay for the study of gene expression .
To address the regulation of genome-wide transcription , we analyzed mouse liver nascent RNA expression , that is , RNA being transcribed by RNA Polymerase II ( Pol II ) prior to 3′ end formation ( from 12 independent samples in LD , 6 time points per day done twice; see analysis of rhythmic transcription in mouse liver section ) . To this end , nascent RNA was extracted from purified nuclei using the high salt/urea/detergent buffer originally described by Wuarin and Schibler ( 1994 ) . A very similar sequencing strategy was recently applied by Smale , Black and colleagues to macrophage nascent RNA ( Bhatt et al . , 2012 ) . We then prepared illumina libraries with standard protocols for high-throughput sequencing ( Nascent-Seq; Carrillo Oesterreich et al . , 2010; Khodor et al . , 2011 ) . Removal of rRNA was unnecessary as approximately 65–70% of the sequences uniquely mapped to the genome ( Table 1 ) . Seventy six percent of these uniquely mapped sequences map to introns ( Figure 1A ) . This contrasts dramatically with more conventional RNA-Seq; it has minimal intron reads as it assays polyadenylated ( pA ) RNA and therefore predominantly mature ( spliced ) mRNA ( Figure 1A ) . Intronic Nascent-Seq reads are also more abundant in mouse compared to Drosophila ( 76% vs 45% ) , reflecting longer intron size and less efficient mouse co-transcriptional splicing ( Khodor et al . , 2012; Khodor et al . , 2011 ) . Many genes exhibit a 5′ to 3′ gradient in the Nascent-Seq dataset , presumably reflecting nascent RNAs of different lengths attached to elongating Pol II ( Figure 1B ) . In addition , Nascent-Seq signals frequently extend past the polyadenylation site , reflecting RNA not yet cleaved by the cleavage/polyadenylation specificity factor ( CPSF ) and/or RNA molecules still associated with Pol II after cleavage but prior to degradation by the 5′ to 3′ exoribonuclease Xrn2 ( Figure 1C ) . These features are absent from standard RNA-Seq data , and indicate that Nascent-Seq predominantly detects nascent RNA molecules attached to elongating Pol II ( Figure 1B , C ) . 10 . 7554/eLife . 00011 . 003Figure 1 . Genome-wide assay of transcription in the mouse liver using Nascent-Seq . ( A ) : Distribution of high-throughput sequencing signal within introns ( green ) , exons ( blue ) and intergenic regions ( grey ) for Nascent-Seq and RNA-Seq datasets . ( B ) : Visualization of Nascent-Seq and RNA-Seq signal at chr4: 40 , 730 , 000–41 , 002 , 500 . Genes above the scale bar are transcribed from left to right and those below the scale bar are transcribed from right to left . Nascent-Seq signal exhibits increased intron signal and a 5ʹ to 3ʹ gradient signal ( arrow ) . Moreover , differences between Nascent-Seq signal and RNA-Seq signal are observed for many genes ( e . g . , Bag1 and B4galt1 ) . ( C ) : Nascent-Seq signal ( brown ) , but not RNA-Seq signal ( red ) , extends past the annotated 3ʹend of the genes B4galt1 and Nfx1 . ( D ) : Gene ontology of genes with high Nascent-Seq and low RNA-Seq signals ( and inversely ) is indicative of RNA with short or long half-lives , respectively ( see ‘Materials and methods’ for details ) . ( E ) : Distribution of the Nascent-Seq/RNA-Seq signal ratio for the classes of genes enriched in ( D ) . ( F ) : Nascent-Seq/RNA-Seq signal ratio significantly correlates with mRNA half-lives ( values from Sharova et al . , 2009 ) , and genes with high ratio display shorter half-lives and inversely . ( G ) and ( H ) : Strategy used to determine the gene signal cut-off threshold used in our analysis . Variation of gene signal coming from the sequencing of a Nascent-Seq library ( G; ZT8 , replicate 1 ) sequenced in two Illumina flow-cell lanes was assessed by calculating the z-score ( H ) . Less than 5% of the genes with a read per base pair superior to three exhibit a 1 . 3-fold gene signal variation . See ‘Materials and methods’ for more details . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 003 Another feature was apparent in the comparison of Nascent-Seq and RNA-Seq: gene expression was often different between the two datasets . Some genes have a high ratio of Nascent-Seq to RNA-Seq signal ( e . g . , B4galt1 , Figure 1B ) , whereas others have a low ratio ( e . g . , Bag1 , Figure 1B ) . Genes with a high ratio ( top 10% of all genes ) are dramatically enriched for specific functions , namely , non-coding RNA ( ncRNA ) , G-coupled protein receptors , regulation of transcription and chromatin organization ( Figure 1D , E ) . Genes with a low ratio are enriched for genes involved in ribosome function and mitochondrial respiration ( Figure 1D , E ) . Because these genes are associated with short or long mRNA half-lives , respectively , we compared the Nascent-Seq to RNA-Seq ratios of genes with their published mRNA half-lives ( Sharova et al . , 2009 ) . Not surprisingly , genes with a high ratio have relatively short half-lives , and genes with low ratios have longer mRNA half-lives ( Figure 1F ) . These data indicates that mRNA stability contributes to the wide range of Nascent-Seq to RNA-Seq ratios . To identify genes that are rhythmically transcribed , we performed two independent six time-points rhythms of mouse liver Nascent-Seq . We found that some genes exhibit very high amplitude rhythms , with no detectable signal at low time points ( e . g . , Npas2; Figure 2A ) . About 15% of expressed genes manifest transcriptional rhythms ( p<0 . 05: 6 . 3% , strong rhythms; 8 . 9% medium-strength rhythms; see ‘Materials and methods’ for analysis details; Figure 2B ) . Phases of maximal transcription are heterogeneous yet not uniformly distributed; very few genes peak at ZT16-20 in the mid-late night ( Figure 2C , D ) . 10 . 7554/eLife . 00011 . 004Figure 2 . Genome-wide analysis of rhythmic transcription in the mouse liver . ( A ) : Visualization of Npas2 Nascent-Seq signal at six time points of the light:dark cycle ( first replicate ) . Npas2 Nascent-Seq signal is rhythmic and peaks at ZT20-ZT0 , contrary to the signal within the adjacent gene Rpl31 . ( B ) : Quantification of the number of genes that are rhythmically transcribed in the mouse liver . Genes with more than three reads per base pair for at least one time point were included for the analysis . Genes are considered to be rhythmically transcribed if signal amplitude ( Amp ) is greater than 1 . 5 , if signals for the 12 time points follow a sinusoid curve ( F24 > 0 . 45 ) and if the F24 value is in the top 5% of all F24 values calculated after time points were permutated 10 , 000 times ( p<0 . 05 ) . A rhythm was considered to be strong ( dark red ) if F24 > 0 . 6 and Ampl > 1 . 75 . ( C ) : Heatmap representation of Nascent-Seq signal for the 963 genes that are rhythmically transcribed in the mouse liver . High expression is displayed in yellow ( z-score > 1 ) , low expression in blue ( z-score < 1 ) . ( D ) : Expression phase of rhythmically expressed nascent RNA ( n = 936 ) was separated by bins of 2 hr . Analysis of their distribution reveals that fewer genes are transcribed at ZT16-20 . ( E ) and ( F ) : Rhythmic Nascent-Seq signal was detected for many precursors of non-coding RNAs such as pri-miRNA ( d , pri-miR122a ) and long non-coding RNA ( e , lin-ncRNAs BC019819 , AK157581 , BC049268 , BC056646 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 00410 . 7554/eLife . 00011 . 005Figure 2—source data 1 . Gene expression values for all UCSC genes from our mouse liver Nascent-Seq datasetDOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 00510 . 7554/eLife . 00011 . 006Figure 2—figure supplement 1 . Rhythmic transcription of lncRNA ENSMUSG00000098984 in the mouse liver . Visualization of Nascent-Seq signal ( brown; six time points of replicate 1 ) for the long non-coding RNA ( lncRNA ) precursor ENSMUSG00000098984 . Genes above the scale bar are transcribed from left to right and those below the scale bar are transcribed from right to left . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 00610 . 7554/eLife . 00011 . 007Figure 2—figure supplement 2 . Rhythmic transcription of lncRNA ENSMUSG00000086813 in the mouse liver . Visualization of Nascent-Seq signal ( brown; six time points of replicate 1 ) for the long non-coding RNA ( lncRNA ) precursor ENSMUSG00000086813 . Genes above the scale bar are transcribed from left to right and those below the scale bar are transcribed from right to left . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 00710 . 7554/eLife . 00011 . 008Figure 2—figure supplement 3 . Rhythmic transcription of lncRNA ENSMUSG00000086771 in the mouse liver . Visualization of Nascent-Seq signal ( brown; six time points of replicate 1 ) for the long non-coding RNA ( lncRNA ) precursor ENSMUSG00000086771 . Genes above the scale bar are transcribed from left to right and those below the scale bar are transcribed from right to left . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 00810 . 7554/eLife . 00011 . 009Figure 2—figure supplement 4 . Rhythmic transcription of pri-miRNA pri-Mir17hg in the mouse liver . Visualization of Nascent-Seq signal ( brown; six time points of replicate 1 ) for the pri-miRNA pri-Mir17hg . Enlargement of pri-miRNA signal reveals that pri-miRNA transcription units are not well annotated , precluding a rigorous quantification of the signal Genes above the scale bar are transcribed from left to right and those below the scale bar are transcribed from right to left . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 00910 . 7554/eLife . 00011 . 010Figure 2—figure supplement 5 . Rhythmic transcription of pri-miRNA ENSMUSG00000077856 in the mouse liver . Visualization of Nascent-Seq signal ( brown; six time points of replicate 1 ) for the pri-miRNA ENSMUSG00000077856 . Enlargement of pri-miRNA signal reveals that pri-miRNA transcription units are not well annotated , precluding a rigorous quantification of the signal . Genes above the scale bar are transcribed from left to right and those below the scale bar are transcribed from right to left . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 01010 . 7554/eLife . 00011 . 011Figure 2—figure supplement 6 . Rhythmic transcription of pri-miRNA ENSMUSG00000093077 in the mouse liver . Visualization of Nascent-Seq signal ( brown; six time points of replicate 1 ) for the pri-miRNA ENSMUSG00000093077 . Enlargement of pri-miRNA signal reveals that pri-miRNA transcription units are not well annotated , precluding a rigorous quantification of the signal . Genes above the scale bar are transcribed from left to right and those below the scale bar are transcribed from right to left . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 011 Protein-coding genes dominate the cycling Nascent-Seq dataset ( >85% , data not shown ) , but it also contains genes encoding ncRNAs , for example , pri-miRNAs and long non-coding RNAs ( lncRNA ) ( Figure 2E , F; Figure 2—figure supplement 1–6 ) . As previously described ( Gatfield et al . , 2009 ) , pri-miRNA 122a is robustly rhythmic ( Figure 2E ) . However , rigorous quantitation of rhythmic ncRNA transcription is precluded by the poor annotation of these transcription units . To address the relationship of cycling transcription to cycling mRNAs , we assayed two independent six time-points profiles of pA RNA by RNA-Seq . The comparison with Nascent-Seq was restricted to genes that were sufficiently expressed in both datasets ( n = 5454 genes; see ‘Materials and methods’ for details ) . Using the identical statistical analysis and cut-offs to determine rhythmicity , the fraction of rhythmic mRNA was higher than the fraction of rhythmic nascent RNA ( 22 . 1% and 15 . 1% , respectively; Figure 3A ) . 10 . 7554/eLife . 00011 . 012Figure 3 . Post-transcriptional events account for a significant fraction of rhythmic gene expression in the mouse liver . ( A ) : Rhythmic gene expression was assessed as in Figure 2B for genes sufficiently expressed in both Nascent-Seq and RNA-Seq datasets . Four categories of rhythmically expressed genes were determined by comparing the Nascent-Seq and RNA-Seq datasets: rhythmic nascent RNA and mRNA ( R-R ) , rhythmic nascent RNA only ( R-AR ) , rhythmic mRNA only ( AR-R ) and arrhythmic nascent RNA and mRNA ( AR-AR ) . ( B ) : Heatmap representation of genes with rhythmic nascent RNA and mRNA expression ( n = 342 ) . Classification is based on the phase of nascent RNA oscillations , and each lane corresponds to one gene . ( C ) : Double-plotted phase distribution of rhythmic nascent RNA expression ( brown ) and rhythmic mRNA expression ( red ) for genes of the R-R gene set . Both phases are highly correlated ( r = 0 . 92 ) . ( D ) : Distribution of the difference between the phase of mRNA expression rhythm and the phase of nascent RNA expression rhythm for the 342 R-R genes . ( E ) : Amplitude of mRNA expression rhythms are correlated with nascent RNA expression rhythms ( r = 0 . 76 ) . ( F ) and ( G ) : Similar representation to ( B ) for rhythmically transcribed genes with no mRNA expression rhythms ( C , n = 480 ) , and genes that exhibit mRNA oscillations but no rhythms of transcription ( D , n = 862 ) . For all three heatmaps , high expression is displayed in yellow ( z-score > 1 ) , low expression in blue ( z-score < 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 01210 . 7554/eLife . 00011 . 013Figure 3—source data 1 . Gene expression values from our Nascent-Seq and RNA-Seq datasetDOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 01310 . 7554/eLife . 00011 . 014Figure 3—source data 2 . Gene expression values for all UCSC genes from our mouse liver RNA-Seq datasetDOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 014 There was also a notably poor overlap between the two rhythmic gene sets: only 41 . 6% of rhythmically transcribed genes also manifest rhythmic mRNA expression ( R-R gene set; 342/822; Figure 3A–C ) . However , the mRNA phase of these R-R genes was highly correlated to the nascent RNA phase ( r = 0 . 92; Figure 3C ) , and more than half ( 57% ) of these genes exhibit a phase difference of less than 2 hr ( 195/342; Figure 3D ) . The amplitudes of the nascent RNA rhythms were also correlated with those of mRNA ( r = 0 . 76; Figure 3E ) , indicating that transcriptional regulation dominates these R-R rhythms . Not surprisingly , almost all clock genes ( Figure 4 ) and well-characterized clock-controlled genes ( e . g . , Nocturnin , Por , Alas1 , Upp2 , Usp2 , Inmt , Nfil3 , etc; Figure 3—source data 1 ) are in this R-R gene set . 10 . 7554/eLife . 00011 . 015Figure 4 . Clock genes nascent RNA and mRNA expression in the mouse liver . Clock genes nascent RNA levels ( brown; time points every 4 hr starting at ZT0 ) and mRNA levels ( red; time points every 4 hr starting at ZT2 ) from the Nascent-Seq and RNA-Seq datasets . Relative levels between nascent RNA and mRNA expression profiles are identical for all genes to allow direct comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 015 The other 58 . 4% of rhythmically transcribed genes ( 480/822 ) do not show robust mRNA expression ( R-AR gene set; Figure 3A , F ) . A simple explanation is that the mRNA half-lives of these R-AR genes are relatively long and therefore mask the transcriptional oscillations . However , these genes do not have altered nascent RNA to mRNA ratios compared to the whole genome ( Figure 5A ) or reported longer mRNA half-lives ( assessed using the dataset from Sharova et al . , 2009; Figure 5B ) . These considerations suggest that other mechanisms account for the poor mRNA oscillations of this gene set ( e . g . , the rhythmic Nascent-Seq signal of 25 R-AR genes results from rhythmic transcription of an adjacent gene that reads into the R-AR gene; Figure 5C ) . 10 . 7554/eLife . 00011 . 016Figure 5 . Analysis of the different classes of rhythmically expressed genes in the mouse liver . ( A ) : Nascent-Seq/RNA-Seq signal ratio ( used as inferred half-life ) is similar for the four categories of rhythmically expressed genes: rhythmic nascent RNA and mRNA ( R-R ) , rhythmic nascent RNA only ( R-AR ) , rhythmic mRNA only ( AR-R ) and arrhythmic nascent RNA and mRNA ( AR-AR ) . ( B ) : Similar as ( A ) , using the RNA half-life values from Sharova et al . , 2009 . ( C ) : Nascent-Seq rhythms of 25 of the 480 R-AR genes can be attributed to the rhythmic transcription of an adjacent gene . This applies to Sphk2 Nascent-Seq rhythm , which likely results from rhythmic Dbp nascent RNA signal that extend the 3ʹend of Dbp gene and read through Sphk2 . Genes above the scale bar are transcribed from left to right and those below the scale bar are transcribed from right to left . ( D ) : Gene ontology of three categories of rhythmically expressed genes: rhythmic nascent RNA and mRNA ( R-R ) , rhythmic nascent RNA only ( R-AR ) , rhythmic mRNA only ( AR-R ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 016 The opposite comparison is based on genes with rhythmic mRNA expression , of which only 28 . 4% ( 342/1204 ) have rhythmic transcription that meets the cycling criteria ( Figure 3A , G ) . This surprising conclusion was similar when the analysis was restricted to genes with the strongest mRNA rhythms ( 121/435 ) and indicates that most cycling mRNAs ( 862/1204 ) likely undergo post-transcriptional regulation . This might include the circadian regulation of nuclear RNA processing , export , translational regulation and/or mRNA turnover , as described for the few circadian genes shown to be regulated post-transcriptionally ( Kojima et al . , 2011; Staiger and Koster , 2011; Zhang et al . , 2011 ) . Gene ontology analysis of this arrhythmic transcription-rhythmic mRNA ( AR-R ) gene set did not reveal any striking enrichment of particular gene functions ( Figure 5D ) . There is a peculiar feature of the large number of genes within this AR-R category: visualization of RNA expression with heatmap indicates that many of these transcriptionally ‘arrhythmic’ genes manifest elevated transcription at times that match their cycling mRNA peaks ( Figure 3G; note that the heatmaps show that the transcription peak of many genes matches the peak phase of mRNA cycling ) . Further inspection of all individual expression profiles confirmed this correlation: many genes have matching peak phases despite large variations in nascent RNA expression ( Figure 6A ) . 10 . 7554/eLife . 00011 . 017Figure 6 . Transcriptional variability of AR-R genes contributes to rhythmic mRNA expression . ( A ) and ( B ) : Nascent RNA levels ( brown; time points every 4 hr starting at ZT0 ) and mRNA levels ( red; time points every 4 hr starting at ZT2 ) from the Nascent-Seq and RNA-Seq datasets for six genes of the AR-R gene set . While the majority of the AR-R genes exhibit variable nascent RNA expression ( A ) , some of them exhibit a relatively constant transcription when compared to mRNA expression ( B ) . ( C ) : Standard deviation ( SD; calculated using the 12 time points and normalized to the mean ) of nascent RNA expression is higher than the SD normalized to the mean of mRNA levels for most AR-R genes . ( D ) and ( E ) : Higher transcriptional variability ( SD ) of arrhythmically transcribed genes is associated with higher occurrence of rhythmic mRNA expression ( D ) , but not to nascent RNA expression levels ( E ) . ( F ) : Higher variability of transcription for the genes of the AR-R group is associated with increase amplitude of rhythms at both Nascent RNA ( brown ) and mRNA ( red ) level . Genes of the AR-R group ( n = 862 ) were binned into five quintiles of equal size ( q1–q5 ) . ( G ) : Heatmap representation of 86 AR-R genes that exhibit high level of transcription at only one time point , and with rhythmic mRNA expression . High expression is displayed in yellow ( z-score > 1 ) , low expression in blue ( z-score < 1 ) . ( H ) : Nascent RNA levels ( brown ) and mRNA levels ( red ) for four AR-AR genes with variable nascent RNA expression that is not associated to rhythmic mRNA expression . ( I ) : Number of predicted miRNA target sites of AR-R genes with high transcriptional variability ( q1 , top 20% of the 826 AR-R genes ) and low transcriptional variability ( q5 , bottom 20% ) . ( J ) : Gene ontology of AR-R genes with high transcriptional variability ( top 25% ) when compared to all AR-R genes . Significant enrichment ( top ) and depletion ( bottom ) of biological functions for these genes are displayed . Values correspond to the number of genes within this top 25% of genes , when compared to all AR-R genes . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 01710 . 7554/eLife . 00011 . 018Figure 6—source data 1 . Peak coordinates for CLK:BMAL1 , BMAL1 only and CLK only DNA binding sitesDOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 018 To substantiate this variability in nascent RNA expression , we calculated the standard deviation normalized to the mean ( SD ) for every gene using the two 6 time-points profiles as 12 independent samples . The reasoning was that low variability expression between time points should result in a low SD . The SD was noticeably higher in the Nascent-Seq dataset than in the RNA-Seq dataset , indicating that the transcription of most AR-R genes is indeed variable compared to the mRNA-Seq dataset ( Figure 6C; note that most points are to the right of the diagonal line ) . Higher variability of nascent RNA expression is also observed for other categories , including the AR-AR set ( data not shown ) , suggesting that it is a common feature of transcription vs mRNA comparisons independent of circadian considerations . Yet a higher variability of transcription ( higher SD ) correlates with rhythmic mRNA expression within all arrhythmically transcribed genes ( n = 4632 ) , suggesting that transcriptional variability generally contributes to the generation of rhythmic mRNA expression ( Figure 6D ) . This relationship is independent of nascent RNA expression levels , indicating that the correlation is not due to sequencing depth ( Figure 6E ) . More variable transcription is also associated with higher amplitudes of rhythmic mRNA expression ( Figure 6F ) . This correlation is valid for most AR-R genes ( ∼80% , Figure 6F ) , suggesting that this transcriptional variability or noise has a significant role in the emergence of rhythmic mRNA expression from arrhythmic transcription . A comparison between the two replicates of rhythmic mRNAs indicates a better overlap than between one replicate and one replicate of rhythmic nascent RNAs . In contrast , the overlap between the two replicates of rhythmic nascent RNAs was no better than a single replicate rhythmic nascent RNA-rhythmic mRNA comparison ( data not shown ) . Although the two comparisons cannot be definitive because of the limited six time point temporal resolution and resultant noise , they also support more pronounced variation at the transcriptional level than at the mRNA level . Manual inspection of AR-R genes with high variability revealed a set of genes with high transcription at only one time point ( 10% of the AR-R gene set; 86/862 ) . This was observed in both replicates and also correlates with rhythmic mRNA expression ( Figure 6G ) . mRNA stability ( half-life ) regulation may contribute to the generation of rhythmic mRNA expression from what is likely a short burst of transcription . This ability of post-transcriptional regulation to generate rhythmic mRNA oscillations is selective , as not all arrhythmically transcribed genes with variable transcription exhibit rhythmic mRNA expression ( Figure 6H ) . Moreover , not all genes exhibit variable transcriptional profiles ( Figure 6B , see below ) . About 20% of the AR-R genes are exceptions and exhibit higher variability at the mRNA than at the nascent level ( Figure 6B , F ) . Because there is evidence that miRNAs can regulate mRNA levels independently of transcription , we examined whether those genes could be preferentially linked to miRNA regulation . Indeed , a higher number of predicted miRNA target sites was found for these genes compared to genes with higher transcriptional variability ( using MirTarget2; p<0 . 01 ) , suggesting that miRNAs contribute to mRNA cycling of genes with low transcriptional variability ( Figure 6I ) . The separation of these AR-R genes between genes with high or low transcriptional variability is therefore likely to be linked to different modes of gene expression regulation . Interestingly , this separation also reflects distinct biological functions , as AR-R genes with highly variable transcription were enriched for responsive genes ( GO: response to stimulus ) and metabolic genes ( Figure 6J ) , indicating that their intrinsic transcriptional responsiveness is linked to their variable transcriptional profiles . Genes within the R-R gene set include clock genes and many well-characterized clock controlled genes ( see above ) . Because a large fraction of them are directly targeted by the core clock , we asked how CLK:BMAL1 regulates the transcription of its target genes at the genome-wide level . We also took advantage of our Nascent-Seq dataset to establish whether the phase differences between rhythmic CLK:BMAL1 DNA binding and rhythmic target gene mRNA expression ( Rey et al . , 2011 ) reflect transcriptional or post-transcriptional regulation . To this end , we first performed a ChIP-Seq analysis of CLK and BMAL1 at a time of high DNA binding ( ZT8 ) . As expected , CLK and BMAL1 target many DNA binding sites in mouse liver ( 759 and 1579 , respectively ) and significantly overlap on 211 of these peaks ( Figure 7A ) . Although highly significant ( chi-square test , p<0 . 0001 ) , the rather low fraction may indicate competition between CLK:BMAL1 and NPAS2:BMAL1 for binding sites . Importantly , about 90% of these 211 peaks have been previously characterized as rhythmic BMAL1 DNA binding sites in mouse liver ( Rey et al . , 2011 ) . 10 . 7554/eLife . 00011 . 019Figure 7 . Characterization of CLK and BMAL1 target genes in the mouse liver . ( A ) and ( B ) : Visualization ( A ) and quantification ( B ) of BMAL1 ChIP-Seq , CLK ChIP-Seq and input signal at BMAL1 and CLK significant peaks ( analysis using MACS algorithm ) . BMAL1 ChIP-Seq , CLK ChIP-Seq and Input signals were retrieved based on the location of the BMAL1 peaks ( center ± 1kb , for CLK:BMAL1 peaks and BMAL1 only peaks ) or the CLK peaks ( center ± 1kb , for CLK only peaks ) . Normalization was performed on the entire datasets by calculating the z-score ( ( x − mean ) /SD ) . Heatmap displays high expression in red and low expression in blue . Quantification ( B ) was performed by averaging the z-score by bins of 25 bp for all CLK:BMAL1 peaks ( n = 211 ) , BMAL1 only peaks ( n = 1368 ) and CLK only peaks ( n = 548 ) . ( C ) : Enrichment of e-boxes ( perfect CACGTG in red , degenerated e-boxes [one nucleotide mismatch , in orange] ) within ±500 bp of CLK:BMAL1 , BMAL1 only and CLK only peak centers . ( D ) : Motifs enriched within CLK:BMAL1 peaks , BMAL1 only peaks and CLK only peaks , as revealed by MEME analysis . ( E ) – ( H ) : Visualization of BMAL1 ChIP-Seq ( blue ) , CLK ChIP-Seq ( green ) and Nascent-Seq ( brown; six time points of replicate 1 ) signals for Rev-Erbα ( E ) , Per1 ( F ) , Cry1 ( G ) and a cluster of 4 lncRNA ( AK079377 , AK007907 , AK036974 , AK087624 ) ( H ) targeted by CLK:BMAL1 . Genes above the scale bar are transcribed from left to right and those below the scale bar are transcribed from right to left . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 019 CLK and BMAL1 sites overlap at their peak center , consistent with binding to DNA as a heterodimer , and they are enriched for the canonical consensus sequence CACGTG ( Figure 7B , C ) The data therefore indicate that almost all of the 184 direct target genes identified by the 211 CLK:BMAL1 DNA binding sites ( Figure 6—source data 1 ) are bona fide direct target genes . They include the expected core clock genes ( Figure 7E–G ) as well as other interesting targets . There are for example 12 ncRNA genes , which include a cluster of four rhythmically transcribed ncRNAs ( Figure 7H ) . These cycling ncRNAs suggest novel mechanisms by which CLK:BMAL1 impact circadian rhythms . Although CLK:BMAL1 target genes are significantly enriched for rhythmically transcribed genes ( chi-square test , p<0 . 0001; Figure 8A , C ) , there is a large discrepancy between the phases of rhythmic BMAL1 DNA binding and those of rhythmic transcription . This is because BMAL1 binding is essentially uniform at ZT3-5 , whereas the transcription peaks are much more broadly distributed ( Figure 8A , B ) . Remarkably , this is also true for the core clock genes . Whereas a small number have the expected phase similar to that of CLK:BMAL1 DNA binding ( Rev-Erbα , Dbp; Figure 7E ) , the transcription of most target genes is significantly delayed . They include Per1 ( Figure 7F ) , Cry1 ( Figure 7G ) and Per2 ( 8E ) . As recently proposed for Cry1 ( Ukai-Tadenuma et al . , 2011 ) , this general delay in the peak of rhythmic transcription may be due to a collaboration of CLK:BMAL1 with other transcription factors ( Rev-erbα and Dbp in the case of Cry 1 ) . 10 . 7554/eLife . 00011 . 020Figure 8 . Disconnect between rhythmic BMAL1 DNA binding and its transcriptional output . ( A ) : Heatmaps representing BMAL1 ChIP-Seq signal ( from Rey et al . , 2011 ) , Nascent-Seq and RNA-Seq signal for CLK:BMAL1 target genes ( six time points in duplicate ) . Genes were classified in four categories: rhythmic nascent RNA and mRNA ( R-R ) , rhythmic nascent RNA only ( R-AR ) , rhythmic mRNA only ( AR-R ) and arrhythmic nascent RNA and mRNA ( AR-AR ) . High expression is displayed in yellow , low expression in blue . ( B ) : Peak phase distribution of rhythmic BMAL1 DNA binding ( blue , from Rey et al . , 2011 ) , of nascent RNA ( black ) and of mRNA ( red ) for the direct target genes that are rhythmically expressed at both the nascent RNA and mRNA levels . ( C ) : Distribution of CLK:BMAL1 target genes within the 4 different classes of rhythmically expressed genes and its comparison to the genome-wide distribution . Rhythmic nascent RNA and mRNA: R-R; rhythmic nascent RNA only: R-AR; rhythmic mRNA only: AR-R; arrhythmic nascent RNA and mRNA: AR-AR . ( D ) : qPCR quantification of Rev-Erbα , Per1 , Per2 and Cry1 pre-mRNA every 4 hr throughout the day in wild-type ( black , n = 4 per time points ) and Bmal1−/− mice ( blue , n = 3 per time points ) . Error bar: s . e . m . ( E ) : Visualization of BMAL1 ChIP-Seq ( blue ) , CLK ChIP-Seq ( green ) , Nascent-Seq ( brown; six time points of replicate 1 ) , Pol II ChIP-Seq signal ( purple ) at ZT10 and ZT22 ( from Feng et al . , 2011 ) and strand-specific Nascent-Seq signal for Per2 ( plus strand , top; minus strand , bottom ) . Per2 is rhythmically transcribed ( minus strand ) with a peak at ZT16 . An antisense transcript is rhythmically transcribed to Per2 RNA ( plus strand ) , peaking at ZT4 . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 020 To validate this interpretation , we assayed the pre-mRNA levels in Bmal1−/− mice of four clock genes that exhibit different phases of transcription in wild-type mice ( Figure 8D ) . Only the levels of Rev-Erbα pre-mRNA in the mutant mice conform to expectation and are at trough levels of wild-type mice . The levels of Per1 , Per2 and Cry1 pre-mRNAs in Bmal1−/− mice are higher than the trough of expression in wild-type mice , indicating that other transcriptional regulators are indeed important for the transcription of these three clock genes ( Figure 8D ) . This notion is also supported by the enrichment of other transcription factor motifs such as HNF3/FOXA1 , SP1 and E4BP4 adjacent to CLK:BMAL1 binding sites ( Figure 7D ) . Notably , Per2 has a prominent anti-sense transcript at times of low Per2 sense transcription ( Figure 8E ) . Moreover , the 5′ end of this transcript coincides with peaks of Pol II ( Figure 8E ) . This suggests that antisense transcription could be an additional mechanism responsible for the disconnect between the phase of CLK:BMAL1 DNA binding and the phase of rhythmic transcription .
Rhythmic mRNA expression is a hallmark of circadian biology and commonly assumed to be a consequence of rhythmic transcription . However , the application here of Nascent-Seq to genome-wide mouse liver transcriptional rhythms indicates that about 70% of the genes that exhibit rhythmic mRNA expression do not have robust transcriptional rhythms ( AR-R category ) , suggesting that post-transcriptional mechanisms are important for the generation of robust mRNA rhythms . Yet the transcription of most AR-R genes is variable , with elevated levels coinciding with the peak of the rhythmic mRNA profile . This suggests that post-transcriptional events buffer variable transcriptional output to generate robust and reproducible rhythms of mRNA expression ( Figure 9 ) . A similar Nascent-Seq vs RNA-Seq strategy for Drosophila head RNA ( Joe Rodriguez and Michael Rosbash , personal communication ) and a very recently published paper based on a different strategy for assessing liver transcriptional rhythms ( Koike et al . , 2012 ) come to a generally similar conclusion , namely , a widespread contribution of post-transcriptional regulation to circadian mRNA cycling . 10 . 7554/eLife . 00011 . 021Figure 9 . Post-transcriptional events contribute to rhythmic mRNA expression in the mouse liver . Although rhythmic transcription plays a major role for approximately 30% of the genes that exhibit rhythmic mRNA expression , post-transcriptional events significantly contribute to the generation of mRNA rhythms for the majority of genes ( ∼70% ) . Many post-transcriptional cyclers exhibit highly variable transcription that is buffered to generate robust rhythmic mRNA expression . Few genes exhibit a relatively constant transcription when compared to mRNA expression . These post-transcriptional events may include roles for RNA binding proteins and miRNAs to regulate RNA stability , 3′ end formation and nuclei export . DOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 021 Relevant mechanisms likely include RNA stability as well as other RNA processing events , as suggested in other systems such as the regulation of gene expression in Arabidopsis mitochondria ( Giege et al . , 2000 ) and after T-cell activation ( Cheadle et al . , 2005 ) . Because the half-lives of nascent transcripts are generally much shorter than those of mRNA ( Griffiths-Jones , 2007; Mattick , 2009; Mercer et al . , 2009; Wang and Chang , 2011 ) , a short burst of transcription can result in elevated mRNA expression that lasts several hours , as has been shown in systems involving an acute inflammatory response ( Cheadle et al . , 2005; Hao and Baltimore , 2009 ) . However , generic mRNA stability cannot account for this buffering , as many genes with arrhythmic but variable transcription do not exhibit rhythmic mRNA expression ( Figure 6H ) . This suggests that the post-transcriptional buffering is clock-controlled and selective for specific genes . Because our experiments were done under LD conditions , it is possible that some cycling RNAs and mechanisms are not circadian but driven by the LD cycle . Nonetheless , it is likely that many of them also occur under DD conditions and that specific and perhaps multiple post-transcriptional mechanisms contribute to rhythmic mRNA expression . They may include 3′ end formation and coupled polyadenylation , splicing , mRNA export as well as cytoplasmic events involving translation , RNA binding proteins ( RBP ) and miRNAs ( Joshi et al . , 2012 ) . Interestingly , a few recent reports highlight the tight coupling between transcriptional regulation and post-transcriptional events that govern mRNA stability ( Bregman et al . , 2011; Trcek et al . , 2011 ) . In these examples , RBPs are recruited by specific transcription factors , which then help load the RBPs onto nascent RNA; they then control cytoplasmic mRNA stability ( Bregman et al . , 2011; Trcek et al . , 2011 ) . A mechanism of this nature could account for the post-transcriptional generation of rhythmic mRNA expression . The AR-R genes with high transcriptional variability are enriched for metabolic functions as well as those involved in ‘response to stimulus’ . The transcription of many metabolic genes is regulated by metabolites and/or hormones ( e . g . , transcription of Sds is induced by glucagon and CREB , Haas and Pitot , 1999 ) . A large fraction of rhythmic mRNAs may therefore result from a transcriptional response , dependent on the cellular environment , as well as post-transcriptional events that stabilize mRNA at an appropriate time of the day . This scenario could also explain the R-AR gene set: despite rhythmic transcription , the lack of rhythmic post-transcriptional regulation would negate the transcriptional oscillations . Importantly , many genes with rhythmic mRNA expression also exhibit robust transcriptional rhythms . They include all well-known clock genes and many well-characterized clock-controlled genes ( see above ) . Their transcriptional profiles suggest that they are under more stringent transcriptional control ( Figure 9 ) , due perhaps to direct regulation by the core clock in combination with additional transcription factors . Our genome-wide characterization of rhythmic transcription also allowed us to directly assay how the rhythmic binding of CLK:BMAL1 to its target gene promoters correlates with transcription . The transcriptional phase of these target genes is heterogeneous and distributed throughout the day , despite a more discrete phase of BMAL1 DNA binding at the beginning of the light phase . This indicates that transcriptional output is not identical for all target genes and suggests that CLK:BMAL1 cooperates with other transcription factors to establish the phase of transcription , as previously shown only for the Cry1 gene ( Ukai-Tadenuma et al . , 2011 ) . In addition , transcription of most of these target genes is arrhythmic but not absent without BMAL1 . This feature of target gene expression as well as the heterogeneity of phase is unlike what is observed in flies: core CLK:CYC target genes exhibit a discrete phase of expression that matches the phase of DNA binding ( Abruzzi et al . , 2011 ) . In summary , the application of Nascent-Seq and RNA-Seq to mammalian circadian gene expression regulation challenges two assumptions of the mammalian circadian field . The first is that rhythmic transcriptional regulation is sufficient to describe the cycling gene expression landscape . The second is that CLK:BMAL1 DNA binding alone sets the phase of , and is essential for , core clock gene transcription . The dramatic , genome-wide disconnect between the phases of rhythmic CLK:BMAL1 DNA binding and rhythmic target gene transcription suggests that other transcription factors and/or mechanisms collaborate with CLK:BMAL1 binding and are critical to determine the phase of clock gene transcription . We anticipate that Nascent-Seq will impact gene expression regulation far beyond the circadian applications shown here .
3- to 6-month-old male mice housed in a 12 hr-light:12 hr-darkness ( LD12:12 ) schedule were used . Wild-type mice ( C57BL/6 strain ) and Bmal1−/− mice ( originally from Christopher A Bradfield; Bunger et al . , 2000 ) were used . All experiments were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Brandeis Institutional Animal Care and Use Committee ( IACUC protocol #0809-03 ) . Mice housed in LD12:12 were sacrificed every 4 hr ( ZT0 , 4 , 8 , 12 , 16 and 20 ) by isoflurane anesthesia followed by decapitation . Mouse liver was then quickly removed and homogenized in 3 . 5 ml of 1× PBS and 3 . 5 ml of homogenization buffer ( 2 . 2 M sucrose , 10 mM Hepes pH 7 . 6 , 15 mM KCl , 2 mM EDTA , 1× protease inhibitor cocktail [Roche , Basel , Switzerland] , 0 . 15 mM spermine , 0 . 5 M spermidine , 0 . 5 mM DTT ) with a dounce homogenizer ( six strokes loose pestle , four strokes tight pestle ) . The liver homogenate was then mixed with 21 . 5 ml of homogenization solution and layered on the top of a 10 ml ice-cold cushion solution ( 2 . 05 M sucrose , 10 mM Hepes pH 7 . 6 , 10% glycerol , 15 mM KCl , 2 mM EDTA , 1× protease inhibitor cocktail [Roche , Basel , Switzerland] , 0 . 15 mM spermine , 0 . 5 M spermidine , 0 . 5 mM DTT ) and centrifuged for 45 min at 2°C at 24 , 000 rpm ( 100 , 000×g ) using a Bechmann SW27 rotor . Nuclei were resuspended in 1 ml of 20 mM Hepes pH 7 . 6 , 150 mM NaCl , 2 mM EDTA , 1× protease inhibitor cocktail ( Roche , Basel , Switzerland ) , 0 . 5 mM PMSF , 1 mM DTT , 0 . 5 U/ml of RNAseOUT/SUPERase-In ( Invitrogen/Ambion ) , homogenized using a 1 ml dounce homogenizer ( three times with loose pestle , two times with tight pestle ) , and divided into three samples of equal volume . One volume of 2× NUN buffer ( 50 mM Hepes pH 7 . 6 , 2 M Urea , 2% NP-40 , 600 mM NaCl , 2 mM DTT , 1× protease inhibitor cocktail [Roche , Basel , Switzerland] , 0 . 5 mM PMSF , 0 . 5 U/ml of SUPERase-In [Ambion , Carlsbad , California] ) was then added drop-by-drop while gently vortexing ( level 2 ) . Samples were left on ice for 20 min , then centrifuged at 24 , 000 rpm for 10 min at 4°C . The supernatant was removed and 1 ml of Trizol ( Invitrogen , Carlsbad , California ) was added to the DNA pellet . Samples were then incubated at 65°C for 15 min , and DNA pellet was then resuspended by gentle pipetting . Nascent RNA was then extracted using standard Trizol RNA extraction ( Invitrogen , Carlsbad , California ) . Mice housed in LD12:12 were sacrificed by isoflurane anesthesia followed by decapitation every 4 hr ( ZT2 , 6 , 10 , 14 , 18 and 22 ) . Mouse liver was then quickly removed and cut into small pieces that were frozen on dry ice . Total RNA was extracted using standard Trizol extraction ( Invitrogen , Carlsbad , California ) . Nascent RNA was first DNase-treated with TURBO DNase ( Ambion , Carlsbad , California ) using manufacturer's recommendation . Polyadenylated RNA was then removed from the nascent RNA using Dynabeads mRNA direct kit ( Invitrogen , Carlsbad , California ) following manufacturer's recommendations , and nascent RNA was precipitated ( ethanol precipitation ) . Sequencing libraries have been made using standard protocols . Briefly , 100 ng of purified nascent RNA were used to generate Illumina libraries . Nascent RNA was first fragmented using Fragmentation reagents ( AM8740; Ambion , Carlsbad , California ) by heating at 70°C for 5 min . Fragmented nascent RNA were then purified and used for standard Illumina library preparation . Following adaptor ligation , libraries of 200–300 bp length were size-selected on a 2% TAE agarose gel , and amplified by PCR for 15 cycles . Strand-specific libraries were processed as above except for the following modifications that have also been described elsewhere ( Levin et al . , 2010 ) . Briefly , after the first strand cDNA synthesis , dNTPs were removed by size-exclusion chromatography columns ( G-50 columns; Amersham , Amersham , UK ) and by ethanol precipitation using ammonium acetate . Second strand synthesis was then performed using a dNTP mixture containing dUTP instead of dTTP . After adaptor ligation and size selection ( i . e . , prior to the amplification ) , libraries were digested using Uracil-Specific Excision Reagent ( USER Enzyme , NEB , Ipswich , Massachusetts ) by incubating 2 units ( 2 μl ) of USER Enzyme with 18 μl of libraries at 37°C for 30 min . Reaction was then heat-inactivated , the libraries were purified and PCR-amplified . RNA-Seq libraries ( e . g . , mRNA ) were made using Truseq RNA sample kit ( Illumina , San Diego , California ) following manufacturer's recommendations . Adult mice housed in LD12:12 were sacrificed at ZT8 by isoflurane anesthesia followed by decapitation . Mouse liver was then quickly removed and homogenized in 3 . 5 ml of 1× PBS supplemented with 1% formaldehyde . After 10 min incubation at room temperature , cross-linking was stopped by mixing liver homogenate with 25 ml of ice-cold quenching solution ( 2 . 2 M sucrose , 150 mM glycine , 10 mM Hepes pH 7 . 6 , 15 mM KCl , 2 mM EDTA , 1× protease inhibitor cocktail [Roche , Basel , Switzerland] , 0 . 15 mM spermine , 0 . 5 M spermidine , 0 . 5 mM DTT ) . Homogenate was then layered on top of a 10 ml ice-cold sucrose cushion ( 2 . 05 M sucrose , 125 mM glycine , 10 mM Hepes pH 7 . 6 , 10% glycerol , 15 mM KCl , 2 mM EDTA , 1× protease inhibitor cocktail [Roche] , 0 . 15 mM spermine , 0 . 5 M spermidine , 0 . 5 mM DTT ) and centrifuged for 30 min at 2°C and 24 , 000 rpm ( 100 , 000×g ) using a Bechmann SW27 rotor . Nuclei were resuspended in 1 ml of 20 mM Hepes pH 7 . 6 , 150 mM NaCl , 2 mM EDTA and sedimented at 1500×g for 1 min . Washed nuclei were resuspended in 1 . 2 ml sonication buffer ( 20 mM Hepes pH 7 . 6 , 1% SDS , 150 mM NaCl , 2 mM EDTA ) and sonicated on ice using a Fisherbrand Sonic Dismembranator at setting 2 ( 57 W ) for 5 × 15 s to obtain chromatin fragments of about 500–1000 bp in length . The resulting chromatin was centrifuged at 15 , 000×g for 10 min and the resulting supernatant was aliquoted in 200 μl samples for immunoprecipitation and 25 μl samples for input . Immunoprecipitation of chromatin was performed by mixing 200 μl of sonicated chromatin with 1 . 8 ml of IP buffer ( 10 mM Hepes pH 7 . 6 , 150 mM NaCl , 2 mM EDTA , 0 . 1% NaDeoxycholate , 1% Triton X-100 ) . Antibodies were added and samples were incubated overnight ( Rabbit anti-BMAL1 antibody: 10 μl , ab3350; Abcam; Rabbit anti-CLK antibody: 10 μl , NB100-126; Novus Biologicals ) . Dynabeads protein G ( 100 μl per sample; Invitrogen , Carlsbad , California ) were blocked in parallel overnight in 0 . 1 mg/ml yeast tRNA and 1 mg/ml BSA in IP buffer . Following the overnight incubation , beads were washed once with IP buffer . The chromatin/antibodies mixture was then added to the beads and incubated at 4°C for an additional 2 hr . Beads were then washed once for 10 min with HSE I ( 0 . 1% SDS , 1% Triton X-100 , 2 mM EDTA , 20 mM Hepes KOH , pH 7 . 6 , 150 mM NaCl ) and twice for 10 min with HSE II ( 0 . 1% SDS , 1% Triton X-100 , 2 mM EDTA , 20 mM Hepes KOH , pH 7 . 6 , 500 mM NaCl ) . Beads were then briefly washed with ice-cold TE and eluted with 200 µl of ChIP Elution buffer ( 50 mM Tris–HCl pH 8 . 0 , 10 mM EDTA , 1% SDS , 1 mM DTT ) . 175 µl of ChIP Elution buffer was also added to the 25 µl input samples . Elution was performed at 65°C for 6–18 hr . The resulting supernatant was removed , supplemented with 200 μl of TE and 8 μl of 1 mg/ml RNAse A ( Ambion Cat #2271 ) and incubated at 37°C for 30 min . Then , 4 μl of 10 mg/ml proteinase K was added and samples were incubated at 55°C for another 2 hr . DNA was then isolated with PCR purification kit ( Qiagen ) and eluted with 40 μl of elution buffer . BMAL1 , CLK and input libraries were made from ChIPs performed from the same mouse liver extract . ChIP-seq libraries were made as described by Schmidt et al . ( 2009 ) . Size-selected libraries of 200–300 bp length were used for Illumina deep-sequencing , whereas libraries with a 300–650 bp length were used for qPCR validation of the quality of the ChIP-seq libraries . High-throughput sequencing has been performed as follow:- ChIP-seq libraries: BMAL1 , CLK and Input ChIP-seq libraries were sequenced using an Illumina Genome Analyzer ( GAII ) with a sequencing length of 36 nt . To increase depth coverage , libraries were sequenced on multiple lanes ( BMAL1: four lanes , CLK: five lanes and Input: three lanes ) . - Nascent-Seq libraries: libraries ( 12 samples corresponding to two independent six-time points rhythms ) were sequenced using an Illumina Genome Analyzer ( GAII ) with a sequencing length of 80 nt . Both replicates of the ZT8 and ZT20 time points were sequenced on two lanes and all other samples on one lane . - Nascent-Seq libraries , strand-specific: libraries ( six samples corresponding to the first replicate of the six-time points rhythm ) were generated using bar-coded adaptors , mixed in an equimolar ratio and sequenced on two lanes using a HiSeq2000 ( Illumina ) with a sequencing length of 101 nt . - RNA-Seq libraries: libraries ( 12 samples corresponding to two independent six-time points rhythms ) were generated using bar-coded adaptors , mixed in an equimolar ratio and sequenced on two lanes using a HiSeq2000 ( Illumina ) with a sequencing length of 101 nt . High-throughput sequencing has been performed following manufacturer recommendations and 8–12 pmol of libraries were hybridized to each lane of the flow-cells . Data were extracted and processed following Illumina recommendations . Sequences were aligned to the mouse genome ( UCSC version mm9 database ) . Number of the sequences obtained for each library can be found in Table 1 . Datasets are deposited on the Gene Expression Omnibus database under the accession number GSE36916 ( GEO , http://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? acc=GSE36916 ) . 10 . 7554/eLife . 00011 . 022Table 1 . Number of sequences and statistics for the different sequencing datasetsDOI: http://dx . doi . org/10 . 7554/eLife . 00011 . 022Index numberBarcodeNumber of sequences ( fastq file ) Number of uniquely mapped sequencesPercentage of uniquely mapped sequencesNormaliz . factorChIP-Seq librariesInput——39 , 214 , 69618 , 846 , 30348 . 1%—CLK——75 , 944 , 49537 , 371 , 04749 . 2%—BMAL1——60 , 952 , 29328 , 920 , 75447 . 5%—Nascent-Seq librariesNorm . 40 mRep1_ZT0——27 , 845 , 32018 , 319 , 01165 . 8%2 . 184Rep1_ZT4——30 , 088 , 98120 , 931 , 03869 . 6%1 . 911Rep1_ZT8——57 , 719 , 17439 , 567 , 60968 . 6%1 . 011Rep1_ZT12——29 , 442 , 24419 , 485 , 10266 . 2%2 . 053Rep1_ZT16——27 , 645 , 10218 , 385 , 66866 . 5%2 . 176Rep1_ZT20——50 , 331 , 24234 , 703 , 72769 . 0%1 . 152Rep2_ZT0——30 , 243 , 85621 , 014 , 08769 . 5%1 . 903Rep2_ZT4——30 , 162 , 51421 , 082 , 49869 . 9%1 . 897Rep2_ZT8——51 , 471 , 47736 , 118 , 06870 . 2%1 . 107Rep2_ZT12——27 , 304 , 92117 , 815 , 97165 . 3%2 . 245Rep2_ZT16——27 , 196 , 80519 , 077 , 43370 . 2%2 . 097Rep2_ZT20——51 , 105 , 23633 , 547 , 43965 . 7%1 . 192RNA-Seq librariesNorm . 40 mRep1_ZT22CGATGT13 , 031 , 4968 , 693 , 55566 . 7%4 . 601Rep1_ZT64TGACCA13 , 197 , 07810 , 214 , 58077 . 4%3 . 916Rep1_ZT105ACAGTG13 , 479 , 6369 , 916 , 77473 . 6%4 . 034Rep1_ZT146GCCAAT10 , 366 , 7027 , 497 , 38672 . 3%5 . 335Rep1_ZT187CAGATC13 , 147 , 6499 , 600 , 12573 . 0%4 . 167Rep1_ZT2212CTTGTA11 , 182 , 7568 , 233 , 81573 . 6%4 . 858Rep2_ZT213AGTCAA14 , 645 , 2639 , 876 , 35967 . 4%4 . 050Rep2_ZT614AGTTCC15 , 836 , 01312 , 270 , 33877 . 5%3 . 260Rep2_ZT1015ATGTCA15 , 123 , 72611 , 507 , 85676 . 1%3 . 476Rep2_ZT1416CCGTCC12 , 127 , 1028 , 594 , 60970 . 9%4 . 654Rep2_ZT1818GTCCGC12 , 903 , 6789 , 512 , 76573 . 7%4 . 205Rep2_ZT2219GTGAAA13 , 438 , 8739 , 592 , 40471 . 4%4 . 170Strand-specific Nascent-Seq librariesNorm . 40 mRep1_ZT02CGATGT34 , 386 , 62215 , 930 , 80146 . 3%2 . 511Rep1_ZT44TGACCA45 , 356 , 90624 , 224 , 15153 . 4%1 . 651Rep1_ZT85ACAGTG44 , 309 , 21624 , 275 , 35754 . 8%1 . 648Rep1_ZT126GCCAAT49 , 118 , 10422 , 882 , 16346 . 6%1 . 748Rep1_ZT167CAGATC49 , 535 , 73821 , 835 , 60544 . 1%1 . 832Rep1_ZT2012CTTGTA54 , 905 , 00532 , 586 , 39659 . 4%1 . 228 Sequences from the different libraries ( fastq format ) were first mapped to the mouse genome ( version mm9 ) using bowtie ( Langmead et al . , 2009 ) with the command line: bowtie –q –a –-best –m 1 . Only those that mapped uniquely to the mouse genome were used for further analysis , and their number has been used for normalization to compare signal difference between libraries . ChIP-seq libraries were analyzed with the MACS algorithm ( Zhang et al . , 2008 ) by comparing the treatment sample ( BMAL1 or CLK ChIP ) to the control sample ( Input ) using the following criteria: effective genome size = 1 . 89 × 109 , tag size = 36 , band width = 80 , model fold = 5 , p-value cutoff = 1 × 10−5 . Significant peaks were computationally assigned to a gene . Briefly , a peak located between the transcription start site and the transcription start end of a gene was assigned to that gene , regardless of the ChIP-Seq peak position . The other peaks , referred as intergenic , were assigned to the gene with the closest transcription start site . Confirmation of this computational gene assignment was then confirmed by manual inspection for the 211 CLK:BMAL1 peaks . Visualization of the ChIP-seq signal was performed using the wig output file ( from the MACS analysis ) and the IGB browser . Overlap between CLK and BMAL1 DNA binding peaks has been determined computationally using all significant peaks coordinates . Any overlap between the two peaks ( even of one nucleotide ) was considered significant . Quantification of the signal has been extracted from the raw data ( number of reads per bp ) normalized to sequencing depth of each library . For most experiments ( e . g . , Figure 7A , B ) , signal was binned using a 25 bp window . Quantification of the number of e-boxes within BMAL1 and CLK DNA binding peaks has been performed computationally using peak fasta sequences . Enrichment has been calculated using the number of e-boxes found at a fixed position from the peak center divided by the expected number of e-boxes . The maximal window size ( difference from the fixed position and the peak center ) was 500 nt , as the number of expected e-boxes dropped to the background at this window size . Motif analysis has been performed using MEME suite ( http://meme . sdsc . edu/meme/intro . html ) , using a 100 bp sequence for each peak ( peak center ± 50 bp ) . Parameters were as follow: -dna -mod anr -nmotifs 20 -minw 6 -maxw 30 . The background model contained the same nucleotide distribution as the input file . Significant motifs were then analyzed using TOMTOM from MEME suite . Gene signal ( reads per base pair ) was averaged for the 12 time points , and the ratio Nascent RNA/mRNA was calculated . Genes with a ratio over 2 SD from the average of all ratios were selected for gene ontology ( GO ) analysis; 302 genes had a Nascent-Seq/RNA-Seq ratio below 2 SD and 463 genes had a ratio over 2 SD . 13595 genes were considered for this analysis . GO analysis has been performed using GOToolBox ( Martin et al . , 2004 ) ( http://genome . crg . es/GOToolBox/ ) , using an hypergeometric test with Benjamini-Hochberg correction . The background model consisted of the entire list of genes . Only genes with more than three reads per base pair for at least one time point of the Nascent-Seq dataset and two read per base pair for the RNA-Seq dataset were further considered for subsequent analysis ( see above ) . Rhythmically expressed genes were determined based on three parameters: amplitude , F24 ( 24-hr spectral power , see below ) and p-value of the F24 . The amplitude was calculated by dividing the highest value of the 12 time points by the lowest value . The F24 score was calculated by Fourier transformation using a R code originally described by Wijnen et al . ( 2005 ) . Briefly , normalized reads per base pair from the two independent six-time points rhythms were concatenated and the 24-hr spectral power ( F24 ) was determined for each gene . The F24 score ( expressed in range 0–1 ) indicates the relative strength of the extracted rhythmic component . The F24 p-value ( pF24 ) represents the probability of observing an F24 score from randomly permuted data that is of equal or greater strength than the extracted Fourier component . It was calculated for each gene after performing 10 , 000 randomized permutations of the rpbp values . A pF24 was considered significant if ( pF24 < 0 . 05 ) if the experimental pF24 was within the top 5% of the 10 , 000 pF24 calculated from randomized permutation . Transcripts were considered to be rhythmically expressed when meeting the three following criteria: ( 1 ) pF24-values < 0 . 05 ( i . e . , experimental pF24-value is within the top 5% of the 10 , 000 pF24-values calculated from randomized permutation ) , ( 2 ) F24 > 0 . 45 and ( 3 ) amplitude ( maximal/minimal experimental values ) > 1 . 5 . A more stringent cut-off was also used to identify strong rhythmic expression: pF24 < 0 . 05 , F24 > 0 . 6 and amplitude > 1 . 75 . The phase information from the Fourier transformation ( which indicates the peak of the cosine curve ) was further used to assess phase difference between rhythmic nascent RNA and mRNA expression ( Figures 3C , 4B ) . Total RNA from wild-type and Bmal1−/− mice was prepared from mouse liver using Trizol reagent ( Invitrogen ) and DNAse-treated using Turbo DNAse ( Ambion ) according to the manufacturer's protocols . cDNA derived from RNA ( using Invitrogen Superscript II and random primers ) was utilized as a template for quantitative real-time PCR performed with the Rotor-Gene 3000 real-time cycler ( Qiagen ) . The PCR mixture contained Platinum Taq polymerase ( Invitrogen ) , optimized concentrations of Sybr-green ( Invitrogen ) and specific primers for either Rev-Erbα , Per1 , Per2 or Cry1 pre-mRNAs . Quantitative PCR using Actg1-specific primers was used as a loading control . Cycling parameters were 95°C for 3 min , followed by 40 cycles of 95°C for 30 s , 55°C for 45 s , and 72°C for 45 s . Fluorescence intensities were plotted vs the number of cycles by using an algorithm provided by the manufacturer . mRNA levels were quantified using a calibration curve based upon dilution of concentrated cDNA . | Many biological processes oscillate with a period of roughly 24 hr , and the ability of organisms as diverse as bacteria and humans to maintain such circadian rhythms , even under conditions of continuous darkness , influences a range of phenomena , including sleep , migration and reproduction . One characteristic of circadian rhythms is that they can adjust to local time ( with humans suffering from jet lag as they wait for this to happen ) . Experiments have shown that the circadian system in mammals relies on feedback loops that operate at the level of individual cells . These loops are controlled by two particular proteins , which comprise the transcription factor complex called BMAL1:CLK . Transcription factors cause particular sequences of bases in the DNA of cells to be transcribed into messenger RNA , thus starting the process by which target genes are expressed as proteins . In the case of BMAL1:CLK , these proteins are then modified , which inhibits any further transcription of the target genes . A reversal of these modifications is then followed by the synthesis of new proteins , which allows a new cycle of the transcription process to begin . The amounts of many messenger RNAs ( mRNAs ) in a cell also increases and decreases with a period of 24 hr , and it was generally assumed that this was due to the changes in the level of transcription . More recently , however , it was suggested that other processes , such as splicing and translation , might also contribute to rhythmic changes in the amount of mRNA associated with particular genes . Such post-transcriptional processes are known to have a role in other areas of cell biology , including aspects of the circadian system , but until very recently this had not been studied in detail for all genes . Now Menet et al . have directly assayed rhythmic transcription by measuring the amount of nascent mRNA being produced at a given time , six times a day , across all the genes in mouse liver cells using a high-throughput sequencing approach called Nascent-Seq . They compared this with the amount of liver mRNA expressed at six time points of the day . Although the authors found that many genes exhibit rhythmic mRNA expression in the mouse liver , about 70% of them did not show comparable transcriptional rhythms . Post-transcriptional regulation must , therefore , have a major role in the circadian system of mice and , presumably , other mammals . Menet et al . also found that the influence of CLK:BMAL1 differed from what was expected , which suggests that it collaborates with a number of other transcription factors to effect transcription of most target genes . In addition to showing that circadian systems of mammals are more complex than previously believed , the results also illustrate the potential of Nascent-Seq as a genome-wide assay technique for exploring a range of questions related to gene expression and gene regulation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"genetics",
"and",
"genomics"
] | 2012 | Nascent-Seq reveals novel features of mouse circadian transcriptional regulation |
The functional repertoire of surface ion channels is sustained by dynamic processes of trafficking , sorting , and degradation . Dysregulation of these processes underlies diverse ion channelopathies including cardiac arrhythmias and cystic fibrosis . Ubiquitination powerfully regulates multiple steps in the channel lifecycle , yet basic mechanistic understanding is confounded by promiscuity among E3 ligase/substrate interactions and ubiquitin code complexity . Here we targeted the catalytic domain of E3 ligase , CHIP , to YFP-tagged KCNQ1 ± KCNE1 subunits with a GFP-nanobody to selectively manipulate this channel complex in heterologous cells and adult rat cardiomyocytes . Engineered CHIP enhanced KCNQ1 ubiquitination , eliminated KCNQ1 surface-density , and abolished reconstituted K+ currents without affecting protein expression . A chemo-genetic variation enabling chemical control of ubiquitination revealed KCNQ1 surface-density declined with a ~ 3 . 5 hr t1/2 by impaired forward trafficking . The results illustrate utility of engineered E3 ligases to elucidate mechanisms underlying ubiquitin regulation of membrane proteins , and to achieve effective post-translational functional knockdown of ion channels .
Integral surface membrane proteins including ion channels , transporters , and receptors are vital to the survival and function of all cells . Consequently , processes that control the surface abundance and composition of membrane proteins are critical determinants of cellular biology and physiology . Impaired surface trafficking of membrane proteins underlies diverse diseases ranging from cystic fibrosis to cardiac arrhythmias ( Gelman and Kopito , 2002; Anderson et al . , 2014 ) , motivating a need to better understand fundamental mechanisms controlling membrane protein surface density . The surface repertoire of membrane proteins is regulated by multi-layered maturation and trafficking processes ( MacGurn et al . , 2012; Foot et al . , 2017 ) . As such , the mechanisms governing diverse aspects of membrane protein fate is an intensely studied research area . Ubiquitination determines membrane protein functional expression by regulating multiple steps in the membrane protein lifecycle . Ubiquitin is a 76-residue protein that can be covalently attached to lysine residues on polypeptide substrates through the sequential action of three enzymes: a ubiquitin activation enzyme ( E1 ) ; a ubiquitin-conjugating enzyme ( E2 ) ; and a ubiquitin ligase ( E3 ) , that catalyzes transfer of ubiquitin to substrates . The human genome encodes 2 E1s , 37 E2s , and >600 E3 ubiquitin ligases . Ubiquitin contains seven lysine residues ( K6 , K11 , K27 , K29 , K33 , K48 , K63 ) that , together with its N-terminus ( Met1 ) , can serve as secondary attachment points to make diverse polyubiquitin chains with different structures and functions ( Komander , 2009 ) . Ubiquitination has classically been ascribed to targeting cytosolic proteins for degradation by the proteasome ( Hershko and Ciechanover , 1998 ) . However , it is now evident that ubiquitination of both cytosolic and membrane proteins can lead to more nuanced outcomes including regulating protein trafficking/sorting , stability , and/or function ( Komander , 2009; Foot et al . , 2017 ) . Nevertheless , precisely how ubiquitination regulates such diverse aspects of protein fate— and membrane protein fate in particular— is often poorly understood . Factors that complicate analyses include: ( 1 ) multiple E3 ligases may ubiquitinate a single substrate; ( 2 ) a particular E3 can typically catalyze ubiquitination of multiple substrates; ( 3 ) distinct E3 ligases can have preference for particular ubiquitination profiles ( e . g . monoubiquitination versus polyubiquitination ) and polyubiquitin chain linkages ( e . g . K48 versus K63 ) ; ( 4 ) lack of temporal control over the ubiquitination process . The elusive nature of ubiquitin signaling is exemplified by its regulation of diverse voltage-gated ion channels . KCNQ1 ( Kv7 . 1; Q1 ) , is a voltage-gated K+ channel which together with auxiliary KCNE1 subunits give rise to the slowly activating delayed rectifier current IKs that is important for human ventricular action potential repolarization ( Barhanin et al . , 1996; Sanguinetti et al . , 1996 ) . Loss-of-function mutations in Q1 lead to long QT syndrome type 1 ( LQT1 ) , a precarious condition that predisposes affected individuals to exertion-triggered cardiac arrhythmias and sudden cardiac death ( Tester et al . , 2005 ) . In heterologous expression studies , NEDD4-2 , a HECT domain E3 ligase , binds a PY motif on Q1 C-terminus; enhances Q1 ubiquitination; down-regulates Q1 expression; and inhibits IKs current ( Jespersen et al . , 2007 ) . Understanding precisely how NEDD4-2 accomplishes these distinctive effects is confounded by the promiscuity of this E3 ligase in targeting many other proteins that contain PY motifs ( Abriel and Staub , 2005; MacGurn et al . , 2012; Goel et al . , 2015 ) , as well as a lack of tight temporal control over its action . This could potentially be resolved if it were possible to target distinct E3 ligase activity to Q1/KCNE1 proteins in a selective and temporally controllable manner . Several studies have applied an approach that utilizes engineered E3 ubiquitin ligases to selectively target cytosolic proteins to direct their degradation by the proteasome ( Zhou et al . , 2000; Hatakeyama et al . , 2005; Caussinus et al . , 2011; Ma et al . , 2013; Portnoff et al . , 2014 ) . The general principle involves replacing the intrinsic substrate-targeting module of an E3 ligase with a motif that directs it to a desired target protein . Here , we sought to determine , for the first time , whether this method could be applied to elucidate mechanisms underlying ubiquitin regulation of ion channel complexes . We engineered E3 ubiquitin ligases to selectively target YFP-tagged Q1 or KCNE1 subunits and assessed the impact on channel surface density , stability , and IKs . We found that targeted ubiquitination of Q1/KCNE1 with distinct engineered ligases dramatically diminished channel surface expression without necessarily affecting total protein expression . We developed a chemo-genetic variation of the approach that enabled controllable targeting of an engineered E3 to the channel using chemical heterodimerization . The temporal control afforded by the chemo-genetic method in combination with fluorescence pulse-chase assays revealed that ubiquitination diminished Q1 surface density by selectively limiting channel delivery to the cell surface , and not by enhancing the rate of endocytosis . To demonstrate the generality of the approach , we used the engineered E3 ligase to selectively eliminate surface expression and currents through voltage-gated L-type Ca2+ ( CaV1 . 2 ) channels . Similar to Q1 , targeted ubiquitination of CaV1 . 2 markedly decreased channel surface density without impacting total expression , emphasizing a fundamental distinction in the impact of targeted ubiquitination between ion channels and previously studied cytosolic proteins . Beyond enabling original mechanistic insights , the approach provides a potent tool to post-translationally manipulate surface expression of ion channel macromolecular complexes in a manner that complements , and provides particular advantages over , well-established and widely used genomic/mRNA interference methods .
The lifecycle of surface ion channels and other membrane proteins involves minimally their genesis and folding in the endoplasmic reticulum ( ER ) ; post-translational maturation in the Golgi; their delivery to and removal from their site of action on the plasma membrane; and ultimately their demise by degradation in lysosomes or via the proteasome ( Figure 1A ) . Ubiquitination looms as a powerful mechanism to control membrane protein fate since it potentially influences multiple steps in their lifecycle ( Figure 1A ) . Ubiquitination is mediated by a step-wise cascade of three enzymes ( E1 , E2 , E3 ) , resulting in the covalent attachment of the 76-residue ubiquitin to lysines of a target protein ( Figure 1B ) . We sought to develop a system that enabled selective ubiquitination of the voltage-gated K+ channel pore-forming subunit , Q1 , to dissect the mechanistic impact of specific post-translational modification of this protein . We took advantage of the modular design of E3 ligases , which typically have distinct substrate-binding and catalytic domains . For example , CHIP ( C-terminus of the Hsp70-interacting protein ) , is a U-box E3 ligase comprised of a catalytic domain that binds E2 and a tetratricopeptide repeats ( TPR ) targeting domain that binds Hsp70 ( Connell et al . , 2001; Murata et al . , 2003; Zhang et al . , 2005 ) . This modular arrangement enables its function for chaperone-mediated ubiquitination of substrate proteins as a quality control mechanism ( Figure 1B ) . We substituted the TPR domain of CHIP with the vhh4 nanobody , which binds GFP/YFP ( but not CFP ) ( Kubala et al . , 2010 ) , creating nanoCHIP . We hypothesized that nanoCHIP would selectively target and catalyze ubiquitination of Q1-YFP , leading to three possible ( but not mutually exclusive ) outcomes of reducing protein stability , altering trafficking , or modulating channel function ( Figure 1C ) . We utilized optical fluorescence assays to conveniently measure surface and total pools of Q1-YFP in a robust and high throughput manner . We introduced a 13-residue high-affinity bungarotoxin binding site ( BBS ) into the extracellular S1-S2 loop of Q1 , enabling detection of surface channels in non-permeabilized cells with Alexa Fluor 647-conjugated bungarotoxin ( BTX647 ) ( Figure 1C and Figure 2 ) ( Aromolaran et al . , 2014 ) . The C-terminal YFP tag provides a fluorescent measure of total Q1 expression ( Figure 2A ) . The nanoCHIP construct was generated in a P2A vector that expressed CFP as a separate reporter protein in a 1:1 ratio with nanoCHIP ( Figure 2B ) . We performed two types of control experiments . First , BBS-Q1-YFP was expressed with nanobody-P2A-CFP alone ( nano ) ( Figure 2A ) . Second , nanoCHIP was co-expressed with BBS-Q1 lacking the C-terminus YFP tag ( Figure 2C ) . We used flow cytometry to rapidly quantify total ( YFP ) and surface ( red; BTX647 ) Q1 expression in ~50 , 000 live cells , with single cell resolution . Control cells ( nano + BBS-Q1-YFP ) displayed robust total Q1 expression ( YFP signal ) in CFP-positive cells ( Figure 2D ) . Test cells expressing nanoCHIP + BBS-Q1-YFP showed little change in YFP fluorescence compared to control ( YFP = 577 ± 11 a . f . u , n = 1727 for nano; YFP = 630 ± 18 a . f . u , n = 783 for nanoCHIP ) , suggesting that the presumed targeted ubiquitination did not substantively affect Q1 stability ( Figure 2D , E ) . By contrast , the surface density of BBS-Q1-YFP between the two conditions revealed an entirely different picture . Whereas control cells ( nano ) displayed a sizable population of surface BBS-Q1-YFP as reported by robust mean red fluorescence signal ( BTX647 = 822 ± 26 a . f . u , n = 4837 ) , this surface pool was almost completely eliminated in cells co-expressing nanoCHIP ( BTX647 = 55 ± 2 a . f . u , n = 2257 ) ( Figure 2F , G ) . To assess specificity , we co-expressed nanoCHIP with BBS-Q1 . In sharp contrast to the result obtained with BBS-Q1-YFP , nanoCHIP had no effect on surface expression of BBS-Q1 channels ( BTX647 = 527 ± 16 a . f . u , n = 6425 for nano; BTX647 = 633 ± 24 a . f . u , n = 3657 for nanoCHIP ) ( Figure 2H , I ) . These data were obtained with a 1:3 transfection ratio of Q1 to nanoCHIP cDNA . Similar results were obtained using transfection ratios of 1:1 and 1:5 ( Figure 2—figure supplement 1 ) . As a further control , co-expression of a CHIP deletion mutant ( nanoCHIP* ) that abolishes E3 ligase activity ( Nikolay et al . , 2004 ) did not alter Q1 surface expression , confirming the requirement of catalytic activity for nanoCHIP-dependent surface modulation ( Figure 2—figure supplement 2 ) . We used a similar strategy to selectively target distinct E3 ligase classes to Q1-YFP , notably two members of the RING family ( nanoNSlmb and nanoMDM2 ) , as well as NEDD4-2 from the HECT family ( nanoNEDD4-2 ) ( Figure 2—figure supplements 3 and 4 ) . We obtained qualitatively similar results with nanoNSlmb and nanoMDM2 to what we observed with nanoCHIP in that they all reduced BBS-Q1-YFP surface density with minimal effects on total expression ( Figure 2—figure supplement 3 ) . Interestingly , nanoNEDD4-2 diminished total Q1-YFP expression concomitant with the decreased surface expression ( Figure 2—figure supplement 4 ) . Both effects on Q1 stability and surface density were abolished with co-expression of a NEDD4-2 catalytic inactive mutant ( nanoNEDD4-2* ) ( Figure 2—figure supplement 4 ) . Given that nanoCHIP had the most robust effect on reducing Q1 surface density ( Figure 2—figure supplement 3 ) we focused the rest of the study on this engineered E3 ligase . Q1 is known to be ubiquitinated and regulated by heterologously expressed wild-type NEDD4-2 ( Jespersen et al . , 2007 ) . As a prelude to determining whether nanoCHIP enhances ubiquitination of Q1-YFP , we first sought to reproduce the previously reported NEDD4-2-mediated ubiquitination of Q1 ( Jespersen et al . , 2007 ) . We transiently expressed Q1-YFP either alone ( control ) or with NEDD4-2 in HEK293 cells . The cells were lysed under denaturing conditions and Q1-YFP was pulled down with anti-Q1 antibody . Western blot of the immunoprecipitated Q1-YFP using anti-Q1 displayed four bands representing the monomeric , dimeric , trimeric , and tetrameric channel species ( Figure 3A ) . Densitometric analyses of the bands indicated that co-expression with NEDD4-2 reduced the total expression of Q1-YFP ( Figure 3A; area under the curve ) , in agreement with previous results ( Jespersen et al . , 2007 ) . Flow cytometry measurements were also consistent with this result ( Figure 3—figure supplement 1 ) . Having confirmed Q1-YFP pulldown , the membrane was stripped and probed with anti-ubiquitin ( Figure 3B ) . Control cells expressing Q1-YFP alone displayed some ubiquitination reflecting the activity of an endogenous E3 ligase ( s ) . Co-expression of NEDD4-2 substantially increased Q1-YFP ubiquitination compared to the control condition ( Figure 3B , C ) . With the method for detecting Q1-YFP ubiquitination validated , we turned to the effect of nanoCHIP in this biochemical assay . Consistent with the flow cytometry experiments , nanoCHIP did not substantively affect Q1-YFP total expression ( Figure 3D ) . Nevertheless , nanoCHIP significantly augmented Q1-YFP ubiquitination compared to the control condition , although the effect was smaller than observed with NEDD4-2 ( Figure 3E , F ) . As such , our findings suggest that modest changes in total ubiquitination intensity as detected by conventional Western blot can result in substantial functional and cell biological effects on ion channel surface trafficking . Physiologically , Q1 is typically associated with auxiliary KCNE subunits that are single transmembrane spanning proteins . There are five distinct KCNE subunits ( KCNE1-KCNE5 ) , each of which can profoundly shape the outward K+ current waveform when associated with Q1 ( Sun et al . , 2012 ) . The interaction of KCNE1 with Q1 transforms the current waveform from one which is small and rapidly activating ( Q1 alone ) to one that is large and slowly activating ( IKs; Q1 + KCNE1 ) ( Barhanin et al . , 1996; Sanguinetti et al . , 1996 ) . The slowly activating kinetics of IKs is crucial to its physiological role in human cardiac action potential repolarization . We determined whether nanoCHIP could regulate the Q1/KCNE1 macromolecular complex . Similar to our observations with Q1 alone , nanoCHIP dramatically reduced surface density of BBS-Q1-YFP co-expressed with KCNE1 ( Figure 4A–C ) , while minimally affecting total protein expression ( Figure 4—figure supplement 1 ) . This effect was selective , as in cells expressing BBS-Q1 + KCNE1 ( lacking a YFP tag ) , nanoCHIP had no effect on channel surface density ( Figure 4D–F ) . Finally , we wondered whether we could manipulate the surface expression of Q1 by targeting nanoCHIP to the auxiliary KCNE1 subunit . Indeed , in cells expressing BBS-Q1 + KCNE1-YFP , nanoCHIP effectively and selectively eliminated surface channels ( Figure 4G–I and Figure 4—figure supplement 1 ) , demonstrating the potential power of the approach to sculpt ion channel macromolecular complexes by targeting accessory proteins . Consistent with these results , nanoCHIP targeted to KCNE1-YFP markedly increased ubiquitination of co-expressed Q1 ( Figure 4J , K ) . We next examined the impact of nanoCHIP on functional IKs currents , under conditions that mirrored those used to examine channel trafficking and total Q1 expression . Control Chinese hamster ovary ( CHO ) cells transfected with Q1-YFP + KCNE1 + nano displayed robust outward K+ currents with the signature slow activation kinetics of IKs , which were essentially eliminated by nanoCHIP ( Figure 5A–C ) . A similar result was observed in HEK293 cells ( Figure 5—figure supplement 1 ) . By contrast , nanoCHIP had no effect on IKs reconstituted with subunits that lacked the YFP tag ( Figure 5D–F ) . Finally , switching the YFP tag to KCNE1 also yielded IKs that was significantly reduced by nanoCHIP ( Figure 5G–I ) . We next sought to exploit the rapamycin-induced heterodimerization system to develop an approach that enables acute temporal control of ubiquitination of specific target proteins ( Crabtree and Schreiber , 1996 ) . We separated the nanoCHIP catalytic ( CHIP ) and substrate-binding ( nano ) domains , and fused them to the rapamycin binding proteins FRB and FKBP , respectively ( Inoue et al . , 2005; Yang et al . , 2007 ) ( Figure 6A ) . FRB-CHIP and FKBP-nano are expected to have a low affinity for each other under basal conditions . Hence , when these two constructs are co-expressed with BBS-Q1-YFP we would not expect any channel ubiquitination by the engineered FRB-CHIP ( Figure 6A ) . Application of the small molecule rapamycin would then facilitate FRB-FKBP interaction , effectively recruiting the CHIP catalytic domain to BBS-Q1-YFP and initiating ubiquitination ( Figure 6A ) . We tested the effectiveness of this inducible nanoCHIP ( iN-CHIP ) approach by measuring the kinetics of rapamycin-induced decrease in surface density of BBS-Q1-YFP ( Figure 6B ) . Utilizing the flow cytometry-based fluorescence assay , we observed a time-dependent decrease in the surface pool of BBS-Q1-YFP after adding rapamycin . A measurable effect was seen within 20 min of rapamycin addition to the transfected HEK293 cells; the half-life for reduction of surface channels was 209 ± 17 mins ( Figure 6B , C ) . In control experiments , rapamycin treatment for 20 hr had no impact on Q1 surface density in cells expressing BBS-Q1 together with FRB-CHIP and FKBP-nano ( Figure 6—figure supplement 1 ) . In principle , the nanoCHIP-mediated reduction in BBS-Q1-YFP surface density could be due to: a decreased rate of delivery of new Q1 channels to the surface; an increased rate of internalization ( or removal ) of Q1 channels from the cell surface; or a combination of the two processes . To distinguish among these possibilities , we employed iN-CHIP along with two optical pulse-chase methods to measure the rates of BBS-Q1-YFP delivery to and removal from the cell surface ( Figure 7 ) . HEK293 cells were transiently transfected with BBS-Q1-YFP + FRB-CHIP + FKBP-nano . To measure rate of BBS-Q1-YFP delivery to the cell surface , we incubated live , non-permeabilized cells at 4°C to halt all trafficking processes , and exposed them to unconjugated BTX to block all extracellular BBS epitopes initially present at the plasma membrane ( pulse ) . Cells were then incubated at 37°C for varying epochs during which trafficking processes resumed , including delivery of new BBS-tagged channels to the surface ( chase ) . Cells were then returned to 4°C and the newly delivered surface channels labeled with BTX647 and quantified by flow cytometry ( Figure 7A ) . When this experiment was conducted in the absence of rapamycin pre-treatment , we observed robust delivery of new BBS-Q1-YFP to the cell surface during the chase period with a half-life of ~36 ± 3 mins ( Figure 7B ) . When cells were pretreated with rapamycin for 3 hr , delivery of new channels to the surface plateaued at a value ~75% lower compared to control ( Figure 7B ) . Thus , iN-CHIP-induced ubiquitination of BBS-Q1-YFP compromises forward trafficking of channels to the cell surface . To evaluate channel removal from the surface , we labeled live , non-permeabilized HEK293 cells ( expressing BBS-Q1-YFP + FRB-CHIP + FKBP-nano ) with biotinylated bungarotoxin ( BTX-biotin ) at 4°C ( pulse ) . Cells were then incubated at 37°C for varying time periods to resume trafficking processes ( chase ) . Following the chase period , cells were labeled with streptavidin-conjugated Alexa Fluor 647 at 4°C . In this paradigm , red fluorescent labeling would only occur on channels that were initially present at the surface and labeled with BTX-biotin during the pulse period . A decrease in fluorescence with increasing chase times would be expected due to internalization of BTX-biotin-labeled channels ( Figure 7C ) . Indeed , control cells ( no rapamycin pre-treatment ) displayed an exponential decline in red fluorescence with increasing chase time ( Figure 7D ) . Surprisingly , pre-activation of iN-CHIP with a 3 hr rapamycin pre-treatment had no impact on the rate of BBS-Q1-YFP internalization ( Figure 7D ) . Together , the results indicate that nanoCHIP reduces BBS-Q1-YFP surface density by selectively reducing forward trafficking of the channel . Ultimately , the general usefulness of the engineered E3 ligase approach to manipulate functional expression of membrane proteins hinges critically on the system performing robustly in native cells and tissues which have a more complex intracellular environment compared to heterologous cells . We tested the ability of nanoCHIP to suppress the surface density of BBS-Q1-YFP expressed in adult rat ventricular myocytes . We generated adenoviral vectors for BBS-Q1-YFP , nano-P2A-CFP , and nanoCHIP-P2A-CFP and used these to infect cultured cardiomyocytes . Control cells expressing BBS-Q1-YFP + nano displayed strong YFP/CFP fluorescence as well as QD655 signal on the sarcolemma , indicating robust cell surface density of the channel ( Figure 8A , C ) . By contrast , cardiomyocytes co-expressing nanoCHIP showed a sharply depressed QD655 signal ( Figure 8B , C ) . YFP fluorescence was not significantly different between the two experimental conditions ( Figure 8D ) . These data demonstrate that nanoCHIP is effective in cardiomyocytes , and selectively down-regulates BBS-Q1-YFP surface pool in this native cellular context . To test the generalizability of the engineered E3 ligase approach for regulating ion channels , we probed whether nanoCHIP could modulate the trafficking of a recombinant voltage-gated calcium channel ( CaV1 . 2 ) . Cav1 . 2 mediates excitation-contraction coupling and excitation-transcription coupling in heart and neurons , respectively , and is comprised of a pore-forming α1C and accessory ( β , α2δ , γ ) subunits ( Catterall , 2000 ) . We attached YFP to the C-terminus of α1C to render it a putative substrate for nanoCHIP , and a BBS epitope tag on an extracellular loop to enable fluorescent detection of surface channels ( Figure 9A ) ( Yang et al . , 2010; Subramanyam et al . , 2013 ) . Similar to our observations with Q1 , nanoCHIP selectively eliminated the surface CaV1 . 2 pool with no impact on total BBS-α1C-YFP expression ( Figure 9B–E ) . Consistent with these results , nanoCHIP essentially abolished CaV1 . 2 currents reconstituted in HEK 293 cells ( Figure 9F , G ) .
The precise mechanisms and signals regulating the dynamic trafficking of ion channels among membrane compartments are not completely understood and difficult to study , in part , due to their complexity and a lack of enabling tools . This is a serious limitation given that a number of ion channelopathies ( e . g . cystic fibrosis , epilepsy , Liddle syndrome , cardiac arrhythmias ) may arise due to dysregulation in ion channel surface expression ( Abriel and Staub , 2005 ) . Ubiquitination is a critical post-translational modification capable of regulating diverse aspects of protein fate including trafficking , sorting , and stability . Several ion channels and transporters are known to be regulated by ubiquitination , including ENaC , ClC-5 , KCNQ1 , and NaV channels , which have all been found to be regulated by NEDD4-like family ubiquitin ligases ( Staub et al . , 1997; Abriel et al . , 1999; Schwake et al . , 2001; Fotia et al . , 2004; van Bemmelen et al . , 2004; Abriel and Staub , 2005; Jespersen et al . , 2007 ) . Many of these channels possess PY motifs ( P-P-X-Y-X-X-ϕ where P is proline , Y is tyrosine , X is any amino acid , and ϕ is a hydrophobic residue ) to which NEDD4-like proteins bind using WW protein interaction modules . In co-expression studies , NEDD4-like proteins have been shown to bind to these membrane proteins , to increase their ubiquitination , and to promote their degradation presumably via targeting to lysosomes ( Staub et al . , 1996; Staub et al . , 1997; Abriel et al . , 1999 ) . Nevertheless , significant questions remain regarding underlying mechanisms due to complexities in the ubiquitin regulatory system that arises at three levels . First , multiple E3 ligases may converge to regulate a single substrate . Second , a single E3 ligase such as NEDD4-2 catalyzes ubiquitination of multiple substrates . Third , distinct E3 ligases can give rise to monoubiquitination or polyubiquitin chains with differing lysine chain linkages and , thus , divergent functional consequences ( Abriel and Staub , 2005; Komander , 2009; MacGurn et al . , 2012; Foot et al . , 2017 ) . The method of specifically targeting E3 catalytic domains to selected membrane proteins offers opportunities to dissect these inherent intricacies of the ubiquitin regulatory system . Using this approach , our results yield new insights into ubiquitin regulation of Q1 and CaV1 . 2 channels . First , we found that nanoCHIP increased ubiquitination of both YFP-tagged Q1 and CaV1 . 2 without significantly impacting total protein expression , demonstrating that mere ubiquitination of these channels is not sufficient to direct their degradation . By contrast , NEDD4-2 both enhances ubiquitination and decreases stability of Q1 ( Jespersen et al . , 2007 ) ( Figure 3A and Figure 3—figure supplement 1 ) . A possible explanation for this difference could lie in the fact that NEDD4-like ubiquitin ligases catalyze ubiquitination of multiple protein substrates , including components of the ESCRT complex involved in sorting membrane proteins on endosomes into multi-vesicular bodies and lysosomes ( MacGurn et al . , 2012 ) . Alternatively , the intrinsic differences in the type of ubiquitination conferred between NEDD4-2 and CHIP E3 ligase could be important . The finding that nanoNEDD4-2 was more effective than nanoCHIP in reducing Q1 stability , mirroring full-length NEDD4-2 , is consistent with this interpretation . Overall , these results emphasize that distinct ligases can differentially impact the stability and subcellular localization of ion channels . Moreover , this work illustrates how engineered E3 ligases can be utilized to systematically and selectively probe the impact of particular E3 ligases on target proteins in the complex cellular environment . The targeted E3 ligase approach is complementary to a recently developed method that uses ubiquitin variants that are selective for distinct HECT ligases , and can either activate or inhibit their cognate E3 ( Zhang et al . , 2016 ) . A second unique observation was that iN-CHIP suppressed Q1 surface density by selectively diminishing forward trafficking of the channel , with no apparent effect on the rate of endocytosis . A common assumption is that ubiquitination of membrane proteins enhances their internalization from the cell surface , although recent results suggest a more complex picture in mammalian cells . For example , the E3 ligase Cbl is important to the internalization of activated epidermal growth factor receptors ( EGFRs ) . Nevertheless , mutation of ubiquitination sites on EGFRs did not abolish their endocytosis ( Huang et al . , 2007 ) . Similarly , a ubiquitin-deficient β2 adrenergic receptor was also internalized to a similar extent as the wild type protein ( Shenoy et al . , 2001 ) . A complication in our experimental system is that Q1 displays a basal level of ubiquitination due to the activity of an endogenous E3 ligase of unknown identity . It is possible that this basal level of ubiquitination is sufficient to set a level of internalization that is not further enhanced by nanoCHIP . The iN-CHIP-induced decrease in forward trafficking is intriguing though questions remain regarding the mechanistic bases of this effect . There are endogenous ubiquitin-mediated quality control mechanisms that would be expected to limit forward trafficking of membrane proteins . For example , the ER-associated degradation ( ERAD ) pathway is a prominent quality control mechanism which is accomplished by a chaperone-mediated ubiquitination of misfolded membrane proteins which are then retrotranslocated to the cytosol and targeted to the proteasome for degradation ( MacGurn et al . , 2012; Foot et al . , 2017 ) . There are also ubiquitin-dependent quality control mechanisms present at the Golgi which results in the diversion of membrane proteins to endosomes where they are sorted by the ESCRT system into multi-vesicular bodies and fusion with lysosomes ( Fire et al . , 1998; MacGurn et al . , 2012; Foot et al . , 2017 ) . Both these quality control pathways result in the degradation of target proteins which is fundamentally different from our observations of the impact of nanoCHIP on Q1 and CaV1 . 2 channels . Ultimately , precise identification of the intracellular compartments in which nanoCHIP arrested Q1 and CaV1 . 2 channels reside will be important for deducing the mechanism of the compromised forward trafficking of these channels . Eliminating protein function by preventing expression or with pharmacological agents is a cornerstone of modern biological research and disease therapy . Several approaches have been developed to eradicate expression of target proteins by interference at the genomic ( knockout , zinc finger nucleases , TALENs , CRISPR/Cas ) or mRNA ( siRNA , shRNA , microRNA ) levels ( Fire et al . , 1998; Gaj et al . , 2013; Doudna and Charpentier , 2014; Boettcher and McManus , 2015 ) . These methods are widely used and powerful , but do have certain limitations that may be addressable with the tools developed here . The temporal control and resolution of mRNA interference methods is relatively poor because they are dependent on the degradation of the targeted native protein . For stable proteins with a long half-life this can adversely impact the efficacy of the mRNA interference approach . This gap can be potentially addressed by post-translational degradation of target proteins using engineered ubiquitin ligases . Indeed , several groups have utilized this approach to target diverse cytosolic proteins for degradation in situ . Yet none to date have applied this approach to ion channels , a specialized class of proteins that rely on a very different post-translational lifecycle of maturation , sorting , and trafficking . The first implementation of targeted ubiquitination exploited SKP1-CUL1-F-box ( SCF ) E3 ligase complexes , in which a peptide motif was fused to a modular F-box-domain-containing protein in order to target the retinoblastoma tumor suppressor protein for degradation by a Cullin/RING E3 ligase complex ( Zhou et al . , 2000 ) . Seeking a more generalizable approach , Caussinius et al fused the N-terminus F-box domain of Slmb ( a Drosophila melanogaster F-box protein ) to vhhGFP4 ( NSlmb-vhhGFP4 , also termed deGradFP ) , which effectively degraded various GFP-tagged proteins in Drosophila ( Caussinus et al . , 2011 ) . Yet because approaches that exploit SCF complexes have the potential to sequester endogenous components by over-expressed engineered F-box proteins ( Hatakeyama et al . , 2005; Portnoff et al . , 2014 ) , other studies have utilized E3 ligases that are not similarly reliant on endogenous scaffold proteins for their mechanism of action . One such approach replaced the substrate binding ( TPR ) domain of CHIP to enable targeted ubiquitination and degradation of the proto-oncogenes c-Myc ( Hatakeyama et al . , 2005 ) and K-Ras ( Ma et al . , 2013 ) . Further proof of concept experiments replaced the TPR of CHIP with intrabodies directed against β-galactosidase and maltose binding protein , which yielded effective destruction of these target proteins in transfected HEK293 cells ( Portnoff et al . , 2014 ) . More recently , a technique referred to as Trim-Away exploits the high affinity of the E3 ligase TRIM21 for the Fc domain of antibodies for selective degradation of target proteins ( Clift et al . , 2017 ) . Against this background , nanoCHIP represents a hybrid of these previously engineered E3 ligases , wherein the TPR region of CHIP is replaced with a vhhGFP4 nanobody . Without exception , all previous renditions of this technology have relied on degradation of target proteins as the ultimate expression of efficacy . In comparison , our results suggest a fundamental difference between ion channels and cytosolic proteins in which targeted ubiquitination with nanoCHIP ( as well as nanoNSlmb and nanoMDM2 ) yielded impaired trafficking and functional inactivation without frank degradation of the protein . As such , our results indicate that absolute degradation is not necessary for potent functional knockdown of Q1 and CaV1 . 2 channels , and highlight the importance of employing functional/cell biological assays to assess the efficacy of post-translational knockdown . Furthermore , we demonstrate here that nanoNEDD4-2 can selectively degrade the ion channel Q1 in situ , emphasizing the potential for customized protein manipulation with engineered E3 ligases . One advantage of this post-translational knockdown approach is the potential to uniquely manipulate ion channel macromolecular complexes . Many ion channels , including Q1 , are multi-protein complexes made up minimally of pore-forming proteins associated with accessory subunits . For example , Q1 can associate with any of five KCNE auxiliary subunits ( KCNE1-KCNE5 ) , each of which confers distinctive functional properties ( Sun et al . , 2012 ) . Moreover , some KCNE subunits may interact with other K+ channel pore-forming subunits ( Abbott et al . , 1999; Abbott , 2016 ) . Cardiac ventricular myocytes express all five KCNE subunits together with Q1 ( Radicke et al . , 2006 ) . Thus , a method to selectively inactivate protein complexes composed of specific Q1/KCNE combinations in heart cells could be powerful in illuminating the physiological logic for having multiple KCNE subunits expressed in heart cells . The capability to inactivate specific macromolecular complexes is beyond the capacity of genomic and mRNA interference approaches since simply knocking out expression of a particular KCNE could have reverberations on multiple channel types , thereby limiting specificity . By contrast , post-translational based methods such as the engineered E3 ligase approach potentially offer a platform to address this blind spot in macromolecular complex inactivation . The observation that targeting nanoCHIP to KCNE1 effectively arrests trafficking of Q1 offers rational strategies in pursuit of this goal . Finally , it is worth commenting on potential therapeutic dimensions of our findings . Gain-of-function mutations in distinct ion channels cause diverse diseases including but not limited to: ( NaV1 . 7 ) inherited erythromelalgia and paroxysmal extreme pain disorder ( Waxman , 2013 ) ; ( CaV2 . 1 ) familial hemiplegic migraine ( Pietrobon , 2010 ) ; ( CaV1 . 2 ) Timothy syndrome ( Splawski et al . , 2004 ) ; ( TrpV4 ) heritable skeletal dysplasia ( Rock et al . , 2008 ) ; ( KCNQ1 and HERG ) short QT syndrome and familial atrial fibrillation ( Giudicessi and Ackerman , 2012 ) . Targeted ubiquitination of overactive channels may provide a viable therapeutic strategy for some gain-of-function channelopathies . Although the nano-E3 ligases reported here are only effective against GFP/YFP-tagged proteins and do not target endogenous proteins , this is addressable by development of antibody mimetic proteins capable of binding specific target peptides in situ with high affinity . Various methods to develop such antibody mimetic proteins have recently emerged , including; nanobodies , single chain variable fragments ( scFv ) , DARPins , and FingRs/monobodies ( Koide et al . , 1998; Pardon et al . , 2014; Plückthun , 2015; Gross et al . , 2016; Sha et al . , 2017 ) . Furthermore , a chemical strategy has been developed that utilizes hetero-bivalent small molecules referred to as PROTACS ( proteolysis-targeting chimeras ) to bridge endogenous substrates to endogenous ubiquitin ligases ( Schneekloth et al . , 2004; Lai and Crews , 2017 ) . Our results suggest that development of PROTACS capable of selectively targeting endogenous E3 ubiquitin ligases to ion channels may be a promising therapy for diverse cardiovascular and neurological diseases .
A customized bicistronic vector ( xx-P2A-CFP ) was synthesized in the pUC57 vector , in which coding sequence for P2A peptide was sandwiched between an upstream multiple cloning site and enhanced cyan fluorescent protein ( CFP ) ( Genewiz , South Plainfield , NJ ) . The xx-P2A-CFP fragment was amplified by PCR and cloned into the PiggyBac CMV mammalian expression vector ( System Biosciences , Palo Alto , CA ) using NheI/NotI sites . To generate nano-xx-P2A-CFP , we PCR amplified the coding sequence for GFP nanobody ( vhhGFP4 ) and cloned it into xx-P2A-CFP using NheI/AflII sites . The nanoCHIP construct was created by gene synthesis ( Genewiz ) , and featured the coding sequence for GFP nanobody ( vhhGFP4 ) ( Kubala et al . , 2010 ) in frame with the minimal catalytic unit of CHIP E3 ligase ( residues 128–303 ) , separated by a flexible GSG linker . This fragment was amplified by PCR and cloned into the xx-P2A-CFP vector using NheI/AflII sites . To create the catalytically inactive nanoCHIP* , we deleted the coiled-coil domain [Δ128–229] as previously described ( Nikolay et al . , 2004 ) , by amplifying the U-box domain of CHIP E3 ligase ( residues 230–303 ) , and cloned this fragment into nano-xx-P2A-CFP using AscI/AflII sites separated by a flexible GSG linker . NSlmb:nano-P2A-CFP was derived from pcDNA3_NSlmb-vhhGFP4 ( Addgene #35579 , Cambridge , MA ) ( Caussinus et al . , 2011 ) . We PCR amplified the NSlmb-vhhGFP4 fragment and cloned it into xx-P2A-CFP using NheI/AflII sites . To generate nanoMDM2 , we PCR amplified the RING domain ( residues 432–491 ) from MDM2 ( Addgene #16233 ) ( Zhou et al . , 2001 ) and cloned this fragment into nano-xx-P2A-CFP using AscI/AflII sites . To create nanoNEDD4L we first PCR amplified the HECT domain ( residues 596–975 ) of NEDD4L ( PCI_NEDD4L; Addgene #27000 ) ( Gao et al . , 2009 ) and cloned this fragment into nano-xx-P2A-CFP using AscI/AflII sites . The resulting construct , nanoNEDD4L-P2A-CFP expressed poorly so we swapped positions of the nanoNEDD4L and CFP . We first generated CFP-P2A-xx and then PCR amplified nanoNEDD4L . The resulting fragment was cloned into CFP-P2A-xx using BglII/NotI sites . To create the catalytically inactive nanoNEDD4-2* , we introduced a point mutation at the catalytic cysteine residue [C942S] by site-directed mutagenesis . KCNQ1/E1 constructs were made as described previously ( Aromolaran et al . , 2014 ) . Briefly , overlap extension PCR was used to fuse enhanced yellow fluorescent proteins ( EYFP ) in frame to the C-terminus of KCNQ1 and KCNE1 . A 13-residue bungarotoxin-binding site ( BBS; TGGCGGTACTACGAGAGCAGCCTGGAGCCCTACCCCGAC ) ( Sekine-Aizawa and Huganir , 2004; Yang et al . , 2010 ) was introduced between residues 148–149 in the extracellular S1–S2 loop of KCNQ1 using the Quik-Change Lightning Site-Directed Mutagenesis Kit ( Stratagene , La Jolla , CA ) according to the manufacturer’s instructions . The inducible nanoCHIP construct ( FRB:CHIP-P2A-CFP-P2A-FKBP:nano ) was created in three parts . First , FRB:CHIP-P2A-CFP was created by PCR amplifying the CHIP catalytic domain and cloning the amplified fragment into FRBxx-P2A-CFP vector using AscI/AflII sites . Second , we used overlap extension PCR to create a P2A-FKBP:nano cassette which was then cloned downstream of CFP in the FRB:CHIP-P2A-CFP construct using BglII/NotI sites , generating FRB:CHIP-P2A-CFP-P2A-FKBP-nano . Adenoviral vectors were generated using the pAdEasy system ( Stratagene ) according to manufacturer’s instructions as previously described ( Subramanyam et al . , 2013; Aromolaran et al . , 2014 ) . Plasmid shuttle vectors ( pShuttle CMV ) containing cDNA for nano-P2A-CFP , nanoCHIP-P2A-CFP , and BBS-Q1-YFP were linearized with PmeI and electroporated into BJ5183-AD-1 electrocompetent cells pre-transformed with the pAdEasy-1 viral plasmid ( Stratagene ) . PacI restriction digestion was used to identify transformants with successful recombination . Positive recombinants were amplified using XL-10-Gold bacteria , and the recombinant adenoviral plasmid DNA linearized with PacI digestion . HEK cells cultured in 60 mm diameter dishes at 70–80% confluency were transfected with PacI-digested linearized adenoviral DNA . Transfected plates were monitored for cytopathic effects ( CPEs ) and adenoviral plaques . Cells were harvested and subjected to three consecutive freeze-thaw cycles , followed by centrifugation ( 2 , 500 × g ) to remove cellular debris . The supernatant ( 2 mL ) was used to infect a 10 cm dish of 90% confluent HEK293 cells . Following observation of CPEs after 2–3 d , cell supernatants were used to re-infect a new plate of HEK293 cells . Viral expansion and purification was carried out as previously described ( Colecraft et al . , 2002 ) . Briefly , confluent HEK293 cells grown on 15 cm culture dishes ( x8 ) were infected with viral supernatant ( 1 mL ) obtained as described above . After 48 hr , cells from all of the plates were harvested , pelleted by centrifugation , and resuspended in 8 mL of buffer containing ( in mM ) Tris·HCl 20 , CaCl2 1 , and MgCl2 1 ( pH 8 . 0 ) . Cells were lysed by four consecutive freeze-thaw cycles and cellular debris pelleted by centrifugation . The virus-laden supernatant was purified on a cesium chloride ( CsCl ) discontinuous gradient by layering three densities of CsCl ( 1 . 25 , 1 . 33 , and 1 . 45 g/mL ) . After centrifugation ( 50 , 000 rpm; SW41Ti Rotor , Beckman-Coulter Optima L-100K ultracentrifuge; 1 hr , 4°C ) , a band of virus at the interface between the 1 . 33 and 1 . 45 g/mL layers was removed and dialyzed against PBS ( 12 hr , 4°C ) . Adenoviral vector aliquots were frozen in 10% glycerol at −80°C until use . Human embryonic kidney ( HEK293 ) cells were a kind gift from the laboratory of Dr . Robert Kass ( Columbia University ) . Cells were mycoplasma free , as determined by the MycoFluor Mycoplasma Detection Kit ( Invitrogen , Carlsbad , CA ) . Low passage HEK293 cells were cultured at 37°C in DMEM supplemented with 8% fetal bovine serum ( FBS ) and 100 mg/mL of penicillin–streptomycin . HEK293 cell transfection was accomplished using the calcium phosphate precipitation method . Briefly , plasmid DNA was mixed with 62 μL of 2 . 5M CaCl2 and sterile deionized water ( to a final volume of 500 μL ) . The mixture was added dropwise , with constant tapping to 500 μL of 2x Hepes buffered saline containing ( in mM ) : Hepes 50 , NaCl 280 , Na2HPO4 1 . 5 , pH 7 . 09 . The resulting DNA–calcium phosphate mixture was incubated for 20 min at room temperature and then added dropwise to HEK293 cells ( 60–80% confluent ) . Cells were washed with Ca2+-free phosphate buffered saline after 4–6 hr and maintained in supplemented DMEM . Chinese hamster ovary ( CHO ) cells were obtained from ATCC ( Manassas , VA ) , and cultured at 37°C in Kaighn’s Modified Ham’s F-12K ( ATCC ) supplemented with 8% FBS and 100 mg/mL of penicillin–streptomycin . CHO cells were transiently transfected with desired constructs in 35 mm tissue culture dishes—KCNQ1 ( 0 . 5 µg ) , KCNE1 ( 0 . 5 µg ) , and nano-P2A-CFP ( 0 . 5 µg ) , and nanoCHIP-P2A-CFP ( 0 . 5 µg ) using X-tremeGENE HP ( 1:2 DNA/reagent ratio ) according to the manufacturers’ instructions ( Roche , Indianapolis , IN ) . Primary cultures of adult rat heart ventricular cells were prepared as previously described ( Colecraft et al . , 2002; Subramanyam et al . , 2013 ) , in accordance with the guidelines of Columbia University Animal Care and Use Committee . Adult male Sprague–Dawley rats were euthanized with an overdose of isoflurane . Hearts were excised and ventricular myocytes isolated by enzymatic digestion with 1 . 7 mg Liberase–TM enzyme mix ( Roche ) using a Langendorff perfusion apparatus . Healthy rod-shaped myocytes were cultured in Medium 199 ( Life Technologies ) supplemented with ( in mM ) carnitine ( 5 ) , creatine ( 5 ) , taurine ( 5 ) penicillin-streptomycin-glutamine ( 0 . 5% , Life technologies ) , and 5% ( vol/vol ) FBS ( Life Technologies ) to promote attachment to dishes . After 5 hr , the culture medium was switched to Medium 199 with 1% ( vol/vol ) serum , but otherwise supplemented as described above . Cultures were maintained in humidified incubators at 37°C and 5% CO2 . Cell surface and total ion channel pools were assayed by flow cytometry in live , transfected HEK293 cells as previously described ( Yang et al . , 2010; Aromolaran et al . , 2014 ) . Briefly , 48 hr post-transfection , cells cultured in 6-well plates gently washed with ice cold PBS containing Ca2+ and Mg2+ ( in mM: 0 . 9 CaCl2 , 0 . 49 MgCl2 , pH 7 . 4 ) , and then incubated for 30 min in blocking medium ( DMEM with 3% BSA ) at 4°C . HEK293 cells were then incubated with 1 μM Alexa Fluor 647 conjugated α-bungarotoxin ( BTX647; Life Technologies ) in DMEM/3% BSA on a rocker at 4°C for 1 hr , followed by washing three times with PBS ( containing Ca2+ and Mg2+ ) . Cells were gently harvested in Ca2+-free PBS , and assayed by flow cytometry using a BD LSRII Cell Analyzer ( BD Biosciences , San Jose , CA , USA ) . CFP- and YFP-tagged proteins were excited at 407 and 488 nm , respectively , and Alexa Fluor 647 was excited at 633 nm . Optical pulse chase assays to monitor rates of channel forward trafficking and internalization were conducted on live , transfected HEK293 cells . For the iN-CHIP treatment groups , cells were pretreated with 1 μM rapamycin for 3 hr prior to the experiments . Cells were placed on 4°C to halt trafficking processes and washed twice with PBS containing Ca2+ and Mg2+ . For forward trafficking experiments , cells were incubated with 5 µM untagged BTX in DMEM/3% BSA at 4°C for 1 hr to block surface channels , and then washed three times with PBS containing Ca2+ and Mg2+ . Cells were incubated with DMEM/3% BSA and placed at 37°C to resume trafficking for different time intervals ( 0 , 5 , 10 , 20 , 40 , 60 min ) . Cells were then returned to 4°C and newly delivered channels were labeled with 1 µM BTX647 in DMEM/3% BSA for 1 hr . Finally , cells were washed three times with PBS containing Ca2+ and Mg2+ , gently harvested in Ca2+-free PBS , and assayed by flow cytometry . For internalization experiments , cells were incubated in DMEM/3% BSA blocking medium for 30 min at 4°C , followed by a pulse of 1 μM biotinylated α-bungarotoxin ( BTX-biotin; Life Technologies ) for 1 hr with gentle rocking at 4°C . Cells were washed three times in PBS containing Ca2+ and Mg2+ and placed in DMEM/3% BSA at 37°C for different time intervals ( 0 , 5 , 10 , 20 , 40 , 60 min ) to resume trafficking . Cells were returned to 4°C , washed once with PBS , and channels remaining at the surface were labeled with streptavidin-conjugated Alexa Fluor 647 ( Life Technologies ) . Finally , cells were washed twice more with PBS with Ca2+ and Mg2+ , harvested in Ca2+-free PBS , and assayed by flow cytometry . At 48 hr post-infection , adult rat cardiomyocytes cultured on 35 mm MatTek dishes ( MatTek Corporation , Ashland , MA ) were gently washed with M199 media ( with 0 . 9 mM CaCl2 , 0 . 49 mM MgCl2 , pH 7 . 4 ) and fixed with 4% formaldehyde for 10 min at room temperature ( RT ) . Cardiomyocytes were washed three times with PBS , and incubated for 30 min in blocking medium ( M199 with 3% BSA ) . Cardiomyocytes were then incubated with 1 μM BTX-biotin in M199/3% BSA at room temperature for 1 hr followed by washing three times with PBS to remove unbound biotinylated BTX . Cells were then incubated with 10 nM streptavidin-conjugated Quantum Dot 655 ( QD655; Life Technologies ) for 1 hr at 4°C in the dark , washed three times with PBS , and imaged with Nikon Ti Eclipse inverted microscope for scanning confocal microscopy . For potassium channel measurements , whole-cell membrane currents were recorded at room temperature in CHO cells using an EPC-10 patch-clamp amplifier ( HEKA Electronics , Lambrecht/Pfalz , Germany ) controlled by the PatchMaster software ( HEKA ) . A coverslip with adherent CHO cells was placed on the glass bottom of a recording chamber ( 0 . 7–1 mL in volume ) mounted on the stage of an inverted Nikon Eclipse Ti-U microscope . Micropipettes were fashioned from 1 . 5 mm thin-walled glass and fire-polished . Internal solution contained ( mM ) : 133 KCl , 0 . 4 GTP , 10 EGTA , 1 MgSO4 , 5 K2ATP , 0 . 5 CaCl2 , and 10 HEPES ( pH 7 . 2 ) . External solution contained ( in mM ) : 147 NaCl , 4 KCl , 2 CaCl2 , and 10 HEPES ( pH 7 . 4 ) . Pipette resistance was typically 1 . 5 MΩ when filled with internal solution . I–V curves were generated from a family of step depolarizations ( −40 to +100 mV in 10 mV steps from a holding potential of −50 mV ) . Currents were sampled at 20 kHz and filtered at 5 kHz . Traces were acquired at a repetition interval of 10 s . For calcium channel measurements , whole-cell recordings were carried out in HEK293 cells at room temperature . Internal solution contained ( mM ) : 135 Cs Methanesulfonate , 5 CsCl , 5 EGTA , 1 MgCl2 , 4 MgATP , 10 HEPES ( pH 7 . 2 ) . External solution contained ( mM ) : 140 tetraethylammonium-methanesulfonate , 5 BaCl2 , 10 HEPES ( pH 7 . 4 ) . Leak and capacitive currents were subtracted using a P/4 protocol . I-V curves were generated from a family of step depolarizations ( −60 to +100 mV in 10 mV steps from a holding potential of −90 mV ) . Currents were sampled at 20 kHz and filtered at 5 kHz . Traces were acquired at a repetition interval of 10 s . HEK293 cells were washed once with PBS without Ca2+ , harvested , and resuspended in RIPA lysis buffer containing ( in mM ) Tris ( 20 , pH 7 . 4 ) , EDTA ( 1 ) , NaCl ( 150 ) , 0 . 1% ( wt/vol ) SDS , 1% Triton X-100 , 1% sodium deoxycholate and supplemented with protease inhibitor mixture ( 10 μL/ mL , Sigma-Aldrich , St . Louis , MO ) , PMSF ( 1 mM , Sigma-Aldrich ) , and PR-619 deubiquitinase inhibitor ( 50 µM , LifeSensors , Malvern , PA ) . Lysates were prepared by incubation at 4°C for 1 hr , with occasional vortex , and cleared by centrifugation ( 10 , 000 × g , 10 min , 4°C ) . Supernatants were transferred to new tubes , with aliquots removed for quantification of total protein concentration determined by the bis-cinchonic acid protein estimation kit ( Pierce Technologies , Waltham , MA ) . Lysates were pre-cleared by incubation with 10 µL Protein A/G Sepharose beads ( Rockland ) for 40 min at 4°C and then incubated with 0 . 75 µg anti-Q1 ( Alomone , Jerusalem , Israel ) for 1 hr at 4°C . Equivalent total protein amounts were added to spin-columns containing 25 µL Protein A/G Sepharose beads , tumbling overnight at 4°C . Immunoprecipitates were washed 3–5 times with RIPA buffer , spun down at 500 × g , eluted with 40 µL of warmed sample buffer [50 mM Tris , 10% ( vol/vol ) glycerol , 2% SDS , 100 mM DTT , and 0 . 2 mg/mL bromophenol blue] , and boiled ( 55°C , 15 min ) . Proteins were resolved on a 4–12% Bis·Tris gradient precast gel ( Life Technologies ) in Mops-SDS running buffer ( Life Technologies ) at 200 V constant for ~1 hr . We loaded 10 μL of the PageRuler Plus Prestained Protein Ladder ( 10–250 kDa , Thermo Fisher , Waltham , MA ) alongside the samples . Protein bands were transferred by tank transfer onto a nitrocellulose membrane ( 3 . 5 hr , 4°C , 30 V constant ) in transfer buffer ( 25 mM Tris pH 8 . 3 , 192 mM glycine , 15% ( vol/vol ) methanol , and 0 . 1% SDS ) . The membranes were blocked with a solution of 5% nonfat milk ( BioRad ) in tris-buffered saline-tween ( TBS-T ) ( 25 mM Tris pH 7 . 4 , 150 mM NaCl , and 0 . 1% Tween-20 ) for 1 hr at RT and then incubated overnight at 4°C with primary antibodies ( anti-Q1 , Alomone ) in blocking solution . The blots were washed with TBS-T three times for 10 min each and then incubated with secondary horseradish peroxidase-conjugated antibody for 1 hr at RT . After washing in TBS-T , the blots were developed with a chemiluminiscent detection kit ( Pierce Technologies ) and then visualized on a gel imager . Membranes were then stripped with harsh stripping buffer ( 2% SDS , 62 mM Tris pH 6 . 8 , 0 . 8% ß-mercaptoethanol ) at 50°C for 30 min , rinsed under running water for 2 min , and washed with TBST ( 3x , 10 min ) . Membranes were pre-treated with 0 . 5% glutaraldehyde and re-blotted with anti-ubiquitin ( VU1 , LifeSensors ) as per the manufacturers’ instructions . Data were analyzed off-line using FlowJo , PulseFit ( HEKA ) , Microsoft Excel , Origin and GraphPad Prism software . Statistical analyses were performed in Origin or GraphPad Prism using built-in functions . Statistically significant differences between means ( p<0 . 05 ) were determined using Student’s t test for comparisons between two groups . Data are presented as means ± s . e . m . | Cells are surrounded by a membrane that separates the outside of the cell from its inside . Proteins called ion channels are embedded within this membrane and allow charged ions to move in and out of the cell . The movement of ions generates electrical currents that are essential for many processes that keep us alive , including our heartbeat and the activity within our brain . Like many other proteins , newly made ion channels undergo several steps before they mature and become active . Cells destroy any proteins that do not mature properly , as well as those that become damaged or are simply no longer needed . A small protein called ubiquitin helps to mark such unwanted proteins for destruction . Enzymes known as E3 ligases attach ubiquitin to target proteins in a process known as ubiquitination . This process regulates both the quality and amount of proteins within cells . To understand the role of a particular protein , it is often necessary to remove it from the cell and then examine the consequences . In the past , researchers have harnessed the ubiquitin system to remove many kinds of proteins , but this approach had not previously been used to target an ion channel . Now , Kanner et al . set out to selectively eliminate ion channels via targeted ubiquitination . The experiments showed that previous approaches that could destroy proteins within the cell were not effective against ion channels . Kanner et al . then engineered a particular E3 ligase so that it could selectively attach ubiquitin to the desired ion channels . This approach successfully prevented the channels from reaching the cell membrane , thereby silencing the electrical currents that they normally generate . Additionally , a new tool was developed to stop ion channels in their tracks , essentially with a flip of a chemical switch . Kanner et al . then used this approach to manipulate ion channels in a highly controlled manner , within their normal environment of heart muscle cells . These new approaches form a toolset that scientists can now exploit to study diverse ion channels . In the future , the toolkit could potentially be used to develop treatments for disorders such as epilepsy , chronic pain , and irregular heartbeats , where too many channels are active or present at the cell membrane . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics",
"tools",
"and",
"resources"
] | 2017 | Sculpting ion channel functional expression with engineered ubiquitin ligases |
We lack detailed knowledge about the spatio-temporal physiological signatures of REM sleep , especially in humans . By analyzing intracranial electrode data from humans , we demonstrate for the first time that there are prominent beta ( 15–35 Hz ) and theta ( 4–8 Hz ) oscillations in both the anterior cingulate cortex ( ACC ) and the DLPFC during REM sleep . We further show that these theta and beta activities in the ACC and the DLPFC , two relatively distant but reciprocally connected regions , are coherent . These findings suggest that , counter to current prevailing thought , the DLPFC is active during REM sleep and likely interacting with other areas . Since the DLPFC and the ACC are implicated in memory and emotional regulation , and the ACC has motor areas and is thought to be important for error detection , the dialogue between these two areas could play a role in the regulation of emotions and in procedural motor and emotional memory consolidation .
A night’s sleep consists of periods of rapid eye movement ( REM ) sleep and periods of non rapid eye movement ( NREM ) sleep . We have a relatively detailed picture of the spatio-temporal pattern of activity that occurs during NREM sleep , and this has allowed detailed hypotheses concerning the circuits and mechanisms at play during NREM sleep . For example , the transfer of memories from the hippocampus to the neocortex during NREM sleep is thought to be mediated by the coordination of oscillatory activity seen in the local field potentials , namely by the interaction of hippocampal ripples ( 80–250 Hz ) with neocortical spindles ( 11–16 Hz ) and slow ( <1 Hz ) oscillations ( Battaglia et al . , 2004; Ji and Wilson , 2007; Lee and Wilson , 2002; Mölle et al . , 2006; Siapas and Wilson , 1998; Sirota et al . , 2003; Skaggs and McNaughton , 1996; Staresina et al . , 2015; Wilson and McNaughton , 1994 ) . The spatio-temporal patterns of activity during REM sleep are not as well characterized . This may be due in part to the fact that the activity patterns seen during REM sleep do not easily lend themselves to neat description like those seen during NREM sleep . During REM sleep field potential recordings are not dominated by prominent oscillations; rather they are irregular and of low voltage , similar to recordings during awake behavior ( Rasch and Born , 2013 ) . During awake activity , the oscillations that emerge during cognitive tasks are thought to play a key role in aspects of cognition such as working memory and communication between brain areas ( Başar et al . , 2001 ) . Thus a detailed understanding of the spatio-temporal characteristics of emergent oscillations during REM sleep may shed light on which areas are interacting during REM sleep , how dreams are generated , and how some of the putative functions of REM sleep , such as memory consolidation , are carried out . Little is known about oscillatory activity during REM sleep with the notable exceptions of theta oscillations in the rodent hippocampus ( Buzsáki , 2002; Louie and Wilson , 2001; Montgomery et al . , 2008 ) , gamma oscillations in the neocortex ( Castro et al . , 2013; Llinás and Ribary , 1993; Steriade et al . , 1996 ) , and pontine-geniculate-occipital ( PGO ) waves ( Calvo and Fernández-Guardiola , 1984; Datta , 1997; Datta and Hobson , 1994; Steriade et al . , 1989 ) . We decided to focus on the oscillatory dynamics of the frontal cortices during REM sleep based on imaging studies and REM sleep’s putative functional roles . Imaging studies suggest that activity patterns in certain regions of the frontal cortices are markedly different during REM sleep than during awake activity , with some areas ( e . g . , the anterior cingulate ( ACC ) ) being more active and some ( e . g . , the dorsolateral prefrontal cortex ( DLPFC ) ) less active ( Braun et al . , 1997; Maquet et al . , 1996 ) . These differences in activity patterns suggest that important REM sleep processes might be taking place in the frontal cortices ( Braun et al . , 1997; Maquet et al . , 1996; Muzur et al . , 2002 ) . At a functional level , REM sleep has been hypothesized to play a role in such processes as the consolidation of emotional and procedural motor memories ( Fischer et al . , 2002; Gilson et al . , 2016; Nishida et al . , 2009; Nitsche et al . , 2010; Rasch and Born , 2013; Smith , 2001 ) and the regulation of emotions . Since the frontal cortices serve memory , emotional regulation ( Buhle et al . , 2014; Etkin et al . , 2011 , 2006 ) , and motor functions , they make good candidate areas for the mediation of procedural motor memory and emotional memory consolidation as well as emotional regulation during REM sleep . We show that the frontal cortices are dominated by bursts of beta activity and theta oscillations , in particular in the ACC and the DLPFC . The prominent oscillatory activity in the DLPFC is especially surprising since current thinking , based on imaging studies ( Braun et al . , 1997; Maquet et al . , 2000; Muzur et al . , 2002 ) , is that the DLPFC is relatively quiet during REM sleep in comparison with awake periods , and it has been postulated that the relative quiescence of this ‘executive’ structure might be responsible for the bizarre nature of our dreams . We also show the novel result that beta bursts and theta oscillations during REM sleep are coherent between the DLPFC and the ACC , two relatively distant structures . This finding is especially intriguing since the ACC and the DLPFC are both implicated in memory and emotional processing ( Buhle et al . , 2014; Etkin et al . , 2011 , 2006; Goldman-Rakic , 1995 ) , and the ACC is interconnected with motor areas ( Bates and Goldman-Rakic , 1993; Dum and Strick , 2002 , 1991; Hatanaka et al . , 2003; Morecraft and Van Hoesen , 1993 , 1992; Picard and Strick , 1996 ) , is thought to have motor regions itself ( Paus , 2001 ) , and is thought to be important in error detection ( Holroyd and Yeung , 2011; Paus , 2001 ) . Therefore the rhythmic activity we observe in the frontal cortices may play a role in some of the hypothesized functions of REM sleep such as the consolidation of emotional and procedural motor memories and the regulation of emotions .
We examined overnight sleep recordings of five patients with intractable epilepsy undergoing invasive recordings to localize the focus of their seizures . Patient information is in Table 1 . We used electroencephalography ( EEG ) , electromyography ( EMG ) , and electrooculography ( EOG ) to score the sleep recordings in 30 s epochs as either N1 , N2 , N3 ( NREM stages ) , REM , or Awake using standard criteria ( Iber et al . , 2007 ) . According to these criteria , REM periods are identified by low EMG power ( the lowest of a night’s sleep ) , irregular EEG of low voltage , and conjugate eye movements ( Figure 1a , b ) . Identified REM periods were selected for further analysis . 10 . 7554/eLife . 18894 . 003Figure 1 . Sleep recordings and electrode localization . ( a ) EMG , delta band , and spindle band power , and the cross-covariance of the EOG leads during a portion of a night’s sleep from one subject ( bin size 30 s ) . The dotted box indicates a period of REM sleep . During this REM period , EMG , delta , and spindle band powers are all relatively low and the sums of the cross-covariance functions over the 30 s bins are negative . The EOG montages are arranged such that rapid eye movements result in voltage deflections of the opposite polarity resulting in negative cross-covariance values during rapid eye movements . ( b ) Voltage traces of the EMG , scalp EEG , and EOG ( purple trace for left eye , red trace for right eye ) during a 30 s period from the REM episode demarcated by the dotted box in ( a ) . Note that the EMG is of relatively low voltage , the EEG is of low voltage and irregular , and there are rapid eye movements in the EOG with the deflections in the right and left EOG of opposite polarity . ( c ) A coronal MRI image showing the locations of the electrode contacts ( red dots ) of a frontal electrode in one of the subjects . ( d ) A sagittal MRI image showing the location of the most medial contact in ( c ) , which is located in the anterior cingulate . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 00310 . 7554/eLife . 18894 . 004Table 1 . Patient Information . Each row provides a patient’s demographic information and diagnostic information about their epilepsy . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 004Subject Gender Age Handedness Diagnosis Imaging 1F45RBilateral temporal lobe epilepsyMRI-normalPET-hypometabolism right side2F55RBilateral temporal lobe epilepsyMRI-mild T2 hyperintensity and volume loss ( right greater than left ) in the mesial temporal lobes3F45RMultifocaltemporal parietal-occipital epilepsyMRI- multifocal FlAIR abnormalities - second lymphoma ? 4F42AmbidextrousMultifocal epilepsy with involvement of bilateral temporal lobesMRI- nonspecific T2 hyperintensitiesPET-without clear lesion5M53LeftFocal epilepsy etiology and localization unknownMRI- normalPET- normal All patients were monitored using intracranial depth electrodes whose distal-most contacts targeted the medial temporal and frontal regions ( Figure 1c , d ) . Many of the frontal electrodes showed a marked change in the spectral content during REM sleep onset . In particular , REM sleep onset resulted in the appearance of prominent bursts of beta activity . Figure 2a depicts the spectrogram from an exemplar frontal electrode . The activity shown here and in other figures ( and the activity included in all analyses ) was detected by bipolar derivations from two neighboring contacts; thus this physiological activity is relatively local in nature . As REM sleep begins there is a change in the spectral content; beta bursts appear and then occur throughout the REM sleep episode . To get an impression of the global occurrence of the beta bursts we identified periods during which there was a prominent beta activity on a given electrode ( three standard deviations above the median power and at least two cycles in length ) . This analysis suggested that the beta bursts indeed tended to occur in the frontal electrodes after onset and throughout the REM sleep episode , but not in temporal lobe electrode leads ( Figure 2b ) . Furthermore , the beta bursts seemed to occur in close proximity to one another across the frontal electrodes . 10 . 7554/eLife . 18894 . 005Figure 2 . Oscillatory activity across scales . ( a ) Spectrogram shows the activity pattern detected from a bipolar derivation of an electrode located in the frontal cortices during REM sleep . The black line indicates the beginning of the REM sleep episode . ( b ) Pattern of beta activity across all the electrode contacts of one subject . There are two frontal depth electrodes ( Frontal 1 and Frontal 2 ) , two temporal depth electrodes ( Temporal 1 and Temporal 2 ) , and one parietal depth electrode ( Parietal ) . Each depth electrode has eight contacts . Each row represents activity from one bipolar derivation . Blue dots indicate periods in which beta power in a given contact was 3 . 0 standard deviations above the median . The black line indicates the beginning of the REM sleep episode . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 005 To determine where exactly the beta bursts were prominent within the frontal cortices , and to determine the prominent spectral features of the frontal electrodes , we focused on specific regions within the frontal cortices . We first examined the activity of contacts in the DLPFC ( See Methods for electrode localization details ) , an area that imaging studies suggest is relatively inactive during REM sleep relative to awake periods ( Braun et al . , 1997; Maquet et al . , 1996; Muzur et al . , 2002 ) . Surprisingly , the DLPFC contacts showed bursts of oscillatory activity throughout each period of REM sleep , with activity occurring most prevalently in the beta and theta bands ( Figure 3a , b ) . All of the DLPFC electrode contacts ( 15/15 ) showed significant theta activity ( t test , p<0 . 05/15 , Bonferroni corrected for number of electrodes ( 15 ) , median 5 . 2 Hz , Figure 3c , d ) , while ~87% ( 13/15 ) of the electrodes showed significant beta activity ( median 20 . 9 Hz , Figure 3c , d ) . The frequency range of the beta activity could be clearly distinguished from the frequency range of the spindling activity that occurs during NREM sleep ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 18894 . 006Figure 3 . Oscillatory activity in the DLPFC . ( a ) Spectrogram shows the activity pattern of an electrode in the DLPFC during a period of REM sleep . The lower panel shows the voltage trace during the period demarcated by the dotted lines in the spectrogram . ( b ) Average power spectrum across all subjects ( green trace ) of electrodes located in the DLPFC . The red trace is the average of the fits of the power spectra to the model a*fb and the black traces are ± the standard error of the mean . ( c ) Whisker plot of the peak frequency in the beta band ( top ) and peak frequency in the theta band ( bottom ) for electrodes showing significant power in those bands . ( d ) Bar graph showing the fraction of total electrodes in the DLPFC with significant beta or theta power . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 00610 . 7554/eLife . 18894 . 007Figure 3—figure supplement 1 . 1/f fit of power spectrum . ( a ) Example power spectrum during REM period . ( b ) Power spectrum ( green trace ) and 1/f fit ( red trace ) , which was determined using a robust least-squares regression . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 00710 . 7554/eLife . 18894 . 008Figure 3—figure supplement 2 . Comparison of REM and NREM power spectra . ( a ) Power spectra during NREM sleep and REM sleep from electrode contacts in the DLPFC . Each panel is from a different subject . Note the clear peak in the spindling band ( 11–16 Hz ) in the NREM power spectrum traces . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 008 We then examined electrode contacts in the ACC , which is thought to be relatively active during REM sleep . They also showed significant activity in the beta and theta bands ( Figure 4a , b ) . All of the ACC contacts ( 22/22 ) showed significant beta activity ( p<0 . 05/22 , Bonferroni corrected for number of electrodes ( 22 ) , median 19 . 7 Hz , Figure 4c , d ) and ~86% ( 19/22 ) showed significant theta activity ( median 4 . 6 Hz , Figure 4c , d ) . Significant activity in the beta and theta bands was seen not only in the DLPFC and ACC areas of the frontal cortices , but also in the inferior frontal gyrus ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 18894 . 009Figure 4 . Oscillatory activity in the ACC . Same as Figure 3a–d , but for electrode contacts in the ACC . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 00910 . 7554/eLife . 18894 . 010Figure 4—figure supplement 1 . Oscillatory activity in the IFG . ( a ) Average power spectrum averaged across all subjects ( green trace ) for electrodes located in the IFG . The red trace is the average of the fits of the power spectra to the model a*f^b and the black traces are ± the standard error of the mean . ( b ) Whisker plots of the peak frequencies in the beta band ( top ) or theta band ( bottom ) for electrodes showing significant power in those bands . ( c ) Bar graph of the fraction of total electrodes in the IFG showing significant beta or theta power . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 010 We next explored whether theta and beta activity is as prevalent outside the frontal cortices . Theta activity was prevalent in the medial temporal regions , but not beta activity . For example , in the medial temporal gyrus we found ~58% ( 11/19 ) of the electrode contacts had prominent theta activity ( ( p<0 . 05/19 , Bonferroni corrected for number of electrodes ( 19 ) , median 6 . 1 Hz , Figure 5a–c ) , but only ~16% ( 3/19 ) had prominent beta activity ( median 17 . 9 Hz , Figure 5a–c ) . 10 . 7554/eLife . 18894 . 011Figure 5 . Oscillatory Activity Outside the Frontal Cortices . ( a–c ) Same as Figure 3b–d , but for electrode contacts in the middle temporal gyrus . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 011 We then examined whether the DLPFC and the ACC interacted , since bursts of oscillatory activity appeared to occur in concert between the two areas ( Figure 6a ) . We found that a large number ( 23/31 , ~74% ) of ACC-DLPFC electrode pairs showed a significant coherence in the beta band ( p<0 . 05 , Bonferroni corrected for number of electrode pairs ( 31 and 34 for beta and theta respectively ) , Figure 6b , c ) and 26/34 ( ~76% ) showed a significant coherence in the theta band ( Figure 6b , c ) . We found the same results whether we employed the confidence levels provided by Chronux or a Monte Carlo simulation method ( see Methods for more details ) . 10 . 7554/eLife . 18894 . 012Figure 6 . Simultaneous oscillatory activity in the DLPFC and ACC . ( a ) Spectrograms showing the simultaneous activity patterns of an electrode in the ACC ( left ) and an electrode in the DLPFC ( right ) during a period of REM sleep . Lower panels show the voltage traces during the periods demarcated by the dotted lines in the spectrograms . ( b ) Coherence during REM sleep between the DLPFC electrode depicted above and the two ACC contacts on the ipsilateral side of the DLPFC contact . The coherence plot on the left is for the ACC contact depicted above . ( c ) Bar graph of the fraction of total ACC-DLPFC electrode pairs showing significant theta or beta coherence . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 012 We then attempted to determine if either the beta or the theta activity occurs first in either the DLPFC or the ACC , on average . We looked at the cross-covariance in theta power and in beta power between all electrode pairs that showed significant coherence . Both theta and beta cross-covariance functions showed a central peak . The theta peak occurred at approximately 22 ms ( Figure 7a ) , suggesting that theta activity on average tended to occur first in the DLPFC . The peak of the beta cross-covariance function occurred very close to zero but was slightly positive ( 2 ms , Figure 7b ) , suggesting that the beta activity occurred nearly simultaneously in the DLPFC and ACC but that on average it occurred very slightly earlier in the DLPFC . To further examine the difference in the timing of the occurrence of beta and theta activity , we quantified the asymmetry in the cross-covariance function in the −100 to 100 ms window . In particular , we subtracted the area under the cross-covariance curve from −100 to 0 ms from that of 0 to 100 ms . A positive value would suggest that DLPFC activity preceded ACC activity and a negative value the opposite . We found that 69% ( 18/26 , Figure 7c ) of electrode pairs had a positive value with respect to theta , while 65% ( 15/23 , Figure 7d ) of electrode pairs had a positive value with respect to beta , suggesting that although on average theta activity and beta activity tended to occur first in the DLPFC , there was a fair amount of variability in timing among the electrodes . 10 . 7554/eLife . 18894 . 013Figure 7 . ACC and DLPC oscillatory power relationships . ( a ) Cross-covariance of theta power across ACC-DLPFC electrode pairs ( solid black trace ) . The dotted red lines correspond to the standard error of the mean and the dotted black line indicates the zero lag time point . ( b ) Same as ( a ) but for beta power . ( c ) Histogram of the difference between the area under the theta power cross-covariance function from −100 ms to 0 ms and the area under the function from 0 ms to 100 ms . ( d ) Same as ( c ) but for the beta power cross-covariance function . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 013
Imaging studies , in particular Positron Emission Tomography ( PET ) studies , suggest that the DLPFC is relatively quiescent during REM sleep in comparison with quiet awake periods ( Braun et al . , 1997; Maquet et al . , 1996; Muzur et al . , 2002 ) . The relative silence of the DLPFC , a structure thought to be important for executive functions , during REM sleep has been postulated to underlie the often bizarre and illogical nature of our dreams ( Muzur et al . , 2002 ) . Our data suggest that the DLPFC is actually active during REM sleep , since it displays significant theta and beta oscillations ( Figure 3a , b , d ) . Furthermore , the DLPFC might be communicating with other structures , since theta and beta bands in the DLPFC are coherent with those in the ACC ( Figure 6a–c ) . The DLPFC might communicate directly with the ACC or it might be driven by a third structure that coordinates the activity of the DLPFC with other areas . A discrepancy in brain activity levels assessed by PET studies and electrophysiological techniques has also occurred with respect to sleep spindles ( Dang-Vu et al . , 2010 ) . PET studies suggest that the thalamus is relatively quiet during sleep spindles ( Hofle et al . , 1997 ) , while a large body of animal studies shows that spindles are generated precisely by thalamocortical interactions ( McCormick and Bal , 1997; Steriade et al . , 1985 , 1987 ) . One possibility is that during REM sleep the periods outside the oscillatory events may show less overall neural activity than is seen during awake periods , so that PET measurements , which average over relatively large periods of time ( on the order of minutes ) , indicate that the DLPFC is less active during REM sleep than during quiet awake periods . Another possibility is that the oscillatory activity observed in the DLPFC during REM sleep may manifest under relatively inhibited conditions; the connectivity patterns between the DLPFC and the ACC suggest that the DLPFC may receive more inhibition from the ACC under the physiological conditions of REM sleep in comparison with awake conditions ( Medalla and Barbas , 2012 ) . What functional role sleep plays has yet to be resolved . There is a large body of evidence suggesting that NREM sleep may play an important role in memory consolidation ( Rasch and Born , 2013 ) , but whether or not REM sleep also plays a role in memory consolidation remains unclear ( Dudai et al . , 2015; Rasch and Born , 2013 ) . Some animal and human studies suggest that REM sleep plays a role in procedural and emotional memory consolidation ( Fischer et al . , 2002; Fu et al . , 2007; Karni et al . , 1994; Louie and Wilson , 2001; Nishida et al . , 2009; Nitsche et al . , 2010; Popa et al . , 2010; Smith , 2001; Wagner et al . , 2001 ) , but other studies , especially human studies , do not corroborate these findings and suggest that REM sleep may be important for other cognitive functions ( Cai et al . , 2009; Walker et al . , 2002b ) such as emotional regulation ( Baran et al . , 2012 ) while stage 2 NREM sleep may be more important for the consolidation of procedural memories ( Mantua et al . , 2016; Tamaki et al . , 2008; Walker et al . , 2002a ) . Whether these differential findings arise because of the different natures of the tasks , differences across species , or simply because REM sleep plays little or no role in memory consolidation remains to be determined . Below we discuss our findings in light of previously known facts about the structures and oscillations at play during REM sleep . Our discussion is driven by the perspective that REM oscillatory activity may play an important role in memory , and we put forward hypotheses for how the rhythmic activity that we have observed during REM sleep could serve a role in memory consolidation , but we note that rhythmic activity during REM sleep could serve some other function or possibly no function at all . We observed rhythmic activity in the beta and theta bands that was coherent between the ACC and the DLPFC , two relatively distant structures that are reciprocally connected and both implicated in memory ( Goldman-Rakic , 1995 ) . It is possible that this activity could be important in the procedural motor memory consolidation since the ACC is interconnected with motor areas ( Bates and Goldman-Rakic , 1993; Dum and Strick , 2002 , 1991; Hatanaka et al . , 2003; Morecraft and Van Hoesen , 1993 , 1992; Picard and Strick , 1996 ) , is thought to have motor regions itself ( Paus , 2001 ) , and is thought to be important in error detection ( Holroyd and Yeung , 2011; Paus , 2001 ) . In humans , beta activity has been shown to follow PGO waves during REM sleep in the subthalamic nucleus of the basal ganglia , so the beta activity observed in our work could result from the transmission of the PGO-related beta activity from the basal ganglia ( Amzica and Steriade , 1996; Fernández-Mendoza et al . , 2009 ) ; pontine wave density has been correlated with sleep-related improvements in memory tasks and the expression of plasticity-related genes ( Datta , 2000; Datta et al . , 2008 ) . This possibility is in line with our findings that the observed beta activity tends to manifest at approximately the same time in both the ACC and the DLPFC , suggesting that it may originate from a subcortical source . Such a scenario leads to the intriguing possibility that PGO waves may trigger an evaluation of a motor plan by a dialogue occurring via the beta band channel ( Figure 8 ) . 10 . 7554/eLife . 18894 . 014Figure 8 . Schematic of Hypothesized Circuit Involved in REM Sleep Memory Consolidation . PGO waves originating from the PPN may trigger beta activity in the basal ganglia . This beta activity in the basal ganglia may then be transmitted to cortical areas including the ACC and DLPFC . The triggering of the beta activity by the PGO waves may begin a dialogue between the basal ganglia , ACC , and DLPC ( as well as other areas ) through the beta band in which a motor plan is evaluated , ultimately leading to motor memory consolidation . Theta activity may be a channel through which the hippocampus provides contextual and spatial information about a given memory to the ACC and other parts of the frontal cortex . ( PPN = Pedunculopontine tegmental nucleus and Hipp = Hippocampus ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18894 . 014 With respect to theta , experimental studies suggest that the hippocampus provides contextual and spatial information to the cingulate and prefrontal cortex via the theta band ( Jones and Wilson , 2005; Remondes and Wilson , 2013 ) and disrupting hippocampal theta activity during REM sleep results in contextual memory deficits ( Boyce et al . , 2016 ) . Therefore , the theta band may be a channel through which spatial content , e . g . , the location at which a procedural motor task was acquired , is provided to the ACC . With respect to emotional memories the theta band may be a channel through which the amygdala provides the emotional content or valence of a particular memory to the ACC since both human and animal studies suggest that prefrontal theta and the interaction of prefrontal theta activity with both the hippocampus and the amygdala may be important for the consolidation of emotional memories during REM sleep ( Fu et al . , 2007; Nishida et al . , 2009; Popa et al . , 2010; Rasch and Born , 2013; Wagner et al . , 2001 ) . Furthermore , emotional information from the limbic system conveyed via the theta band may bias the processing and consolidation of procedural memories with high emotional content ( Popa et al . , 2010 ) . Note that many different functions have been attributed to the ACC ( Paus , 2001 ) , so its activation during REM sleep does not necessarily mean that it serves the function of emotional or procedural memory consolidation during REM sleep . We generally lack detailed knowledge about the physiological characteristics of REM sleep , especially in humans . Recent intracranial studies in humans have begun to provide us with more detailed information about the physiological processes underlying REM sleep ( Andrillon et al . , 2015; Cantero et al . , 2003 ) . These studies have primarily focused on REM-related processes ( both at the unit and LFP level ) in the medial temporal lobe . We build on this knowledge base by examining the spatio-temporal characteristics of rhythmic activity in the frontal cortices . Here we have made significant inroads toward better understanding the physiological processes underlying REM sleep through the discovery of a frontal beta-theta network and we hypothesize that this network might mediate memory consolidation during REM sleep .
Data were collected from five patients ( four females and one male ) between the ages of 42 and 55 years ( mean 48 . 0 ± 5 . 6 ( SD ) years ) undergoing invasive monitoring using intracranial depth electrodes for intractable epilepsy ( see Table 1 ) . Each depth electrode had 6 to 8 contacts . Subject 1 had 10 depth electrodes ( two anterior temporal ( bilateral ) , two posterior temporal ( bilateral ) , and six frontal [bilateral] ) , subject 2 had eight depth electrodes ( two anterior temporal ( bilateral ) , two posterior temporal ( bilateral ) , four frontal [bilateral] ) , subject 3 had five depth electrodes ( one anterior temporal , one posterior temporal , two frontal , and one parietal ) , subject 4 had 10 depth electrodes ( two anterior temporal ( bilateral ) , two posterior temporal ( bilateral ) , and six frontal [bilateral] ) , and subject 5 had 10 depth electrodes ( two anterior temporal ( bilateral ) , two posterior temporal ( bilateral ) , and six frontal [bilateral] ) . The placement of the electrodes was determined strictly by clinical criteria . The patients gave informed consent . The research protocol was approved by the Partners Human Research Committee . The contacts on the depth electrodes were spaced by 10 mm and had an impedance of ~100 Ω . The electrophysiological signals were sampled at 500 Hz ( four subjects ) or at 512 Hz ( one subject ) . The locations of the depth electrode contacts were determined using a preoperative high-resolution T1-weighted magnetic resonance imaging ( MRI ) scan and a postoperative computerized tomography ( CT ) scan . The preoperative MRI scan was used to generate a 3D rendering of the brain using FreeSurfer ( Dale et al . , 1999; Fischl et al . , 1999 ) ; this rendering was placed in a 3D coordinate system ( RAS ) . The postoperative CT scan was then coregistered with the 3D rendering , thus providing a RAS coordinate for each electrode ( Dykstra et al . , 2012 ) . FreeSurfer automatically generated subcortical and cortical parcellations with anatomical labels using the Desikan-Killiany Atlas ( Desikan et al . , 2006; Fischl et al . , 2002 ) . Using these parcellations and the RAS coordinates , we determined the anatomical location of each electrode contact . We considered contacts labeled as being in rostral middle frontal by FreeSurfer according to the Desikan-Killiany Atlas ( Desikan et al . , 2006 ) to be in the DLPFC . We considered the inferior frontal gyrus to consist of contacts labeled as parstriangularis , parsopercularis , and parsorbitalis . Contacts labeled as middle temporal gyrus were categorized as such . The locations of the contacts were also determined by a trained neurologist . In the case of the midline contacts ( in relation to this paper , the anterior cingulate contacts ) there was some discrepancy between the locations of the contacts determined by the neurologist and by the FreeSurfer labeling scheme . The locations determined by the neurologist were used . We employed bipolar derivations using two abutting contacts so that we could be sure the activity being analyzed was relatively local in nature . Only electrode derivations for which both contacts had labels within the same defined region and such that at least one of the contacts was in the gray matter were used for subsequent analyses . We did not use bipolar derivations with contacts in two different regions since in these cases it was not possible to determine from which region the activity originated . In addition , contacts that showed epileptic activity were eliminated from subsequent analyses ( 36 bipolar derivations were eliminated by this criterion ) . After eliminating bipolar derivations based on these criteria , we were left with 15 bipolar derivations in the DLPFC , 22 in the ACC , 12 in the inferior frontal gyrus , and 19 in the medial temporal gyrus . In other defined regions , there were relatively few bipolar derivations that met the criteria for inclusion in our analyses , and not all subjects had bipolar derivations in these regions . Scalp electroencephalography ( EEG ) , electromyography ( EMG ) , and electrooculography ( EOG ) were used to score sleep . Sleep was first scored using custom written software ( https://github . com/svijayan9/Sleep-Viewer ) that incorporated the criteria specified by the American Association of Sleep Medicine for sleep scoring ( Iber et al . , 2007 ) . Those periods specified as REM sleep were subsequently examined by eye and verified as REM sleep epochs by experienced sleep scorers ( SSC ) . We also verified via video recordings that the patients were asleep during identified REM sleep epochs . The subjects spent 73 . 18% ± 6 . 82% ( SEM ) of the time in NREM sleep , 16 . 23% ± 2 . 85% of the time in REM sleep , and 10 . 59% ± 1 . 78% of the time awake after sleep onset ( WASO ) , with a total sleep time of 378 . 13 ± 22 . 67 min . We used conservative measures in identifying REM periods for the purposes of further analysis . We chose relatively long ( >5 min ) REM periods that were not interrupted by non-REM sleep stages . Furthermore , we selected the beginning and the end of each REM period ( first and last 10 s ) such that there were no other stages or transition stages present . Thus , our results were not influenced by the spectral content from any other sleep stages . Those electrode contacts identified to be in the epileptic focus or identified as having interictal activity by the epileptologists were excluded from further analysis . In addition , nights during which seizures occurred were excluded from further analysis . All analyses were done using a bipolar montage , specifically by subtracting adjacent contacts . REM episodes were divided into 10 s windows . For each electrode , a power spectrum estimate , Se , j ( f ) , was calculated for each of the ten-second windows using the Chronux ( Bokil et al . , 2010 ) toolbox for Matlab ( MathWorks ) . Here , e is the electrode number ( e∈{ 1 , … , n } ) and j specifies the trial number ( j∈ {1 , … , J} ) . For each 10 s window , the power spectrum estimate was fit as a function of frequency , f , to the model afb , using a robust least-squares regression to obtain Se , jr ( f ) , the robust fit for a given electrode and trial ( see Figure 3—figure supplement 1 ) . For each frequency band of interest ( e . g . , f ∈ [4 , 8] for theta ) , the mean values S¯e , j and S¯e , jrwere calculated for each 10 s window for both the spectrum estimates Se , j ( f ) and the fitted curves Se , jr ( f ) . For each electrode , a t-test was used to determine whether the set of mean values of the spectrum estimates for a given band was significantly different ( p<0 . 05 , Bonferroni corrected for number of electrodes ) from the set of mean values of the fitted curves . We used this method rather than a comparison with awake activity since during awake conditions the calculated power spectrum would depend on the cognitive state and the task being performed . See Figure 3—figure supplement 2 for power spectrum comparisons between NREM sleep and REM sleep , which demonstrate that the spectral content is very different between the two states and show the clear distinction between spindling activity during NREM sleep and beta activity during REM sleep . Note that all subjects had at least 21 min of REM sleep , resulting in over 125 10 s windows per electrode in each subject . Only electrode contacts showing significant power in the band of interest were used for subsequent analysis . However , this resulted in the elimination of very few electrodes . In the DLPFC no electrodes were eliminated from the theta band analyses ( 0/15 ) and two electrodes were eliminated from the beta band analyses ( 2/15 ) . In the ACC no electrodes were eliminated from the beta band analyses ( 0/22 ) and three electrodes were eliminated from the theta band analyses ( 3/22 ) . We estimated coherence values using the Chronux toolbox . Coherence was examined for 34 ACC-DLPFC electrode pairs in the theta band ( 3 ACC electrodes were eliminated because theta power was not significant ) and 31 ACC-DLPFC electrode pairs in the beta band ( 2 DLPFC electrodes were eliminated because beta power was not significant ) . We estimated the coherence during REM sleep by first dividing all REM sessions of a subject into 2 s windows . All 2 s intervals of data were included in a single coherence calculation for that subject using the coherence formula employed by Chronux ( see equations 7 . 77 and 7 . 80 in Mitra and Bokil , 2008 ) . The time-bandwidth was set such that the coherence estimate for the center band ( e . g . , 6 Hz for theta ) integrated the entire band of interest . We assessed significance using two methods: by using the confidence level provided by Chronux and by doing Monte Carlo simulations ( 2000 shuffles ) . The confidence level provided by Chronux is based on the distribution of the statistic ( i . e . , the coherence measure ) given the null hypothesis of zero coherence ( Bokil et al . , 2010; Brillinger , 2001; Jarvis and Mitra , 2001 ) . The distribution of the estimator is given by the following formula . p ( |C|2 ) =1m−1 ( 1−|C|2 ) m−2 In the formula m is the degrees of freedom , which is equal to the number of 2 s intervals multiplied by the number of data tapers used in the calculation . See Brillinger ( 2001 ) for details regarding the asymptotic convergence of the coherence estimator under the hypothesis of zero coherence . For the Monte Carlo simulations , a shuffle was done by permuting the 2 s windows of one of the electrode pairs and then calculating the coherence . Using the distribution of coherences from the shuffles we determined the probability of getting the coherence value we got using our data . Both methods identified the same set of DLPFC-ACC electrode derivation pairs as having significant ( p<0 . 05 , Bonferroni corrected for number of electrode pairs ) beta or theta coherence . The cross-covariance of the power was calculated by first filtering the raw signal in the band of interest . The instantaneous amplitude was then calculated using the Hilbert transform of the filtered signal . Using these calculated instantaneous amplitudes the cross-covariance of the power ( amplitude ) of an ACC-DLPFC electrode pair was calculated for each two-second REM sleep window . For a given electrode the cross-covariance functions for all two-second windows were averaged to obtain an average cross-covariance function . Only those REM windows in which both electrodes had power in the top quartile in the band of interest with respect to the REM episode were used for the analysis . We also restricted our analysis to those electrode pairs that showed a significant coherence in the band of interest . Finally the average cross-covariance functions for all electrodes were averaged to obtain an overall average cross-covariance function . The max lag reported is simply the time lag at the maximum value of the overall average cross-covariance function . We used 23 ACC-DLPFC bipolar electrode pair derivations for the beta band cross-covariance analysis and 26 ACC-DLPFC bipolar electrode pair derivations for the theta band cross-covariance analysis ( we omitted eight electrode pairs from the theta and beta analyses ) . | Over the course of a night we cycle through several different stages of sleep . During one of these stages , our eyes move rapidly from side to side behind our closed eyelids . This movement gives this stage its name: rapid eye movement sleep , or REM sleep for short . Most other muscles are paralyzed during REM sleep , possibly to prevent us from acting out the vivid dreams that also occur during this stage of sleep . But despite the distinctive properties of REM sleep , relatively little is known about about why we need it or how the brain generates it . Vijayan et al . have now obtained new insights into the brain activity that underlies REM sleep by recording from the brains of human patients with epilepsy . The patients all had electrodes temporarily inserted into their brains to help neurologists identify the area of the brain that was responsible for their seizures . By recording from these electrodes overnight , Vijayan et al . were able to study the activity of individual brain regions while the patients slept . Analysis of the recordings revealed rhythmic waves of neuronal activity in areas at the front of the brain during REM sleep . Two types of brain waves dominated: theta waves , which are relatively slow waves with a frequency of 4–8 cycles per second ( Hertz ) , and beta waves , which are faster with a frequency of 15–35 Hertz . These theta and beta waves were especially pronounced in two subregions of the frontal lobe of the brain , called the dorsolateral prefrontal cortex ( DLPFC ) and the anterior cingulate cortex ( ACC ) . The discovery of prominent rhythmic activity in the DLPFC was unexpected . This is because previous studies had shown that this region , which is involved in decision-making and planning , was relatively inactive during REM sleep . Indeed it had been suggested that the limited activity of the DLPFC subregion might be responsible for the often bizarre and illogical nature of our dreams . Instead , Vijayan et al . showed that the ACC and the DLPFC coordinate their activity during REM sleep . The next challenge is to find out whether this dual activity helps support other roles that the two regions share in common , such as the strengthening of memories and the regulation of emotions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Frontal beta-theta network during REM sleep |
Sleep deprivation has marked effects on food intake , shifting food choices toward energy-dense options . Here we test the hypothesis that neural processing in central olfactory circuits , in tandem with the endocannabinoid system ( ECS ) , plays a key role in mediating this relationship . We combined a partial sleep-deprivation protocol , pattern-based olfactory neuroimaging , and ad libitum food intake to test how central olfactory mechanisms alter food intake after sleep deprivation . We found that sleep restriction increased levels of the ECS compound 2-oleoylglycerol ( 2-OG ) , enhanced encoding of food odors in piriform cortex , and shifted food choices toward energy-dense food items . Importantly , the relationship between changes in 2-OG and food choices was formally mediated by odor-evoked connectivity between the piriform cortex and insula , a region involved in integrating feeding-related signals . These findings describe a potential neurobiological pathway by which state-dependent changes in the ECS may modulate chemosensory processing to regulate food choices .
Sleep deprivation profoundly impacts food choices . When individuals are sleep-deprived , their dietary behavior shifts toward increased consumption of foods high in sugar and fat , leading to weight gain ( Markwald et al . , 2013; Nedeltcheva et al . , 2009 ) . These effects on ingestive behavior are likely related to sleep-dependent changes in appetite-regulating compounds , including ghrelin ( Rihm et al . , 2019; Spiegel et al . , 2004b ) , leptin ( Spiegel et al . , 2004a ) , and endocannabinoids ( Hanlon et al . , 2016 ) . Indeed , the endocannabinoid system ( ECS ) exerts strong effects on food intake ( Bellocchio et al . , 2010; Di Marzo et al . , 2001 ) , and levels of the endocannabinoid 2-arachidonoylglycerol ( 2-AG ) and its structural analog 2-oleoylglycerol ( 2-OG ) are enhanced in sleep-deprived individuals ( Hanlon et al . , 2016 ) . While previous studies have tested the effects of sleep deprivation on the human brain ( Greer et al . , 2013; Krause et al . , 2017; Muto et al . , 2016; Rihm et al . , 2019 ) , the neural pathways through which sleep-dependent alterations in the ECS influence food intake have not been investigated in humans . One likely target for sleep-dependent neuromodulation of food intake is the olfactory system . Odors serve as powerful signals for the initiation and termination of feeding behavior ( Saper et al . , 2002; Shepherd , 2006 ) , and animal studies have shown that olfactory processing is modulated in a state-dependent manner ( Julliard et al . , 2007; McIntyre et al . , 2017; Murakami et al . , 2005 ) . In rodents ( Aimé et al . , 2007; Aimé et al . , 2014 ) and humans ( Hanci and Altun , 2016; Stafford and Welbeck , 2011 ) , olfaction is altered by hunger and satiety , and satiety reduces neural activity in olfactory brain regions in parallel with a suppression of feeding behavior ( Boesveldt , 2017; Gervais and Pager , 1979; O'Doherty et al . , 2000; Prud'homme et al . , 2009; Soria-Gómez et al . , 2014 ) . Moreover , recent work across different species suggests a link between the ECS , olfactory processing , and food intake , such that endocannabinoids may directly modulate neural activity in olfactory circuits ( Breunig et al . , 2010; Soria-Gómez et al . , 2014 ) . However , whether odor-evoked responses in the human olfactory system are similarly modulated by the ECS , and whether this accounts for the effects of sleep deprivation on food intake , is not known . We hypothesized that sleep deprivation is associated with a cascade of metabolic and olfactory changes , ultimately steering food choices toward energy-dense options ( Simon et al . , 2015 ) . We predicted that after a night of restricted sleep , relative levels of circulating ECS compounds will be increased ( Hanlon et al . , 2016 ) , leading to changes in how olfactory brain regions in the medial temporal and basal frontal lobes respond to food odors ( Soria-Gómez et al . , 2014 ) . We expected that such sleep-dependent changes in olfactory processing would manifest in odor-evoked activity patterns in piriform cortex ( Howard and Gottfried , 2014; Howard et al . , 2009 ) , and that effects on food intake would involve interactions with areas downstream of piriform cortex , such as the insula . Olfactory , gustatory , homeostatic , and visceral signals are integrated in the insula ( Craig , 2002; de Araujo et al . , 2003; Johnson et al . , 2000; Livneh et al . , 2017; Small et al . , 2008 ) , optimally positioning this region to regulate ingestive behavior in a state-dependent manner ( Dagher , 2012; de Araujo et al . , 2006 ) .
To test these hypotheses , we utilized a within-subject crossover design with a partial sleep-deprivation protocol and pattern-based functional magnetic resonance imaging ( fMRI ) of food and non-food odors ( Figure 1A ) . The experiment was designed to simultaneously measure the effects of sleep deprivation on ECS signaling , neural responses to food odors , and real-life food choices . After one week of sleep stabilization ( 7–9 hr sleep/night between 10:30 pm and 7:30 am ) , healthy-weight participants ( N = 25 , 10 male , age mean ± SEM: 26 . 6 ± 0 . 98 years ) were randomly assigned to one night of deprived sleep ( DS , 4 hr sleep between 1 am and 5 am ) or non-deprived sleep ( NDS , 8 hr sleep between 11 pm and 7 am ) . All subjects participated in both DS and NDS sessions , which were separated by 4 weeks to allow for sufficient recovery time and to control for potential effects of menstrual phase in female participants . Actigraphy-monitored sleep times did not differ between DS and NDS sessions during the 7-nights of sleep stabilization ( NDS: 6 . 63 ± 0 . 18 hr , DS: 6 . 75 ± 0 . 19 hr; T22=−1 . 40 , p=0 . 174 ) . However , sleep times did differ during the night of sleep manipulation ( NDS: 6 . 8 ± 0 . 12 hr , DS: 3 . 8 ± 0 . 18 hr; T24 = 14 . 70 , p=1 . 68×10−13 , Figure 1B , Figure 1—figure supplement 1 , and Supplementary file 1 ) , confirming that subjects complied with the sleep deprivation protocol . The functional imaging sessions occurred in the evening after the night of sleep manipulation . In the 24 hr period leading up to these critical testing sessions , subjects received individually standardized isocaloric diets to ensure identical food intake in both sessions . Subjective states of sleep deprivation were assessed in the morning , and standardized sleepiness scores ( Stanford sleepiness scale ) were obtained upon arrival at the imaging center . Compared to the NDS session , participants in the DS session felt subjectively sleep-deprived , as indicated by reduced self-reported sleep quality , levels of alertness and well-restedness ( Figure 1D ) , and increased sleepiness ( Figure 1E ) . To account for potential effects of sleep deprivation-related stress and anxiety on food intake and associated brain responses ( Maier et al . , 2015 ) , subjects completed the State Anxiety Inventory ( SAI ) . We also measured serum cortisol levels as a physiological marker of stress . There were no significant changes in SAI scores ( T24=−1 . 49 , p=0 . 148 ) or cortisol ( T24 = 0 . 70 , p=0 . 584 ) between DS and NDS sessions ( Figure 1—figure supplement 2 ) , suggesting that stress and anxiety levels were not altered by sleep restriction . Before fMRI scanning , subjects received a standardized isocaloric dinner based on individually estimated energy demands . Hunger ratings decreased immediately after dinner , and returned to pre-dinner levels 45 min after the meal ( Figure 1C ) . Importantly , there were no sleep-dependent differences in hunger ratings made at any time point , indicating that measures of odor-evoked brain activity and food intake cannot be driven by differences in subjective levels of hunger . During fMRI scanning , subjects intermittently smelled a set of energy-dense food odors and non-food control odors ( Supplementary file 2 ) . There were no sleep-dependent differences in rated odor pleasantness and intensity , or respiratory behavior during fMRI scanning ( Figure 1—figure supplement 3 ) . Immediately after scanning , subjects were given ad libitum access to food items in a buffet-style setting to measure effects of sleep deprivation on food choices and intake . In the DS session , participants consumed food items with a significantly higher energy density ( 6 . 0 ± 2 . 43% change from NDS; T24 = 2 . 48 , p=0 . 021 , Figure 1F ) , with no significant difference in the overall calories consumed ( 18 . 6 ± 16 . 93% change from NDS; T24 = 1 . 10 , p=0 . 282 ) . Importantly , effects of sleep deprivation on dietary behavior persisted into the next day ( after a night of unrestricted recovery sleep ) , with a higher percentage of calories consumed as fat ( DS: 36 . 5 ± 1 . 28% , NDS: 30 . 6 ± 1 . 28%; T24 = 2 . 34 , p=0 . 028 ) , indicating that a single night of restricted sleep can have relatively long-lasting effects on food choices . Based on previous reports that circulating levels of 2-AG and 2-OG are enhanced after sleep restriction ( Hanlon et al . , 2016 ) , we next tested whether these ECS compounds were increased in the present study . Partially replicating the previous findings , circulating levels of 2-OG collected during fMRI scanning were increased in the DS relative to the NDS session ( 35 . 02 ± 17 . 82% change from NDS; T24 = 1 . 96 , p=0 . 031 , one-tailed , Figure 2A ) . However , although sleep-dependent changes in 2-AG and 2-OG were significantly correlated ( r = 0 . 55 , p=0 . 004 ) , relative increases in 2-AG were not significant ( 5 . 29 ± 5 . 20% change from NDS; T24 = 1 . 02 , p=0 . 16 , one-tailed ) . We also observed no significant changes in other appetite-regulating hormones , including ghrelin , leptin , and insulin ( Figure 1—figure supplement 2B ) . Interestingly , sleep-dependent increases in 2-OG correlated significantly with increases in the energy density of food consumed at the post-scanning buffet ( robust regression , β = 0 . 47 , p=0 . 027; permutation test , p=0 . 018; Figure 2B ) . Although correlative in nature , this finding suggests that the ECS may play a role in modifying dietary behavior after sleep deprivation . Having established a link between sleep-dependent changes in the ECS and food choices , we next analyzed the fMRI data to examine whether this relationship was mediated by effects of sleep deprivation on central olfactory responses to odors . Based on previous rodent work showing that endocannabinoids affect feeding-related changes in olfactory processing ( Soria-Gómez et al . , 2014 ) , we hypothesized that elevated levels of 2-OG would be accompanied by enhanced representations of odors in olfactory sensory cortices . Both animal ( Barnes et al . , 2008; Illig and Haberly , 2003; Stettler and Axel , 2009 ) and human studies ( Howard et al . , 2009; Zelano et al . , 2011 ) have shown that odors are encoded in piriform cortex by sparsely distributed patterns of ensemble activity , with no apparent topographical organization . Such distributed representations cannot be detected by univariate fMRI analyses , in which activity is typically averaged across voxels , thus blurring information contained within fine-grained patterns of activity . To examine such distributed responses to food odors ( Figure 3A ) , here we used a searchlight-based multi-voxel pattern analysis ( MVPA ) , which enables unbiased whole-brain decoding based on activity patterns ( Haynes et al . , 2007; Kahnt et al . , 2011; Kriegeskorte et al . , 2006 ) . Specifically , we used a support vector machine ( SVM ) classifier to decode information about food vs . non-food odors from patterns of odor-evoked fMRI activity ( Figure 3B ) . Across sleep sessions , we found significant decoding of odor information in the piriform cortex ( x = 16 , y=−2 , z=−14 , T24 = 4 . 20 , PFWE-SVC = 0 . 012 , Figure 3—figure supplement 1 ) and insula ( left x=−32 , y=−4 , z = 16 , T24 = 4 . 75 , PFWE-SVC = 0 . 021; right x = 44 , y = 8 , z = 10 , T24 = 4 . 84 , PFWE-SVC = 0 . 017 ) . Importantly , comparing odor encoding between DS and NDS sessions , we found significantly higher decoding accuracy in the piriform cortex in the DS session ( x = 20 , y = 8 , z=−12 , T24 = 5 . 91 , PFWE-SVC = 0 . 001; Figure 3C ) , suggesting that sleep deprivation enhances encoding of odor information in olfactory brain areas . In contrast , a univariate analysis of odor-evoked fMRI responses in piriform cortex showed no sleep-dependent effects or interactions ( Figure 4 ) . Enhanced encoding of odor information after sleep deprivation could be the mediating neural factor behind the observed relationship between sleep-dependent changes in the ECS and food choices . However , decoding accuracy in piriform cortex was not correlated with either sleep-dependent changes in 2-OG ( r=−0 . 24 , p=0 . 248 ) or energy-dense food choices ( r=−0 . 05 , p=0 . 80 ) , suggesting that the relationship between the ECS and food intake was not directly mediated by changes in odor encoding . In the next step , we therefore considered the possibility that changes in the propagation of olfactory signals from piriform cortex to downstream regions may account for the effects of sleep deprivation on food choices . One downstream region particularly relevant for integrating chemosensory , interoceptive , and homeostatic signals to guide food intake is the insula ( Dagher , 2012; de Araujo et al . , 2006; Livneh et al . , 2017 ) . To test whether sleep deprivation altered the functional connectivity between piriform and insular cortex , we utilized a psychophysiological interaction ( PPI ) model , with piriform cortex as the seed region and odor presentation ( food and non-food odors > clean air ) as the psychological variable . We found that sleep-dependent changes ( DS >NDS ) in piriform connectivity ( odorized >clean air ) with the right insula correlated with changes in the energy density of food consumed immediately after the fMRI session ( x = 40 , y = 6 , z = 0 , T23 = 6 . 04 , PFWE-SVC = 0 . 005; Figure 5A ) , such that reduced odor-evoked connectivity was associated with enhanced intake of energy-dense food ( Figure 5B ) . A similar effect in the left insula was present but did not survive correction for multiple comparisons ( x=−44 , y = 4 , z = 0 , T23 = 4 . 20 , PFWE-SVC = 0 . 17 ) . In addition , while not part of our main set of hypotheses , we found a positive correlation between sleep-dependent changes in energy-dense food choices and changes in odor-evoked connectivity between the piriform cortex and the right anterior hippocampus ( x = 28 , y=−10 , z=−18 , T23 = 5 . 36 , Puncorr = 1 . 0×10−5 ) . In contrast , sleep-dependent changes in piriform connectivity for food vs . non-food odors did not show a significant relationship with changes in food choices . These results suggest that the connectivity between olfactory signals in the piriform cortex and downstream areas may play a role in linking sleep-dependent changes in olfactory processing to changes in food choices . Given that energy-dense food choices were associated with changes in 2-OG levels , it is possible that the observed connectivity effects were also related to the ECS . Indeed , we found that piriform-insula connectivity was significantly correlated with sleep-dependent changes in 2-OG levels ( r=−0 . 58 , p=0 . 002; Figure 5C ) , raising the possibility that the association between sleep-dependent changes in ECS and food choices is mediated by piriform-insula connectivity . To directly test this hypothesis , we employed a formal mediation analysis ( Baron and Kenny , 1986 ) , including a direct path from sleep-dependent changes in 2-OG ( x ) to changes in the energy density of the food consumed after fMRI ( y ) , and an indirect path with changes in piriform-insula connectivity as mediator ( m , Figure 5D ) . Importantly , the direct path between 2-OG levels and food intake was fully explained by the indirect path through piriform-insula connectivity , establishing a significant mediation effect ( Sobel test , z = 2 . 97 , p=0 . 003 ) . This suggests that sleep deprivation affects the ECS , which then modulates the connectivity between piriform and insular cortex , and in turn shifts food choices toward energy-dense options .
Clinical and epidemiological studies have linked reduced sleep to elevated food intake and weight gain ( Kant and Graubard , 2014; Markwald et al . , 2013; Patel and Hu , 2008 ) . This relationship has been confirmed in studies using experimentally induced sleep deprivation , demonstrating that sleep restriction increases the desire for foods high in sugar and fat content ( Cain et al . , 2015; Greer et al . , 2013; Hogenkamp et al . , 2013; Simon et al . , 2015 ) , and leads to excessive consumption of such food options ( Brondel et al . , 2010; Nedeltcheva et al . , 2009 ) . Several different factors have been proposed to account for this relationship ( Patel and Hu , 2008 ) , including changes in hunger induced by appetite-regulating hormones such as ghrelin and leptin ( Spiegel et al . , 2004a; Spiegel et al . , 2004b ) , and the ECS ( Hanlon et al . , 2016 ) . In addition , previous imaging studies have reported sleep-dependent activity changes in response to food cues ( Benedict et al . , 2012; Greer et al . , 2013; Rihm et al . , 2019; St-Onge et al . , 2014 ) , but whether and how these neural changes are related to actual food choices and the ECS has remained unclear . In the current study , we tested the hypothesis that central olfactory mechanisms , in conjunction with the ECS , play a role in mediating the effects of sleep deprivation on dietary choices . This hypothesis was based on previous findings that experimentally induced sleep deprivation elevates relative levels of ECS compounds ( Hanlon et al . , 2016 ) , and animal work suggesting that ECS activity drives changes in food intake through modulatory effects on olfactory circuits ( Breunig et al . , 2010; Soria-Gómez et al . , 2014; Wang et al . , 2012 ) . In line with this idea , we found that sleep deprivation increased consumption of energy-dense foods , proportional to relative increases in 2-OG , and that it enhanced pattern-based encoding of odor information in the piriform cortex . Finally , we found that the effects of the ECS on food intake were mediated by changes in connectivity between the piriform cortex and the insula . Previous animal studies have established an important role for the ECS in regulating feeding behavior ( Bellocchio et al . , 2010; Di Marzo et al . , 2001; Rodríguez de Fonseca et al . , 2001 ) . More recently , it has been shown that levels of endocannabinoids in humans are elevated after sleep restriction ( Hanlon et al . , 2016 ) , suggesting that this may drive altered dietary choices . However , unlike previous studies ( Hanlon et al . , 2016 ) , we did not find a significant increase in 2-AG that paralleled increases in 2-OG . This may be due to the shorter duration of sleep restriction used in our study , but may also indicate different roles of the two compounds . Whereas 2-AG stimulates food intake and lipogenesis by activating CB1 receptors ( DiPatrizio and Simansky , 2008; Osei-Hyiaman et al . , 2005 ) , the exact role of 2-OG and its relation to 2-AG is not fully understood ( Murataeva et al . , 2016 ) . Our study shows that sleep-dependent increases in 2-OG are associated with changes in the consumption of energy-dense foods , providing novel evidence for a link between sleep , ECS , and dietary behavior . We found that sleep restriction induced qualitative changes in food intake , biasing choices toward energy-dense options , without altering total calorie intake . Although some studies have shown increases in calorie intake with sleep deprivation ( Al Khatib et al . , 2017; Broussard et al . , 2016; Markwald et al . , 2013; Patel and Hu , 2008 ) , our findings are in line with several other studies ( Cain et al . , 2015; Nedeltcheva et al . , 2009; Simon et al . , 2015 ) and suggest that sleep deprivation induces nuanced changes in food-based decision making , rather than simply increasing hunger or the motivation to eat . Our results further elaborate on the effects of sleep deprivation on the human brain , suggesting that neural processing of odors is enhanced in primary olfactory brain areas after sleep restriction . Although decoding accuracy can be influenced by factors other than the strength of neural encoding , such as reduced variability ( noise ) , our results indicate that encoding of food vs . non-food odors was more robust in the piriform cortex in a sleep-deprived state . In theory , such enhanced encoding of olfactory information could facilitate odor-evoked approach and consummatory responses ( Aimé et al . , 2007; Soria-Gómez et al . , 2014 ) . However , we did not observe a direct correlation between encoding in piriform cortex and food intake , and changes in odor information were not directly related to changes in circulating levels of 2-OG . It is possible that nonlinear effects or interactions among different ECS compounds ( Ho et al . , 2008; Murataeva et al . , 2016 ) may have obscured a direct linear relationship between 2-OG and encoding in piriform cortex . In any case , our findings indicate that sleep-dependent increases in odor information may not directly mediate the relationship between the ECS and food intake . Instead , they suggest that interactions between olfactory cortex and downstream areas may translate altered chemosensory encoding into changes in food intake . We found that changes in piriform-insula connectivity were correlated with the effects of sleep deprivation on food choices , suggesting a relationship between sleep-dependent food intake and neural processing in extended olfactory pathways . A formal mediation analysis showed that relative increases in 2-OG were related to reductions in odor-evoked piriform-insula connectivity , which in turn was correlated with increased choices of energy-dense food options . Although these findings need to be confirmed in future studies , they suggest a broader role for neural processing of chemosensory signals along piriform-insula pathways in the regulation of food intake . There are several possible ways by which reduced piriform-insula connectivity could promote choices of energy-dense foods in the presence of enhanced odor information in piriform cortex . Previous studies show that sensory and visceral-homeostatic signals are integrated in the insula , and that this integration is critical for guiding food intake ( de Araujo et al . , 2003; Livneh et al . , 2017; Small , 2012 ) . Reduced piriform-insula connectivity could reflect a diminished integration of chemosensory and homeostatic-visceral information , and a failure to adequately integrate elevated olfactory signals with homeostatic information may drive excess intake of energy-dense food . Alternatively , reduced piriform-insula connectivity could indicate an aberrant assignment of value to energy-dense foods ( Balleine and Dickinson , 2000; Gottfried et al . , 2003 ) . Finally , to the degree that sleep deprivation decreases activity in other cortical areas ( Greer et al . , 2013; Muto et al . , 2016 ) , it is possible that reduced piriform-insula connectivity is related to diminished top-down control over elevated olfactory signals in piriform cortex , which may promote impulsive behavior in response to energy-dense food ( Cedernaes et al . , 2014; Krause et al . , 2017 ) . In the current study , we compared brain responses to food odors with high palatability and non-food items with low palatability . As expected , the food odors were rated as higher in pleasantness than non-food odors . In principle , it is therefore possible that our observed effects for food vs . non-food odor encoding in the brain were fully explained by this corresponding pleasantness difference . However , our results remained significant when including pleasantness as a covariate in the statistical models , indicating that in this case pleasantness does not account for our results . Taken together , our findings show that sleep-dependent changes in food choices are associated with changes in an olfactory pathway that is related to the ECS . This pathway is likely not restricted to sleep-dependent changes in food intake but may also account for dietary decisions more generally . In this regard , our current findings may help to guide the identification of novel targets for treatments of obesity .
We consented and screened 41 healthy , right handed , non-smoking , and non-obese men and women between the ages of 18–40 year and body mass index ( BMI ) between 18 . 5 and 24 . 9 kg/m2 , with no history of neurological disorders . We included individuals with a self-reported habitual sleep duration of 7–9 hr , and regular sleep time between 9 pm and midnight . The 7–9 hr habitual sleep inclusion criteria was based on guidelines set by National Sleep Foundation for healthy adults ( Hirshkowitz et al . , 2015 ) . This ensured that all participants had sleep duration and sleep timing that are within the normal range for this age group . A regular sleep time between 9 pm and midnight ensured circadian rhythms were aligned across all participants ( Burgess and Eastman , 2004 ) , minimizing between-subject variance . Additional exclusion criteria were daytime nap , variable sleep habits , regular night work , travel across time zone during the study , use of medications affecting sleep , and caffeine intake of >300 mg/day . We also administered the Center for Epidemiologic Studies Depression Scale ( cut off 16 ) ( Radloff , 1977 ) , and the Pittsburgh Sleeping Quality Index ( cut off 5 ) ( Buysse et al . , 1989 ) . Only individuals with scores below the cut off were included in the study . Only non-pregnant women were included . Of the 41 individuals screened , 29 proceeded with the experimental procedures . Of those , three participants dropped out of the study due to discomfort inside the scanner , and one was excluded from the analysis because of a large number of missed responses . This resulted in a final sample of N = 25 participants ( 10 male ) , whose data are reported here . The main outcome measures did not differ between male and female participants ( energy density: T23 = 0 . 618 , p=0 . 542; 2-OG: T23=−0 . 826 , p=0 . 416; piriform encoding: T23 = 0 . 095 , p=0 . 924 ) and did not correlate with body weight ( energy density: β = 0 . 019 , p=0 . 865; 2-OG: β=−1 . 334 , p=0 . 103; piriform encoding: β = 0 . 024 , p=0 . 611 ) . All experimental procedures of this study ( STU00203395 ) were approved by the Institutional Review Board of Northwestern University . In a randomized within-subject crossover protocol , all subjects participated in two sleep sessions that included one night of deprived sleep ( DS; 4 hr in bed 1 am – 5 am ) and one night of non-deprived sleep ( NDS; 8 hr in bed 11 pm – 7 am ) at home . There was a washout period of 19 days between the last day of the first session ( 7 days of sleep stabilization phase , 1 day sleep manipulation , 1 day fMRI session ) and the first day of second session ( Figure 1A ) . This ensured that the two fMRI days were separated by 28 days for all participants , such that female participants were tested in the same phase of the menstrual cycle . In addition , we recorded self-reported menstrual cycle phase and compared sleep-dependent changes in our primary outcome measures between female participants in the follicular and luteal phase . No significant differences were found between the two menstrual phases ( energy-dense food intake: T10=−0 . 76 , p=0 . 465; 2-OG: T10=−1 . 41 , p=0 . 188; piriform encoding: T10=−0 . 05 , p=0 . 960 ) . During the week preceding each session , participants were instructed to maintain a standardized sleep schedule of 7–9 hr sleep ( between 10:30 pm and 7:30 am ) in order to align the phase of the circadian rhythm across participants . Compliance with the sleep schedule during the sleep stabilization and sleep manipulation phase was monitored using wrist-worn actigraphy and a self-reported sleep diary . Subjects also rated their subjective sleep quality , alertness , and restfulness every morning using an online questionnaire . For the ratings , subjects used a 5-point scale , where one indicated ‘poor’ or ‘least’ and five indicated ‘excellent’ or ‘highest’ . No naps were allowed during both sleep stabilization and sleep manipulation periods . Participants were also instructed to abstain from alcohol , caffeine , and drugs , including all recreational drugs , to avoid interference with sleep and hormone levels . To ensure that participants limited their caffeine intake to <300 mg/day on sleep stabilization days , they were instructed to consume not more than one small caffeinated drink per day . Foods and drinks high in caffeine ( e . g . , coffee , chocolate , soda , most tea , including ice tea , traditional black tea , and green tea ) were listed in the instruction sheet provided to participants . Subjects were verbally reminded of the caffeine restriction at every study visit . In addition , subjects were instructed to not consume any caffeinated drinks on the scanning day . On the evening following the night of the sleep manipulation , fMRI scanning was performed after subjects consumed a standardized isocaloric dinner ( subjects received exactly the same meal in both sessions ) . We collected imaging data and blood samples after dinner in the evening following the night of sleep manipulation because previous studies found that experimentally induced sleep deprivation affects ECS compounds and the desire for and consumption of energy-dense food most prominently in this time window ( Hanlon et al . , 2016; Nedeltcheva et al . , 2009 ) . During fMRI scanning , participants were presented with food odors , non-food odors , and clean air , and rated odor pleasantness and intensity . Before the fMRI scan in both sessions , participants also rated their subjective sleepiness using the Stanford Sleepiness Scale . To equate food intake leading up to the two fMRI sessions , isocaloric meals were provided for the 24 hr preceding both sessions . Our study used an in-home setting to render the sleep-deprivation protocol as ecologically valid as possible without the additional distractions and stressors of being in an unfamiliar hospital laboratory environment . However , because in-home settings come with potential limitations related to non-compliance , we took several measures to reinforce and monitor compliance with the sleep stabilization and sleep manipulation schedule . For both sleep conditions , participants were instructed to strictly follow the sleep protocol ( DS: sleep from 1:00 am to 5:00 am; NDS: Sleep between 11:00 pm and 7:00 am ) . To encourage compliance with the instructions , research staff discussed strategies to stay awake , such as watching TV , standing up , etc . Sleep and wake-up times were monitored for 8 days ( 7 days of stabilization , 1 day of manipulation ) using a wrist-worn 3-axis accelerometers ( ActiGraph GT9X Link , ActiGraph , LLC , Pensacola , FL ) ( Ancoli-Israel et al . , 2003; Marino et al . , 2013 ) . Due to a technical failure , actigraphy data from two participants collected during the sleep stabilization phase of the NDS session was lost ( data from both critical nights of sleep manipulation were not affected ) . Data from one additional participant recorded during one day of the sleep stabilization phase was also lost . Actigraphy data were classified as sleep or awake using Cole-Kripke algorithm , as implemented in the Actilife software . Total sleep time ( TST ) , time in bed ( TIB ) , sleep efficiency ( SE ) , and wake after sleep onset ( WASO ) were also calculated using the same algorithm . TIB began at sleep onset and ended at awakening , and TST was defined as time sleeping within TIB . WASO was defined as the wake time within TIB , and SE was computed as the ratio between TST and TIB . In addition to wearing the actigraphy device , subjects also logged their bed time , sleep duration , and sleep quality immediately after scheduled sleep hours in an online sleep diary . Information entered in the sleep diary was time stamped , and study personnel cross-referenced these time stamps and entries with the actigraphy data . Participants were also instructed to avoid daytime naps , and to not consume alcoholic or caffeinated drinks . Participants who failed to follow the sleep protocol were excluded from the study . During the 24 hr period before each fMRI session , participants were provided with an isocaloric diet . All meals were planned and packaged by a registered dietician at the Clinical Research Unit ( CRU ) at Northwestern Memorial Hospital , and were based on individually estimated energy requirements according to height , weight , age , and sex . Estimated calorie requirements ranged from 1400 to 2600 kcal/day . Meals were composed of 55–60% carbohydrate , 15–20% protein , and 25–30% fat . On the evening preceding the sleep manipulation , participants arrived at the laboratory and consumed a dinner at 6:00 pm . They were also provided with a take-out breakfast and lunch to be consumed on the following day at 8:00 am and 12:00 pm , respectively . Participants were instructed to not consume any additional foods or drinks , other than water , during the 24 hr period preceding the fMRI session . Breakfast consisted of ~30% of total caloric needs , while the lunch and dinner each consisted of ~35% of the estimated calorie requirement . The standardized pre-scan dinner consisted of an entrée ( e . g . , hamburger , veggie burger , grilled chicken with dinner roll , ham/turkey sandwich , etc . ) , a fruit/snack ( e . g . , apple , banana , granola bar , etc . ) , and a small non-alcoholic and non-caffeinated drink . The total calorie content of this dinner ranged from 490 to 910 kcal ( ~35% of estimated calorie requirements ) and was composed of 55–60% carbohydrate , 15–20% protein , and 25–30% fat . All participants consumed the entire meal . For each participant , the same meals were provided in both fMRI sessions . All subjects reported that they did not consume any other foods or caffeinated and caloric drinks . Upon arrival at the imaging center , an MRI safe catheter was placed in participant’s left arm and samples were collected through antecubital venipuncture . Blood plasma samples were collected at baseline ( before dinner ) and at four time points following the dinner . Blood serum samples for ECS analysis were only collected at 7:30 pm while subjects were inside the MRI scanner , 90 min after the initiation of dinner ( in between the 2nd and 3rd fMRI run ) . Prior to drawing each sample , 1 . 5–2 mL of blood were drawn off to remove potentially diluted blood from the dead space of the catheter . Samples were put on ice , centrifuged , aspirated , divided into aliquots , and stored at −80°C until assay . Serum levels of 2-arachidonoylglycerol ( 2-AG ) and 2-oleoylglycerol ( 2-OG ) were extracted using the Bond Elut C18 solid-phase extraction columns ( 1 ml; Varian Inc , Lake Forest , CA ) . Serum samples were processed and the two compounds were quantified using chemical ionization liquid chromatography/mass spectrometry ( LC-ESI-MS; Agilent LC-MSD 1100 series , Ramsey , MN ) , as previously described ( Patel et al . , 2005 ) . We used enzyme-linked immunosorbent assay ( ELISA ) kits for total plasma ghrelin ( human ghrelin , Millipore ) , plasma leptin ( human leptin , Millipore ) , and cobas e411 analyzer ( Roche ) for plasma insulin , and serum cortisol . On the day of fMRI scanning , subjects completed both paper and computerized versions of visual analogue scales to measure motivation to eat at baseline ( before dinner ) and 30 min , 60 min , 90 min , and 120 min after dinner initiation and after the ad libitum buffet ( see below ) . At each timepoint , sensation of hunger , fullness , satisfaction , and prospective food consumption was assessed with following questions: ( 1 ) How hungry do you feel ? ( 2 ) How full do you feel ? ( 3 ) How satisfied do you feel ? ( 4 ) How much do you think you can eat ? Ratings were made on a 10 cm visual analogue scale with text at each end indicating the most positive and most negative rating . After completion of the fMRI scan , participants were presented with excessive portion sizes of energy-dense sweet ( cinnamon roll , donut holes , chocolate chip cookies , mini muffins ) and savory food items ( hash browns , garlic bread , pizza bites , potato chips ) in an all-you-can-eat buffet-style setting . In both sessions participants were instructed to wait in a separate room for 30 min before filling out another questionnaire . This waiting room contained the buffet of food options , and participants were told that they could consume the food freely while they waited , if they wanted . Food items were weighed before and after to determine the amount of food consumed . Total calorie and energy density ( kcal/g ) of consumed food was calculated from the product nutrition labels . Participants were presented with identical buffet options in both testing sessions . To track food intake on the day following the experiment ( after a night of unrestricted sleep ) , participants were asked to record their food intake for 24 hr following the fMRI scan using a food diary . To estimate total calories and fat consumed , nutritional information for each consumed food item was obtained from the United States Department of Agriculture ( USDA ) Food Composition Databases ( ndb . nal . usda . gov/ndb/ ) . Percent of calories consumed as fat was calculated by multiplying total grams of consumed fat by nine kcal/g and dividing this number by the total number of calories consumed . During initial screening , participants rated the pleasantness of six energy-dense food odors in randomized order ( pot roast , potato chips , garlic bread , cinnamon roll , caramel , and gingerbread , provided by International Flavors and Fragrances [New York City , NY] and Kerry [Tralee , Ireland] ) . Ratings were made on visual analog scales using a scroll wheel and mouse button press . Anchors were ‘most-liked sensation imaginable’ ( 10 ) and ‘most disliked sensation available’ ( −10 ) . Based on these ratings , two savory and two sweet odors were chosen for each participant such that they were matched in pleasantness , and these odors were used for the remainder of the experiment . After odor selection , participant also rated the pleasantness , intensity ( anchors ‘strongest sensation imaginable’ and ‘weakest sensation imaginable’ ) , edibility ( anchors ‘certainly edible’ and ‘certainly inedible’ ) , and quality ( anchors ‘clearly savory’ and ‘clearly sweet’ ) of the four selected food odors and two non-food control odors ( fir needle and celery seed ) . Odor ratings collected during screening are summarized in Supplementary file 2 . Most importantly , edibility ratings for food and non-food odors collected during the screening session differed significantly between food and non-food odors ( T24 = 12 . 69 , p=3 . 87×10−12 ) . For all odor ratings and the experimental task , odors were delivered directly to subjects’ nose using a custom-built MR-compatible olfactometer ( Howard and Kahnt , 2017; Howard and Kahnt , 2018; Suarez et al . , 2019 ) , capable of redirecting medical grade air with precise timing at a constant flow rate of 3 . 2 L/min through the headspace of amber bottles containing liquid solutions of the odors . The olfactometer is equipped with two independent mass flow controllers ( Alicat , Tucson , AZ ) , allowing for dilution of odorants with odorless air . At all times throughout the experiment , a constant stream of odorless air is delivered to participants’ noses , and odorized air is mixed into this airstream at specific time points , without changing in the overall flow rate . Thus , odor presentation does not involve a change in somatosensory stimulation induced by the airstream . On the evening following the sleep manipulation , all participants arrived at the scanning center after a 6 hr fast and consumed dinner at 6:00 pm before entering the scanner at 6:45 pm . Immediately before the fMRI scan , participants repeated rated the pleasantness , edibility , and quality of the four selected food odors and the two non-food control odors . Each scanning session consisted of four fMRI runs , and each run consisted of 63 pseudo-randomized trials of olfactory stimulation . On each trial , after a 2 s countdown , the white crosshair in the center of the screen turned blue , cuing participants to sniff the odor for 2 . 5 s . The sniff cue was followed by a rating scale ( pleasantness or intensity , counterbalanced ) for 5 s and a 1–8 s inter-trial interval . Each run consisted of nine presentations of the two sweet food odors , the two savory food odors , the two non-food control odors , and clean air ( totaling 63 trials/run ) . Functional MRI data were acquired on a Siemens 3T PRISMA system equipped with a 64-channel head-neck coil . In each scanning run , 382 Echo-Planar Imaging ( EPI ) volumes were acquired with a parallel imaging sequence with the following parameters: repetition time , 2 s; echo time , 22 ms; matrix size , 104 × 96; field-of-view , 208 × 192 mm ( resulting in an in-plane resolution of 2 × 2 mm2 ) ; flip angle , 90o; multi-band acceleration factor , 2; slice thickness , 2 mm; 58 slices; no gap; acquisition angle , ~30o rostral to intercommissural line to minimize susceptibility artifacts in piriform cortex ( Deichmann et al . , 2003; Weiskopf et al . , 2006 ) . Figure 3—figure supplement 2 shows a normalized group average EPI . A high-resolution ( 1 mm isotropic ) T1-weighted structural scan was also acquired at the beginning of the fMRI session . To support the co-registration of functional and structural images , we also collected 10 whole-brain EPI volumes using the same parameters as the functional EPIs , except for 96 slices and a repetition time of 3 . 22 s . Respiration , as an indirect measure of nasal sniffing , was measured using a MR-compatible breathing belt ( BIOPAC Systems Inc , Goleta , CA ) affixed around the participant’s torso , and recorded at 1 kHz using PowerLab equipment ( ADInstruments , Dunedin , New Zealand ) . Respiratory traces for each fMRI run were temporally smoothed using a moving window of 500 ms , high-pass filtered ( cutoff , 50 s ) to remove slow-frequency drifts , normalized by subtracting the mean and dividing by the standard deviation across the run trace , and down-sampled to 0 . 5 Hz for use as nuisance regressors in fMRI data analyses ( see below ) . For quantification of respiratory peak amplitude and respiratory latency , trial-specific respiratory traces were baseline corrected by subtracting the mean signal across the 0 . 5 s window preceding sniff cue onset , and then normalized by dividing by the maximum respiratory amplitude of all trials in the run . Respiratory traces were sorted by sleep session and odor , and averaged across trials . Respiratory amplitude was then calculated as the max signal within 5 s of sniff cue onset , and respiratory latency was calculated as the time from sniff cue onset to max amplitude . Pre-processing of fMRI data was performed using SPM12 ( http://www . fil . ion . ucl . ac . uk/spm/software/spm12/ ) . For each participant , we aligned all functional volumes for both sleep sessions to the first acquired functional volume to correct for head motion . We then realigned and averaged the ten whole-brain EPI volumes , and co-registered the mean whole-brain EPI to the anatomical T1 image . The mean functional volume was then co-registered to the mean whole-brain EPI , and this transformation was applied to all functional volumes . Spatial normalization was performed by normalizing the T1 anatomical images to the MNI ( Montreal Neurological Institute ) space using the six tissue probability map provided by SPM12 . For multivariate analysis , the resulting deformation fields were applied to searchlight-based maps of decoding accuracy ( see below ) . The normalized decoding accuracy maps were spatially smoothed with a 6 × 6 × 6 mm full-width half-maximum ( FWHM ) Gaussian kernel before group-level statistical testing . For functional connectivity and univariate analyses ( see below ) , the motion-corrected and co-registered functional images were normalized to MNI space using the previously estimated deformation fields and spatially smoothed with a 6 × 6 × 6 mm FWHM Gaussian kernel . To quantify and compare head motion between sessions we computed the average ( across scans ) of the absolute volume-by-volume displacements for each of the six realignment parameters for each session . We then computed composite scores for translation and rotation parameters and compared them between sessions . There were no sleep-dependent differences in these composite head motion parameters ( translation , T24 = 0 . 25 , p=0 . 803; rotation , T24=−0 . 14 , p=0 . 891 ) . We implemented a searchlight-based multi-voxel pattern analysis ( MVPA ) ( Howard and Kahnt , 2018; Kahnt et al . , 2011 ) to decode information about food vs . non-food odors . We first estimated general linear models ( GLM ) for each subject , separately for each session , using the non-normalized and un-smoothed functional images . The GLM included three regressors of interest specifying onset times for the following conditions: 1 ) food odors , 2 ) non-food odors , 3 ) clean air . We also included the following nuisance regressors: the smoothed and normalized respiratory trace , down-sampled to scanner temporal resolution ( 0 . 5 Hz ) ; the six realignment parameters ( three translations , three rotations ) , calculated for each volume during motion correction; the derivate , square , and the square of the derivative of each realignment regressor; the absolute signal difference between even and odd slices , and the variance across slices in each functional volume ( to account for fMRI signal fluctuation caused by within-volume head motion ) ; additional regressors as needed to model out individual volumes in which particularly strong head motion occurred ( absolute difference between odd and even slices >5 SD or slice variance >4 SD ) . The parameter estimates from the first two regressors of this GLM reflect the voxel-wise response amplitudes for food and non-food odors , separately for each run and sleep session . Next , we used these voxel-wise parameter estimates in a searchlight-based , leave-one-run-out cross-validated decoding approach . We decoded food vs . non-food odors from patterns of odor-evoked activity , separately for each of the two sleep sessions . We used The Decoding Toolbox ( TDT ) to implement the searchlight ( Hebart et al . , 2014 ) and LIBSVM ( Chang and Lin , 2011 ) for the linear support vector machine ( SVM ) classifier . To test for brain regions that encoded food vs . non-food odors , at each searchlight ( sphere with 8 mm radius ) , we trained a SVM to discriminate between activity patterns evoked by food vs . non-food odors in three of the four runs per session ( DS or NDS ) , and tested it on activity patterns evoked by food vs . non-food from the fourth ‘left out’ run of the same session . The procedure was repeated four times leaving a different run out , and decoding accuracies were averaged and mapped to the center voxel of the searchlight . This procedure was repeated for every voxel within a 10% gray-matter mask ( based on SPMs tissue probability map that was inverse-normalized into the individual native space , as described in Howard and Kahnt , 2018 ) . The resulting accuracy maps for food vs . non-food odors for DS and NDS sessions were subtracted ( DS >NDS ) , normalized , and smoothed ( 6 mm FWHM ) . We tested for significant differences between DS and NDS sessions at the group level using voxel-wise one-sample t-tests . Statistical thresholds were set to p<0 . 05 , family-wise error ( FWE ) small-volume corrected for multiple comparisons at the voxel-level in a functional mask of piriform cortex that was obtained from a one-sample t-test of decoding accuracy for food vs . non-food odors , averaged across sleep sessions ( p<0 . 001 , see Figure 3—figure supplement 1 ) . We used the generalized psycho-physiological interaction ( PPI ) model ( McLaren et al . , 2012 ) to test for regions in which sleep-dependent changes in functional connectivity with the piriform cortex correlated with sleep-dependent changes in food intake . The seed region in the piriform cortex was defined from significant voxels ( p<0 . 001 ) in the contrast DS >NDS for decoding food vs . non-food odors . We first specified session-wise ( DS or NDS ) PPI models at the single-subject level using normalized and smoothed functional images . Odor presentation ( odor vs . clean air ) was used as ‘psychological variable’ and mean piriform cortex activity as ‘physiological variable’ . The PPI models also included the same nuisance regressors as described above for the GLM for the MVPA analysis . Estimated connectivity parameters for odor vs . no-odor were contrasted between DS and NDS sessions , and entered into a group-level model with changes in energy-dense food intake as regressor of interest . We tested for regions in which sleep-dependent changes in odor-evoked functional connectivity ( odor >clean air ) correlated significantly with sleep-dependent changes in food intake . Statistical thresholds were set to p<0 . 05 , FWE small-volume corrected for multiple comparisons at the voxel-level in an anatomical mask of insula cortex obtained using the Automated Anatomical Labeling ( AAL ) atlas . We conducted a univariate analysis using the traditional GLM approach on normalized and spatially smoothed functional images . The session-wise GLM included three regressors of interest specifying onsets for the following conditions: 1 ) food odors , 2 ) non-food odors , 3 ) clean air . The GLM included the same nuisance regressors as the GLM described in the MVPA section . To test for odor-evoked fMRI activity in the piriform cortex , contrast images for food and non-food odors > clean air trials were created at the single-subject level and averaged across sessions . Group-level analyses were carried out using voxel-wise one-sample t-tests thresholded at p<0 . 05 , FWE whole-brain corrected . To test for sleep-dependent differences in odor-evoked activity in the piriform cortex , we extracted parameter estimates for the three odor conditions per sleep session from a piriform cortex region of interest ( defined using the odor >clean air contrast , at p<0 . 001 , see Figure 4 ) . We computed a two-way ANOVA on the parameter estimates to test for sleep-dependent main effects and interactions . | People who do not get enough sleep often start to favor sweet and fatty foods , which contributes to weight gain . While the exact mechanisms are still unknown , lack of sleep seems to change food preferences by influencing the levels of molecules that regulate food intake . In particular , it could have an effect on the endocannabinoid system , a complex network of molecules in the nervous system that controls biological processes such as appetite . The sense of smell is also tightly linked to how and what organisms choose to eat . Recent experiments indicate that in rodents , endocannabinoids enhance food intake by influencing the activity of the brain areas that process odors . However , it is still unclear whether the brain regions that process odors play a similar role in humans . To investigate , Bhutani et al . examined the impact of a four-hour night’s sleep on 25 healthy human volunteers . Blood analyses showed that after a short night , individuals had increased amounts of 2-oleoylglycerol , a molecule that is part of the endocannabinoid system . When sleep-deprived people were given the choice to eat whatever they wanted , those with greater levels of 2-oleoylglycerol preferred food higher in energy . Bhutani et al . also imaged the volunteers’ brains to examine whether these changes were connected to modifications in the way the brain processed smells . This revealed that , in people who did not sleep enough , an odor-processing region called the piriform cortex was encoding smells more strongly . The piriform cortex is connected to another region , the insula , which integrates information about the state of the body to control food intake . Lack of sleep altered this connection , and this was associated with a preference for high-energy food . In addition , further analysis showed that changes in the amounts of 2-oleoylglycerol were linked to modifications in the connection between the two brain areas . Taken together , these results suggest that sleep deprivation influences the endocannabinoid system , which in turn alters the connection between piriform and insular cortex , leading to a shift toward foods which are high in calories . In the United States alone , one in three people sleep less than six hours a night . Learning more about how sleep deprivation affects brain pathways and food choice may help scientists to develop new drugs or behavioral therapies for conditions like obesity . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2019 | Olfactory connectivity mediates sleep-dependent food choices in humans |
A major feature of embryogenesis is the specification of stem cell systems , but in contrast to the situation in most animals , plant stem cells remain quiescent until the postembryonic phase of development . Here , we dissect how light and metabolic signals are integrated to overcome stem cell dormancy at the shoot apical meristem . We show on the one hand that light is able to activate expression of the stem cell inducer WUSCHEL independently of photosynthesis and that this likely involves inter-regional cytokinin signaling . Metabolic signals , on the other hand , are transduced to the meristem through activation of the TARGET OF RAPAMYCIN ( TOR ) kinase . Surprisingly , TOR is also required for light signal dependent stem cell activation . Thus , the TOR kinase acts as a central integrator of light and metabolic signals and a key regulator of stem cell activation at the shoot apex .
Light is the sole energy source of plants and therefore one of the most important environmental factors influencing their development and physiology . Consequently , several of the core developmental decisions during the lifecycle of a plant from germination to seedling development and flowering are strongly influenced by light conditions . After germination , higher plants undergo two distinct developmental programs depending on the availability of light , termed skotomorphogenesis and photomorphogenesis . Skotomorphogenesis , the dark adaptation program , is characterized by an etiolated phenotype , including an elongated hypocotyl , closed cotyledons , the formation of an apical hook and etioplast development . Importantly , stem cells at the shoot and root tip remain dormant and thus growth in etiolated seedlings is mainly dependent on cell elongation rather than cell division . In contrast , photomorphogenesis , the developmental program triggered in light , leads to seedlings with short hypocotyls , unfolded cotyledons and development of chloroplasts . In the light , shoot and root meristems are activated , leading to root growth and development of the first leaves by cell division and expansion ( reviewed in Nemhauser and Chory 2002 ) . Based on evolutionary evidence , photomorphogenesis is the default pathway , since gymnosperms for example do not follow a strict skotomorphogenic development in darkness ( Wei , 1994 ) . With the advance of land plants and resulting new environmental challenges , such as growth in dense canopy and germination in soil , the evolution of the dark-adapted skotomorphogenesis program ensued an advantage: It allowed plants to allocate the limited energy sources supplied by the seed to maximally grow by elongation , in order to reach favorable light conditions that will provide energy for further growth and development . To faithfully execute these opposing developmental programs , plants have evolved complex mechanisms to perceive light quality and quantity through a whole range of photoreceptors that are mainly absorbing in the blue , red and far-red range of the spectrum . Activation of the blue absorbing CRYPTOCHROMES ( crys ) and/or the red and far-red absorbing PHYTOCHROMES ( phys ) overrides the skotomorphogenic program and plants undergo photomorphogenesis within minutes after perception of a light stimulus ( reviewed in Chory , 2010 ) . On the molecular level , activated light receptors inhibit the function of the core repressor of photomorphogenesis , CONSTITUTIVE PHOTOMORPHOGENESIS 1 ( COP1 ) , an E3 ubiquitin ligase that targets positive regulators of photomorphogenesis for degradation in darkness ( Yi and Deng , 2005 ) . At the same time , a group of potent transcription factors , the PHYTOCHROME INTERACTING FACTORS ( PIFs ) , which promote skotomorphogenesis in darkness , are degraded upon light perception through the PHYTOCHROMES ( Leivar and Quail , 2010 ) . The activities of these pathways converge on the differential regulation of thousands of genes resulting in a massive reprogramming of the transcriptome in response to light ( Ma et al . , 2002; Tepperman et al . , 2004; Peschke and Kretsch , 2011; Pfeiffer et al . , 2014 ) . Light not only activates photoreceptors , it also fuels photosynthesis and therefore leads to the production of a number of energy rich metabolites including sugars . Plants are able to monitor their metabolic state with several signaling systems ( Lastdrager et al . , 2014 ) and recent studies have focused on the evolutionary conserved TARGET OF RAPAMYCIN ( TOR ) kinase complex ( Dobrenel et al . , 2016 ) . In other eukaryotes , TOR functions as a central integrator of nutrient , energy , and stress signaling networks and consistently , TOR regulates cell growth and proliferation , ribosome biogenesis , protein synthesis , cell wall integrity and autophagy ( Díaz-Troya et al . , 2008; Henriques et al . , 2014; Lastdragere et al . , 2014; Xiong and Sheen , 2014 ) . While other eukaryotes possess two TOR complexes , so far only a single complex has been identified in plants . It is comprised of TOR , FKBP12 , LST8 and RAPTOR ( Mahfouz et al . , 2006; Moreau et al . , 2012 ) and thus resembles the mammalian TOR complex 1 ( mTORC1 ) . AtTOR is expressed in the embryo and endosperm and in meristematic regions of the adult plant ( Menand et al . , 2002 ) . While tor null mutants show premature arrest of embryo development ( Menand et al . , 2002 ) , knock down of TOR leads to growth reduction and affects the carbohydrate and amino acid metabolism ( Caldana et al . , 2013 ) . In contrast , the presence of sugars in general promotes TOR kinase activity ( Ren et al . , 2012; Dobrenel et al . , 2013; Xiong et al . , 2013 ) . So far , the only known direct downstream targets of AtTOR kinase are S6 kinase 1 ( S6K1 ) ( Schepetilnikov et al . , 2011 , 2013; Xiong et al . , 2013 ) , TAP46 ( Ahn et al . , 2011 , 2014 ) and E2 promoter binding factor a ( E2Fa ) ( Xiong et al . , 2013 ) . S6K1 plays an important role in reinitiating translation ( Schepetilnikov et al . , 2011 ) as well as in the regulation of the cell cycle ( Henriques et al . , 2010; Shin et al . , 2012 ) . Similarly , E2Fa is associated with cell cycle control through the expression of S-phase genes ( Polyn et al . , 2015 ) . Though little is known about how TOR is activated on a molecular level in plants , reports from the past decade suggest that TOR functions as a central regulator of protein synthesis , cell proliferation and metabolism in response to metabolic signals . Several photomorphogenic responses , like the inhibition of hypocotyl elongation , unfolding of the hypocotyl hook and cotyledons , as well as chloroplast development can be triggered by a light signal alone , as displayed in dark-grown cop1 and pif1/pif3/pif4/pif5 quadruple mutants ( pifq ) ( Deng and Quail , 1991; Leivar et al . , 2009 ) . However , root growth is not induced in cop1 mutants unless sucrose is supplied with the growth medium . Photosynthetic assimilates dominantly promote growth in the root where they can synergistically interact with photoreceptor-triggered light signaling ( Kircher and Schopfer , 2012 ) . Recently , Xiong et al . showed that this photosynthesis-driven growth and proliferation in the root is mediated by the TOR kinase ( Xiong et al . , 2013 ) . Here , we analyzed the role of light and nutrients for post-germination stem cell activation in the shoot apical meristem ( SAM ) of young seedlings . Stem cell control in the SAM of Arabidopsis thaliana is based on the activity of the homeodomain transcription factor WUSCHEL ( WUS ) , which is expressed in the organizing centre and necessary and sufficient to non-cell- autonomously induce stem cell fate by protein movement . Stem cells in turn express CLAVATA3 ( CLV3 ) , a short secreted peptide , that acts via the CLV / CORYNE receptor system to limit the expression of WUS in the organizing center ( Schoof et al . , 2000; Daum et al . , 2014 ) . The use of a reporter system based on the regulatory regions of WUS and CLV3 allowed us to quantitatively trace behavior of stem cells ( pCLV3:mCHERRY-NLS ) as well as cells of the underlying organizing center ( pWUS:3xVENUS-NLS ) . With the help of these tools , we were able to genetically dissect the individual contribution of light signaling and photosynthesis-driven nutrient sensing on the stem cell system of the SAM . We show that both pathways ultimately converge at the level of TOR kinase activation , revealing a role for TOR as a central regulator of stem cell activation in response to environmental cues .
The SAM of Arabidopsis seedlings remains dormant during skotomorphogenesis and therefore , plants are unable to advance to the organogenic stage in the absence of light . However , since light acts as signal and energy source alike , we first asked which of the two roles is dominant for SAM development . While supplementation of sugar to wild-type seedlings grown in the dark is known to be inefficient to trigger development ( Figure 1A ) , activation of the light pathway alone , either physiologically by low level illumination , or genetically by introduction of the cop1 mutation , was shown to induce photomorphogenic development of the hypocotyl and cotyledons in darkness ( Deng and Quail , 1991 ) . Despite this stark developmental transition , SAMs of cop1 mutants were unable to produce organs when grown in the dark . However , the SAM was activated and organogenesis initiated in 100% of the dark-grown cop1 mutants when supplemented with sucrose as external energy source ( Figure 1B , see also McNellis et al . , 1994; Nakagawa and Komeda , 2004 ) . Conversely , the SAM of light-grown wild-type seedlings remained dormant when photosynthesis was compromised by the carotenoid biosynthesis inhibitor norflurazon . In line with our observation of cop1 mutants , supplementing the growth medium of these plants with sucrose rescued the dormant phenotype in approximately every third seedling ( Figure 1C ) . Thus , neither the availability of energy metabolites , nor light perception alone was sufficient for SAM activation . In contrast , light and energy , likely in the form of photosynthetic products , seemed to be sensed independently , and both factors need to act cooperatively to trigger SAM development . 10 . 7554/eLife . 17023 . 003Figure 1 . SAM development depends on light and sugar . ( A–C ) Five week old plants grown on media with ( + ) or without ( - ) sucrose . ( A ) Wild-type plants grown in darkness , ( B ) cop1 mutant plants grown in darkness and ( C ) wild-type seedlings grown in light in the presence of 0 . 5 µM photosynthesis inhibitor norflurazon . ( D–I ) Maximum projections of SAMs of four day old seedlings; scale bar 20 µm . ( D–F ) pCLV3:mCHERRY-NLS ( red ) and pWUS:3xVENUS-NLS ( green ) . ( G–I ) pCLV3:mCHERRY-NLS ( red ) and pWUS:WUS-linker-GFP ( green ) . Quantification of pWUS:3xVENUS-NLS ( J ) and pCLV3:mCHERRY-NLS ( K ) expression by fluorescence intensity under different growth conditions ( gray = darkness , red = red light ( 30 µmol*m−2*s−1 ) , solid box = w/o sucrose , hatched box = 1% sucrose , dag = days after germination ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17023 . 00310 . 7554/eLife . 17023 . 004Figure 1—figure supplement 1 . Expression of CLV3 and WUS during seedling development . ( A , B ) CLV3 reporter signal in seedlings exposed to light and/or sucrose increases in size , but not intensity . Quantification of pCLV3:mCHERRY-NLS expression under different growth conditions ( gray = darkness , red = red light ( 30 µmol*m−2*s−1 ) , solid box = w/o sucrose , hatched box = 1% sucrose , dag = days after germination ) . ( A ) Quantification of the relative area above a consistent pCLV3:mCHERRY-NLS signal meristems above a consistent gray value . ( B ) Quantification of the mean gray value within the area determined in ( A ) . ( C , D ) WUS mRNA in situ hybridization on meristems of six day old wild-type seedlings grown in darkness and white light ( 150 µmol*m−2*s−1 ) , respectively ( scale bar = 40 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17023 . 004 Our phenotypic analysis suggested that light and metabolic signals synergize to activate SAM development , and thus we asked which of the known components underlying stem cell homeostasis might be the relevant cellular and molecular targets . By using transcriptional reporters for stem cells ( pCLV3:mCherry-NLS ) and niche cells ( pWUS:3xVENUS-NLS ) we found that stem cell identity was actively maintained independently of growth conditions and was even observed in the dormant state mediated by germination in the dark ( Figure 1D ) . In contrast , expression of the reporter for the stem cell inducing WUS transcription factor was critically dependent on environmental signals and preceded meristem activity and the initiation of organogenic development ( Figure 1E ) . To test if our WUS reporter faithfully recapitulated the behavior of the endogenous gene , we used in situ hybridization and were able to confirm strong light dependent induction of WUS mRNA ( Figure 1—figure supplement 1C and D ) . Since WUS protein exhibits complex movement and a short lifetime ( Daum et al . , 2014 ) , we furthermore analyzed the behavior of WUS-GFP protein in vivo by recording the GFP signal in our rescue line ( pWUS:WUS-linker-GFP in wus mutant background [Daum et al . , 2014] ) . Again , we observed a strong light- and sucrose-dependency of the WUS-GFP signal in line with the observed activation of the WUS promoter and accumulation of the endogenous WUS mRNA under these conditions ( Figure 1G–I ) confirming that the simple pWUS:3xVenus-NLS reporter represents a faithful and quantitative readout for WUS activity . Taken together , these findings on the one hand suggested that CLV3 expression is at least partially independent of WUS and on the other hand that the environmentally dependent transcriptional activation of WUS is the trigger to overcome stem cell dormancy . Using seedlings carrying both reporters grown under wave-length specific LED illumination and image quantification we found that the WUS reporter ( pWUS:3xVENUS-NLS ) was below detection level in dark-grown seedlings . In contrast , GFP signals were readily detectable in light-grown plants from three days after germination onwards with the signal steadily increasing over time ( Figure 1J ) . Interestingly , WUS expression was also induced in the absence of light , when plants were grown on sucrose-supplemented medium ( Figure 1F , J ) and when sucrose was supplied to light-grown seedlings , the effect of light and sucrose on WUS expression was additive ( Figure 1J ) . Light and sucrose also had a similar effect on the regulation of CLV3 expression ( Figure 1D–I , K ) , however , since the CLV3 reporter was already detectable in dark-grown seedlings , the induction of expression by light and sucrose was less pronounced and mainly due to an enlargement of the CLV3 domain rather than an increase of signal intensity in individual cells ( Figure 1—figure supplement 1A and B ) . In sum , development of the SAM required both , light signal transduction and the availability of photosynthetic products , whereas WUS expression was induced also by each signal individually . Thus , tracing WUS expression in the SAM of young seedlings represented a sensitive model to decipher the contribution of upstream signals to stem cell activation in a developmentally and physiologically relevant setting . Since the expression of the transcriptional WUS reporter showed an early and dynamic response to environmental stimuli that mimicked both endogenous WUS mRNA , as well as WUS-GFP protein , we used the intensity of the reporter signal in four day-old seedlings as an easily quantifiable proxy for stem cell activation . First , we wanted to elucidate the molecular players involved in stem cell activation by light . To this end , we irradiated seedlings with monochromatic light of low intensities ( 30 µmol*m−2*s−1 ) to analyze the effect of light signaling with minimal influence of photosynthesis-derived metabolites . Even at low intensities , blue , as well as red light were sufficient to robustly induce WUS expression ( Figure 2A ) . In line with the well-documented biochemistry of the photoreceptors , red-light-induced WUS activation was specifically reduced in the phyB mutant background , while blue-light-induced reporter activity was impaired in the cry1/cry2 double mutant background ( Figure 2A ) . We thus concluded that light perceived through phyB as well as the crys influences the developmental fate of the SAM . 10 . 7554/eLife . 17023 . 005Figure 2 . Light induced WUS expression depends on photoreceptors and is repressed by COP1 . Quantification of pWUS:3xVENUS-NLS expression by fluorescence intensity was measured in four day old wild-type ( WT ) or mutant seedlings ( WT ) or mutant background under different growth conditions ( gray = darkness , red = red light ( 30 µmol*m−2*s−1 ) , blue = blue light ( 30 µmol*m−2*s−1 ) , solid box = w/o sucrose , hatched box = 1% sucrose ) . ( A ) 0 . 5 mM lincomycin and 5 µM norflurazon , respectively were applied to the growth media of wild-type seedlings . ( B ) Impact of cop1−4 mutation on WUS expression . DOI: http://dx . doi . org/10 . 7554/eLife . 17023 . 00510 . 7554/eLife . 17023 . 006Figure 2—figure supplement 1 . Light regulation of WUS expression . ( A ) Induction of WUS expression in far-red light depends on phyA . ( B ) Growth in CO2-deficient environment does not affect light induced WUS expression in four day old seedlings , but inhibits the development of the first leaves unless 1% sucrose is added to the growth medium ( C ) . Seedlings in ( C ) were grown in white light ( 150 µmol*m−2*s−1 ) and in absence of CO2 for 11 days . ( D and E ) Quantification of WUS expression by qRT-PCR in seven day old seedlings grown in darkness ( gray ) , 150 µmol*m−2*s−1 white light ( yellow ) or darkness in the presence of 1% sucrose ( hatched gray ) . Expression levels were normalized to PP2A expression . Error bars show standard error of the mean of two biological replicates . ( F ) The hy5 mutation does not affect light induction of WUS expression . Tissue specific activation of light signaling outside the SAM can induce WUS expression in the meristem ( G ) and photomorphogenic growth ( H , J ) in dark-grown seedlings . Constitutively active phyB Y276H was expressed under different promoters: AP19: pAt1g26680; AP20: pAt3g59270; AP21: pSUC2; AP22: pUBQ10; AP87: pCAB3; AP88: pML1 . Two independent lines for each construct were quantified . ( I ) Phenotypic differences of pML1:PHYB Y276H lines correlate with the expression level of PHYB . Expression of endogenous PHYB was determined by using a primer pair annealing in the 3' UTR of PHYB . Expression levels were normalized to PP2A expression . pWUS:3xVENUS-NLS expression ( A , B , F and G ) and hypocotyl length ( H ) were quantified in four day old seedlings exposed to different growth conditions ( gray = darkness , red = red light ( 30 µmol*m−2*s−1 ) , far red = far red light ( 30 µmol*m−2*s−1 ) , blue = blue light ( 30 µmol*m−2*s−1 ) , yellow = white light ( 150 µmol*m−2*s−1 ) , solid box = w/o sucrose , hatched box = w/ 1% sucrose ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17023 . 006 We also tested WUS expression under far-red light , which is sensed by phyA and found that reporter activity was only weakly induced by far-red light . Interestingly , phyA mutants showed similar WUS promoter activity under far-red light and in darkness ( Figure 2—figure supplement 1A ) . However , when we supplemented the growth media with 1% sucrose we observed a clear induction of WUS expression in response to far-red light , which was dependent on a functional copy of PHYA . Still , phyA mutants displayed a basal level of WUS promoter activity already in darkness even when grown on plates containing sucrose , suggesting a complex and so far unknown regulatory role for phyA under these conditions . The fact that plants grown in far-red light are photosynthetically inactive and required an exogenous energy source for WUS activation , while plants under blue and red light did not , raised the question whether minimal levels of photosynthetically derived sugars might contribute to WUS expression in blue and red light , despite the low fluence . Therefore , we tested whether the availability of photosynthetic products is a prerequisite for light-induced WUS expression by chemical interference . However , the inhibition of photosynthesis by either norflurazon or lincomycin , did not affect WUS promoter activity in red or in blue light ( Figure 2A ) . In the presence of lincomycin , WUS expression was even slightly increased under both light conditions . To avoid potential side effects of the pharmacological treatments we also tested the effect of CO2 withdrawal on seedling development and WUS expression . Preventing photosynthetic assimilation in a CO2-deficient atmosphere inhibited development of seedlings even when grown in light . This phenotype could be rescued in one third of the plants by adding 1% sucrose to the media , similar to what we observed using norflurazon treatment ( compare Figure 1C and Figure 2—figure supplement 1C ) . Importantly , WUS induction by red light was unaffected by CO2 reduction in the atmosphere ( Figure 2—figure supplement 1B ) . Thus , photosynthetically derived metabolites produced in a low light environment were not required for activation of stem cells , confirming that light signaling alone was sufficient for WUS expression . We next asked how the light signal perceived by PHYTOCHROMES and CRYPTOCHROMES is relayed to the nucleus by testing the contribution of known downstream signaling components , such as COP1 and HY5 . The E3-ubiquitin ligase COP1 , which targets HY5 but also other factors for degradation in darkness , showed robust inhibitory effects on WUS expression . Cop1-4 mutants displayed photomorphogenic development in darkness , which was accompanied by WUS expression ( Figure 2B ) . Furthermore , the repressive function of COP1 was prominent under all conditions tested and cop1-4 seedlings displayed strongly elevated WUS promoter activity compared to wild-type when grown in dark with or without sugar , and also under low light conditions ( Figure 2B ) . To confirm that these effects were not caused by second site mutations present in the cop1-4 background or specific to the allele tested , we used qRT-PCR to assay WUS expression in seedlings carrying other cop1 loss-of-function alleles . However , since this approach lacked the spatial resolution provided by microscopic quantification of the WUS reporter , it proofed to be much less sensitive . Still , we were able to detect accumulation of endogenous WUS mRNA in response to light in 7d old wild-type seedlings , as well as in cop1-4 mutants in the dark ( Figure 2—figure supplement 1D ) . Importantly , all three cop1 mutant alleles tested showed robust elevation of WUS mRNA levels when grown in the dark ( Figure 2—figure supplement 1E ) , demonstrating that loss of COP1 function leads to activation of WUS . One of the main functions of COP1 is to target the transcription factor HY5 , a positive master regulator of photomorphogenesis , for degradation . Thus , we analyzed the role of HY5 working under the hypothesis that in contrast to cop1 mutants , which had shown elevated WUS reporter expression , hy5 mutants should suffer from a much reduced meristem activity due to the absence of an important photomorphogenesis stimulating activity . However , hy5 mutants were unaffected in activation of WUS expression ( Figure 2—figure supplement 1F ) , suggesting that SAM stem cell activation is dependent on another COP1-targeted transcriptional transducer , such as HY5 HOMOLOG ( HYH ) , or a so far unknown regulator . Since the SAM is shielded from the environment especially in etiolated seedlings , where it is buried between the closed cotyledons and protected by the apical hook of the hypocotyl , it seemed questionable that the meristem itself is the site of light perception . We therefore tested the competence of different tissues to perceive light signals and translate them into a stem cell activating output . To this end , we expressed a constitutive active form of phyB ( Su and Lagarias , 2007 ) under different tissue specific promoters ( Figure 2—figure supplement 1C , D ) . Expression of phyB Y276H under an ubiquitous promoter ( pUBI10 ) caused strong cop1-like phenotypes and a substantial activation of the WUS promoter in the SAM showing that transgenic activation of light signaling is sufficient to trigger stem cell activation in darkness ( Figure 2—figure supplement 1G , H , J ) . In line with our hypothesis that light is likely perceived by cells outside the SAM , vascular specific expression of phyB Y276H by the pSUC2 , or mesophyll specific expression by pCAB3 promoters ( Ranjan et al . , 2011 ) initiated constitutive photomorphogenic phenotypes and WUS expression in dark-grown seedlings . Similar results were also obtained for the epidermal pML1 promoter , in lines showing high expression levels of phyB Y276H ( Figure 2—figure supplement 1G–J ) . These results suggested that the stimulus downstream of light perception can be transmitted between tissues by a mobile signal and raised the question whether the SAM itself even has the ability to respond to light . To explore this , we expressed phyB Y276H specifically in the SAM under the promoter of At3g59270 ( Yadav et al . , 2009 ) , but in contrast to expression outside of the SAM , we observed fully etiolated seedlings without detectable WUS expression when these plants were grown in the dark . Even when we drove PHYB Y276H expression in cells surrounding the organizing center by the promoter of At1g26680 ( Yadav et al . , 2009 ) , we only observed a minor reduction in hypocotyl elongation in darkness compared to wild-type and marginal WUS expression ( Figure 2—figure supplement 1G , H , J ) . We therefore concluded that light is perceived by cells outside of the SAM , likely in the cotyledons or the hypocotyl and that this stimulus is transmitted to the SAM by a so far unidentified mobile signal . Amazingly , the SAM does not possess the competence to perceive and/or translate the light stimulus into stem cell activation , but rather is limited to responding to the signals that are transmitted from distant plant organs . Since we had shown that light is perceived at a distance from the SAM , which also for energy rich metabolites is not a source , but a sink tissue , we next asked how the information for both environmental inputs is relayed to the stem cell system . Obvious candidates for inter-regional signaling components are plant hormones and there are a number of studies demonstrating their importance in regulating the shoot stem cell niche , especially for cytokinin ( CK ) and auxin ( reviewed in Murray et al . , 2012 ) . A previous study had analyzed the environmental influence on organ initiation at the SAM using transfer of light grown tomato plants to darkness as a model and found that light is required for CK signaling and polarized membrane localization of the auxin export carrier PIN1 ( Yoshida et al . , 2011 ) . However , these studies could not distinguish whether light was perceived as informational cue or energy source . We therefore analyzed CK signaling activity using the pTCSn:GUS cytokinin output sensor ( Zürcher et al . , 2013 ) as well as auxin flux directionality using polarization of pPIN1:PIN1-GFP ( Benková et al . , 2003 ) as a proxy ( Figure 3A–K ) . In line with the findings of Yoshida et al . , CK signaling was strongly activated by light when compared to etiolated seedlings ( compare Figure 3H , G ) . Furthermore , we also found that PIN1 polarly localized to the plasma membrane in a light dependent manner ( Figure 3A and B ) . Interestingly , sucrose treatment of etiolated seedlings did not affect the localization of PIN1 ( Figure 3C ) but lead to a mild activation of the TCS reporter also in the absence of light ( Figure 3I ) , suggesting that there is specificity in the hormonal response . 10 . 7554/eLife . 17023 . 007Figure 3 . Hormonal control of the SAM . ( A–F ) Confocal images of four day old seedlings expressing pPIN1:PIN1-GFP in WT ( A–D ) or cop1-4 ( E , F ) background under diverse growth conditions . The lower row shows a magnification of the meristem shown in the picture above . ( G–K ) GUS staining of four day old plants expressing pTCSn:GUS ( light = white light ( 150 µmol*m−2*s−1 ) , + suc = 1% sucrose , nor = 5 µM norflurazon , scale bar = 20 µm ) . ( L ) Wild-type seedlings after 20 days on plates containing CK ( 75 µM benzyladenine ) supplemented with ( + ) or without ( - ) sucrose . DOI: http://dx . doi . org/10 . 7554/eLife . 17023 . 00710 . 7554/eLife . 17023 . 008Figure 3—figure supplement 1 . pTCSn:GUS activity in four day old seedlings grown under different conditions . scale bar = 20 µm; light = white light ( 150 µmol*m−2*s−1 ) ; -CO2 = CO2-deficient; + suc = 1% sucroseDOI: http://dx . doi . org/10 . 7554/eLife . 17023 . 008 To test the light signaling response independently from impeding effects of photosynthesis , we treated plants grown in light with the photosynthesis inhibitor norflurazon . While PIN1 localization was still light responsive , no activity of the TCS reporter was detectable under these conditions ( Figure 3D , J ) . We also observed a reduction of TCS signal when plants were grown in a CO2-deficient environment ( Figure 3—figure supplement 1C ) . Since in both cases TCS reporter activity could be restored by sucrose supplementation ( Figure 3K and Figure 3—figure supplement 1D ) , we concluded that CK signaling output is dependent on the availability of energy metabolites . However , light signaling and photosynthesis together had a much stronger effect on CK output than nutrient availability alone , suggesting that both signals synergize to stimulate CK signaling at the SAM . In contrast , PIN1 localization to the plasma membrane was fully dependent on light perception and could not even be restored by the cop1 mutation ( Figure 3E , F ) . If CK signaling indeed integrates energy status and light perception , it may be sufficient to activate SAM development . In line with this idea , the importance of CK in light dependent SAM activation had already been demonstrated ( Chory et al . , 1994; Skylar et al . , 2010 ) and Yoshida et al . had shown that the application of CK to tomato apices can induce organogenesis in the dark ( Yoshida et al . , 2011 ) . Consistently , etiolated Arabidopsis seedlings treated with CK produced leaf like structures even in darkness ( Figure 3L and Chory et al . 1994 ) . However , this developmental transition was strictly dependent on the presence of an external energy source , similar to the behavior of cop1 mutants and in the absence of sucrose , CK treated seedlings failed to develop leaves in the dark ( Figure 3L ) . Thus , our experiments were consistent with CK being an important component of , but not the sole signal for environmental stem cell activation . Since we had found WUS expression to be a much more sensitive readout for SAM activation than organ development , we made use of our reporter system to dissect the role of CK for overcoming stem cell dormancy . Using hormone treatment assays we found that CK alone was sufficient to induce low levels of WUS expression even in darkness in line with its known role in SAM regulation ( Gordon et al . , 2009; Buechel et al . , 2010 ) . Interestingly , we observed the strongest stimulation of reporter activity in plants on sucrose medium , whereas the light response was largely unaffected by CK treatment ( Figure 4A ) . These results suggested that application of CK can at least partially replace the perception of light and thus supported a role for CK as a mobile transducer for light signals upstream of WUS expression . However , as demonstrated by our results using the TCS reporter , in the absence of energy metabolites downstream CK signaling cannot be fully activated , resulting in a lack of organ development . Having established an environment specific role for CK in the initial steps leading up to stem cell activation , we wondered about the mechanism of regulating endogenous hormone levels . We therefore mined the literature for light responsive genes , expressed in the meristem and functionally related to CK metabolism , perception , or signaling . Only CYTOKININ OXIDASE 5 ( CKX5 ) , which codes for one of seven homologous cytokinin dehydrogenases involved in CK catabolism in Arabidopsis met all criteria ( Frebort et al . , 2011 ) . CKX5 is a direct transcriptional target of several PHYTOCHROME INTERACTING FACTORs ( PIFs ) and is highly expressed in etiolated seedlings as well as under shade conditions ( Hornitschek et al . , 2012; Zhang et al . , 2013; Pfeiffer et al . , 2014 ) . Similar to cop1 mutants , also pifq mutants form leaves in darkness when supplied with sugars externally , therefore the PIFs are likely to play an important role in suppressing SAM activation in darkness ( Figure 4—figure supplement 1A ) . CKX6 , a close homologue of CKX5 , had already been described to limit primordia growth in response to shade treatment ( Carabelli et al . , 2007 ) and both CKX5 and CKX6 had been shown to be expressed around the shoot apex ( Motyka et al . , 2003; Bartrina et al . , 2011 ) . When we tested the contribution of CKX5 and CKX6 to SAM activation by crossing the single mutants to our reporter line , we found that loss of either did not promote WUS activity in the dark as CK treatment . In contrast , the responses to sucrose and light were robustly enhanced ( Figure 4B ) , revealing that genetically removing etiolation specific antagonists of CK accumulation is sufficient to replace one of the essential environmental signals , but not both . Because CKX genes act partially redundantly ( Bartrina et al . , 2011 ) , we decided to remove CKX6 in the ckx5 mutant reporter background by CRISPR/Cas9 . Of 27 plants tested in T1 , three were homozygous ckx6 mutants , with either insertions or deletions at the 5´end of the CKX6 locus leading to a shift in the reading frame ( Figure 4—figure supplement 1E ) . Seedlings of all three independent ckx5/ckx6 mutant reporter lines were analyzed in the T2 generation and showed comparable effects on WUS expression ( Figure 4B includes data from one representative ckx5/ckx6 line , AP101 . 9 ) . In contrast to either single ckx mutant , ckx5/ckx6 double mutants showed basal WUS reporter activity in darkness , similar to its behavior under CK treatment . Furthermore , we observed a dramatic enhancement of the effect of sucrose on WUS activity and also light dependent reporter induction was increased two-fold over either single mutant , which was also detectable by qRT-PCR ( Figure 4—figure supplement 1B ) . Based on these results we concluded that CKX5 and CKX6 repress activation of the SAM by degradation of the plant hormone CK in darkness . Interestingly , the level of WUS expression in the ckx5/ckx6 double mutant background was almost comparable to one found in cop1 mutant plants ( Figure 2B ) . However , in striking contrast to cop1 , the ckx5/ckx6 double mutant displayed a fully etiolated phenotype in darkness ( Figure 4—figure supplement 1C and D ) , demonstrating that stem cell activation was fully uncoupled from photomorphogenesis in these plants . Thus , CK can act as an interregional transmitter of light signals specifically for WUS expression , but not for general regulators of photomorphogenic growth , suggesting that the developmental integration of light and metabolic signals might occur downstream of CK and locally at the SAM . 10 . 7554/eLife . 17023 . 009Figure 4 . Role of cytokinin signaling in WUS activation . Quantification of pWUS:3xVENUS-NLS expression in four day old seedlings . ( A ) Cytokinin was applied as 75 µM benzyladenine to wild-type ( WT ) . ( B ) Ckx mutant lines display enhanced expression of the pWUS:3xVENUS-NLS reporter construct ( growth conditions: gray = darkness , red = red light ( 30 µmol*m−2*s−1 ) , solid box = w/o sucrose , hatched box = 1% sucrose ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17023 . 00910 . 7554/eLife . 17023 . 010Figure 4—figure supplement 1 . Role of PIF and CKX genes for stem cell activation . ( A ) PIFs repress development in darkness similar to COP1 . Four-week-old pifq seedlings grown on 1% sucrose develop leaves in darkness . ( B ) Quantification of WUS expression by qRT-PCR normalized to PP2A in seven day old seedlings grown in darkness on 1% sucrose . Error bars show standard error of the mean of two biological replicates . ( C ) CKX5 and CKX6 do not affect hypocotyl growth in darkness . Hypocotyl length of four day old etiolated ckx mutants analyzed in Figure 4B were compared to the corresponding background line . ( D ) ckx5/ckx6 double mutant grown on 1% sucrose in darkness for two weeks ( scale bar = 20 µm ) . ( E ) Sequencing analysis of CKX6 locus in three homozygous T1 mutants induced by CRISPR-Cas9 ( Reference genome in top row shows bases 1–60 and 380–440 of the CKX6 coding sequence . gRNA and PAM in capital letters are highlighted in turquoise and blue , respectively . Deletions are represented as dots . All mutations are highlighted in magenta ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17023 . 010 In addition to light , seedling development and WUS expression both showed a strong dependency on sucrose . Since sugars can not only act as energy source , but also as signaling molecules , we first aimed to disentangle these functions for SAM activation . To this end , plants were grown on plates supplemented with equimolar amounts ( 15 mM ) of different sugars with diverse energy content for four days in darkness ( Figure 5A ) . While mannitol , a non-metabolizable sugar , did not affect WUS activity , WUS expression could be observed in etiolated seedlings in the presence of glucose and sucrose , with sucrose being approximately twice as effective as glucose . Conversely , palatinose , a non-metabolizable sugar structurally related to sucrose and able to induce the sugar-dependent bud burst of roses in vitro ( Rabot et al . , 2012 ) , was not sufficient to induce WUS ( Figure 5A ) . These findings strongly suggested that in the context of stem cell activation , sugars do not act as signaling molecules directly , but rather as energy source . In turn , the metabolic status of the plant seemed to be sensed and translated into appropriate cell behavior by the SAM . 10 . 7554/eLife . 17023 . 011Figure 5 . The TOR pathway integrates metabolic and light signals upstream of WUS . ( A ) Activation of pWUS:3xVENUS-NLS in four day old dark-grown seedlings grown with 15 mM of diverse sugars ( +sug ) . ( B–D ) Activation of pWUS:3xVENUS-NLS in seedlings grown in liquid culture . ( B ) Effect of AZD-8055 TOR inhibitor on glucose treatment . ( C ) Effect of AZD-8055 TOR inhibitor on light treatment . ( D ) Effect of AZD-8055 TOR inhibitor on cop1-4 mutant seedlings . ( growth conditions: gray = darkness , red = red light ( 30 µmol*m−2*s−1 ) , solid box = w/o sugars added , hatched box = with sugar ) ( E ) Quantification of S6K phosphorylation relative to total S6K levels based on western blots shown in Figure 5—figure supplement 1B . ( F–H ) Metabolite measurements from four day old seedlings . Gray bars represent untreated seedlings; hatched bars represent 24 hr treatment with 15 mM glucose; green bars represent 24 hr light treatment ( 60 µmol*m−2*s−1 of red and blue light each ) ; yellow bars represent cop1-4 mutant seedlings . Error bars show standard error of the mean; asterixs indicate significance tested by unpaired two-tailed oneway ANOVA Student-Newman-Keuls test: *p<0 . 05 , **p<0 . 01 , ***p<0 . 001; Glyox = glyoxylate; KG = ketoglutarate . ( I ) PCA based on metabolite measurements shown in Figure 5F–H and Figure 5—figure supplement 1D , E . DOI: http://dx . doi . org/10 . 7554/eLife . 17023 . 01110 . 7554/eLife . 17023 . 012Figure 5—figure supplement 1 . The TOR pathway is activated by light signal transduction . ( A ) The experiment shown in Figure 5C was repeated in the presence of 5 µM norflurazon ( growth conditions: gray = darkness , red = red light [30 µmol*m−2*s−1] ) . ( B , C ) Western blot of WT and cop1 mutant seedlings using Phospho-p70 S6 kinase ( Thr ( P ) -389 ) antibody to detect S6K phosphorylation ( top row ) and S6K1/2 antibody to detect total S6K1 and S6K2 ( middle row ) . Lowest row shows Coomassie stained membrane to visualize total protein loaded . The same growth conditions as for the metabolite measurements were used ( see below ) . AZD-8055 was applied at a concentration of 2 µM in the growth medium while plants were treated with light for the last 24 hr only ( C ) . ( D–F ) Metabolite measurements from four day old seedlings . Gray bars represent untreated seedlings; hatched bars represent 24 hr treatment with 15 mM glucose; green bars represent 24 hr light treatment ( 60 µmol*m−2*s−1 of red and blue light each ) ; yellow bars represent cop1-4 mutant seedlings . Error bars show standard error of the mean; asterixs indicate significance tested by unpaired two-tailed oneway ANOVA Student-Newman-Keuls test: *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 17023 . 012 A key sensor of nutrient availability in plants is the TOR ( TARGET OF RAPAMYCIN ) kinase and photosynthesis-mediated activation of the root meristem has recently been described to be under the regulation of TOR ( Xiong et al . , 2013 ) . Characteristic expression changes that were described in response to glucose-TOR signaling ( Xiong et al . , 2013 ) and E2Fa overexpression ( Vandepoele et al . , 2005; López-Juez et al . , 2008 ) , namely affecting genes involved in ribosome biogenesis , protein translation and cell proliferation have also been identified in microarray analyses of shoot apex tissue derived from young seedlings ( López-Juez et al . , 2008 ) . These genes were rapidly and synchronously induced by photosynthetically active light preceding organ growth , which lead us to hypothesize that stem cell activation in the SAM might also be under control of the TOR kinase . Since mutations in TOR are lethal , we used chemical interference to functionally test its contribution to SAM activation . To efficiently inhibit TOR activity , we applied the ATP-competitive TOR kinase inhibitor AZD-8055 ( Montané and Menand , 2013 ) on seedlings grown in liquid culture ( Figure 5B–D ) . After an initial germination and growth phase of three days in darkness , sugar , light and inhibitor treatments were applied for another three days followed by microscopic analysis of the seedlings . In line with our hypothesis that TOR is required for energy sensing in the SAM , AZD-8055 inhibited glucose-induced WUS expression in a dose dependent manner ( Figure 5B ) . Already 2–5 µM of AZD-8055 resulted in a considerable repression of WUS promoter activity when compared to mock controls , and almost completely suppressed the positive effect of glucose . Since TOR kinase has such a central function in the regulation of growth and development , we wondered whether besides the well documented nutrient availability pathway other stimuli , such as light signaling , might also depend on TOR activity . Using the same experimental setup as described above for glucose we indeed found that light-induced WUS expression was efficiently inhibited by AZD-8055 ( Figure 5C ) . This suggested that TOR kinase might play a specific role in light signaling independent of energy production , since the plants were exposed to a rather low light intensity of 30 µmol*m−2*s−1 . To further test this , we analyzed TOR dependency in the absence of photosynthesis and found that light dependent WUS expression was repressed by the TOR inhibitor even under norflurazon treatment ( Figure 5—figure supplement 1A ) . Consistently , even genetic activation of the light signaling pathway through the cop1 mutation in the absence of a physiological stimulus was sensitive to TOR inhibition by AZD-8055 ( Figure 5D ) . Since we could not exclude that our observations were caused by unspecific side effects of the AZD-8055 inhibitor , we wanted to monitor TOR activity in response to sugars and light directly . Working under the hypothesis that TOR activity should correlate with energy and light signals if the inhibitor results were meaningful , we quantified the phosphorylation level of the direct TOR target S6K by a phospho-specific antibody ( Figure 5E and Figure 5—figure supplement 1B ) . After transfer of three-day-old etiolated seedlings to either light ( 60 µmol*m−2*s−1 of blue and red light each ) or 15 mM glucose for 24 hr , we detected a substantial increase in phosphorylation of S6K compared to seedlings that were kept in darkness ( Figure 5E ) . Surprisingly , light treatment was even more efficient in activating TOR kinase than supplying plants with external glucose . Consistently , cop1-4 mutant seedlings displayed high S6K phosphorylation levels under all conditions , including the dark-grown control . Also in this case we were able to detect a similar activation of the TOR pathway in other cop1 alleles ( Figure 5—figure supplement 1C ) , supporting our previous observation that the TOR pathway is repressed by the negative light signaling component COP1 in darkness . Taken together , TOR kinase seems to play a central and so far underestimated role not only for metabolic , but also for light dependent activation of WUS and ultimately stem cells in the SAM . After having shown that light and energy status converge to activate stem cells , we wondered about the underlying metabolic changes in response to these environmental cues . We were most interested to analyze metabolic re-programming in response to light signaling because we hypothesized that TOR could be indirectly activated following at least two scenarios: First , since in the wild light signaling and photosynthesis usually go hand in hand , we were wondering whether signaling alone would trigger a metabolic shift in expectation of photosynthesis derived energy metabolites . And second , since COP1 regulates more than 20% of the Arabidopsis genome ( Ma et al . , 2002 ) , we speculated that this could also extend to the activity of metabolic enzymes . In both cases the metabolic state of the seedlings , especially with respect to glucose levels , could be affected by light signaling and this in turn could indirectly trigger TOR . We therefore analyzed metabolites in seedlings that were subjected to the same experimental workflow as for studying TOR activity described above , namely wild-type and cop1-4 seedlings grown in darkness compared to seedlings treated with either 15 mM glucose or light for 24 hr before harvest ( Figure 5F–H; Figure 5—figure supplement 1D–F; Supplementary file 1 ) . A Principal Component Analysis ( PCA ) plot clearly identified light signaling triggered by the cop1 mutation ( PC1 ) and energy availability caused by glucose supplementation ( PC2 ) as the two major components in our sample set , with light treated samples being influenced by both components . Interestingly , glucose treatment only had a mild effect on seedling metabolism . We detected slight increases of glucose , fructose and sucrose as well as a clear increase of mannose levels in glucose treated seedlings , while raffinose levels were reduced ( Figure 5F , H ) . The fact that under these conditions TOR activity is markedly increased and WUS expression is robustly detectable demonstrates that the energy sensing machinery must be exquisitely sensitive to small changes in metabolite content , most likely glucose . To our surprise , cop1 and light treated seedlings showed similar metabolic profiles that were clearly different from the glucose treated samples . Since cop1 seedlings were grown in darkness , the metabolic re-programming in cop1 and light treated seedlings is likely caused by light signaling dependent transcriptional regulation of metabolic enzymes , rather than simple photosynthesis dependent accumulation of sugars . We detected a clear reduction of fructose and sucrose levels and a decrease in the cell wall monosaccharides fucose , arabinose and galactose in both samples ( Figure 5H ) . In addition , ADP and ATP levels were slightly increased by light and cop1 mutation ( Figure 5G ) . Cop1 seedlings also displayed an accumulation of glyoxylate , an intermediate of the glyoxylate cycle that allows plants to use lipids as carbon source ( Figure 5H ) . Another noteworthy exception from the similarity between cop1 and light treated seedlings was glucose . Importantly , only the light- and glucose-treated samples , but not cop1 seedlings showed an increase in glucose levels that might be responsible for the observed activation of TOR kinase . However , since it is not known how the TOR complex is activated in plants , we cannot exclude that the obvious metabolic re-programming indirectly triggered TOR kinase activity in cop1 seedlings . Nevertheless , since cop1 seedlings showed an overall reduction of energy rich metabolites , it seemed unlikely that the nutrient state was exclusively responsible for activation of the TOR pathway in the cop1 background . Interestingly , we saw a significant increase in the levels of specific amino acids and their derivatives especially in cop1 ( Glu , Gln , His , Arg , ornithin , spermidine and citrulline ) that were all previously described to be reduced in amiR-tor seedlings ( Figure 5—figure supplement 1F and Supplementary file 1 ) . Conversely , time amino acids whose levels were reported to be increased upon TOR repression ( Thr , Tyr , Val , Ile and Leu ) were unchanged or reduced in cop1 ( Caldana et al . , 2013 ) . The inverse correlation of amino acid levels in cop1 and amiR-tor lines was not apparent in the other samples , suggesting that the permanent de-regulation of the TOR pathway during either cop1 or amiR-tor seedling development strongly affected the metabolome while the short term nature of the glucose or light treatments was insufficient for such a profound change .
The life of a plant begins in many cases with skotomorphogenesis , which is characterized by the elongation of the hypocotyl , formation of an apical hook , unfolded cotyledons and the dormancy of the SAM . While most characteristics of skotomorphogenesis can be revoked by the perception of light through the photoreceptors alone , we showed that activation of the SAM in addition requires the presence of energy metabolites . Thus , for stem cell activation , light not only acts as a signal , but also needs to fuel the photosynthetic apparatus to produce sugars and despite the dual role of a single environmental factor , both inputs are sensed independently . Amazingly , it is not stem cell fate that is dependent on these signals , but rather the expression of WUS , which defines the niche and at the same time acts as a mobile stem cell activator . Thus , stem cell fate as defined by CLV3 promoter activity can exist without WUS in a physiologically and developmentally relevant setting , strongly suggesting that WUS is not the primary stimulating input for CLV3 expression . For the sensing , transmission and integration of light and metabolic signals by the stem cell system , evolution seems to have co-opted well studied regulators into a novel and so far unsuspected regulatory network . On the one hand , the roles of the phyB and cry photoreceptors for stem cell activation are fully in line with text book photomorphogenesis , on the other hand phyA exhibits novel positive and negative functions in far-red light regulation of WUS expression , which point to a re-wiring of this core component of light signaling . Similarly , in light signal transduction , the etiolation regulator COP1 plays a prominent part and cop1 mutations can substitute for light in essentially all experiments , apart from PIN1 membrane localization . In contrast , HY5 , another core component of the photomorphogenesis network , does not seem to have any apparent role in light dependent stem cell activation . This is even more striking when taking into account the recently discovered role in shoot-to-root signaling of HY5 ( Chen et al . , 2016 ) , and our observation that the meristem region itself does not possess the competence to perceive and process light signals . However , we cannot exclude that the HY5 homolog HYH with partially overlapping function but higher expression level in the shoot might mask the effect of hy5 on WUS expression ( Holm et al . , 2002; Sibout et al . , 2006 ) . Since hy5/hyh mutants showed deficiencies in development of the first leaves future experiments analyzing such a double mutant in the context of our double reporter are required . Interregional transmission of the light signal seems to depend on CK and with the cytokinin dehydrogenases CKX5 and CKX6 , we identified two potent regulators of meristem activity that are highly responsive to environmental light conditions . Both are direct targets of PIFs and have already been shown to accumulate in etiolated seedlings ( CKX5 ) and shade conditions ( CKX6 ) , respectively ( Carabelli et al . , 2007; Pfeiffer et al . , 2014 ) , which ultimately leads to an inactivation of cytokinin under unfavorable light conditions . Under open sun light PIFs are degraded , thus liberating cytokinin from CKX mediated degradation , which in turn results in the activation of stem cells . CKX5 is expressed in the rib zone below the stem cell niche and CKX6 additionally in the vasculature ( Motyka et al . , 2003; Bartrina et al . , 2011 ) , which might explain the surprisingly high efficiency of our pSUC2:PHYB Y276H line in activating WUS expression ( Figure 2—figure supplement 1G ) . We did not specifically activate light signaling in the rib zone below the stem cell niche in our experiments and thus the possibility remains that light perception in the immediate vicinity to the meristem affects stem cell activity by short range CK signaling . In parallel to light signals , WUS expression was strongly responsive to energy availability . The effect of sucrose on cop1 and ckx5/ckx6 double mutants was much stronger than the cumulative effect of light and sucrose treatments ( Figures 2B and 4B ) . Also the activity of the CK output reporter pTCSn:GUS strictly depended on sucrose , but could be further stimulated by light , suggesting that CK signaling mainly acts to transmit light signals , but that elevation of CK levels genetically or by treatment are insufficient to elicit the full developmental response . Based on these results we suggest that light enhances CK levels by reducing the expression of CKX genes , however we cannot rule out other explanations , such as stimulation of CK biosynthesis . Interestingly , the WUS paralog WOX9/STIMPY ( STIP ) also plays an important role in light , sugar and CK crosstalk at the shoot apex . Meristems of weaker stip mutants arrest at seedling stage but can be rescued by addition of sucrose to the medium ( Wu et al . , 2005 ) . STIP was further shown to integrate CK signals at the meristem and STIP over-expression can partially overcome the deficits of CK perception mutants ( Skylar et al . , 2010 ) . In contrast to WUS , however , sugar acts downstream of STIMPY , and it would be interesting to investigate whether the same is true for light signals . Our experiments suggested that light and metabolic signals converge downstream of CK and locally at the SAM and consistently , we identified the TOR kinase to be an integrator of both signaling pathways . Earlier reports had shown that glucose-TOR signaling regulates photosynthesis-driven activation of the root meristem ( Xiong et al . , 2013 ) and we demonstrated here that also shoot stem cell activation by metabolizable sugars is dependent on TOR kinase . Intriguingly , TOR activity was not only required for the response to energy availability , but also for WUS stimulation by light . This activity was independent of photosynthesis , because the TOR pathway was also activated under norflurazon treatment , in the absence of CO2 or the suppression of COP1 function . The direct regulation of TOR kinase by light signaling elegantly explains the recently discovered impact of phytochromes on the metabolic state of the plant ( Yang et al . , 2016 ) , but also represents another striking example of a so far undiscovered re-wiring of a core regulatory component within the stem cell network . Consistently , upstream regulators of TOR kinase described in other organisms are not well conserved in plants ( Dobrenel et al . , 2016 ) and thus evolution seems to have found alternative mechanisms to activate the TOR pathway in a highly context dependent manner . Coupling light and energy sensing via TOR could help plants to prepare for the dramatic developmental transition from skotomorphogenesis to photomorphogenesis , which involves transcriptional re-programming of almost a quarter of the genome ( López-Juez et al . , 2008 ) . Light dependent activation of TOR kinase could allow etiolated seedlings to build up the photosynthetic apparatus , initiate ribosome biogenesis and prime stem cells via the expression of WUS to efficiently shift gear towards organogenic growth and development , once energy becomes available .
All used plant lines were in the Col-0 background . Arabidopsis Col-0 was transformed by floral dip ( Clough and Bent , 1998 ) using the A . tumefaciens strain ASE ( pSOUP+ ) carrying the pGREEN-IIS and GreenGate based plasmids . The Agrobacteria strain GV3101 was used for the transformation of plants with the plasmid pAP101 . To select for the presence of the corresponding resistance markers , plants were grown on plates supplemented with 5 mM D-Ala , 20 µg/ml Hygromycin B or 50 µg/ml Kanamycin or grown on soil and sprayed with the herbicides Inspire ( Syngenta Agro AG , Dielsdorf , Switzerland ) ( 7 . 6 µl/l ) ( Rausenberger et al . , 2011 ) or Basta ( 0 . 02% ) 1 week after germination . The plasmids pMD149 and pTS81 were used to generate the reporter lines pCLV3:mCHERRY-NLS and pWUS:3xVENUS-NLS , respectively . The double reporter line is comprised of a cross of both lines . The WUS-GFP rescue line ( pWUS:WUS-linker-GFP in wus mutant background [Daum et al . , 2014] ) was also crossed to the pCLV3:mCHERRY-NLS line . Both crossed lines were homozygous for all loci . The mutant cop1-4 ( McNellis et al . , 1994 ) was crossed to pPIN1:PIN1-GFP ( Benková et al . , 2003 ) and the double reporter line was crossed to the mutants cop1-4 ( McNellis et al . , 1994 ) , phyA-211 ( Reed et al . , 1994 ) , phyB-9 ( Reed et al . , 1993 ) , cry1-304/cry2-1 ( Mockler et al . , 2003 ) , hy5 ( SALK_096651C ) , ckx5-1 ( SALK_064309 ) and ckx6-2 ( SALK_070071 ) ( Bartrina et al . , 2011 ) . All of these crossed lines used in the manuscript were homozygous mutants and screened to be also homozygous for carrying the pWUS:3xVENUS-NLS construct . Additional cop1 alleles , cop1-6 ( McNellis et al . , 1994 ) and cop1-19 ( Favory et al . , 2009 ) , were used for qRT-PCR and the TOR activity assay . A homozygous double reporter/ckx5 line was transformed with the CRISPR/CAS9 plasmid pAP101 to create a ckx5/ckx6 double mutant in the double reporter background . Genomic DNA of T1 plants was PCR-amplified using the oligos A05465 ( ATCAAAAACCCTTTTCCATCCT ) and A05466 ( AGCCAACTTAAAGGCTATGCAG ) and the PCR product was digested with Eco81I to screen for homozygous ckx6 mutants at the locus of the first gRNA . The genomic region of T1 plants that produced undigested PCR fragments was amplified with the oligos A05465 and A05468 ( ACTTGAGGGTCTCATGCAAAAT ) , and sequenced after subcloning the PCR product into pGEM-T Easy ( Promega , Madison , WI ) ) to confirm the mutation of the CKX6 locus . T2 plants of homozygous T1 mutants were used in this manuscript . Seeds were sterilized with 70% ethanol and 0 . 1% Triton for 10 min and afterwards washed twice with autoclaved water . All seeds were imbibed in water for three days at 4°C in darkness before plating 30–40 seeds on 0 . 5x MS ( Duchefa , Haarlem , The Netherlands ) , 0 . 8% Phytoagar in vented petri dishes that were sealed with micropore tape ( 3 M , Two Harbors , MN ) . Germination was induced by 150 µmol*m−2*s−1 of white light for 6 hr . Afterwards plants were either kept in white light , transferred to darkness or to LED cabinets equipped with red ( 673 nm ) , far-red ( 745 nm ) and blue ( 471 nm ) LEDs ( floralLEDs StarterKit 2 , CLF Plant Climatics , Wertingen , Germany ) . Unless otherwise noted , constant red , far-red or blue light was applied with an intensity of 30 µmol*m−2*s−1 . All white light treatments were carried out at 150 µmol*m−2*s−1 of fluorescent white light with a 16-h-light/8-h-dark cycle . Starting with the light induction of germination , plants were kept at 21–22°C . 0 . 5x MS plates were supplemented with 1% ( 30 mM ) sucrose or 15 mM glucose only if mentioned explicitly . Growth in a CO2-deficient environment was accomplished by growing unsealed vented petri dishes in a sealed plastic bag with 5 g NaOH and 5 g CaO ( Kircher and Schopfer , 2012 ) . Four day old seedlings that were harvested for protein extracts and metabolite measurements were grown vertically on top of 100 µm nylon meshes ( nitex 03/100–44 , Sefar , Heiden , Switzerland ) . For the glucose treatment in these experiments seedlings were transferred with the mesh to plates containing 15 mM glucose . Light treatments entailed irradiation with 60 µmol*m−2*s−1 of blue and red light each . Both treatments were started 24 hr before and continued until the harvest of the material . To exclude ungerminated seeds and empty seed shells form the metabolite measurements , only above root tissue was harvested . All seedlings were rinsed with distilled water prior to harvest . About 30–40 seeds , that were imbibed as described above , were sown in 3 ml 0 . 5x MS in petri dishes of 35 mm diameter . Plants were kept in darkness for three days after the induction of germination by 6 hr light treatment . The medium of three day old etiolated seedlings was supplemented with 15 mM glucose and/or 0 . 5–2 µM AZD-8055 ( Santa Cruz Biotechnology , Dallas , TX ) . Stock solutions of 1000x concentrated AZD-8055 were diluted in DMSO , therefore control plants were mock treated with the same volume of DMSO . A detailed protocol of the in situ hybridization procedure was provided previously ( Medzihradszky et al . , 2014 ) . Total RNA was extracted from 100 mg seven day old Arabidopsis seedlings with the Plant RNA Purification Reagent ( Invitrogen , Carlsbad , CA ) according to the instructions of the manufacturer , digested with TURBO DNAse ( Ambion/ Thermo Fisher , Waltham , MA ) and purified with RNeasy Mini Kit ( Quiagen , Hilden , Germany ) . Equal amounts of RNA were used for oligo dT primed cDNA synthesis with the RevertAid First Strand cDNA Synthesis Kit ( Thermo Fisher , Waltham , MA ) . The qPCR reaction was set up using the SG qPCR Master Mix ( EURx , Gdansk , Poland ) and run on a Chromo4 Real-Time PCR System ( Bio-Rad , Hercules , CA ) with technical duplicates each . The relative expression levels were calculated using the ddCt method with PP2A expression as a reference . Results shown are the means of 2 independent biological replicates . The following oligos were used: PP2A: A01067: TAA CGT GGC CAA AAT GAT GC and A01068: GTT CTC CAC AAC CGC TTG GT; WUS: A00317: TTA TGA TGG CGG CTA ACG AT and A00318: TTC AGT ACC TGA GCT TGC ATG; PHYB total: A05986: AGC AAA TGG CTG ATG GAT TC and A05987: GCT TGT CCA CCT GCT GCT AT; PHYB 3'UTR: A05984: GCG ACC ATT GTC AAC TGC TA and A05985: CTC CGA CGT CGT TAG ACA CA . Four day old seedlings were harvested in 90% acetone and incubated at −20°C for at least 1 hr . Seedlings were washed with PBS and incubated in substrate buffer ( 1x PBS ( pH 7 . 0 ) , 1 mM K3Fe ( III ) ( CN ) 6 , 0 . 5 mM K4Fe ( II ) ( CN ) 6 , 1 mM EDTA , 1% Triton X-100 , 1 mg/ml X-gluc ) at 22°C over night . After staining , the seedlings were incubated with 60% and subsequently in 95% ethanol to remove chlorophyll . To image the fluorescent reporter activities in the SAM , seedlings were split in half by pulling one cotyledon away from the SAM with forceps . The exposed meristem was imaged with a Zeiss Imager M1 , the Plan-APOCHROMAT 20x/0 . 8 objective ( Zeiss , Oberkochen , Germany ) and YFP- and mCHERRY-specific filter sets . For the quantification of VENUS and mCHERRY signal intensities the settings for the intensity of the fluorescent lamp and exposure times were unchanged for each channel . 16-bit B/W pictures of at least 20 SAMs per sample were analyzed by FIJI ( Schindelin et al . , 2012 ) , using the background subtraction ( 100 pixel rolling ball radius ) prior to measuring the mean gray value of a circular area surrounding the SAM with a diameter of 51 µm ( 100 pixels ) for WUS and 41 µm ( 80 pixles ) diameter for CLV3 . Quantifications in each figure were normalized to the median of the fluorescence levels of wild-type plants grown in red light for four days . Only exception: in Figure 5A and B we used the glucose treated plants as a reference and in Figure 5D the cop1 mutant plants ( second box in each box plot ) . For all these experiments plants of one experimental set were always grown and analyzed in parallel to the untreated ( dark-grown ) and the corresponding reference sample . In situ sections were analyzed with the same microscope and a 40x/0 . 95 Plan-APOCHROMAT objective ( Zeiss , Oberkochen , Germany ) . Equipment and settings used for confocal microscopy was described earlier ( Daum et al . , 2014 ) . Proteins were extracted from 50 mg materials in 250 µl 2x Laemmli buffer ( 0 . 25 mM Tris-HCL pH 6 . 8 , 8% SDS , 5% ß-mercaptoethanol , 20% glycerol ) supplemented with 1 . 5% phosphatase inhibitor cocktail 2 ( Sigma-Aldrich , St . Louis , MO ) . After adding extraction buffer , samples were briefly mixed and heated at 95°C for 10 min . Cellular debris was removed by two centrifugation steps ( 10 min , 14 , 000 rpm , 4°C ) . 20 µg protein were separated on a 10% SDS gel and transferred to PVDF membrane . Phospho-p70 S6 kinase ( Thr ( P ) -389 ) polyclonal antibody ( No . 9205 , Cell Signaling Technology , Cambridge , UK ) was used to detect S6K phosphorylation . S6K1/2 antibody ( AS12-1855 , Agrisera AB , Vännäs , Sweden ) was used to detect total S6K1 and S6K2 . Three biological replicates were harvested for and analyzed as described ( Poschet et al . , 2011 ) by the Metabolomics Core Technology Platform at the University of Heidelberg . PCA was performed with statistical language R ( version 3 . 3 . 1 ) . For the analysis all metabolite data except the amino acid measurements were used . | Plants are able to grow and develop throughout their lives thanks to groups of stem cells at the tips of their shoots and roots , which can constantly divide to produce new cells . Energy captured from sunlight during a process called photosynthesis is the main source of energy for most plants . Therefore , the amount and quality of light in the environment has a big influence on how plants grow and develop . An enzyme called TOR kinase can sense energy levels in animal cells and regulate many processes including growth and cell division . Plants also have a TOR kinase , but it is less clear if it plays the same role in plants , and whether it can respond to light . Plant stem cells only start to divide after the seed germinates . In shoots , a protein called WUSCHEL is required to maintain stem cells in an active state . Here , Pfeiffer et al . studied how shoot stem cells are activated in response to environmental signals in a plant known as Arabidopsis . The experiments show that light is able to activate the production of WUSCHEL independently of photosynthesis via a signal pathway that depends on TOR kinase . The stem cells do not directly sense light; instead other cells detect the light and relay the information to the stem cells with the help of a hormone called cytokinin . Further experiments show that information about energy levels in cells is relayed via another signal pathway that also involves the TOR kinase . Therefore , Pfeiffer et al . ’s findings suggest that the activation of TOR by light allows plant cells to anticipate how much energy will be available and efficiently tune their growth and development to cope with the environmental conditions . Future challenges are to understand how TOR kinase is regulated by light signals and how this enzyme is able to act on WUSCHEL to trigger stem cell division . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"plant",
"biology",
"developmental",
"biology"
] | 2016 | Integration of light and metabolic signals for stem cell activation at the shoot apical meristem |
The detection of foreign antigens in vivo has relied on fluorescent conjugation or indirect read-outs such as antigen presentation . In our studies , we found that these widely used techniques had several technical limitations that have precluded a complete picture of antigen trafficking or retention across lymph node cell types . To address these limitations , we developed a ‘molecular tracking device’ to follow the distribution , acquisition , and retention of antigen in the lymph node . Utilizing an antigen conjugated to a nuclease-resistant DNA tag , acting as a combined antigen-adjuvant conjugate , and single-cell mRNA sequencing , we quantified antigen abundance in the lymph node . Variable antigen levels enabled the identification of caveolar endocytosis as a mechanism of antigen acquisition or retention in lymphatic endothelial cells . Thus , these molecular tracking devices enable new approaches to study dynamic tissue dissemination of antigen-adjuvant conjugates and identify new mechanisms of antigen acquisition and retention at cellular resolution in vivo .
Depending on the route of infection , vaccination mode , and ability of antigens to traffic , different dendritic cell ( DC ) subsets are required to initiate T cell priming . Upon subcutaneous immunization , small soluble proteins and virus particles pass through the lymphatics to the lymph node ( LN ) , where LN-resident DCs acquire and present antigen ( Manolova et al . , 2008; Gerner et al . , 2017 ) . For larger antigens and/or pathogens that are too large to pass through the lymphatic capillaries , dermal DCs migrate to the LN for presentation of processed antigens to naive T cells ( Manolova et al . , 2008; Bonneau et al . , 2006; Hampton and Chtanova , 2019 ) . Most adaptive immune responses require antigen processing and presentation by conventional DCs in either the draining LN or at the site of infection or vaccination ( migratory cutaneous or dermal DCs ) ( Eisenbarth , 2019 ) . Previous studies have shown that viral antigens persist in the LN beyond the time frame of infectious virus ( Jelley-Gibbs et al . , 2005; Kim et al . , 2010; Kim et al . , 2011; Takamura et al . , 2010; Woodland and Kohlmeier , 2009; Zammit et al . , 2006 ) . We recently found that lymphatic endothelial cells ( LEC ) store antigens from viral infection and vaccination ( Kedl et al . , 2017; Kedl and Tamburini , 2015; Tamburini et al . , 2014 ) . Using a vaccine formulation that elicits robust cell-mediated immunity comprising antigen , a Toll-like receptor ( TLR ) agonist , and an agonistic αCD40 antibody ( TLR/αCD40 vaccination ) or viral infection ( Ahonen et al . , 2004; Ahonen et al . , 2008; Badovinac et al . , 2002; Corbin and Harty , 2004; Kaech and Ahmed , 2001; Kurche et al . , 2010; Kurche et al . , 2012; Lucas et al . , 2018; McWilliams et al . , 2010; Sanchez and Kedl , 2012; Sanchez et al . , 2007; Tamburini et al . , 2012 ) , we discovered that antigens were durably retained in the LN ( Kedl et al . , 2017; Kedl and Tamburini , 2015; Tamburini et al . , 2014 ) . Antigen storage was dependent on the presence of a TLR agonist ( e . g . polyI:C alone [TLR3/MDA5/RIGI or Pam3cys ( TLR1/2 ) + αCD40] ) , but also occurred with antigen conjugated to a TLR agonist ( e . g . 3M019 [TLR7] ) [Tamburini et al . , 2014] . We named this process ‘antigen archiving’ and showed it is important to poise memory T cells for future antigenic encounters ( Tamburini et al . , 2014 ) . Prior to these studies , the only non-hematopoietic cell type thought to retain antigens were follicular DCs , which harbor antigens in antigen-antibody complexes for extended periods of time and for the benefit of B cell memory ( Zammit et al . , 2006; Heesters et al . , 2013 ) . Fibroblasts and non-endothelial stromal cells ( SCs ) comprise a large portion of the LN stroma and are capable of presenting peripheral tissue antigens , but their capacity to acquire and present foreign antigens is not yet well understood ( Fletcher et al . , 2010; Fletcher et al . , 2011; Turley et al . , 2010 ) . We were unable to detect antigen archiving by blood endothelial cells ( BECs ) or fibroblasts in our initial studies ( Kedl et al . , 2017; Kedl and Tamburini , 2015 ) . While LECs have been shown to present antigens in the absence of inflammation to induce T cell tolerance ( Cohen et al . , 2010; Cohen et al . , 2014; Nichols et al . , 2007; Rouhani et al . , 2015; Tewalt et al . , 2012; Dubrot et al . , 2014; Hirosue et al . , 2014; Lund et al . , 2012 ) , we showed that presentation of archived antigen occurs only after exchange of the archived antigen from an LEC to a migratory DC; changing the stimulus from tolerizing to immunostimulatory ( Kedl et al . , 2017; Kedl and Tamburini , 2015 ) . Soluble antigens are exchanged via two distinct mechanisms: ( i ) direct exchange between LECs and migratory DCs and ( ii ) LEC death . Antigen transfer from LECs to both migratory conventional ( c ) DC1s and cDC2s is required for archived antigen presentation to antigen-specific memory T cells ( Kedl et al . , 2017; Kedl and Tamburini , 2015 ) . After viral infection , archived antigen is transferred to Batf3-dependent migratory DCs as a result of LEC death during LN contraction ( Kedl et al . , 2017 ) . Limitations of current approaches have precluded sensitive and quantitative measures of antigen levels across cell types , providing only a glimpse of the cell types and molecular mechanisms that control antigen acquisition , processing , and retention in the LN . Studies of antigen in the LN and peripheral tissues have mainly relied on antigen-fluorophore conjugates or indirect measurement of antigen uptake and presentation ( Gerner et al . , 2017; Jelley-Gibbs et al . , 2005; Kim et al . , 2010; Zammit et al . , 2006; Kedl et al . , 2017; Tamburini et al . , 2014; Jelley-Gibbs et al . , 2007 ) , which defined antigen acquisition by specific DC subsets and trafficking of antigens using live imaging ( Gerner et al . , 2017 ) . However , antigen archiving has been difficult to study because antigen-fluorophore conjugates suffer from low microscopic detection sensitivity , yielding weak signals that diminish over time . Moreover , detection of antigen in the LN and other tissues has relied on flow cytometric analysis using cell surface markers , restricting analysis to specific cell types . To address these limitations and better understand antigen archiving , we developed a new approach to track an antigen-phosphorothioate DNA . The phosphorothioate DNA contained a tracking device for detection using single-cell mRNA sequencing and initiated a robust immune response when conjugated to the protein antigen . Here , we outline the tissue distribution in vivo of this antigen-DNA conjugate by utilizing the conjugated phosphorothioate DNA as an adjuvant and tracking device .
To quantify the dissemination and uptake of antigen in the draining LN after vaccination , we developed a vaccination strategy to measure antigen levels using single-cell mRNA sequencing . Many prior studies have used the model antigen , ovalbumin ( ova ) , conjugated to a fluorophore to track antigen in vivo . Here , we conjugated ova to DNA oligonucleotides with barcodes suitable for analysis by single-cell mRNA sequencing ( Figure 1a ) . The ~60 nt DNA tag contains a unique sequence barcode and PCR primer binding sites , similar to CITE-seq tags ( Stoeckius et al . , 2017; Figure 1—source data 1 ) . We measured the stability of unconjugated DNA and ova-DNA conjugates in which the conjugated DNA either had normal phosphodiester linkages ( pDNA ) or was protected throughout by phosphorothioate linkages ( psDNA ) . Quality control of these conjugates indicated a 1:1 stoichiometry of protein to DNA ( Figure 1b ) . To measure the stability of the antigen-DNA conjugate , we added antigen-DNA conjugates to cultures of bone marrow-derived dendritic cells ( BMDCs ) and quantified the amount of DNA in cell lysates and media over time using the PCR handle to detect the DNA by quantitative PCR . Amount of DNA was quantified as a ratio of DNA detected relative to the amount of protein acquired from the cell lysate . We found significantly higher levels of ova-psDNA in cells relative to ova-pDNA ( approximately fourfold at day 1; p=0 . 002 and approximately sevenfold at day 3; p=0 . 004 ) , indicating that psDNA is more stable than pDNA ( Figure 1c ) . In addition , ova conjugation was required for phagocytosis by BMDCs as we detected limited amounts of unconjugated pDNA or psDNA ( values <1 at days 1–7 ) ( Figure 1c ) . To determine if the BMDCs had both the ova and DNA within each cell , we used flow cytometry and immunofluorescence using an antibody to detect ova and streptavidin to detect the biotinylated DNA tag . We detected both ova and DNA within the same cells by flow cytometry ( Figure 1d , Figure 1—figure supplement 1a ) and co-localization by immunofluorescence ( Figure 1e ) . We also measured conjugate stability in mouse LECs , a cell type that retains foreign proteins for long periods ( Tamburini et al . , 2014 ) , and found that ova-psDNA conjugates were stable over 7 days of culture , whereas ova-pDNA was rapidly degraded ( Figure 1f ) . In the endothelial cells , we detected both the ova protein and the barcode within the same cell and co-localized to same location ( Figure 1—figure supplement 1b , c ) . Furthermore , the ova-psDNA retention within the LECs was similar to a vaccine strategy using an ova protein-fluorophore conjugate with polyI:C and anti-CD40 , which we previously demonstrated induces antigen archiving ( Tamburini et al . , 2014; Figure 1—figure supplement 1d–f ) . Using a more phagocytic cell , bone marrow-derived macrophages , we observed nearly all macrophages phagocytosed the ova-psDNA at day 1 and found the ova and psDNA within the same cell ( Figure 1—figure supplement 1g ) . In macrophages given ova-psDNA 7 days prior , we detected only ova protein ( Figure 1—figure supplement 1g ) , potentially resulting from high levels of endonucleases found within the lysosome of macrophages ( Krieser et al . , 2002; Nagata , 2007 ) . To determine whether conjugation of psDNA to ova affected ova processing and presentation , we measured BMDC presentation of ova-derived SIINFEKL peptide by co-culture with SIINFEKL-specific OT1 T cells . BMDCs given ova-psDNA induced significantly more proliferation of OT1 T cells than unconjugated ova ( Figure 1g , h ) , suggesting enhanced activation of BMDCs upon encounter with ova-psDNA conjugates . Furthermore , we detected pDNA and psDNA in BMDC culture media at 1 day after addition but not at later time points , confirming that ova-psDNA conjugates are processed and not released by BMDCs after phagocytosis ( Figure 1—figure supplement 2a , b ) . Finally , ova-psDNA conjugates led to increased OT1 proliferation relative to ova plus psDNA ( unconjugated ) , showing that ova-psDNA conjugates are immunostimulatory ( Figure 1g , h ) and consistent with studies showing conjugation of antigens to RNA or DNA induce TLR7 ( RNA ) or TLR9 ( DNA ) signals that lead to prolonged antigen presentation ( Xu and Moyle , 2018 ) . Addition of polyI:C and anti-CD40 to BMDCs with ova also elicited robust OT1 proliferation , demonstrating that TLR activation on the BMDCs is required for efficient cross-presentation to T cells ( Figure 1—figure supplement 2c ) . We next asked whether vaccination with ova-psDNA conjugates elicits a T cell response in vivo . We compared antigen-specific T cell responses in mice vaccinated with a mixture of ova-psDNA and polyI:C/αCD40 to its individual components ( ova , psDNA , polyI:C , and polyI:C/αCD40; Figure 2a , Figure 2—figure supplement 2a , b ) and—consistent with the differences in OT1 proliferation we saw in vitro—found that T cell responses to ova-psDNA were greater than either ova with polyI:C , ova with polyI:C/αCD40 , or a mixture of unconjugated ova and psDNA ( Figure 2b ) . Interestingly , ova-psDNA conjugate combined with polyI:C/αCD40 did not significantly enhance the T cell response beyond ova-psDNA alone ( Figure 2b ) . T cells stimulated by ova-psDNA produced significantly more IFNγ than any other vaccination strategy even in the absence of ex vivo SIINFEKL peptide stimulation , indicating prolonged and active presentation of ova-psDNA ( Figure 2c , d ) . Together , these data show that ova-psDNA conjugates elicit antigen-specific T cell responses independent of polyI:C/αCD40 . These findings are consistent with TLR9-dependent immune responses elicited by psDNA ( Baek et al . , 2001; Coffman et al . , 2010; Vollmer et al . , 2004 ) , similar to DC presentation of conjugates of ova demonstrated with other TLR agonists ( van Montfoort et al . , 2009 ) and other subcutaneously administered ova-TLR conjugate vaccine platforms ( Xu and Moyle , 2018 ) . We previously showed that a vaccination strategy comprising soluble antigen and vaccinia virus ( VV; Western Reserve ) induced robust antigen archiving that lasts longer than those using polyI:C/αCD40 adjuvant ( Kedl et al . , 2017 ) . To evaluate antigen-psDNA performance during an active infection , we determined T cell responses after vaccination by comparing individual components with mixtures of ova , VV , ova-pDNA , or ova-psDNA . Subcutaneously administered ova-psDNA alone again elicited a T cell response ( Figure 2 , Figure 2—figure supplement 2a ) , and addition of VV to ova-psDNA conjugate moderately increased T cell responses compared to ova-psDNA alone , similar to what we observed with ova-psDNA/polyI:C/αCD40 ( Figure 2 , Figure 2—figure supplement 2b ) . Finally , we examined the cell-type specificity of ova-psDNA dissemination in vivo . Mice were vaccinated with mixtures of ( i ) ova-psDNA and VV or ( ii ) ova-psDNA and polyI:C/αCD40 , and levels of ova-psDNA were quantified by PCR in both leukocytes and SCs ( fractionated by CD45 expression ) in the draining LNs . We found that CD45- SCs had high amounts of ova-psDNA , but not ova-pDNA , corresponding to increased inflammation ( Tamburini et al . , 2014 ) , whereas CD45+ leukocytes had very low levels of ova-psDNA or ova-pDNA 7 days after vaccination ( Figure 2—figure supplement 2c ) . These data recapitulate our previous demonstration of durable antigen retention by CD45- SCs ( Kedl et al . , 2017; Tamburini et al . , 2014 ) , confirming that ova-psDNA , but not ova-pDNA , is a faithful tracking device for antigen archiving in vivo . Given the ability of the antigen-psDNA conjugates to induce a robust immune response in vivo ( Figure 2 ) and our ability to use the psDNA as a measure of protein antigen levels ( Figure 1 ) , we used the antigen-psDNA conjugate as a ‘molecular tracking device’ to understand the distribution of the protein antigen in the LN following this vaccination . To determine whether we could identify if cells acquire and archive ( Tamburini et al . , 2014 ) antigens following antigen-psDNA , we vaccinated mice subcutaneously with an equimolar mixture of uniquely barcoded ova-psDNA conjugate , unconjugated psDNA , and unconjugated pDNA ( unprotected phosphodiester backbone ) with VV ( as in Figure 2—figure supplement 2c ) , and evaluated antigen distribution ( via psDNA abundance ) in the LN at early ( 2 days ) and late ( 14 days ) time points . At each time point , single-cell suspensions were prepared from draining popliteal LNs and divided into SC ( by depleting CD45+ cells ) or lymphocyte populations ( by flow sorting for CD11c , CD11b , and B220 markers; Figure 2—figure supplement 1b ) . To enrich for myeloid cell populations but maintain representation of other cell types , CD11c+ , CD11b+ , B220+ , and ungated live cells were mixed at a 4:4:1:1 ratio , respectively . These cell populations were analyzed by single-cell mRNA sequencing , measuring both mRNA expression and the quantity of psDNA in each cell using unique molecular identifiers ( Islam et al . , 2014; Figure 3 ) . We recovered a total of 800 cells in the CD45- fraction and 8187 cells in the CD45+ fraction at the 2-day time point . We recovered more CD45- cells ( 6372 CD45-; 4840 CD45+ ) at the 14-day time point likely due to expansion and proliferation of the LN stroma ( Tamburini et al . , 2014; Lucas et al . , 2018; Lucas and Tamburini , 2019 ) . We classified cell types using an automated approach ( Fu et al . , 2020 ) , comparing measured mRNA expression patterns to reference data sets for DCs ( Brown et al . , 2019; Miller et al . , 2012 ) , fibroblastic reticular cells ( FRC ) s ( Rodda et al . , 2018 ) , and LECs ( Fujimoto et al . , 2020; Kalucka et al . , 2020; Xiang et al . , 2020; Figure 3—source data 1 ) . As expected , the CD45+ fraction contained DCs , monocytes , T cells , and B cells ( Figure 3a , b , d , e ) , while the CD45- fraction contained SCs , including LECs , BECs , epithelial cells , and fibroblasts ( Figure 3g , h , j , k ) . We did not recover VV mRNAs in cells at either time point , possibly due to viral clearance or a failure to recover infected , apoptotic cells in the live/dead selection ( Figure 2—figure supplement 1b ) . We first examined the dynamic changes of myeloid populations in the LN . We detected conventional DCs , including cDC1 and cDC2 ( Figure 3a–c ) , which develop from a common DC precursor upon expression of FMS-like tyrosine kinase 3 ligand ( Flt3L ) ( Guilliams et al . , 2014 ) . LN-resident and migratory cDCs can be distinguished by expression of cell-type-specific transcription factors including basic leucine zipper transcription factor ( Batf3 ) and interferon regulatory factor ( IRF8 ) ( cDC1 ) ( Aliberti et al . , 2003; Hildner et al . , 2008; Tsujimura et al . , 2003 ) or IRF4 and Notch ( cDC2 ) ( Lewis et al . , 2011; Schlitzer et al . , 2013 ) . These cDC types are also typically classified based on expression of CD11c , Zbtb46 , and chemokine XC receptor 1 ( cDC1 are XCR1+ , cDC2 are XCR1- ) ( Guilliams et al . , 2014; Bachem et al . , 2012 ) . cDC2s are further categorized as either Tbet-dependent and anti-inflammatory ( cDC2A ) or RORγt-dependent and pro-inflammatory ( cDC2B ) ( Brown et al . , 2019 ) . As expected , at day 2 we identified a large population of LN-resident cDC2B ( cDC2 Tbet- ) cells harboring ova-psDNA ( Brown et al . , 2019 ) . However , we did not find any cDC2A ( cDC2 Tbet+ ) cells , consistent with their role in anti-inflammatory processes ( Brown et al . , 2019 ) . The myeloid populations contained CCR7hi cDCs ( n = 3432; 42% of total ) , which we classified as migratory DCs . This migratory DC population included Langerhans cells ( n = 285; 3 . 5% of total ) , migratory cDC1s ( n = 593; 7 . 2% of total ) , and migratory cDC2s ( n = 2554; 31% of total ) ( Miller et al . , 2012 ) , migrating from the dermis ( Figure 3b ) . At day 14 , we identified a population of LN-resident cDC2 Tbet+ cells ( Figure 3e ) consistent with resolution of the immune response ( Brown et al . , 2019 ) . As cDC2 Tbet+ cells are thought to be anti-inflammatory , these data suggest that the immune response is being quelled ( Figure 3e ) . We also found a group of Siglec-H+ DCs , a cDC progenitor population ( Brown et al . , 2019; Figure 3d , e ) . Using unique barcodes , we quantified the amount of ova-psDNA , psDNA , and pDNA across cell types . Levels of ova-psDNA molecules spanned four orders of magnitude , ranging up to 104 unique molecules and depending on the cell types and time point ( Figure 3c , f , i , l ) . In contrast to the large range of ova-psDNA across cell types , unconjugated psDNA and pDNA were largely undetectable , indicating that antigen conjugation is required for cell acquisition ( Figure 3—figure supplement 1 ) . Consistent with our previous studies ( Kedl et al . , 2017 ) , we did not detect antigen-psDNA at appreciable levels in T cells or B cells ( Figure 3 ) , and because these cell types were captured in both our CD45- and CD45+ samples , we used their median antigen levels to normalize antigen counts in other cell types across captures . We considered the trivial case wherein variation in antigen levels is explained by total mRNA abundance; these variables are uncorrelated in SC types and weakly correlated in cDC subtypes , possibly reflecting activation status ( Figure 3—figure supplement 2 ) . At the early day 2 time point , LN-resident cDC2s contained high levels of antigen-psDNA , consistent with studies of antigens administered with alum ( Gerner et al . , 2017; Figure 3b , c ) . In addition , we found significantly higher levels of antigen in cDC2 Tbet- , migratory CCR7hi cDC2s , and migratory CCR7hi cDC1s ( Figure 3b , c , Figure 3—source data 1 ) , with an average of approximately sevenfold more antigen than T/B cells . At the later time point , migratory cDC1 cells contained the most antigen , consistent with previous studies ( Kedl et al . , 2017; Figure 3e ) . In addition , Tbet- and CCR7hi migratory cDC2s contained moderate levels of antigen , up to threefold more than T/B cells , but had lower amounts of antigen relative to day 2 ( Figure 3b , c , e , f , Figure 3—source data 1 ) . At the late time point , we did not detect significant amounts of antigen in LN-resident cDC1s , Tbet+ cDC2s , Siglec-H+ cells , or monocytes ( Figure 3e ) . We next examined antigen levels in the LN SC populations ( Figure 3g–l , Figure 3—source data 1 ) . Endothelial cells in the LN are classified by their association with blood or lymphatic vasculature; both are required for circulation and trafficking of immune cells to the LN . The blood vasculature circulates naive lymphocytes to the LN , and the lymphatic vasculature transports immune cells from the peripheral tissue including dermal DCs and memory T cells . We used an automated approach ( Fu et al . , 2020 ) that uses correlation between reference and measured gene expression profiles to assign unknown cell types to subtypes defined by previous studies . While strong correlation reflects a good match between reference and query profiles , high correlation between multiple reference LEC subtypes ( Fujimoto et al . , 2020; Kalucka et al . , 2020; Xiang et al . , 2020 ) and changes in expression induced by antigen acquisition made definitive cell-type assignments challenging ( Figure 3—figure supplement 3a–c ) . Notwithstanding these issues , we classified LEC subsets based on the highest correlation values to reference cell types ( Figure 3—figure supplement 3d , e; Xiang et al . , 2020 ) and identified three LEC subtypes ( Fujimoto et al . , 2020; Kalucka et al . , 2020; Xiang et al . , 2020 ) including Ptx3 LECs , ceiling LECs , and Marco LECs with high levels of antigen at the early time point ( Figure 3—figure supplement 3 ) . At the late time point , expansion and proliferation of LN SCs contributed to larger populations of cells including floor LECs , collecting LECs , ceiling LECs , Ptx3 LECs ( Kalucka et al . , 2020 ) , and BECs ( Figure 3h; Malhotra et al . , 2012 ) . At the day 14 time point , several LEC subtypes maintained high antigen levels ( Figure 3h , Figure 3—source data 1 ) . Floor LECs had uniformly high amounts of antigen . Median levels of ova-psDNA were detected in collecting , Ptx3 , and ceiling LEC populations that were significantly higher than B/T cells; however , cells in these groups contained a range of antigen with both high and low populations . We hypothesized that this variability stems from the physical location of the LECs within the LN and their access to trafficking antigen . Using a fluorescently labeled ova with polyI:C/αCD4012 , we confirmed that fluorescent antigen amounts are highest on subcapsular LECs as identified by surface expression of PD-L1 and ICAM1 2 weeks after immunization , similar to ova-psDNA vaccination ( Lucas et al . , 2018; Cohen et al . , 2014; Figure 3—figure supplement 4 ) . Together , our findings suggest that antigen first passes through the sinus followed by the cortex and medulla . These data also suggest that populations of LECs with less antigen could be a result of how the antigen travels through the LN or mechanisms of antigen release over time . Similar to the endothelial cell population , the number and types of non-endothelial SCs increased at the later time point after immunization . Non-endothelial SCs in the LN are classified by their location in the LN into T-zone reticular cells ( TRC ) , marginal reticular cells ( MRCs ) , follicular dendritic cells ( FDCs ) , and perivascular cells ( PvCs ) ( Rodda et al . , 2018 ) . Recently , additional subsets were identified including Ccl19lo TRCs located at the T-zone perimeter , Cxcl9+ TRCs found in both the T-zone and interfollicular region , CD34+ SCs found in the capsule and medullary vessel adventitia , indolethylamine N-methyltransferase+ SCs found in the medullary chords , and Nr4a1+ SCs ( Rodda et al . , 2018 ) . At the early time point , the Cxcl9+ TRCs and CD34+ SCs ( Rodda et al . , 2018 ) had high amounts of antigen ( ~10-fold relative to T/B cells ) ( Figure 3—figure supplement 5 ) . At the late time point , we detected CD34+ SCs , Nr4a1+ SCs , FDCs , and PvCs ( Figure 3k ) . Only the CD34+ and Nr4a1+ SCs contained significant amounts of antigen ( Figure 3k , Figure 3—source data 1 ) . Interestingly , the CD34+ SCs are adjacent to ceiling LECs and the Nr4a1+ SCs are found in the medullary chord and medullary sinus , which are lined by medullary LECs . These findings may suggest potential antigen exchange mechanisms between LECs and SCs that have yet to be defined . We found little antigen in PvCs or FDCs ( Figure 3k , Figure 3—source data 1 ) . Finally , these data provided insight into antigen transfer between SCs and DCs , a process important for enhanced protective immunity ( Kedl et al . , 2017; Tamburini et al . , 2014 ) . We previously showed that archived antigen obtained from the polyI:C/anti-CD40-based vaccine is transferred from LECs to migratory Batf3-dependent cDC1s 2 weeks after infection ( Kedl et al . , 2017 ) . Here , we confirm that with the ova-psDNA vaccine CCR7hi migratory cDC1s had the highest amount of antigen 2 weeks after vaccination ( Figure 3e , Figure 3—source data 1; Kedl et al . , 2017 ) . Together , these data validate the use of molecular tracking devices by corroborating previous studies of antigen trafficking with other vaccination strategies and identify new cells types that dynamically acquire antigen during infection . We next leveraged the variation in antigen levels across cell types ( Figure 3b , e , h , k ) to identify gene expression signatures associated with high levels of antigen that would validate our approach . We classified cells as ‘antigen-high’ and ‘antigen-low’ using a two-component mixture model and identified marker genes associated with each class ( Figure 4a , b ) . To validate this approach , we evaluated the DC populations as genes associated with phagocytosis and activation have been established ( Miller et al . , 2012; Breuilh et al . , 2007; Bune et al . , 2001; Figueiredo et al . , 2018; Gschwandtner et al . , 2019; Hirano et al . , 2007; Jin et al . , 2020; Lämmermann and Kastenmüller , 2019; Mancardi et al . , 2008; PrabhuDas et al . , 2017; Sinclair , 1999 ) . DC populations generally contained lower antigen levels that were variable across subtype ( Figure 3 ) . We classified antigen-low and antigen-high cells for each subtype . Among the subtypes with significant amounts of antigen , Tbet- cDC2 cells had the highest antigen levels and largest differences in gene expression ( 277 genes in antigen-high cells , Figure 4 , Figure 4—source data 1 ) , consistent with cDC2s acting as the primary cell type of antigen uptake following protein ( Gerner et al . , 2017 ) . At the early time point , genes upregulated in antigen-high DCs confirmed DC activation ( Figure 4—source data 1 ) . Antigen-high cDC2 Tbet- cells upregulated genes Ccl2 and Cxcl2 ( consistent with active recruitment of inflammatory cells; Gschwandtner et al . , 2019; Lämmermann and Kastenmüller , 2019 ) , Msr1 ( consistent with antigen scavenging; PrabhuDas et al . , 2017 ) , as well as Pkm , Lgals3 , and Mif ( consistent with DC-T cell responses and DC differentiation during inflammation; Breuilh et al . , 2007; Figueiredo et al . , 2018; Jin et al . , 2020; Figure 4c , d ) . At the late day 14 time point , the highest antigen counts were found in the migratory cDC1 population , consistent with a role for migratory cDC1s in archived antigen acquisition from LECs ( Kedl et al . , 2017; Figure 3e , f ) . Among the genes highly expressed by the antigen-high CCR7hi migratory cDC1 population were Ccl5 and Fscn1 ( Figure 4—source data 1 ) . Consistent with these DCs being involved in archived antigen presentation , Ccl5 ( also known as RANTES ) regulates CD8 T cell responses during chronic viral infection ( Crawford et al . , 2011 ) and Fscn1 , an actin binding protein , regulates cell migration of mature DCs via podosome formation ( Yamakita et al . , 2011 ) . Similar to the day 2 time point , among subtypes with significant amounts of antigen , Tbet- cDC2 populations showed the greatest differences in gene expression between antigen-high and -low cells ( 230 genes in antigen-high cells; Figure 4e , f , Figure 4—source data 1 ) . Genes upregulated in antigen-high Tbet- cDC2s included Fcgr4 , which is involved in phagocytosis , antigen presentation , and proinflammatory cytokine production ( Hirano et al . , 2007; Mancardi et al . , 2008 ) , and CD72 and Acp5 , which are important for the inflammatory response and pathogen clearance ( Bune et al . , 2001; Sinclair , 1999; Figure 4g , h ) . Collectively , these genes evoke specific processes in DC subsets required for the immune response; it remains to be determined whether they are specifically associated with LEC-DC antigen exchange or storage of antigens within DCs . We next evaluated the LEC population to determine whether our classification approach could identify genes involved in antigen archiving . We applied the classifier to LECs as a population and found large numbers of antigen-high-floor , collecting , and ceiling LECs ( Figure 5c ) . Ptx3 LECs comprised a mixture of antigen-low and antigen-high cells , but there was a larger fraction of Ptx3 LECs with low antigen ( Figure 5c ) . There were less antigen-low LECs compared to antigen-high LECs overall ( 34% of total ) , suggesting that antigen archiving may be specific to LECs in general rather than attributable to a specific LEC subset ( Figure 5 ) . Using this classification approach , we identified 142 mRNAs that were significantly changed in antigen-high or antigen-low LECs ( Figure 5—source data 1 ) . Prox1 , while expressed by all LECs identified , was highly expressed in antigen-high LECs , independent of the LEC type ( Figure 5d , e ) . Prox1 is a transcription factor required for LEC differentiation from BECs and defines LEC identity via regulation of Vegfr3 , Pdpn , and Lyve-1 ( Harvey et al . , 2005; Hong et al . , 2002; Wigle and Oliver , 1999 ) . Prox1 upregulation in antigen-high LECs indicates it may also transcriptionally regulate processes involved in antigen archiving . Upregulation of Cavin1 and Cavin2 by antigen-high LECs suggested that caveolar endocytosis may contribute to antigen acquisition by LECs , consistent with LEC dynamin-mediated transcytosis in vitro ( Triacca et al . , 2017; Figure 5d , e ) . Cavin2 appears more specific to LECs than Cavin1 , which is also upregulated by BECs , suggesting that Cavin2 mediates endocytosis specifically in endothelial cells of the lymphatic lineage . Based on Cavin2 gene expression , it appears that this process may be most active in ceiling LECs ( Figure 5e ) . To confirm this finding , we asked whether inhibition of the caveolin pathway with nystatin impaired endocytosis of fluorescent antigen in mice vaccinated with polyI:C/αCD40 . We found a significant decrease in antigen acquired by LECs in the nystatin treatment group 24 hr after administration of fluorescent antigen with this vaccine regimen ( Figure 5f ) , affirming the utility of molecular tracking devices for identifying genes involved in the process of antigen acquisition or archival that are not necessarily specific to antigen-psDNA conjugates . Finally , expression of Stabilin-1 ( Stab1 ) and Stabilin-2 ( Stab2 ) is increased in antigen-high LN endothelial cells , suggesting that scavenging pathways are required for the acquisition of antigen-psDNA conjugates after vaccination . Stab2 is uniquely expressed by LECs in the LN and not by BECs ( Malhotra et al . , 2012 ) , and Stab1 and Stab2 act as receptors for internalization of antisense oligonucleotides with phosphorothioate linkages in liver endothelial cells and Kupffer cells ( Miller et al . , 2016 ) . However , we did not find significant amounts of unconjugated psDNA in LECs ( Figure 3—figure supplement 1 ) , indicating that Stab1/Stab2 are upregulated as part of an antigen scavenging or trafficking program initiated in LECs upon antigen acquisition during infection .
Our development of a ‘molecular tracking device’ enabled tracking of antigen throughout the LN to specific cell types that acquire and archive antigens following subcutaneous immunization . Previous studies used canonical surface markers to track antigen by microscopy and flow cytometry; instead , our approach simultaneously defines cell type by gene expression and quantifies the acquired antigen . The molecular tracking device includes phosphorothioate DNA conjugation , which provides a combined TLR-antigen delivery system to study antigen distribution at time points beyond the lifetime of antigen-fluorophore conjugates and provided a map of cell types involved in antigen-psDNA acquisition and retention . Our approach expands upon our previous studies with other vaccine regimens that induce antigen archiving and cell types that enhance protective immunity . Both here and in our previous studies , we found that whereas LECs archive antigen , migratory DCs passing through the lymphatic vasculature are required to retrieve and present archived antigen to memory CD8 T cells derived from the initial infection or immunization ( Eisenbarth , 2019 ) . Using an antigen/polyI:C/αCD40 vaccine regimen , we determined that antigen exchange from LECs to DCs and subsequent DC presentation yields memory CD8 T cells with robust effector function during infectious challenge . The studies included here predict the same outcome as both LECs and migratory DCs were detected with ova-psDNA at the late time point . Several recent reports defined LEC and non-endothelial SC subsets within the LN ( Rodda et al . , 2018; Fujimoto et al . , 2020; Kalucka et al . , 2020; Xiang et al . , 2020 ) . By combining our molecular tracking device with these reference cell types , we found that non-endothelial SC types acquire foreign antigens including CD34+ SCs , which neighbor subcapsular sinus LECs in the tissue ( Rodda et al . , 2018 ) . These findings suggest that the interstitial pressure created by subcutaneous vaccination allows antigens to pass through the tissue directly to the LN capsule , bypassing the lymphatic capillaries . Intriguingly , bypass of lymphatic capillaries may still lead to LEC acquisition of antigens from the CD34+ SCs via SC-LEC exchange . Such a mechanism would encourage future LEC-DC interactions and provide a benefit to protective immunity . Molecular tracking devices provide a measure of cell state orthogonal to gene expression , which we leveraged to identify candidate pathways involved in antigen-psDNA acquisition ( Figure 4 ) . We show that the caveolin pathway is upregulated in antigen-high LECs and demonstrate this pathway is involved in antigen acquisition in vivo following vaccination with fluorescent ova/polyI:C/αCD40 via pharmacological inhibition of caveolar endocytosis ( Figure 4f ) . Genes uniquely expressed by LECs such as Prox1 , Cavin2 , and Stab2 ( Miller et al . , 2012; Heng et al . , 2008; Malhotra et al . , 2012 ) represent targets for further manipulation of antigen archiving by LECs . The psDNA component of the tracking device elicits an immune response similar to other TLR-antigen conjugate vaccines ( Oh and Kedl , 2010; Oh et al . , 2012 ) , likely due to antigen-psDNA stability within DCs that causes prolonged antigen presentation in the cells that acquire the antigen ( van Montfoort et al . , 2009; Xu and Moyle , 2018 ) . This effect is illustrated by increased IFNγ production in the absence of ex vivo peptide stimulus ( ova-psDNA compared to unconjugated ova; Figure 2 ) . Prolonged antigen presentation better replicates an infection wherein levels of viral or bacterial antigen rise over the duration of infection . However , in other applications it may be helpful to limit the immunoreactivity of the antigen-psDNA via cytosine methylation ( Hemmi et al . , 2000 ) or backbone modification ( Lange et al . , 2019 ) . While many of the experiments we performed with the ova-psDNA were consistent with our experiments using antigen-TLR conjugates or TLR/CD40-based vaccines , it is likely that this type of vaccine interacts with different cell types and utilizes different mechanisms for antigen acquisition and retention . These mechanisms are currently under active investigation and may be more generalizable in the absence of TLR9 . A caveat of our studies concerns the dynamic stability of molecular tracking devices in tissue . Multiple detection methods showed that the protein and DNA components of our conjugates co-localize in LECs ( Figure 1f , Figure 1—figure supplement 1c ) and bone marrow-derived DCs ( Figure 1c , d , g , h ) , and unconjugated psDNA was untrackable both in vitro or in vivo ( Figure 1c , f , Figure 3—figure supplement 1 ) , indicating that psDNA is not readily taken up by cells . However , flow cytometry analysis of conjugates in BMDM indicates that DNA degradation may precede protein degradation ( Figure 1—figure supplement 1g ) . With that said , it remains possible that acquisition of molecular tracking devices by certain cell types leads to decoupling of the individual components after which they could be independently transferred to other cells via trogocytosis or other mechanisms of membrane transfer ( Alegre et al . , 2010 ) . Closer evaluation of the protein-DNA complex in vivo over time will be important to determine how accurately detection of the DNA via single-cell sequencing reflects the movement of the protein-DNA complex . Future experiments will address the dynamics of conjugate stability across multiple cell types to quantify the low levels of unconjugated components and better define the limitations of molecular tracking devices in studying protein degradation intermediates . Molecular tracking devices will enable new approaches to study molecular dissemination in vivo . To date , protein-DNA conjugates have been deployed in single-cell mRNA sequencing experiments for ex vivo staining applications ( e . g . , CITE-seq; Islam et al . , 2014 ) . Our study lays the groundwork for molecular tracking devices involving protein , antibody , drug , or pathogens conjugated to nuclease-resistant , barcoded oligonucleotides that are stable during transit through mouse tissues . The approach naturally extends to understanding how multiple different antigens might be processed ( using unique DNA barcodes ) and enables new studies to manipulate antigen archiving to improve vaccines , vaccine formulations , and prime-boost strategies . Moreover , the oligonucleotide portion of the tracking device should enable analysis of its distribution in cells by in situ hybridization or intact tissue by spatial transcriptomics ( Eng et al . , 2019; Rodriques et al . , 2019; Ståhl et al . , 2016 ) , obviating the need for antibody-mediated detection of antigen .
5-6 week-old mice were purchased from Charles River or Jackson Laboratory , unless otherwise stated , bred and housed in the University of Colorado Anschutz Medical Campus Animal Barrier Facility . Wild type and OT1 mice were all bred on a C57BL/6 background . OT1 mice are a TCR transgenic strain specific to the SIINFEKL peptide of ova ( OVA257-264 ) in the context of H-2Kb . All animal procedures were approved by the Institutional Animal Care and Use Committee at the University of Colorado . Oligonucleotides were synthesized by Integrated DNA Technologies ( IDT ) and contained a 5ʹ amine for conjugation , primer binding site , barcode , 10x Genomics Gel Bead Primer binding site for capture sequence 2 , and a 3ʹ biotin . Phosphorothioate oligonucleotides contained a phosphorothioate modification at every linkage . All oligonucleotide sequences can be found in Figure 1—source data 1 . Oligonucleotides were conjugated to ova by iEDDA-click chemistry ( van Buggenum et al . , 2016 ) . Oligonucleotides were derivatized with trans-cyclooctene ( TCO ) in 10× borate buffered saline ( BBS; 0 . 5 M borate , 1 . 5 M NaCl , pH 7 . 6; sterile filtered ) . Dilution of this buffer to 1× results in a final pH of 8 . 5 . A reaction containing 40 nmol of amine-modified oligo ( 0 . 5 mM ) , 1× BBS , 10% DMSO , 8 µL of 100 mM TCO-PEG4-NHS in DMSO ( 10 mM final; Click Chemistry Tools , A137 ) , pH 8 . 5 was rotated at room temperature for 15 min . A second aliquot containing the same amount of TCO-PEG4-NHS in DMSO was added , and the reaction was rotated at room temperature for another hour . Excess NHS was quenched by adding glycine , pH 8 . 5 to a final concentration of 20 mM and rotated at room temperature for 5 min . Modification was confirmed by analysis on an 8% denaturing TBE PAGE gel . Samples were precipitated by splitting the reaction into 20 µL aliquots and adding 280 µL of nuclease-free water , 30 µL of 3 M NaCl , and 990 µL of 100% ethanol . The precipitation reaction was incubated at −80°C overnight , followed by centrifugation at >10 , 000 , 000 ×g for 30 min . The supernatant was discarded , the pellet was washed with 100 µL of 75% ethanol , and centrifuged at >10 , 000 , 000 ×g for 10 min . The supernatant was removed , and the pellets were dried for 5 min at room temperature . The pellets were recombined by resuspension in 50 µL of 1× BBS . Samples were quantified by A260 . To conjugate methyltetrazine to ova , detoxified ova ( Sigma-Aldrich , St . Louis , MO ) ( using a Triton X-114 lipopolysaccharide detoxification method; Anis et al . , 2007 ) was buffer exchanged into 1× BBS , pH 8 . 5 . To an Amicon 0 . 5 mL 30 kDa filter ( Millipore , UFC5030 ) was added 1 mg of ova and 1× BBS to a volume of 450 μL . The filter was centrifuged at 14 , 000 ×g for 5 min . The flow through was discarded and the sample washed twice with 400 µL of 1× BBS . The product-containing column was inverted into a clean collection tube and centrifuged at 1000 ×g for 2 min . Assuming no loss , the volume of the sample was adjusted to 2 mg/mL with 1× BBS . 400 µL of 1× BBS was added to the Amicon filter and stored at 4°C for later use . A 500 µL labeling reaction containing 0 . 5 mg of ova in 1× BBS and 50 µL of 2 mM mTz-PEG4-NHS in DMSO ( 0 . 2 mM final; Click Chemistry Tools , 1069 ) , pH 8 . 5 was rotated at 4°C overnight . Excess NHS was quenched by adding glycine , pH 8 . 5 to a final concentration of 20 mM and rotated at room temperature for 10 min . The previously stored Amicon filter was centrifuged at 14 , 000 ×g for 5 min and the flow through discarded . 400 µL of reaction mixture was added to the filter and centrifuged at 14 , 000 ×g for 5 min . This was repeated until all 1 mg of protein had been added to the filter and was supplemented with 1× BBS as needed . Samples were washed 1× with 400 µL of 1× BBS . The product-containing column was inverted into a clean collection tube and centrifuged at 1000 ×g for 2 min . Assuming no loss , the volume of the sample was adjusted to 5 mg/mL with 1× BBS . For the final antigen-DNA conjugation , a 100 µL reaction containing 300 µg of ova-mTz and 6 nmol of oligonucleotide-TCO ( 1:1 equivalents ) in 1× BBS was rotated at 4°C overnight . Excess mTz was quenched with 10 µL of 10 mM TCO-PEG4-glycine and rotated at room temperature for 10 min . TCO-PEG4-glycine was prepared by reaction of 10 mM TCO-PEG4-NHS with 20 mM glycine , pH 8 . 5 in 1× BBS for 1 hr at room temperature and stored at −20°C . Products were analyzed by 10% TBE PAGE . For purification , excess ova and DNA were removed by filter centrifugation . 200 µL of 1× PBS was added to an Amicon 0 . 5 mL 50 kDa filter ( Millipore , UFC5050 ) followed by 300 µL of sample . The filter was centrifuged at 14 , 000 ×g for 5 min and the flow through discarded . Samples were washed five times with 400 µL of 1× PBS and centrifuged at 14 , 000 ×g for 5 min . The product-containing column was inverted into a clean collection tube and centrifuged at 1000 ×g for 2 min . Purified products were analyzed by 10% TBE PAGE and total protein quantified with Bio-Rad protein quantification reagent ( Bio-Rad , 5000006 ) . LPS contamination after conjugation was below 0 . 5 EU/mg as mentioned in the 'Vaccinations' section . Both left and right tibia and femur were isolated under sterile conditions . Bone marrow was extracted from femurs of 6–8-week-old C57BL/6 mice by decollating the top and bottom of the bone and releasing the marrow with 27 gauge syringe and 5 mL of Modified Essential Medium ( MEM ) ( Cellgro ) . Suspension was strained through 100 μm filter , pressed with the back of a syringe and washed . Cells were spun 1500 RPM , 5 min then suspended in minimum essential medium ( MEM ) with 10% FBS , 20 ng/mL of Granulocyte-macrophage colony-stimulating factor ( GM-CSF ) from the supernatant of the B78hi-GM-CSF cell line . Every 2 days , dead cellular debris was spun , supernatant collected and combined 1:1 with new 40 ng/mL GM-CSF 20% FBS ( 2× ) in MEM . After 7 days of culturing at 37°C , 5% CO2 cells were harvested for respective assays . Mouse LECs ( Cell Biologics , C57-6092 ) were cultured in Endothelial Cell Media ( Cell Biologics , M1168 ) with kit supplement . T75 Flasks were coated with gelatin for 30 min 37°C , washed with PBS , and then inoculated with mLEC . Cells were passaged with passive trypsin no more than 3–6 times and split at density of 1:2 . SVEC4-10 ( ATCC CRL2181 ) , an SV40 transformed endothelial cell line , was purchased from ATCC and mycoplasma tested before use . SVECs have been characterized to be similar to LECs ( Xiong et al . , 2017 ) , and CD31 and PDPN expression were validated prior to use . SVEC were cultured in RPMI with 10% FBS and passaged with passive trypsin and split at a density of 1:3 . For BMDMs , whole bone marrow was isolated and red blood cells were lysed . Cells were then cultured in M-CSF ( 50 ng/mL ) for 6 days in complete media . Cells were harvested via cell scraper and plated for treatment . Dendritic cells ( BMDC ) , endothelial cells ( mLEC ) , or SV-40 transformed endothelial cells ( SVECs ) or BMDM cultures were stimulated with 5 µg of either ova-psDNA or ova with or without 20 µg of anti-CD40 , 20 µg Poly I:C in a 6-well format . 24 hr post treatment , cells were washed and refreshed with new media . At designated time points , cells were harvested , counted , and transferred into micro-centrifuge tubes , spun at 350 g , and both supernatant and pellets were frozen at −80°C . Cell pellets were lysed in 50 µL of Mammalian Protein Extraction Reagent ( Thermo Scientific , 78503 ) . Conjugate DNA was measured by qPCR amplification from 1 µL of lysate in a 10 µL reaction containing 5 µL of iTaq Universal SYBR Green Supermix ( Bio-Rad , 1725125 ) and 5 pmol of each primer ( Figure 1—source data 1 ) . Quantification was measured using an external standard curve and normalized to lysate protein content . To visualize within ova-psDNA acquisition by cells , cells were fixed with 10% formalin for 10 min at room temperature in the dark , washed with PBS , and spun for 10 min at 2000 rpm . Cells were then permeabilized with 100% ice-cold methanol for 20 min at −20°C . Cells were then washed with PBS and spun as above . Cells were stained with the anti-ova antibody as above for at least 2 hr at room temperature and then washed with 1% bovine serum albumin ( BSA ) with sodium azide ( FACS buffer ) and spun as above . Cells were then incubated with an anti-rabbit secondary in PE for 1 hr at room temperature and then washed with FACS buffer . Cells were then stained with streptavidin conjugated to BV421 in PBS for 15 min at room temperature and then washed twice with FACS buffer prior to acquiring cells on a FACS CANTO II flow cytometer . Analysis was performed using FlowJo software . Immunofluorescence was performed as above except cells were grown on glass coverslips and stained on cover slips using an anti-rabbit dylight 649 and streptavidin-FITC . Coverslips were mounted with Vectashield with DAPI and imaged on a Zeiss LSM780 confocal microscope . The imaging experiments were performed in the Advanced Light Microscopy Core part of the NeuroTechnology Center at University of Colorado Anschutz Medical Campus supported in part by the Rocky Mountain Neurological Disorders Core Grant Number P30 NS048154 and by the Diabetes Research Center Grant Number P30 DK116073 . Contents are the authors' sole responsibility and do not necessarily represent official NIH views . CD8 T cells were isolated from an OT1+ mouse using the mojosort mouse CD8 T cell isolation kit ( Biolegend ) and labeled with violet proliferation dye ( BD Biosciences cat# 562158 ) . For DC-T cell co-culture , BMDCs were treated with psOVA ( 5 μg ) , or ova+psDNA ( 5 μg ) for 1 , 3 , or 7 days . BMDCs were washed and then co-cultured with labeled OT1s for 3 days at a 1:10 ratio of BMDC:OT1 . Cells were then stained and run on a flow cytometer . OT1 division ( percent dividing cells ) was calculated as previously described ( Roederer , 2011 ) using the equation fraction diluted =∑1iNi2i /∑0iNi2i , where i is the generation number ( 0 is the undivided population ) , and Ni is the number of events in generation i . 6–8-week-old C57BL/6 ( CD45 . 2 ) mice were immunized with 1E3 or 1E4 colony-forming units ( CFU ) of Vaccinia Western Reserve or 5 µg of poly I:C ( Invivogen ) with or without 5 µg of anti-CD40 ( FGK4 . 5 , BioXcell ) and 10 µg of ova-psDNA or ova in 50 μL volume by footpad injection . Endotoxin levels were quantified using the Pierce Limulus Amebocyte Lysate Chromogenic Endotoxin Quantitation kit ( Thermo Scientific ) to be less than 0 . 5 EU/mg for either ova or ova conjugated to psDNA . Nystatin ( Sigma N4014 ) was resuspended in DMSO to a concentration of 10 mg/mL . Mice were injected with 50 μL of 10 mg/mL nystatin per footpad 1 hr prior to injection with ova conjugated to Alexa 488 ( 5 μg ) in a mixture with polyI:C and anti-CD40 ( 2 . 5 μg each ) . LNs were harvested and digested as below ( preparation of single-cell suspensions ) and stained with CD45 brilliant violet 510 ( Biolegend clone 30F11 , 1:300 ) , PDPN APC ( Biolegend clone 8 . 1 . 1 , 1:200 ) , CD31 PercP Cy5 . 5 ( Biolegend clone 390 , 1:200 ) , and PD-L1 pacific blue ( Biolegend clone 10F . 9G2 , 1:200 ) . Draining LNs were processed by glass slide maceration 7 days after injection , washed , and suspended in FACS ( 2% FBS in PBS ) buffer containing Tetramer ( SIINFEKL ) -PE ( 1:400 ) ( NIH tetramer core facility ) , CD8 APC-Cy7 ( Biolegend clone 53-6 . 7 1:400 ) for 1 hr at 37C . Cells were washed and stained for 30 min in CD44 PerCP Cy5 . 5 ( Biolegend clone IM7 , 1:400 ) , B220 BV510 ( Biolegend clone RA3-6B2 , 1:300 ) . Samples were ran on the FACS Canto II flow cytometer ( BD ) . 2 days or 2 weeks following vaccination with 1E3 CFU of VV-WR with 10 μg of ova-psDNA per footpad , popliteal LNs were removed from 15 mice and LNs were pulled apart with 22-gauge needles . Tissue was digested with 0 . 25 mg of Liberase DL ( Roche , Indianapolis , IN ) per mL of EHAA media with DNAse ( Worthington , Lakewood , NJ ) at 37° . Every 15 min media was removed , cells spun down , and new digestion media added to the undigested tissue until no tissue remained , ~1 hr . Following digestion , cells were filtered through a screen and washed with 5 mM EDTA in EHAA . LN cells were then divided into thirds where one-third underwent staining with CD11c ( N418 ) , CD11b and B220 , and a live/dead dye ( Tonbo ) . Live cells were then sorted into four tubes on a FACS Aria Cell Sorter ( BD ) : sorted CD11c-APC Cy7 ( Biolegend clone N418 1:400 ) + cells , sorted CD11b PE-Cy7 ( Biolegend clone M1/70 ) + cells , sorted B220 PE ( Biolegend clone RA3-6B2 ) + cells and Fixable Viability Stain 510 ( BD Biosciences Cat # 546406 ) ungated live cells , which were recombined at a 4:4:1:1 ratio , respectively . For the remaining two-thirds of cells , cells were stained with CD45 PE followed by magnetic bead isolation using the Miltenyi bead isolation kit . CD45-negative cells that passed through the column were then washed . Both sorted and selected ( CD45+ and CD45- ) cells were then washed with PBS in 0 . 1% BSA as described in the Cell Prep Guide ( 10x Genomics ) and counted using a hemacytometer . Final concentration of cells was approximately 1000 cells/µL and approximately 10–20 µL were assayed . Cells were assayed using the 10x Genomics single-cell 3ʹ expression kit v3 according to the manufacturer’s instructions ( CG000183 Rev B ) and CITE-seq protocol ( cite-seq . com/protocol Cite-seq_190213 ) with the following changes: All libraries were sequenced on a Illumina NovaSeq 6000 with 2 × 150 base pair read lengths . Briefly , FASTQ files from the gene expression and antigen tracking libraries were processed using the feature barcode version of the cellranger count pipeline ( v3 . 1 . 0 ) . Reads were aligned to the mm10 and vaccinia virus ( NC_006998 ) reference genomes . Analysis of gene expression and antigen tracking data was performed using the Seurat R package ( v3 . 2 ) . Antigen tracking and gene expression data were combined into the same Seurat object for each sample ( CD45-/day 2 , CD45+/day 2 , CD45-/day 14 , CD45+/day 14 ) . Cells were filtered based on the number of detected genes ( >250 and <5000 ) and the percent of mitochondrial reads ( <15% ) . Gene expression counts were log-normalized ( NormalizeData ) , and relative ova signal was calculated by dividing ova-psDNA counts by the median ova-psDNA counts for all T and B cells present in the sample . To allow for the values to be log-transformed for visualization , a pseudo count was added ( smallest non-zero value * 0 . 5 ) . Gene expression data were scaled and centered ( ScaleData ) . 2000 variable features ( FindVariableFeatures ) were used for PCA ( RunPCA ) , and the first 40 principal components were used to find clusters ( FindNeighbors , FindClusters ) and calculate uniform manifold approximation and projection ( UMAP ) ( RunUMAP ) . Cell types were annotated using the R package clustifyr ( https://rnabioco . github . io/clustifyr ) ( Fu et al . , 2020 ) along with reference bulk RNA-seq data from ImmGen ( available for download through the clustifyrdata R package , https://rnabioco . github . io/clustifyrdata ) . To annotate cell subtypes , the samples were divided into separate objects for DCs , LECs , and FRCs and reprocessed ( FindVariableFeatures , ScaleData , RunPCA , RunUMAP , FindNeighbors , FindClusters ) . Cell subsets were annotated using clustifyr with reference bulk RNA-seq data for DCs ( Brown et al . , 2019; Miller et al . , 2012 ) , FRCs ( Rodda et al . , 2018 ) , and LECs ( Fujimoto et al . , 2020; Kalucka et al . , 2020; Xiang et al . , 2020 ) . After assigning DC , LEC , and FRC subtypes , the other cell types ( T/B cells , epithelial cells , NK cells ) were added back to the objects and reprocessed as described above . Identification of ova-low and -high populations was accomplished using a two-component Gaussian mixture model implemented with the R package mixtools ( https://cran . r-project . org/web/packages/mixtools/index . html ) . All LECs were used when identifying ova-low and ova-high cells ( Figure 4 ) . For DCs ( Figure 3—figure supplement 5 ) , ova-low and -high populations were identified independently for each DC cell type . For ova-low and ova-high populations , differentially expressed genes were identified using the R package presto ( wilcoxauc , https://github . com/immunogenomics/presto ) . Differentially expressed genes were filtered to include those with an adjusted p-value<0 . 05 , log fold-change > 0 . 25 , area under the receiver operator curve ( AUC ) > 0 . 5 , and with at least 50% of ova-high cells expressing the gene . Raw and processed data for this study have been deposited at NCBI GEO under accession GSE150719 . A reproducible analysis pipeline is available at https://github . com/rnabioco/antigen-tracking http://doi . org/10 . 5281/zenodo . 4615724 ( Sheridan and Hesselberth , 2021; copy archived at swh:1:rev:f7f6c0696f08aeeac6ad88c39975197a0791e30d ) . Statistical analysis was done using either a non-parametric two-tailed Mann–Whitney t-test or multiple t-tests with a two-stage step-up method of Benjamini , Krieger , and Yekutieli without assuming consistent standard deviations . A biological replicate was considered a measurement of a biologically distinct sample ( such as a separate mouse ) , and a technical replicate was considered a repeated measurement of the same sample . Each in vivo analysis was performed with 3–6 mice per group as determined by a power calculation using the assumption ( based on prior data ) that there will be at least a twofold change with a standard deviation of less than 0 . 5 . To calculate numbers , we performed a power calculation with an α of 0 . 5 and a 1-β of 0 . 80 to determine at least three mice per group are evaluated . Error bars indicate the standard error of the mean ( SEM ) , and all analyses were blinded . | The lymphatic system is a network of ducts that transports fluid , proteins , and immune cells from different organs around the body . Lymph nodes provide pit stops at hundreds of points along this network where immune cells reside , and lymph fluid can be filtered and cleaned . When pathogens , such as viruses or bacteria , enter the body during an infection , fragments of their proteins can get swept into the lymph nodes . These pathogenic proteins or protein fragments activate resident immune cells and kickstart the immune response . Vaccines are designed to mimic this process by introducing isolated pathogenic proteins in a controlled way to stimulate similar immune reactions in lymph nodes . Once an infection has been cleared by the immune system , or a vaccination has triggered the immune system , most pathogenic proteins get cleared away . However , a small number of pathogenic proteins remain in the lymph nodes to enable immune cells to respond more strongly and quickly the next time they see the same pathogen . Yet it is largely unclear how much protein remains for training and how or where it is all stored . Current techniques are not sensitive or long-lived enough to accurately detect and track these small protein deposits over time . Walsh , Sheridan , Lucas , et al . have addressed this problem by developing biological tags that can be attached to the pathogenic proteins so they can be traced . These tags were designed so the body cannot easily break them down , helping them last as long as the proteins they are attached to . Walsh , Sheridan , Lucas et al . tested whether vaccinating mice with the tagged proteins allowed the proteins to be tracked . The method they used was designed to identify individual cell types based on their genetic information along with the tag . This allowed them to accurately map the complex network of cells involved in storing and retrieving archived protein fragments , as well as those involved in training new immune cells to recognize them . These results provide important insights into the protein archiving system that is involved in enhancing immune memory . This may help guide the development of new vaccination strategies that can manipulate how proteins are archived to establish more durable immune protection . The biological tags developed could also be used to track therapeutic proteins , allowing scientists to determine how long cancer drugs , antibody therapies or COVID19 anti-viral agents remain in the body . This information could then be used by doctors to plan specific and personalized treatment timetables for patients . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"tools",
"and",
"resources",
"immunology",
"and",
"inflammation"
] | 2021 | Molecular tracking devices quantify antigen distribution and archiving in the murine lymph node |
As exome sequencing gives way to genome sequencing , the need to interpret the function of regulatory DNA becomes increasingly important . To test whether evolutionary conservation of cis-regulatory modules ( CRMs ) gives insight into human gene regulation , we determined transcription factor ( TF ) binding locations of four liver-essential TFs in liver tissue from human , macaque , mouse , rat , and dog . Approximately , two thirds of the TF-bound regions fell into CRMs . Less than half of the human CRMs were found as a CRM in the orthologous region of a second species . Shared CRMs were associated with liver pathways and disease loci identified by genome-wide association studies . Recurrent rare human disease causing mutations at the promoters of several blood coagulation and lipid metabolism genes were also identified within CRMs shared in multiple species . This suggests that multi-species analyses of experimentally determined combinatorial TF binding will help identify genomic regions critical for tissue-specific gene control .
The combinatorial binding of transcription factors to DNA define the gene regulatory regions that are essential for achieving spatial and temporal gene expression ( Zinzen et al . , 2009; Gerstein et al . , 2012; Hardison and Taylor , 2012 ) . The rapid increase in empirically determined TF bound motifs ( Badis et al . , 2009; Jolma et al . , 2013; Weirauch et al . , 2013 ) , sequenced genomes ( Goode et al . , 2010; Lindblad-Toh et al . , 2011; 1000 Genomes Project Consortium et al . , 2012 ) , and genome-wide profiling of DNA–protein interactions has given us unprecedented insight into the location of gene regulatory regions in multiple tissue and cell types . In particular , experimental results obtained by chromatin immunoprecipitation ( ChIP ) , FAIRE , and DNase I footprinting assays in combination with high-throughput sequencing have unmasked what was previously a hidden landscape of active DNA regions ( Rhee and Pugh , 2011; Furey , 2012; Neph et al . , 2012 ) . The compendium of ChIP-seq determined DNA-binding for 119 different proteins in 72 cell experiments produced by the Encyclopedia of DNA Elements ( ENCODE ) consortium alone has revealed that the number of TF binding events greatly exceeds the number of genes in the genome and that over 8% of the genome can be bound by at least one TF ( ENCODE Project Consortium , 2012 ) . The large number of TF bound genomic regions highlights the growing need for rational strategies for distilling these protein–DNA interactions into functional and non-functional categories . It has recently been shown that unlike TF binding events with high TF occupancy levels ( measured by ChIP signal ) , genomic regions with low TF occupancy levels are not responsible for patterned reporter gene expression in Drosophila ( Fisher et al . , 2012 ) . It remains to be seen how TF occupancy levels relate to functional gene expression in other species . Comparing DNA between species has long been employed to identify transcription factor ( TF ) binding sites that comprise gene regulatory regions ( e . g . , Tagle et al . , 1988; Lindblad-Toh et al . , 2011 ) . Indeed , functional reporter gene expression assays have shown that many highly conserved mammalian non-coding regions serve as developmental limb and nervous system enhancers ( Pennacchio et al . , 2006 ) . In contrast , other tissues including the heart ( Blow et al . , 2010; May et al . , 2012 ) , liver ( Kim et al . , 2011 ) , and adult brain ( Visel et al . , 2013 ) possess many functional enhancers that do not show such deep phylogenetic preservation at the DNA level . An increasingly used way to identify tissue and species-specific gene regulatory regions is to compare experimentally determined TF–DNA interactions or histone modifications between species ( Kunarso et al . , 2010; Mikkelsen et al . , 2010; Schmidt et al . , 2010 , 2012; Xiao et al . , 2012; Cotney et al . , 2013; Paris et al . , 2013 ) . For example , we previously established that the target genes of CEBPA and HNF4A , as identified from gene expression studies of conditional liver TF knockout mice , were enriched for TF binding shared in multiple species ( Schmidt et al . , 2010 ) . Similarly , functional Drosophila enhancers are more likely to be found in regions with conserved TF binding events detected by ChIP ( Paris et al . , 2013 ) . Associating common genetic variation with complex traits is another powerful way to identify functional regulatory DNA in the human genome . Over 80% of the most significant single nucleotide polymorphisms ( SNPs ) associated with human phenotypes and disease occur within non-coding regions of the genome ( Hindorff et al . , 2009 ) . Recent integrative analyses have shown that open chromatin regions obtained for a specific cell type ( e . g . , DNase I hypersensitivity sites in T-cells ) are enriched for reported GWAS SNPs . Importantly , this GWAS enrichment appeared most significant when the DNAse data was ascertained in a cell type relevant to the phenotype studied ( Maurano et al . , 2012; Reddy et al . , 2012; Schaub et al . , 2012 ) . Examples of regulatory DNA mutations that explain differences in disease gene function are increasingly being discovered ( e . g . , Musunuru et al . , 2010 ) and there is tremendous interest in methods that can predict which non-coding variants are of functional consequence ( Schaub et al . , 2012; Ward and Kellis , 2012a , 2012b ) . To test whether evolutionary conservation of cis-regulatory modules ( CRMs ) gives insight into human gene regulatory function , we determined transcription factor ( TF ) binding locations of four liver-enriched TFs in liver tissue from: two primates ( human and macaque ) estimated to have diverged 29 million years ago; two rodents ( mouse and rat ) estimated to have diverged 25 million years ago; and dog which diverged during the mammalian radiation along with primate and rodent lineages ( Hedges et al . , 2006 ) . The liver is a suitable tissue for studying vertebrate gene regulation . It is a relatively homogenous tissue with approximately 75% of the nuclei in the liver coming from hepatocytes ( Marcos et al . , 2006 ) . Both the relative homogeneity and the large cell numbers that can be isolated from diverse organisms under physiologically optimal conditions lend itself well to comparative studies . We focus on four TFs required for liver cell specification and gene function ( HNF4A , CEBPA , ONECUT1 , and FOXA1 ) ( Kyrmizi et al . , 2006 ) . Together , several studies have demonstrated that these four TFs work together directly and indirectly to drive liver-specific function ( Plumb-Rudewiez et al . , 2004 ) . Using liver as a model tissue , we demonstrate how a combinatorial analysis of TF occupancy across multiple species can highlight conserved and species-specific biological processes , as well as potential mechanistic actions of disease variants .
The genome-wide occupancy of four transcription factors ( HNF4A , CEBPA , ONECUT1 , and FOXA1 ) was determined in primary liver in five species ( Homo sapiens [Hsap] , Macaca mulatta [Mmul] , Canis familiaris [Cfam] , Mus musculus [Mmus] , and Rattus norvegicus [Rnor] ) using chromatin immunoprecipitation followed by high-throughput sequencing ( ChIP-seq ) ( Figure 1 , Figure 1—figure supplement 1A , Figure 1—source data 1 ) . The antibodies used for the four TFs have been raised against conserved epitopes and have previously been validated in ChIP experiments in mouse and human ChIP studies ( Figure 1—source data 1D ) . As expected from previous multi-species ChIP study of CEBPA and HNF4A ( Schmidt et al . , 2010 ) , the known binding motifs for the four TFs was virtually identical between species and occurred close to the ChIP-seq binding summit ( Figure 1—figure supplement 1B , C ) . 10 . 7554/eLife . 02626 . 003Figure 1 . Overview of ChIP-seq , CRM construction , and multiple-species comparisons . ChIP-seq peaks were determined for four liver TFs in five mammals . ( A ) CRMs were constructed by merging ChIP-seq peaks whose summits occurred within 300 bp and consisted of at least two distinct TFs . Remaining peaks were designated as singletons . ( B ) Whole genome 9-way EPO multiple sequence alignments ( MSA ) were used to project CRMs/Singletons across the five species . A CRM was considered shared if its position in the EPO MSA overlapped a CRM in a second species by a minimum of 10 bp . Neither the content nor order of TFs within the CRM was required to be classified as a ‘Shared’ CRM . A singleton in one species was considered ‘Shared’ if it overlapped the same TF in a second species . ( C ) Relative to human , the average % of shared CRMs is shown . Human CRMs ( comprised of any two TFs ) that overlap a CRM from a second species are shown with empty circles . Human CRMs containing at least one of each TF ( all 4 TFs ) were compared to all identified CRMs in a second species ( purple circles ) . ( D ) The percentage of human CRMs and singletons in different phylogenetic categories that can be found aligned within the EPO MSAs for each of the five species is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 00310 . 7554/eLife . 02626 . 004Figure 1—source data 1 . Quality control for ChIP-seq , CRM construction , and multi-species comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 00410 . 7554/eLife . 02626 . 005Figure 1—figure supplement 1 . Summary of ChIP-seq peak number and TF motif enrichment . ( A ) Number of Peaks ( B ) Conservation of TF binding motifs . DNA binding specificities of CEBPA , HNF4A , FOXA1 , and ONECUT1 are highly conserved . The known sequence motifs were identified de novo in each species interrogated . ( C ) Central position of motifs under TF binding summits were observed for each factor . Number of motifs identified using the PWM from B ( y-axis ) vs distance from TF binding summit obtained using the SWEMBL peak caller ( x-axis ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 00510 . 7554/eLife . 02626 . 006Figure 1—figure supplement 2 . Pairwise analysis of individual TFs using EPO multiple sequence alignment . Shared binding events were initially identified in a pair-wise fashion using the 9-way EPO-MSA . Average % overlap was calculated and the total number of peaks that could be aligned to a second species is shown . The total peak number for each species shown in brackets . A region was identified as shared if its DNA sequence overlapped in a second species by 10 or more base pairs . ( A ) Average % overlap for the four individual TFs in this study . ( B ) For each pair of species the average % of CRMs ( comprised of any two TFs ) that overlap another CRM , regardless of its composition from a second species is shown . The number of CRMs present within the EPO-MSA is given on the right with the total number of CRMs shown in brackets . ( C ) For each species , CRMs containing at least one of each TF type ( all 4 TFs ) were compared to all identified CRMs in a second species ( Figure 1—source data 1F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 00610 . 7554/eLife . 02626 . 007Figure 1—figure supplement 3 . A majority of the four liver-enriched TFs cluster into CRMs . ( A ) The number of TF binding event clusters ( y-axis ) vs the distance between the summits of TF binding events used to generate the clusters ( x-axis ) is plotted . ( B ) Number of CRMs ( y-axis ) vs CRM width ( x-axis ) is shown for our chosen 300 bp distance-between-summit criteria we used to build CRMs . ( C ) Average width in base pairs for: all CRMS; CRMs containing all four TFs; and individual peaks are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 007 Similar to what was observed for previous CEBPA and HNF4A ChIP-seq experiments , only a minority of ONECUT1 and FOXA1 bound regions overlapped orthologous , TF-bound regions in a second species , a relationship we refer to here as “shared” TF binding ( see Figure 1A , B , Figure 1—figure supplement 2A ) . The rapid evolution of TF binding is further supported by comparisons within primate and rodent orders that are separated by less than 25 million years ( Springer et al . , 2003 ) . For example , on average , between 21 and 37% of TF binding events in human are found in the orthologous location in macaque and 21–31% between mouse and rat for each of the TFs assayed ( Figure 1—figure supplement 2A ) . Tissue-specific TFs are known to bind in close proximity to form cis regulatory modules ( CRMs ) . Similar to what has been done for multi-TF binding analyses in Drosophila ( Zinzen et al . , 2009 ) and mouse ( Stefflova et al . , 2013 ) , we defined CRMs by clustering at least two proximal heterotypic TF binding events ( Figure 1A ) . The number of liver TFs forming clusters falls off sharply when the distance between them is greater than 150 bp , which is less than the average width of the TF bound regions we detected by ChIP-seq ( Figure 1—figure supplement 3 ) . We built CRMs by merging TF binding events whose summits were within 300 bp of each other ( Figure 1 ) . Using this summit-based clustering , we found that approximately two thirds of the human liver TF binding events were incorporated into CRMs ( Figure 1—figure supplement 1A ) . We found that the shared CRM categories were robust to using a more permissive peak caller or calling peaks on individual biological replicates ( Figure 1—source data 1E ) . As we found for individual TFs , the location of CRMs appears to have evolved rapidly ( Figure 1C ) . For example , we found that only ∼35% of human CRMs had a CRM in the orthologous macaque genomic region . Similarly , ∼32% of mouse CRMs were found as CRMs in the orthologous location in the rat genome ( Figure 1—figure supplement 2C ) . This divergence of CRM occupancy was consistent between different lineages separated by the similar evolutionary distances ( Figure 1—source data 1F ) , robust to the multiple sequence alignments ( MSA ) used to detect orthologous CRMs , and also robust to different overlap methods chosen to infer CRM conservation between species ( Figure 1—source data 1G ) . Figure 1D shows that most ( >93% ) of human CRMs and singletons we detect are found in the EPO MSA ( Paten et al . , 2008 ) with macaque , which suggests that the rapid turnover observed between human and macaque CRMs is not due to characteristics of the multiple alignment . CRMs containing all four TFs are on average more highly shared with a CRM from a second species ( e . g . , 53% of human CRMs with all four TFs are shared with a macaque CRM ) , indicating increased selection pressure on higher order combinatorial TF binding ( Figure 1C , Figure 1—figure supplement 2C ) . TF binding events shared in multiple species are more likely to be found within CRMs ( 72% of shared human TF binding events are in CRMs vs 27% that are classified as singletons; hypergeometric test , p = 8 . 48 × 10−238 ) . For example , 32 of the 35 CEBPA binding events previously found to be bound in orthologous regions in five vertebrate species ( Schmidt et al . , 2010 ) fell within CRMs identified in this study . To test how combinatorial binding and TF binding conservation relate to liver gene function , we classified our set of human CRMs ( n = 31 , 765 ) and singletons ( n = 43 , 824 ) into phylogenetic categories ( Figure 2 , Figure 2—source data 1 ) . CRMs were categorized as one of the following: shared only in human and macaque ( Primates only , n = 4672 ) ; shared in human plus at least one non-primate ( Beyond primates , n = 7631 ) ; and shared in at least three species ( Deeply shared n = 5046 ) ( Figure 2 ) . The 43 , 824 singletons not residing in CRMs ( 44% ) were categorized in the same manner ( Figure 2 ) . 10 . 7554/eLife . 02626 . 008Figure 2 . Annotation of human regulatory regions using interspecies combinatorial transcription factor binding . ( A ) Human liver ChIP-seq data from ONECUT1 , HNF4A , FOXA1 , and CEBPA were assembled into CRMs consisting of at least 2 of the 4 TFs . The CRMs or single TFs were then broken down into categories based on their overlap with ChIP-seq data in macaque , dog , mouse , and rat . Singletons and CRMs were considered shared if they overlapped at least 10 bp with another TF bound region in the EPO multiple sequence alignment ( MSA ) . ( B ) Experimentally determined combinatorial binding at the blood coagulation F7 locus . Raw sequencing reads from ChIP-seq experiments: CEBPA ( red ) , HNF4a ( green ) , ONECUT1 ( yellow ) , and FOXA1 ( green ) are overlaid and called peaks are displayed for each species . ChIP-seq determined TF binding events were assembled into CRMs ( black bars ) underneath the enriched regions ( peaks ) . Grey lines are drawn to illustrate shared CRMs using the EPO-MSA . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 00810 . 7554/eLife . 02626 . 009Figure 2—source data 1 . Table of CRMs and singletons along with the phylogenetic categories they were assigned . File coordinates are for the hg18 assembly . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 009 TF binding events contained in CRMs were enriched for their respective TF's DNA binding motif , in addition to other liver TF binding motifs , including those profiled in this study ( Supplementary file 1 ) . Supporting this observation , we find that both Deeply shared and human only CRMs overlap significantly with relevant ENCODE genome-wide experimental data sets , including ChIP peaks for HNF4A , FOXA1 , and CEBPB in the liver cancer cell line HepG2 ( e . g . , p < 10−149 and p < 10−212 for HNF4 respectively; Supplementary file 1 ) . The fold enrichment was higher for Deeply shared CRMs than for human only CRMs ( e . g . , 20 . 5 vs 11 . 8 for HNF4A ) . We also found significant overlap with TFs not tested in our study . For example , 51% of the Deeply shared CRMs and 20% of the human only CRMs overlap binding peaks for SP1 in HepG2 cells ( p < 10−149 and p < 10−212 respectively; Supplementary file 1 ) . Again the enrichment for these additional TFs was higher in the Deeply shared CRM category than the human only category ( e . g . , 17 . 6 vs 8 . 4-fold ) . SP1 and HNF4A have previously been shown to cooperatively regulate gene expression in HepG2 cells ( Sugawara et al . , 2007 ) . In this manner , Deeply shared CRMs can be used to enrich for additional TFs that might play global combinatorial roles in liver gene regulation when used in conjunction with other data sets from related cell types . A comparison of shared CRMs to human-specific CRMs reveals an increase in the number of liver-related biological pathways , diseases , and known target genes of liver enriched TFs ( Figure 3 ) . We used the enrichment tool GREAT ( McLean et al . , 2010 ) to perform functional enrichments . GREAT's default setting assigns TF binding events to a basal region around every gene ( 5 kb upstream , 1 kb downstream ) . ChIP-seq peaks that fall within the basal regulatory region of each gene , as well as the genomic sequence that spans between the basal region of that gene and the nearest gene's basal region ( within a maximum of 1 Mb ) are used to generate functional enrichments . The most significant enrichments that were unique to the shared liver CRMs include: liver disease ( Binomial FDR q-value = 4 . 53 × 10−130 ) from the Disease Ontology database and metabolism of lipids and lipoproteins ( q = 2 . 96 × 10−73 ) from MSigDB Pathway ( Figure 3A , B , Figure 3—source data 1A ) . 10 . 7554/eLife . 02626 . 010Figure 3 . Phylogenetic filtering of experimentally determined liver TF binding events yield distinct functional enrichments . Results were obtained using the programming interface for the online enrichment tool GREAT version 2 . 02 ( McLean et al . , 2010 ) and plotted with custom R scripts . Up to five of the most significant enrichments obtained for each of the six analyses are listed on the left . The −log10 of binomial Q values for Disease ontology , HGNC gene family , and MSigDB are shown along the x-axis . Bars with a black asterisk indicate significant enrichments using GREAT default parameters ( binomial and hypergeometric FDR Q-value significance at P ≤ 0 . 05 with at least twofold region enrichment ) . The size of the asterisk is proportional to the fold enrichment obtained for the given database . See Figure 3—source data 1 for complete list of Q-values , fold enrichments , genes giving the enrichments along with results from additional databases . ( A ) Enrichment analysis of any CRM shared in human plus at least one additional species is shown on the left and human only CRMs are shown on the right ( Figure 3—source data 1A ) . ( B ) Human CRMs ( left panel ) shared in human and at least one non-primate ( Beyond Primates ) is shown vs Human CRMs ( right panel ) shared in human and macaque but no other species ( Primate only ) ( Figure 3—source data 1B ) . ( C ) Enrichment analysis of shared CEBPA CRMs and singletons ( Figure 3—source data 1C ) . ( D ) Enrichment analysis of shared HNF4A CRMs and singletons ( Figure 3—source data 1D ) . ( E ) Enrichment analysis of shared FOXA1 CRMs and singletons ( Figure 3—source data 1E ) . ( F ) . Enrichment analysis of shared ONECUT1 CRMs and singletons ( Figure 3—source data 1F ) . ( G ) Human TFs in CRMs and Singletons were categorized by the number of species in which they are shared with . Profiles of constrained elements ( sequence conservation ) in a 1-kb window around CRMs or singletons were calculated using GERP scores from the 29-way multiple sequence alignments . ( H ) Genomic location of CRMs and Singletons . Proportion of single TFs located near transcription start sites ( TSS ) increases to >50% , but remains stable for CRMs at ∼20% . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 01010 . 7554/eLife . 02626 . 011Figure 3—source data 1 . Functional enrichment results obtained for CRMs and singletons using GREAT . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 011 The most significant liver-related enrichments obtained using human only CRMs were for biological oxidations ( q = 1 . 95 × 10−39; in MSigDB Pathway ) . These enrichments were driven by genes involved in the metabolism of xenobiotics by the cytochrome P450 gene family ( q = 2 . 24 × 10−19; HGNC gene family database ) ( Figure 3A ) . Given the liver's major contribution to the detoxification of xenobiotics and the well established species-specificity of the proteins involved in the process ( Gonzalez and Nebert , 1990 ) , these results suggest that evolutionary filtering of CRMs has the potential to enrich for both conserved and species-specific biological pathways . Singleton TF binding events were predominantly enriched for their respective motif , but were not enriched for the motifs from the other three TFs profiled in this study ( Supplementary file 1 ) . Supporting this , comparisons against all ENCODE TF binding data show that for HNF4A , CEBPA , and FOXA1 singletons , the top ChIP-seq peak association in HepG2 cells corresponded to the TF assayed . HNF4A singletons were enriched for FOX family motifs , albeit not the same FOXA1 motif obtained from CRMs and singleton FOXA1 peaks . Comparing normalized sequence read counts in the HNF4A singletons , and HNF4A-containing CRMs lacking FOXA1 peaks , it is clear that pervasive weak FOXA1 ChIP-seq signal occur at HNF4A binding sites ( Figure 4A ) . Further supporting this hypothesis is the similarity of FOXA1 to a portion of the HNF4A motif ( Figure 4B ) , and a recent study that showed a close association of HNF4A with FOXA1 motifs ( Guo et al . , 2012 ) . 10 . 7554/eLife . 02626 . 012Figure 4 . Comparison of TF occupied regions classified as CRMs and singletons . ( A ) Regions of ±5 kb are represented around the center of CRMs or singletons . Reads centered on the summit of each TF are counts subtracted by input reads in 100 bp bins plus and minus 5 kb from the summit . Colored boxes indicate CRMs or singletons where a peak was called for a given factor: CEBPA ( red ) , HNF4A ( blue ) , ONECUT1 ( orange ) , and FOXA1 ( green ) . Looking at read counts for all four factors reveal that many of the HNF4A singleton in fact have weak FOXA1 signal . ( B ) Alignment of FOXA1 de novo ChIP-seq motif to the HNF4A motif . Motif comparison ( alignment ) was performed using compare-matrices from RSAT . The program calculates the correlation between two matrices shifting positions; the correlation is normalized based on the width of the alignment to avoid high correlation based on few flanking positions . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 01210 . 7554/eLife . 02626 . 013Figure 4—source data 1 . Comparison of motif matches between CRMs and singletons . Chi-square test for differences between the number of peaks associated with CRMs and singletons , for each TF , that contained at least one predicted motif using three different p-value thresholds for motif scanning: stringent ( 10−4 ) , moderate ( 10−3 ) and lenient ( 10−2 ) . Blue shadows highlight siginficnat p-values . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 01310 . 7554/eLife . 02626 . 014Figure 4—figure supplement 1 . Comparison of stringent motif matches between CRMs and singletons . Each TF peak set was scanned using the RSAT tool matrix-scan with PWMs for the four TFs . A stringent motif cutoff ( 10−4 ) was applied . The figure shows the percentage of peaks with at least one detected site for each TF below the corresponding p-value threshold ( y-axis ) and the motif PWM scanned ( x-axis ) . ( A ) Motif scanning in CEBPA peaks; ( B ) Motif scanning in FOXA1 peaks; ( C ) Motif scanning in ONECUT1 peaks; and ( D ) Motif scanning in HNF4A peaks . For the stringent thresholds , significant adjusted p-values comparing singletons and CRMs are given . See all X2 test p-values for lenient ( 10−2 ) and moderate cutoffs ( 10−3 ) in Figure 4—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 014 We asked whether CRMs or singletons differ with regards to the quality of their TF binding motifs . Peaks for each TF were scanned using the RSAT tool matrix-scan with the best position weight matrices ( PWM ) for each TF . We set three p-value threshold cut offs ( stringent:10−4 , moderate:10−3 and lenient 10−2 ) based on the comparison between the theoretical and empirical PWM weight score distribution observed for each peak collection as previously described ( Medina-Rivera et al . , 2011 ) . As expected , motifs were identified in the vast majority of their corresponding peak set using the lenient motif threshold . Similarly , for both singletons and CRMs , the moderate and stringent motif searches returned the highest fraction motifs in their corresponding peak set . Interestingly , for both the moderate and stringent motif searches , the singleton TF binding sites had a significantly greater fraction of high quality motifs than they did for CRMs ( e . g . , 56% for singletons vs 33% for CRMs for the stringent cutoff ) . This trend was observed for all four TFs in this study ( Figure 4—figure supplement 1 , Figure 4—source data 1 ) . Our results in primary liver tissue are supported by an integrative analysis of ENCODE cell lines , which showed that active chromatin states are depleted of regulatory motif instances relative to all regions bound by a given TF ( Ernst and Kellis , 2013 ) . After collapsing the four TFs into CRMs , there were still over 40 , 000 singleton TF binding events . We asked if shared singletons show any distinct genomic properties . Shared singleton TF binding events show high DNA constraint ( Figure 3G ) , and a larger fraction are found close to the transcription start site of annotated genes compared to the CRMs shared in the equivalent number of species ( Figure 3H ) . Unlike the equivalently shared CRMs , shared singleton TF binding events gave fewer and less significant enrichments that than the equivalently shared CRMs ( Figure 3C–F ) . Relative to CRMs , the enrichments unique to shared singletons , did not appear to be overtly liver specific . For example , no significant liver disease ontology enrichments were found for shared singleton CEBPA binding events; however , several cancer disease enrichments from various tissues , such as in situ carcinoma ( q = 1 . 91 × 10−5 ) , were obtained ( Figure 3C ) . One consideration about comparing singletons to CRMs is that singleton TF binding events are likely to become CRMs as more factors are tested and more peaks are called ( see Figure 1—source data 1E for a comparison of the stability singleton and CRM categories ) . Nonetheless , by focusing on the singletons that remain singletons in orthologous regions in two or more species , we have been able to detect distinct genomic properties that warrant future study . To compare the functional properties of shared and species-specific CRMs/singletons , we then looked at how combinatorial binding and evolutionary constraint correlated with gene expression ( Figure 5 ) . Human TF binding events in CRMs and singletons were categorized by the number of species they were shared in and then associated with the nearest gene . Human liver mRNA expression level of the nearest gene was determined by RNA-seq ( Kutter et al . , 2011 ) . Genes nearest to human-specific singletons and CRMs were not significantly different in their expression levels ( p = 0 . 221 ) . In contrast , gene expression levels near TFs in shared CRMs were significantly higher than those near shared singleton-associated genes ( p = 2 . 4 × 10−16 ) ( Figure 5A ) . This striking p-value is due to several CRMs being found close to highly expressed liver genes including albumin , fibrinogen ( FGA , FGB , FGG ) , and several acute phase response genes ( e . g . , CRP , SAA1 etc ) . We therefore broke down each CRM and singleton by transcription factor , which still revealed a significant difference between genes close to Deeply shared CRMs relative to singletons ( p < 1 × 10−3; Figure 5—figure supplement 1 ) . Using a reference transcription data set that comprises RNA-seq data for liver and 15 additional human tissue types ( E-MTAB-513 ) , we confirmed the above observation and found that the gene expression association with liver CRMs , and to a lesser extent singletons , was tissue-specific ( Figure 5B ) . In sum , Deeply shared CRMs are associated with genes that are highly expressed in a distinctly liver-specific manner . 10 . 7554/eLife . 02626 . 015Figure 5 . TFs in Deeply shared CRMs are near genes highly expressed in a tissue-specific manner . ( A ) Association of shared TFs in CRMs and Singletons with human gene expression obtained by RNA-seq in human liver ( Kutter et al . , 2011; E-MTAB-424 ) . TFs in CRMs or Singletons were assigned to the nearest gene , and the FPKM ( Fragments Per Kilobase of exon per Million reads ) was recorded . In contrast to Singletons , TFs in Deeply CRMs are associated with highly expressed genes ( adjusted p-values shown ) . The numbers of target gene associations for the singletons and CRMs in categories 1 to 5 are: 19354 ( S ) , 32706 ( CRM ) ; 6325 ( S ) , 14669 ( CRM ) ; 1935 ( S ) , 5755 ( CRM ) ; 1005 ( S ) , 3292 ( CRM ) ; and 459 ( S ) , 2530 ( CRM ) . ( B ) Comparison with CRMs and Singletons to a reference mRNA-seq data from 16 human tissues ( E-MTAB-513 ) further shows that relative to singletons , liver-specific CRMs are highly expressed in liver , and that each TF contributes to this specificity . The number of gene associations for each category in the liver data is shown in white text within the heat map . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 01510 . 7554/eLife . 02626 . 016Figure 5—figure supplement 1 . Association of shared TFs in CRMs and Singletons with human gene expression obtained by RNA-seq in human liver ( Kutter et al . , 2011; E-MTAB-424 ) broken down by the transcription factor in the CRM: ( A ) HNFA; ( B ) CEBPA; ( C ) FOXA1 and ( D ) ONECUT1 . TFs in CRMs or singletons were assigned to the nearest gene and the FPKM ( Fragments Per Kilobase of exon per Million reads ) was recorded for CRMs and TFs broken down by transcription factor . Adjusted p-values are shown and like Figure 5 , the CRMs were associated with more highly expressed genes in the 5-way and 4-way–shared categories . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 016 The number of reproducibly bound human regions obtained by ChIP-seq often exceeds the number of genes in the genome and so may require prioritization before experimental validation . Ranking ChIP-seq peak regions based on peak enrichment scores is one logical way to prioritize ChIP-seq peaks ( Fisher et al . , 2012 ) . We compared the pathway enrichments for all shared CRMs containing a specific TF ( e . g . , the 6278 HNF4A-containing CRMs shared between human and at least one non-primate ) vs the equivalent number of CRMs ranked by the best ChIP-seq peak enrichment score of that specific TF ( e . g . , the top 6278 HNF4A CRMs ranked by HNF4A peak score ) . As expected , the HNF4A CRMs ranked by peak intensity showed higher read counts than the shared set of CRMs and both CRM sets showed strong , centralized motif enrichments ( Figure 6A ) . However , by using overlap of CRMs in the EPO multiple sequence alignment as a filter , we found that shared CRMs give an increased number of significant enrichments using the ChIP-seq enrichment analysis tool , GREAT ( Figure 3—source data 1 ) . As observed for our collection of shared CRMs , the most significant enrichments for the shared CRMs are related to liver metabolic processes and disease ( Figure 6B ) . Similar results were obtained by performing this comparison from the perspective of the other three TFs ( Figure 6B ) . 10 . 7554/eLife . 02626 . 017Figure 6 . Shared HNF4A CRMs unravel more liver related functional classes than do the equivalent number of CRMs with the best peak enrichment scores . CRMs containing each TF were analyzed separately . ( A ) Read count and motif binding weight scores were calculated for: ( 1 ) all CRMs ( All ) ; ( 2 ) CRMs shared in human and at least one additional non-primate ( Beyond primates ) ; ( 2 ) human CRMs shared in macaque only ( Primates ) ; and ( 3 ) the equivalent number of CRMs ( equal to the number of Beyond primate CRMs ) ranked by the SWEMBL peak intensity score for the TF in question ( Top ) . ( B ) Functional enrichments were performed using GREAT comparing the Beyond primate category to the top ranked category . The top five enrichments for all comparisons performed were collected and the enrichments , if available are plotted . Databases used for GREAT enrichment analyses are indicated by color and are ranked according to the −log10 binomial FDR q-values plotted on the x-axis . Significant enrichments are labelled with an asterisk which is sized according to fold enrichment of the given database category . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 017 In order to determine whether shared CRMs associate with human diseases and phenotypes , we asked: ( 1 ) if the phenotype-associated single nucleotide polymorphisms ( SNPs ) in a curated collection of GWAS ( described here as ‘lead SNPs’; Hindorff et al . , 2009 ) overlap our CRMs/singletons in a liver-related manner; and ( 2 ) whether curated collections of these SNPs were enriched for shared CRMs ( Figure 7 , Figure 7—figure supplement 1 ) . 10 . 7554/eLife . 02626 . 018Figure 7 . Using a window of ±2 . 5 kb , lead SNPs obtained by GWAS were enriched for shared CRMs in a tissue/disease specific manner . Heatmap representation of the −log10 of Bonferroni corrected p-values from hypergeometric testing for enrichment of CRMs or single TFs ( broken down into categories related to their degree of conservation ) by lead GWAS SNPs obtained from the NHGRI catalog ( Hindorff et al . , 2009 ) . The NHGRI catalog disease traits were summarized into 25 categories prior to enrichment . Each GWAS lead SNP was given a ±2 . 5-kb window prior to identifying overlapping CRMs/singletons . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 01810 . 7554/eLife . 02626 . 019Figure 7—source data 1 . Full tables of 2 . 5 kb and LD GWAS enrichments performed in Figure 7 and Table 1 . The file also includes the categories used to annotate the NHGRI catalog . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 01910 . 7554/eLife . 02626 . 020Figure 7—figure supplement 1 . Using linkage disequilibrium ( r2 ≥ 0 . 8 ) , GWAS lead SNPs were enriched for shared CRMs in a tissue/disease specific manner . ( A ) Number of lead SNPs reported in the NHGRI catalog for each of the summarized GWAS categories . ( B ) CRM regions were classified depending on their conservation pattern . These categories were then intersected with summarized disease traits obtained from the NHGRI catalogue ( Hindorff et al . , 2009 ) . Linkage disequilibrium measures were calculated for lead SNPs from the NHGRI GWAS catalog up to a maximum of 100 kb flanking the lead SNP . Windows around each SNP were created based on the largest genomic interval between two SNPs in linkage disequilibrium ( r2 ≥ 0 . 8 ) . The heatmap shows the −log10 of Bonferroni corrected p-values for the enrichments of CRMs and singletons broken down into categories related to their degree of conservation determined by our multi-species ChIP-seq comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 02010 . 7554/eLife . 02626 . 021Figure 7—figure supplement 2 . Super-enhancer enrichments obtained by GWAS lead SNPs . ‘Super-enhancers’ were directly obtained from Hnisz et al . ( 2013 ) and overlapped with lead GWAS SNPs annotated in the NHGRI catalogue ( Hindorff et al . , 2009 ) . Each lead GWAS SNP was given a 5-kb window prior to identify overlapping ‘Super Enhancers’ . The heatmap shows hierarchical clustering of the −log10 of corrected p-values for the enrichments of super-enhancers . A Hypergeometric test with Bonferroni correction was employed to assess the significance of enrichments in all categories . The heatmap shows significant enrichment of lead SNPs by immune cell super-enhancers . Super-enhancers from the liver cancer cell line HepG2 were not enriched for liver-related GWAS lead SNPs but were enriched for blood lipid related SNPs . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 021 Using hypergeometric testing , we found that liver-related GWAS lead SNPs were enriched for nearby ( ±2 . 5 kb from a SNP ) Deeply shared CRMs . For example , we found lead GWAS SNPs related to liver ( p = 8 . 38 × 10−6; 3 . 7-fold enriched ) , blood lipid ( p = 9 . 68 × 10−5; 3 . 3-fold enriched ) , and drug response ( p = 1 . 43 × 10−4; 5 . 1-fold enriched ) categories were all enriched for shared CRMs ( Figure 7 , Figure 7—source data 1A ) . Repeating this analysis using linkage disequilibrium ( LD ) measures ( r2 ≥ 0 . 8 ) to define the boundaries of each GWAS SNP gave similar results . For example , liver-related GWAS SNPs were enriched for Deeply shared CRMs when LD was taken into consideration ( p = 3 . 20 × 10−9; 2 . 4-fold enriched; Figure 7—figure supplement 1 , Figure 7—source data 1B ) . These enrichments were not found when using two distinct null models ( Figure 7—source data 1C , D ) . We then asked if any specific disease traits , as written in the NHGRI GWAS catalog , were enriched for shared CRMs . We found that for the 2 . 5 kb window analysis , only LDL cholesterol significantly enriched for Deeply shared CRMs ( p = 0 . 037; 4 . 54-fold enriched; Figure 7—source data 1E ) , whereas the LD window analysis revealed 11 disease traits that were enriched for Deeply shared CRMs ( see Table 1; Figure 7—source data 1F ) . For example , enrichments driven by lead SNPs for LDL cholesterol ( blood lipid category , p = 4 . 84 × 10−5; 3 . 7-fold enriched ) involved several loci including TRIB1 , ABCG8 , APOB , SORT1 , TOMM40 , APOA5 and HNF1A . Lead SNPs for fibrinogen ( liver category , p = 3 . 40 × 10−3; 3 . 4-fold enriched ) occurred in LD with the fibrinogen locus . C-reactive protein ( liver category , p = 1 . 17 × 10−4; 4 . 0-fold enriched ) enriched for Deeply shared CRMs near CRP itself in addition to RORA , MLXIPL , HNF1A , BAZ1B , and IRF1 loci ( Table 1; Figure 7—source data 1G ) . 10 . 7554/eLife . 02626 . 022Table 1 . Table of GWAS Disease Traits that significantly enriched for Deeply shared CRMsDOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 022Disease trait ( NHGRI ) Category ( this study ) Number of Deeply shared CRMsDeeply shared CRM enrichment ( adjusted p-value ) Fold enrichmentClosest genesOther metabolic traitsother measurement114 . 42E-056 . 42CRP , HNF1A , PANK1LDL cholesterolblood lipid204 . 84E-053 . 71TRIB1 , ABCG8 PSRC1 , DOCK7 , APOB , HNF1A , LDLR , TOMM40 , HNF1A , APOA5C-reactive proteinliver171 . 17E-043 . 97MLXIPL , RORA , CRP , HNF1A , TOMM40 , BAZ1B , IRF1D-dimer levelsliver116 . 87E-044 . 99NME7 , FGG , EDEM2 , FGAFibrinogenliver153 . 40E-033 . 40FGB , FGA , FGGLung cancercancer144 . 39E-033 . 46C2 , CRP , HSPA1A , TP63Mean corpuscular hemoglobinblood88 . 57E-035 . 02GCDH , USP49 , RCL1 , SLC17A1 , TFRC , MPSTProtein quantitative trait lociother measurement142 . 49E-022 . 93CRP , IFT81 , BCO2Serum markers of iron statusother measurement132 . 67E-023 . 03TCP1 , MRPL18 , TF , SLC17A1 , HIST1H4C , MPST , GHRTriglyceridesblood lipid122 . 84E-023 . 16TRIB1 , MLXIPL , DOCK7 , BAZ1B , GALNT2 , TRIB1 , APOA5Select biomarker traitsother measurement83 . 89E-024 . 08CRP , OR10J5Lead GWAS SNPs and their associated Disease Traits were obtained directly from the NHGRI catalog ( Hindorff et al . , 2009 ) . An LD window ( r2 ≥ 0 . 8 ) around each SNP was obtained and regions with identical Disease Traits were collapsed into a single interval . These Disease-Trait–associated intervals were then intersected with all CRMs and Singleton categories as in Figure 7 . This table shows Disease Traits that were significantly enriched for Deeply shared CRMs . The summarized disease category used in Figure 7 , the number of Deeply shared associated CRMs , the Bonferroni corrected p-values from the hypergeometric test , fold enrichment of Deeply shared CRMs , and the nearest gene to the Deeply shared CRM , if it is protein coding , are shown . Figure 7—source data 1G contains detailed enrichment information including SNP ID and primary GWAS publication ( PMID ) . In order to explore the functional relevance of these findings , we looked for annotated regulatory SNPs in the RegulomeDB database ( Boyle et al . , 2012 ) within the 1020 CRM or single TF regions we found to be within 2 . 5 kb of a lead GWAS SNP . Of these 1020 regions , 753 contained at least one variant , 90% of which showed evidence of regulatory potential in RegulomeDB ( Supplementary file 2 ) . In particular , 317 of these 753 regions had TF binding in orthologous regions in additional species , making them rational candidates for future functional exploration . We also asked whether the collection of recently identified ‘super-enhancers’ ( Hnisz et al . , 2013 ) , which were enriched for disease loci in a tissue-specific manner , would also enrich for liver-related lead GWAS SNPs . Our analysis supported the association between super-enhancers and immune system related GWAS SNPs reported by Hnisz et al . ( with the highest enrichment for immune cell GWAS lead SNPs found in ‘super-enhancers’ in cell line CD20 , p = 3 . 24 × 10−6 ) . However , unlike what we found for shared CRMs in liver , super-enhancers obtained from the liver cancer cell line HepG2 were not enriched by the liver-related lead GWAS SNPs ( Figure 7—figure supplement 2 ) . This observation may be related to biological differences between primary liver tissue and HepG2 . To further investigate the functional role of shared CRMs and to test the hypothesis that disruption of conserved combinatorial binding can lead to human disease , we overlapped our CRMs and singletons data with manually curated regulatory mutations directly linked to human disease ( Human Genome Mutation Database Professional version; HGMD ) ( Stenson et al . , 2012 ) . This database contains the most comprehensive set of curated and functionally validated mutations that have occurred in human regulatory DNA regions and have led to a change in disease gene expression . A total of 157 genes associated with regulatory mutations overlapped our human ChIP-seq data , 106 of which were in our CRMs ( Table 2; Supplementary file 3 ) . The Deeply shared CRMs overlapped a set of 47 genes associated with regulatory mutations . These 47 genes were clearly enriched for two liver-related biological pathways: coagulation and complement factors ( p = 9 . 34 × 10−6; 24-fold enrichment ) and lipid homeostasis ( p = 6 . 93 × 10−5 , 36-fold enrichment ) ( Figure 8—source data 1A , B ) . We found multiple disease-causing regulatory mutations overlapping our CRMs at promoters of seven critical genes in the coagulation pathway ( FGA , FGB , F7 , F9 , F10 , F11 , F12; Figure 8 , Figure 8—source data 1 ) . While many of these mutations have been individually known for decades , this is the first time they have been put in context of a regulatory network consisting of these liver-enriched TFs . Furthermore , repeated observation of rare mutations in the promoter regions of several of the blood coagulation proteins correspond to critical positions in predicted DNA binding motifs for the liver-enriched TFs . For example , Mutations leading to Factor VII deficiency and severe bleeding disorders occur in a promoter region harboring a CRM shared in all five species ( Figure 8—figure supplement 1A ) . Several mutations ( including SNP variant rs561241 ) within a conserved HNF4A motif have been shown to perturb F7 transcription leading to hemophilia ( Zheng et al . , 2011 ) . A critical highly conserved CRM localizes to the F9 promoter ( Figure 8—figure supplement 1B ) . Multiple mutations within this region are associated with defective expression of F9 and clinical hemophilia ( Giannelli et al . , 1998 ) . Several of these mutations have previously been shown to disrupt the binding of CEBPA and HNF4A ( Crossley and Brownlee , 1990; Crossley et al . , 1992; Reijnen et al . , 1993; Giannelli et al . , 1998 ) . We recently demonstrated that ONECUT1 binds to the −6 site of F9 in human and mouse ( Funnell et al . , 2013 ) . Inspection of the multiple species alignment suggests the same arrangement and spacing of CEBPA , HNF4A , and ONECUT1 motifs in the human and macaque CRMs , whereas the mouse and rat ONECUT1 motifs are predicted to begin three base pairs upstream . 10 . 7554/eLife . 02626 . 023Table 2 . HGMD disease variants falling within motifs of shared CRMs and singletonsDOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 023CoordinatesGene nameHGMD ( regulatory , disease mutations ) Disease mutationsShared CRM3:30622781-30623121TGFBR2Marfan syndrome II13:37009728-37010105MLH1Colorectal cancer , non-polyposis83:95175129-95175686PROS1Protein S deficiency23:172227012-172227583SLC2A2Diabetes14:187423808-187424193F11Factor XI deficiency15:176769019-176769427F12Factor XII deficiency27:75769704-75769860HSPB1Amyotrophic lateral sclerosis19:35647685-35648108RMRPCartilage-Hair hypoplasia*609:103237817-103238177ALDOBFructose intolerance111:57121333-57121845SERPING1Angioneurotic oedema311:116213420-116213833APOA1Apolipoprotein A1 deficiency; Atherosclerosis with coronary artery disease212:119900361-119900931HNF1ADiabetes913:112807935-112808278F7Factor VII deficiency1517:39777939-39778164GRNAmyotrophic lateral sclerosis; Frontotemporal dementia219:11060834-11061300LDLRHypercholesterolaemia2319:40465042-40465277HAMPHaemochromatosis219:50140788-50141367APOC2Apolipoprotein C2 deficiency120:42417527-42417906HNF4ADiabetes13X:138440311-138440689F9Haemophilia B22Shared singleton1:55277551-55277787PCSK9Hypercholesterolaemia , autosomal dominant12:47483486-47483635MSH2Colorectal cancer , non-polyposis15:147191404-147191657SPINK1Pancreatitis511:107598900-107599060ATMAtaxia telangiectasia117:3486175-3486485CTNSCystinosis217:27840752-27841062CDK5R1Mental retardation119:54160105-54160315FTLCataract , bilateral1X:66680386-66680665ARProstate cancer1Human only CRM1:113300254-113300535SLC16A1Exercise-induced hyperinsulinism11:224075148-224075491EPHX1Hypercholanaemia13:170965492-170965762TERCAplastic anaemia; Dyskeratosis congenita; Myelodysplastic syndrome38:64161127-64161334TTPAAtaxia , isolated vitamin E deficiency110:27429312-27429620ANKRD26Thrombocytopaenia1213:59636050-59636246DIAPH3Auditory neuropathy1X:146800975-146801150FMR1Fragile X mental retardation syndrome3X:153643911-153644318DKC1Dyskeratosis congenita , X-linked1HGMD disease variants falling within shared CRMs , shared singletons and human only CRMs . The number of unique regulatory mutations designated as ‘disease-mutations’ in the HGMD database recorded within each CRM or singleton is shown . *RMRP is a non-coding RNA that is found in HGMD associated to several related diseases . 10 . 7554/eLife . 02626 . 024Figure 8 . Shared CRMs link TF binding with disease-causing variants in coagulation and lipid regulation in the liver . Human CRMs that had TF binding in syntenic regions in at least two additional species ( n = 5046 ) were intersected with the HGMD database . All protein coding genes associated with a regulatory mutation were analysed . Relationships among these genes were investigated and a representative analysis obtained using GeneMANIA ( ‘Materials and methods’ ) . Genes ( large grey circles ) are connected by pathways and protein–protein interactions are shown . The smaller white circles are genes predicted by GeneMANIA to be in the network . The 47 unique genes were associated into 35 clusters using DAVID ( Huang da et al . , 2009 ) . Eight Gene Ontology terms from the 35 clusters had an adjusted p-value of less than 0 . 005 ( Figure 8—source data 1A; Supplementary file 3 ) . 4 of the 8 significant GO categories containing the most genes are illustrated: response to wounding ( open red circle , p = 3 . 16 × 10−9; 9 . 9-fold enriched ) ; blood coagulation ( red dot , p = 9 . 34 × 10−6; 22 . 0-fold enriched ) ; response to organic substance ( open yellow circle , p = 1 . 05 × 10−5; 6 . 4-fold enriched ) ; and lipid homeostasis ( yellow dot , p = 6 . 93 × 10−5; 36 . 2-fold enriched ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 02410 . 7554/eLife . 02626 . 025Figure 8—source data 1 . Table of DAVID enrichments used to annotate Figure 8 and the HGMD genes that overlapped our CRMs and singletons in the different phylogenetic categories . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 02510 . 7554/eLife . 02626 . 026Figure 8—figure supplement 1 . CRMs link TF binding with disease-causing variants in blood coagulation . Raw human ChIP-seq signals ( bed graph format ) are shown at the proximal promoters of three blood coagulation genes: ( A ) F7; and ( B ) F9 . TF reads and significantly bound peak regions are shown as colored rectangles according to the TF ( CEBPA = red; HNF4A = blue; FOXA1 = green; and ONECUT1 = yellow ) . All unique regulatory mutations in the HGMD overlapping human CRMs are shown with black arrows . Asterisks are shown above the position weight matrices for each TF at the position homologous to a unique human regulatory mutation ( i . e . , an A → T is only counted once ) . For each species , in silico predicted motif positions for each TF are shown above the multiple sequence alignment at each promoter when they overlap regions that have evidence of ChIP signal . DOI: http://dx . doi . org/10 . 7554/eLife . 02626 . 026 In this analysis of known functional and disease causing regulatory mutations from the HGMD database , it is worth noting that most of the examples we found overlapping Deeply shared liver CRMs resided close to the TSS . As most TF binding of the factors we profiled occur outside of proximal promoters , it is likely that many more human mutations that regulate genes through long-range interactions remain to be found . Just as gene sequencing has uncovered a diverse array of mutations for most disease genes , we expect that disruption of conserved TF bound regions will be found to have pathological consequences .
There is a great interest in identifying and mechanistically characterizing regulatory SNPs . Such approaches require limiting the genomic search space , finding the function of identified mutations , associating the mutation with target gene ( s ) and addressing the tissue-specificity inherent in transcriptional regulation . To limit the search space , the genome must be distilled to the active functional regions where mutations are likely to be pathological . To determine the function of mutations in gene control regions , the trans-acting molecules ( e . g . , transcription factors ) whose regulation of a particular locus is disrupted need to be identified . Finally , because transcriptional regulatory networks are highly tissue-specific , regulatory regions can only be accurately done within the tissue concerned . In this study , we demonstrated that genomic regions harboring shared combinatorial TF binding were enriched in a tissue-specific manner for essential biological pathways , common DNA variants associated with complex traits , and regulatory DNA mutations associated with rare diseases . In addition to the requisite measures of ChIP-seq data quality and reproducibility , it could be argued that combinatorial binding of transcription factors , relevant to the cell type of interest , increases the likelihood that the region identified is biologically important . However , more than half of the human liver TF binding sites detected in our study are in CRMs . In many cases , this makes CRM clustering on its own an insufficient filter for prioritizing TF binding sites for further study . Given the rapid turnover of TF binding within a mammalian lineage ( Stefflova et al . , 2013 ) , requiring combinatorial binding to be shared in multiple mammals is a strict filter that provides evidence that these biochemical interactions are under selection , even if the percent identity in the sequence alignments themselves is not exceptional . Supporting this idea , we found the percentage of CRMs shared between human and macaque ( ∼35% ) and mouse and rat ( ∼34% ) is considerably lower than the average human-rhesus sequence identity of ∼90% ( Rhesus Macaque Genome Sequencing and Analysis Consortium et al . , 2007 ) or the mouse-rat identity of ∼93% ( Gibbs et al . , 2004 ) . Although insightful , the costs and challenges of performing comparative combinatorial TF binding in multiple species , tissues , developmental stages , and environmental conditions limits its widespread use as a method for finding enhancers . Computational strategies that forgo strict DNA constraint for more flexible criteria , such as shared clusters of motifs , show great promise ( Gordân et al . , 2010 ) . For example , a recent computational strategy that relied upon conservation of clusters of motifs , rather than conserved DNA sequence , was able to fine map regulatory SNPs in select GWAS loci ( Claussnitzer et al . , 2014 ) . The results and the strong functional enrichments , we observe with shared combinatorial binding further support such approaches . It is also becoming clear that integrative approaches to enhancer discovery can outperform predictions made using single criteria ( Erwin et al . , 2014 ) . We compared our CRM and singleton data to enhancer predictions made by a recent integrative approach ( EnhancerFinder; Erwin et al . , 2014 ) . EnhancerFinder is trained on experimentally verified human VISTA enhancers and utilizes evolutionary conservation , DNA motifs , and functional genomics data ( such as p300 and histone modifications ) as classifiers . We compared the 84 , 301 human developmental enhancers predicted by EnhancerFinder to our data . While all phylogenetic categories overlapped significantly ( Heger et al . , 2013 ) , the Deeply shared CRMs gave the highest overlap ( ∼30% ) with the EnhancerFinder predictions . The evidence we present here for the functional relevance of Deeply shared combinatorial TF–DNA interactions in primary liver tissue suggests that results from such empirical studies in primary tissues will be another valuable source of information which can be utilized in integrative methods for enhancer prediction . Just as using DNA constraint as a sole criteria for enhancer finding has its limitations , using filtering approaches that require conserved combinatorial binding will also miss important regulatory events near genes of interest . For example , the common SNP rs2279744 ( also known as ‘SNP 309’ ) is found in the first intron of the MDM2 and contributes to carcinogenesis in humans by increasing levels of MDM2 , a negative regulator of TP53 ( Bond et al . , 2004 ) . This SNP has been shown to enhance the binding of the SP1 transcription factor at a site that does not have a clear orthologous mouse motif . For this reason , transgenic mice with human intron 1 alleles have been created and show a cancer phenotype ( Post et al . , 2010 ) . This example illustrates where strictly using sequence conservation alone would fail to identify a functional regulatory variant . Interestingly , in our data we found a Deeply shared human HNF4A singleton TF binding event spanning rs2279744 . Although the role of HNF4A at the MDM2 intron 1 has yet to be characterized , rs2279744 also serves as an example where excluding a presumptive singleton TF binding events from functional analyses would have missed an important regulatory variant . Our results suggest that conserved liver regulatory regions reside near SNPs or rare mutations associated with liver related phenotypes . The most striking disease associations we observed involved the blood coagulation pathway . This pathway is perhaps one of the best-studied biological pathways and has long served as a model system for understanding human disease gene mutations . This rich history allowed us to observe , perhaps for the first time , how recurrent TF binding mutations found within conserved combinatorially occupied TF binding sites can be afflicted on several members of the same pathway . Conserved combinatorial control regions were also found near proteins that are part of the coagulation and complement system ( e . g . , C1R , C1S , C2 , C4B , C4B , C4BPA , C4BPB , C5 , C6 , C7 , C8A , C8B , C8G , C9 , CFB , CFH , and CPB2 ) . The interplay between the coagulation and complement system proteins has long been appreciated ( reviewed by Markiewski et al . , 2007 ) , and our results suggest that their gene expression in the liver is coordinated by conserved cis-regulatory modules . Overall , the observation of recurrent phenotype-causing regulatory mutations in a single pathway is likely a phenomenon that occurs in other tissues and biological pathways . Our study suggests that identifying sites of shared combinatorial binding will be relevant criteria for assigning pathological significance to candidate disease variants uncovered in whole genome sequencing studies . | Stretches of DNA called cis-regulatory modules ( or CRMs for short ) could help researchers to identify the regions of DNA that are most important for controlling genes . CRMs are regions where multiple transcription factors—proteins that control when and how genes are expressed—bind to DNA . As important biological pathways are often regulated by more than one transcription factor , CRMs are therefore a good target when looking for DNA regions that , if mutated , are likely to cause disease . If a stretch of DNA performs an important role , it is often conserved throughout evolution . This is often observed for genes that make proteins . Indeed , DNA regions that specify critical amino acids that make up proteins are often conserved across distantly related species . However , unlike the changes made to the amino acid encoding parts of genes , it is currently a challenge to predict which changes in the rest of the genome will affect gene expression . One reason for this challenge is that transcription factor binding sites are rapidly evolving . This rapid evolution means that strictly comparing DNA sequences between species may fail to identify where transcription factors like to bind in the genome . Numerous experimental efforts have therefore been made to map these sites . These have revealed that there are a huge number of regions in the human genome that can bind transcription factors: hundreds of thousands of sites , far more than there are genes . For this reason , there is a great interest in revealing which of these regulatory regions are critical for maintaining normal levels and timings of gene expression . Ballester et al . compared the binding sites of four transcription factors responsible for regulating liver function in humans , macaques , mice , rats , and dogs . About two-thirds of these binding sites were found in CRMs . Less than half of the CRMs in humans were also CRMs in another species—but Ballester et al . found that these shared CRMs are predominantly in charge of regulating the essential biological pathways that allow the liver to function correctly . In addition , Ballester et al . identified several examples of disease-causing DNA mutations in shared CRMs that affected the expression of genes that make up pathways such as the blood clotting cascade . Genome-wide association studies also uncovered common variants for liver-related traits that were enriched for the CRMs found in more than one species , further supporting their importance . As transcription factors work in different ways in different tissues , further studies are now required to expand these observations to organs other than the liver . Future work is also needed to investigate the function of thousands of conserved CRMs whose role in liver gene regulation remains unknown . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"genetics",
"and",
"genomics"
] | 2014 | Multi-species, multi-transcription factor binding highlights conserved control of tissue-specific biological pathways |
Genomic sequencing has implicated large numbers of genes and de novo mutations as potential disease risk factors . A high throughput in vivo model system is needed to validate gene associations with pathology . We developed a Drosophila-based functional system to screen candidate disease genes identified from Congenital Heart Disease ( CHD ) patients . 134 genes were tested in the Drosophila heart using RNAi-based gene silencing . Quantitative analyses of multiple cardiac phenotypes demonstrated essential structural , functional , and developmental roles for more than 70 genes , including a subgroup encoding histone H3K4 modifying proteins . We also demonstrated the use of Drosophila to evaluate cardiac phenotypes resulting from specific , patient-derived alleles of candidate disease genes . We describe the first high throughput in vivo validation system to screen candidate disease genes identified from patients . This approach has the potential to facilitate development of precision medicine approaches for CHD and other diseases associated with genetic factors .
A major problem confronting large-scale genomic sequencing studies of disease patients is to identify those genetic variants that actually represent disease-causing mutations . This is particularly true for candidate genes that have not previously been evaluated in an experimental model system for involvement in disease-relevant biological processes . Functional gene validation systems in model organisms are therefore essential . While a definitive assignment of disease association should be established through the analysis of mutant gene phenotypes in a mammalian model , this approach is not practical for rapid high-throughput initial screening of large numbers of candidate genes . Drosophila , by contrast , possesses the attributes of a simple animal model system that make it ideally suited for this role ( Bier , 2005; Chien et al . , 2002 ) . Use of Drosophila as a model to elucidate molecular disease mechanisms is well established and amply documented ( Cagan , 2011; Na et al . , 2015; Zhang et al . , 2013; Diop and Bodmer , 2015; Bier and Bodmer , 2004; Owusu-Ansah and Perrimon , 2014 ) . It has been established that 75% of human disease associated genes are represented in the fly genome by functional homologs ( Reiter et al . , 2001 ) . Although it is difficult to directly link Drosophila developmental defects to patient symptoms , the Drosophila system can serve as a testing platform for gene function in development , which can be used to test disease association for a large number of candidate genes identified from patient genomic sequencing . Congenital heart disease ( CHD ) affects 0 . 8% of live births ( Hoffman , 1990 , 2002; Reller et al . , 2008 ) . Genetic factors are strongly implicated in CHD pathogenesis , but the great majority of genes ( accounting for an estimated 75% of cases ) remain unidentified ( Gelb et al . , 2013 ) . Drosophila has served as a model to study genes related to CHD for over 20 years , based on the evolutionarily conserved genetic basis of heart development ( Bier and Bodmer , 2004; Olson , 2006; Yi et al . , 2006 ) . The Drosophila heart ( Figure 1A ) is a rhythmically beating linear tube composed of parallel rows of fused contractile cardiac cells , flanked by adherent pericardial cells ( Vogler and Bodmer , 2015 ) . The heart functions to maintain circulation of hemolymph ( insect ‘blood’ ) throughout the body cavity . In a typical beating cycle , hemolymph enters the posteriorly-located heart chamber in the abdomen and is pumped anteriorly through the aorta towards the head and brain . Although structurally of relatively low complexity , Drosophila and human heart development are both largely directed by the same highly conserved regulatory networks ( Vogler and Bodmer , 2015 ) . To demonstrate that the fly is the ideal in vivo animal model for cost-effective , rapidly informative , and efficient screening of candidate disease genes , we have developed a Drosophila validation system to screen candidate genes identified in a large-scale genomic sequencing study of congenital heart disease patients . 10 . 7554/eLife . 22617 . 003Figure 1 . Drosophila adult heart structure and tissue visualization , evaluation of 1X vs . 4X Hand enhancer constructs driving heart specific silencing of gene expression . ( A ) Drosophila heart structure and visualization . Left: the adult heart is depicted schematically in green . Supporting lateral alary muscles are depicted in red . Middle: fluorescence microscopy of dissected heart tissue spanning abdominal segments A2–A5 . Myocardial actin filaments ( cardiac myofibers ) were visualized by Phalloidin staining ( red ) , which also stained somatic muscle fibers in segments A2 , A3 , and A4 and alary muscle fibers in A5 . GFP ( green ) labels cardiomyocytes and pericardial nephrocytes . Expression of a nuclear-localized GFP transgene was controlled by the cardioblast specific Hand gene enhancer element . Scale bar = 100 µ . Right: higher magnification of segment A4 heart tissue . Scale bar = 50 µ . In this example , dissection and preparation for microscopy involved removal of a layer of longitudinal muscle , which resulted in the loss of some cardiomyocytes and pericardial nephrocytes . In situ , the heart tube is composed of parallel , symmetrical rows of cardiomyocytes ( small nuclei ) , flanked by pericardial nephrocytes ( large nuclei ) . ( B ) The 4XHand enhancer promotes significantly greater Gal4 mRNA production than the single Hand enhancer . The Gal4 mRNA level in the dissected adult heart was determined by qRT-PCR . Statistical significance ( * ) was defined as p<0 . 05 . w1118 was used as a negative control since it does not express any Gal4 . ( C ) Compared to the single Hand enhancer , 4XHand induced significantly greater knockdown of CG8184 ( the homolog of human HUWE1 ) mRNA when driving expression of the CG8184-IR RNAi silencing transgene . The CG8184 mRNA level in the dissected adult heart was determined by qRT-PCR . Statistical significance ( * ) was defined as p<0 . 05 . ( D ) Compared to the single Hand enhancer , 4XHand induced significantly greater knockdown of Lid ( the homolog of human KDM5A and KDM5B ) mRNA when driving expression of the Lid-IR RNAi silencing transgene . The Lid mRNA level in the dissected adult heart was determined by qRT-PCR . Statistical significance ( * ) was defined as p<0 . 05 . In C and D , Control flies were the progeny of a cross between homozygous 4XHand-Gal4 and w1118 , which has one copy of 4xHand-Gal4 but does not carry a UAS-RNAi silencing construct . DOI: http://dx . doi . org/10 . 7554/eLife . 22617 . 00310 . 7554/eLife . 22617 . 004Figure 1—figure supplement 1 . 4XHand-Gal4 efficiently silences cardiac gene expression . ( A ) Mortality Index , expressed as the percentage of flies of the indicated genotype that die during pre-adult stages . Male and female flies of the appropriate genotypes were crossed to produce progeny flies carrying the indicated UAS-gene silencing ( i . e . RNAi ) interfering RNA ( IR ) construct , the expression of which is directed specifically in cardioblast cells by the indicated Hand-Gal4 driver . Hand-Gal4 and 4X-Hand-Gal4 constructs incorporate either one copy or four tandem copies of a Hand gene enhancer element , respectively . Significant developmental lethal effects of heart-specific gene silencing were only observed in progeny flies carrying the 4XHand-Gal4 driver . DOI: http://dx . doi . org/10 . 7554/eLife . 22617 . 004 In 2013 the Pediatric Cardiac Genomics Consortium ( PCGC ) published a large-scale sequence analysis of sporadically occurring CHD cases ( Zaidi et al . , 2013 ) , providing an opportunity to conduct systematic model system based experimental studies into the genetic basis of CHD , informed from the outset by clinical data . Such studies will ultimately facilitate the development of diagnostic , therapeutic , and preventive approaches , and precision medicine interventions ( Ashley , 2015 ) for CHD . The initial report ( Zaidi et al . , 2013 ) of 249 protein-altering de novo mutations from 362 severe cases of CHD identified 223 candidate disease genes involved in diverse biochemical pathways , with 26 genes selected as particular genes-of-interest based upon bioinformatics criteria ( Zaidi et al . , 2013 ) . We used Drosophila developmental genetics to quantitatively evaluate the phenotypes associated with heart-specific RNAi mediated gene silencing of fly homologs of mutated candidate disease genes . 174 genes ( 78% ) had clear fly homologs , consistent with the published estimate of 75% conservation of disease gene homologs ( Reiter et al . , 2001 ) . Interestingly , the ‘top 26’ genes identified by the PCGC study ( Zaidi et al . , 2013 ) all have conserved Drosophila homologs . Some of these genes were previously identified as having roles in Drosophila heart development , in the context of a global in vivo RNAi screen for candidate fly heart genes . ( Neely et al . , 2010 ) The current study used a stronger heart-specific driver to validate candidate CHD genes . We developed a highly efficient cardiac-targeted gene silencing approach in flies , and used this to examine effects on heart structure and function for fly homologs of 134 candidate disease genes published by the PCGC . We employed a quantitative phenotypic screening protocol that evaluated developmental lethality ( pre-adult mortality ) , heart structure ( Figure 1A; including morphology , cardiac myofibrillar density , cardiac collagen deposition , and cardioblast cell number ) and adult longevity . We found that 52% of the tested fly homologs are required for cardiac development and function in flies . We also developed a gene replacement testing strategy involving simultaneous heart-specific silencing of an endogenous fly gene homolog and expression of either wild type or patient-derived mutant alleles of the candidate human disease gene .
To achieve a highly efficient heart-specific gene knock down , we generated a strong cardiac cell Gal4 driver featuring 4 tandem repeats of the Drosophila Hand gene cardiac enhancer ( Han and Olson , 2005 ) . We compared the levels of Gal4 RNA in heart tissues of transgenic flies carrying the new 4XHand-Gal4 driver construct versus flies carrying the original , established Hand-Gal4 driver ( featuring a single Hand enhancer element ) . As anticipated , 4XHand promoted significantly higher heart cell expression of Gal4 mRNA ( Figure 1B ) . We then compared the Hand-Gal4 and 4XHand-Gal4 drivers together with UAS-Gene-IR RNAi based silencing constructs ( Ni et al . , 2008 , 2011 ) for gene knock down efficiency when used to silence expression ( i . e . lower heart cell RNA levels ) of Drosophila genes CG8184 ( Figure 1C ) and Lid ( Figure 1D ) . Consistent with the higher Gal4 expression , 4XHand was significantly more effective at silencing target gene expression through RNAi . To further confirm the utility of 4XHand-Gal4 for screening purposes , we comparatively evaluated the Hand-Gal4 and 4XHand-Gal4 constructs driving heart-specific expression of UAS-RNAi transgenes targeting eight Drosophila genes for developmental lethal ( death at pre-adult stage ) effects . The Hand enhancer is active during development from embryonic stages , and we reasoned that Hand enhancer-driven silencing of genes known to be essential for heart development would cause developmental lethality . For this assay , male and female flies of the appropriate genotypes were crossed to produce progeny carrying the indicated UAS-gene silencing ( i . e . RNAi ) interfering RNA ( IR ) construct , the expression of which was driven by Hand-Gal4 or 4XHand-Gal4 . Progeny embryos were collected and allowed to develop under standard conditions . The results were quantitatively expressed as a Mortality Index ( MI ) , the percentage of flies that died before adult emergence from the pupa stage . We found that for all eight tested genes , significant developmental lethal effects of heart-specific gene silencing were only observed in flies carrying a 4XHand-Gal4 driver ( Figure 1—figure supplement 1 ) . We combined the new 4XHand-Gal4 driver with available fly lines carrying UAS-Gene-IR RNAi based silencing constructs to evaluate for heart specific essential function ( i . e . MI ) the homologs of all genes from the PCGC study for which multiple independent UAS-RNAi lines are available ( 134 total genes , Supplementary file 1 ) . We extracted the data for the PCGC top 26 genes , identified as most-promising candidate disease genes on the basis of multiple bioinformatics-based criteria ( Figure 2A ) . We reasoned that if these criteria alone were truly informative , the majority of these genes would yield higher MI scores , overall , than observed among all 134 tested genes representing the PCGC total list . As shown in Figure 2 we observed MI values ranging across a spectrum spanning Normal ( ≤6% , equivalent to nonbiased control genes ) , Low ( 7–30% ) , Medium ( 31–60% ) , to High ( 61–100% ) . Importantly , we found that the distribution of MI values across all ranges was not significantly different for the top 26 genes compared to the entire cohort of 134 candidate genes tested ( Figure 2A ) . Thus , selection of candidate CHD genes on the basis of bioinformatics criteria alone did not enrich for genes inducing more severe developmental mortality phenotypes ( indicative of more genes with essential roles in either heart development , tissue maintenance , or function ) . This observation emphasizes the importance of screening and selecting most-promising candidate genes on the basis of functional assays . To determine if the set of 134 candidate genes in itself can be demonstrated to be enriched for heart-essential genes , we examined previously obtained data from a screen of approximately 800 Drosophila genes silenced using the 4XHand-Gal4; UAS-RNAi system . These genes were randomly chosen to screen for effects on pericardial nephrocyte function ( cardioblast lineage cells in which the Hand enhancer is active ) , so they are unbiased with respect to essential heart roles . We found that only 6% of these genes showed any level of developmental lethality when silenced ( i . e . 94% normal ) . Thus , the PCGC genomic association study , though not gene selection based upon bioinformatics criteria , did enrich significantly for heart-essential genes as assayed in Drosophila . That such a small number of unbiased genes conferred a phenotype of silencing-induced developmental lethality also suggests that overexpression of 4xHand-Gal4 itself did not sensitize flies to RNAi mediated knockdown effects . 10 . 7554/eLife . 22617 . 005Figure 2 . Top 26 candidate CHD genes and developmental lethality induced by heart specific RNAi-based silencing of Drosophila gene homologs . ( A ) 26 de novo mutated genes from CHD study participants selected as being of particular interest ( Zaidi et al . , 2013 ) based upon sequence quality , mutation type , the expression level of the mouse homolog during heart development on embryonic day 14 . 5 , and previously reported involvement in CHD or heart development . The 26 corresponding Drosophila homologs are shown with protein function ( Flybase ) , and Mortality Rate ( Mortality Index ) . ( B–D ) The proportions of Drosophila gene homologs that , when silenced by cardiac cell specific RNAi expression , induce developmental lethality at normal/background levels ( blue ) ; low levels ( green ) ; medium levels ( orange ) ; high levels ( red ) based on Mortality Index values . The Mortality Index is determined by crossing homozygous UAS-RNAi transgenic flies with flies bearing a 4XHand-Gal4 ‘driver’ ( four repeats of the cardioblast cell-specific Hand enhancer element 5’ of Gal4 ) balanced over CyO . Progeny flies that emerge as adults with curly wings ( CyO , no transgene expression ) vs . straight wings ( expressing 4XHand-Gal4 driven UAS-RNAi transgene in cardioblasts ) are recorded and the developmental mortality attributable to RNAi heart expression ( Mortality Index ) is calculated as ( Curly – Straight ) / Curly X 100 . Divergence from 1:1 ratio ( Normal , blue ) ≥ 7% was considered a lethal phenotype . A Normal range of divergence from a 1:1 ratio of <6% based on analysis of 400 progeny from control crosses . Varying degrees of phenotype severity were observed ( Low = 7–30% , green; Medium = 31–60% , orange; High = 61–100% , red ) . ( B ) Left: chart ( PCGC top 26 ) shows proportions of RNAi silencing effects on lethality for Drosophila homologs of 26 genes identified as being of particular interest based exclusively on bioinformatics-based criteria . Right: chart ( PCGC total list ) shows the results of 134 fly homologs of all de novo mutated genes ( with available RNAi silencing lines ) identified in pediatric CHD study participants ( Zaidi et al . , 2013 ) ( Supplementary file 1 ) . ( C ) Comparison of silencing-induced lethality for Drosophila homologs of 134 candidate CHD-associated genes as a function of high ( HHE ) versus low ( LHE ) levels of expression of murine homologs in embryonic mouse heart . ( D ) Comparison of silencing-induced lethality for Drosophila homologs of 134 candidate CHD-associated genes as a function of fly-to-human gene conservation ( High Conservation , score 6 to 10; Low Conservation , score 2 to 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22617 . 005 We also stratified and compared MI values we obtained for all of the 134 Drosophila genes as functions of expression level of murine homologs during mouse embryonic heart development , as reported in the PCGC study ( and one basis for selection of top 26 genes ) ( Zaidi et al . , 2013 ) . As shown in Figure 2C , high heart expression ( HHE ) during development did not correlate with greater severity of developmental lethality associated with gene silencing in Drosophila heart . Contrary to expectation , low heart expression ( LHE ) of murine homologs was associated with relatively higher severity of Drosophila gene silencing . These observations again emphasize the importance of applying functional screening assays to the results of genomic sequencing studies . We noted nevertheless that the top 26 genes all have Drosophila homologs , including 22 that are highly conserved . When mortality was assayed as a function of degree of fly-human conservation among all 134 tested genes ( Figure 2D ) we did observe greater mortality among highly conserved genes compared to less conserved genes . Not surprisingly , phylogenetic conservation indeed correlated positively with MI , a result that further supports using Drosophila as a model for functional evaluation of candidate genes for disease association . In order to confirm that RNAi based phenotypes were indeed associated with target gene mRNA reduction , we measured RNA levels in heart tissue from flies expressing silencing constructs targeting six different genes from the top 26 group . These genes were chosen for analysis because they represent a range of different cellular functions , and include three genes for which silencing did not induce developmental lethality . MI values for the six genes ranged from normal to 43% . As shown in Figure 3 , heart RNA levels of target genes were all significantly reduced . Furthermore , we tested developmental lethal phenotypes for all top 26 genes using two or more independent RNAi silencing fly lines for each gene , and obtained consistent MI results in all cases ( data not shown ) . 10 . 7554/eLife . 22617 . 006Figure 3 . Target gene mRNA levels in adult fly heart tissues . ( A ) Drosophila Cul3 homolog of human CUL3 encoding ubiquitin ligase ( Mortality Index 43% ) . ( B ) Drosophila Scny homolog of human USP44 encoding ubiquitin protease ( Mortality Index 25% ) . ( C ) Drosophila Bre1homolog of human RNF20 encoding ubiquitin ligase E3 ( Mortality Index 13% ) . ( D ) Drosophila Rab10 homolog of human RAB10 encoding GTPase ( Mortality Index normal ) . ( E ) Drosophila Med20 homolog of human MED20 ( Mortality Index normal ) . ( F ) Drosophila CG15445 homolog of human NUB1 ( Mortality Index normal ) . mRNA levels in dissected adult hearts were determined by qRT-PCR . Statistical significance ( * ) was defined as p<0 . 05 . Control flies were the progeny of a cross between homozygous 4XHand-Gal4 and w1118 , which has one copy of 4xHand-Gal4 but does not carry a UAS-RNAi silencing construct . DOI: http://dx . doi . org/10 . 7554/eLife . 22617 . 006 Eight of the top 26 gene homologs encode products involved in histone H3K4 and H3K27 methylation ( Figures 2A and 4 ) . Epigenetic activating ( H3K4me ) and inactivating ( H3K27me ) marks are associated with selective activation of promoters and enhancers , and chromatin marks are associated with specification and differentiation of cardiac cells ( Wamstad et al . , 2012 ) , murine heart development ( He et al . , 2012; Delgado-Olguín et al . , 2012 ) , and cardiac pathogenesis ( Kaneda et al . , 2009; Zhang and Liu , 2015 ) . Two additional top 26 genes ( SUV420H1 and HUWE1 ) also encode conserved histone modifying proteins . Thus 42% of the top candidate genes from the PCGC study are involved in histone modification pathways , and all but one are highly conserved . Because chromatin modification pathway associated genes represent the major functional class among the ‘top 26’ , we chose to examine in more detail the roles of these genes in Drosophila heart morphogenesis and function . 10 . 7554/eLife . 22617 . 007Figure 4 . Genes involved in H3K4 and H3K27 methylation , mutated in CHD patients , affect heart structure , developmental mortality , and adult survival . ( A ) Depiction of nucleosome showing H3K4 and H3K27 methylation , and ubiquitination of H2BK120 ( required for H3K4 methylation ) . Drosophila homologs involved in the production , removal , or interpretation of modifications are shown ( human genes shown in red ) . ( B ) Adult heart phenotype induced by cardioblast-specific expression ( driven by 4XHand-Gal4 ) of UAS-RNAi transgenes targeting Kismet and Trx . Cardiac actin ( myofibers ) was visualized by Phalloidin staining . Hand-GFP expression ( nuclear ) labels cardioblast cells . Pericardin was immune-labeled . Dotted lines delineate the heart tube outline . Red arrow points to remnant cardioblast cell in Trx-silenced heart . Scale bar = 50 µ . ( C ) Quantitation of adult heart cardiac myofibrillar density ( as % of control; N = 10 for Control , N = 6 for indicated silenced gene ) . Statistical significance ( * ) was defined as p<0 . 05 . Scale bar = 50 µ . ( D ) Quantitation of adult heart cardiac collagen ( Pericardin ) deposition ( as % of control; N = 10 for Control , N = 6 indicated silenced gene ) . Statistical significance ( * ) was defined as p<0 . 05 . ( E ) Quantitation of adult heart cardioblast cell numbers ( cardioblasts expressing nuclear GFP; N = 10 for Control , N = 10 for indicated silenced gene ) . Statistical significance ( * ) was defined as p<0 . 05 . ( F ) Percentage of adult male flies dead at day 25 post-emergence ( N = 50 flies per genotype ) . ( G ) Larva ( third instar ) heart phenotypes induced by cardioblast-specific expression ( driven by 4XHand-Gal4 ) of UAS-RNAi transgenes targeting Wds , Kismet , and Trx . Cardiac actin ( myofibers ) was visualized by Phalloidin staining . The pericardin was immune-labeled . Dotted lines delineate heart tube outline . ( H ) Quantitation of larval heart cardiac myofibrillar density ( as % of control; N = 10 for Control , N = 6 for indicated silenced gene ) . Statistical significance ( * ) was defined as p<0 . 05 . ( I ) Quantitation of larva heart cardiac collagen ( Pericardin ) deposition ( as % of control; N = 10 for Control , N = 6 for indicated silenced gene ) Statistical significance ( * ) was defined as p<0 . 05 . Control flies were the progeny of a cross between homozygous 4XHand-Gal4 and w1118 , which has one copy of 4xHand-Gal4 but does not carry a UAS-RNAi silencing construct . DOI: http://dx . doi . org/10 . 7554/eLife . 22617 . 00710 . 7554/eLife . 22617 . 008Figure 4—figure supplement 1 . Genes involved in H3K4 and H3K27 methylation , mutated in CHD patients . ( A ) Adult heart phenotype induced by 4XHand-Gal4 driven expression of UAS-RNAi transgenes targeting UbcD6 and Lid . Left to right panels: cardiac myofibers visualized by Phalloidin staining , cardioblast cells labeled by Hand-GFP expression ( nuclear ) , Pericardin ( Type IV collagen ) immuno-labeled , merged images . Scale bar = 50 µ . ( B ) Adult fly survival curves ( N = 50 flies per genotype ) . Control flies were homozygous for w1118 and carried the 4XHand-Gal4 transgene , but not a UAS-RNAi silencing construct . DOI: http://dx . doi . org/10 . 7554/eLife . 22617 . 008 We analyzed heart phenotypes resulting from cardiac-specific silencing of the eight genes involved in H3K4 and H3K27 methylation ( Figure 4 ) . Silencing kismet yielded an adult heart phenotype in which the heart tube was severely affected ( Figure 4B ) with very low cardiac myofibrillar density ( Figure 4C ) , elevated Pericardin ( collagen ) deposition ( Figure 4D ) , and reduced numbers of cardioblasts ( Figure 4E ) . Adult fly longevity was significantly affected , with approximately 50% of flies dead by day 25 ( Figure 4F ) . During development , kismet silencing caused 54% mortality ( Figure 2A ) . In larvae , cardiac myofibrillar density was reduced with pronounced heart tube filament disruption ( Figure 4G , H ) , and Pericardin over-expression with disorganized ECM deposition ( Figure 4G , I ) . Trx-silenced adult flies lacked cardiac actin and heart tube ( Figure 4B , C ) , exhibited severely reduced Pericardin ( Figure 4D ) , almost no cardioblast cells ( Figure 4E ) , and all were dead at day 25 after emergence , compared to 20% mortality among control flies ( Figure 4F ) . Trx silencing caused 35% developmental lethality ( Figure 2A ) . In larvae , knock down of trx induced abnormal heart tube structure ( Figure 4G ) with reduced cardiac myofibrillar density ( Figure 4G , H ) and reduced Pericardin ( Figure 4G , I ) . Silencing of wds caused complete developmental lethality ( Figure 2A ) and a dramatic larva heart phenotype ( Figure 4G ) featuring abnormal cardiac actin organization and fiber density ( Figure 4H ) , and extremely high levels of Pericardin ( Figure 4G , I ) . Taken together , these results strongly suggest that H3K4 methylation is critical for normal heart development and function , a conclusion supported by the observation that knockdown of the H3K4 de-methylase Lid had no phenotypic consequences ( Figure 4C–F , H , I; Figure 4—figure supplement 1 ) . Similarly , silencing of SMOX ( involved in H3K27 de-methylation ) resulted in minimal developmental mortality ( MI 7% , Figure 2A ) but normal cardiac myofibrillar density and Pericardin levels in larvae ( Figure 4H , I ) . The adult heart phenotype was essentially normal , as was adult mortality at day 25 ( Figure 4 ) . Ubiquitination of H2BK120 ( Figure 4A ) is required for H3K4 methylation , and we observed that silencing of the UbcD6 gene resulted in 84% developmental mortality ( Figure 2A ) and reduced adult longevity ( Figure 4F; Figure 4—figure supplement 1 ) . The adult heart was severely affected ( Figure 4—figure supplement 1 ) with greatly reduced cardiac myofibrillar density ( Figure 4C ) , elevated Pericardin ( Figure 4D ) , and reduced cardioblast cell numbers ( Figure 4E ) . Silencing of Bre1 and Scny also caused developmental mortality and cardiac myofibrillar density was somewhat reduced in larvae and adults ( Figure 4C , H ) . Pericardin levels , cardioblast cell numbers , and adult mortality at day 25 were essentially normal ( Figure 4D , E , F , I ) . By contrast , H3K4 de-methylation did not appear critical as evidenced by the essentially normal heart phenotype , normal MI , and normal adult longevity observed upon Lid silencing . ( Figure 4 and Figure 4—figure supplement 1 ) . These results further demonstrate the utility of the Drosophila system to functionally test large-scale clinical genomics data and gain insights into fundamental pathways contributing to disease pathogenesis . We used dye angiography to measure effects on heart function of cardiac-specific silencing of the eight genes involved in H3K4 and H3K27 methylation ( Figure 5 ) . Briefly , fluorescent dye was injected into the posterior abdomen of pharate ( pre-eclosion ) adult flies , where it entered the heart chamber and was pumped anteriorly . Dye accumulation in the head over a 30 s time interval provided a measurement of cardiac efficiency ( Figure 5A , B ) ( Drechsler et al . , 2013 ) . We observed that in normal control flies fluorescent dye is readily detectable in the head within 15 s of injection , and significantly accumulated by 30 s . By contrast , silencing of candidate CHD gene homologs involved in the methylation of H3K4 and H3K27 profoundly impaired cardiac function ( Figure 5B , C ) . Silencing of Smox and Lid , however , did not measurably impair heart function as measured in this assay . These observations indicate that H3K4 and H3K27 methylation , and not demethylation , is essential for heart function , consistent with our findings with regard to cardiac tissue morphology and effects on developmental lethality and adult longevity . 10 . 7554/eLife . 22617 . 009Figure 5 . Direct measurement of heart function . ( A ) The adult heart is depicted schematically in green . Supporting lateral alary muscles are depicted in red . Fluorescent dye injected into the posterior body cavity ( arrow ) enters the heart tube lumen and is pumped anteriorly into the brain region . ( B ) Time course of injected fluorescent dye entering the brain region ( dotted outline ) . In normal control flies dye is easily detectable 15 s ( s ) after injection , and the brain is highly fluorescent within 30 s . In flies expressing heart-specific Wds or Kismet gene targeting RNAi silencing constructs , by contrast , dye does not reach the brain by 30 s after injection . Scale bar = 100 µ . ( C ) Quantitative analysis of brain region fluorescence , relative to control fly levels , as a function of heart-specific gene silencing . Experiments were performed in triplicate ( 3 independent dye injection experiments ) . Statistical significance ( * ) was defined as p<0 . 05 . Control flies were the progeny of a cross between homozygous 4XHand-Gal4 and w1118 , which has one copy of 4xHand-Gal4 but does not carry a UAS-RNAi silencing construct . DOI: http://dx . doi . org/10 . 7554/eLife . 22617 . 009 Gene silencing using available and well-established UAS-RNAi fly lines allows rapid and efficient functional screening of large numbers of candidate genes . We also established an in vivo platform to analyze the cardiac phenotypes of specific patient-derived alleles of candidate CHD genes in Drosophila , employing a ‘gene replacement’ strategy . As a proof-of-concept , we tested the ability of a wild type human WDR5 allele to rescue heart phenotypes induced by silencing of the endogenous wds Drosophila homolog . Heart-specific silencing of wds caused 100% developmental lethality and abnormal heart morphology in late larvae characterized by reduced cardiac myofibers and severe over-deposition of Pericardin ( Figure 6 ) . We found that simultaneously overexpressing a wild type human WDR5 allele reduced developmental lethality almost 7-fold , significantly restored cardiac myofibrillar density , and reduced Pericardin levels essentially to normal ( Figure 6 ) . By contrast , when the endogenous wds expression was ‘replaced’ by a CHD patient-derived WDR5-K7Q mutant allele , developmental lethality remained quite elevated , cardiac myofibrillar density remained abnormally low ( relative both to Control and wild type WDR5 rescue ) , and Pericardin levels were lowered but still significantly higher than both Control and wild type WDR5 rescue . These observations illustrate the structure-function homologies between human and fly ‘heart genes’ that support use of Drosophila for functional validation of candidate CHD genes . Moreover , they demonstrate that the ‘gene replacement’ strategy can be used to quantitatively assess phenotypes induced by patient-specific alleles of candidate disease genes . 10 . 7554/eLife . 22617 . 010Figure 6 . Wds silencing-induced lethality and heart phenotypes rescued by wild type human WDR5 but not by a patient derived WDR5-K7Q mutant allele . ( A ) Developmental lethality ( Mortality Index ) for flies in which endogenous Wds heart expression was silenced , and attempted rescue by either wild type WDR5 or mutant WDR5-K7Q overexpression ( OE ) . ( B ) Larva ( third instar ) heart phenotype induced by cardioblast-specific expression ( driven by 4XHand-Gal4 ) of UAS-RNAi transgene targeting Wds , and attempted rescue by WDR5 or WDR5-K7Q overexpression . Cardiac actin ( myofibers ) visualized by Phalloidin staining . Pericardin was immune-labeled . Dotted lines delineate the heart tube outline . ( C ) Quantitation of larval heart cardiac myofibrillar density ( as % of control; N = 10 for Control , N = 6 for Wds-IR , Wds-IR+WDR5-OE , or Wds-IR+WDR5-K7Q-OE ) . Statistical significance ( * ) was defined as p<0 . 05 . For each strain , fiber density was significantly lower than control . Rescue by wild type WDR5 significantly rescued fiber density , to levels significantly greater than achieved by the overexpression of the WDR5-K7Q mutant allele . ( D ) Quantitation of larva heart cardiac collagen ( Pericardin ) deposition ( as % of control; N = 10 for Control , N = 6 for Wds-IR , Wds-IR+WDR5-OE , or Wds-IR+WDR5-K7Q-OE ) Statistical significance ( * ) was defined as p<0 . 05 . Wild type WDR5-OE fully rescued up-regulated collagen deposition induced by Wds-IR . Mutant WDR5-K7Q-OE partially rescued the Wds-IR phenotype , but collagen deposition levels remained upregulated compared to the control . The control flies were the progeny of a cross between homozygous 4XHand-Gal4 and w1118 , which has one copy of 4xHand-Gal4 but does not carry a UAS-RNAi silencing construct . DOI: http://dx . doi . org/10 . 7554/eLife . 22617 . 010
Our results demonstrate the value of the Drosophila model system to functionally screen and validate candidate disease genes identified through large scale sequencing efforts . The high-throughput testing platform described here can extend gene association data to quantitative , in vivo morphological and functional criteria . On the basis of this screening , the most promising genes can justifiably be advanced to more costly , time consuming , and technically challenging validation platforms required to definitively identify and confirm disease-causing gene variants . Such advanced testing , we presume , will encompass gene function analysis in a mammalian model system . The advantages of functional screening using RNAi-based silencing lines are to some extent offset by certain limitations . From the outset , of human disease genes , approximately 25% are not represented by Drosophila homologs . False negatives may occur if RNAi expression does not sufficiently reduce mRNA levels . However , we tested a number of RNAi lines by qRT-PCR and found no examples of ineffective gene silencing . False negatives may also occur when missense mutations produce gain-of-function ( GOF ) alleles , and we cannot accurately predict the degree to which this could confound our analysis . This problem , we note , is mitigated to a degree by the fact that genes altered by GOF mutations would in many instances normally encode proteins with cardiac cell functions , and thus yield silencing induced phenotypes in our functional screening system . False positives can arise from so-called ‘off target’ effects of a given RNAi targeting construct . We controlled against misleading results from both RNAi silencing inefficiency and off target effects by employing the latest-generation available RNAi lines , and testing multiple such lines for each of the top 26 candidate genes . This precaution against false negative and false positive results further limited the total number of candidate genes we could screen to 134 ( 60% of the CHD candidate genes published by the PCGC ) . Nevertheless , our in vivo screening system rapidly and efficiently confirmed roles in heart development , tissue maintenance , or cardiac function for 72 candidate genes . Phenotypic analysis based on gene silencing , while effective for high throughput screening , represents a first step in validation and characterization of candidate mutations . We also demonstrated an initial ‘gene replacement’ strategy that takes further advantage of the available resources and sophisticated genetics of the Drosophila model system to quantitatively characterize the phenotype induced by expression of a specific patient-derived mutant allele . This approach may prove highly valuable as a testing platform for precision medicine based therapeutic drugs . Ultimately , the knowledge that a specific mutation contributes to disease pathogenesis opens the door to the application of precise gene-editing interventions , strategies for which are currently being developed based on CRISPR/Cas-9 mediated genome editing ( Long et al . , 2016; Carroll et al . , 2016; Long et al . , 2014 ) . The high throughput candidate gene validation studies reported here represent the essential first step in the functional confirmation , phenotypic characterization , and ultimate precision treatment of disease-associated mutations .
Drosophila lines were obtained from the Bloomington Stock Center ( Bloomington , IN; NIH P40OD018537 ) . Transgenes were overexpressed with the UAS-GAL4 system ( Brand and Perrimon , 1993 ) . To increase the penetrance of heart-specific UAS-RNAi mediated gene silencing we developed a powerful Gal4 driver incorporating four repeats of the Hand gene ( Han et al . , 2006 ) enhancer . 4XHand-Gal4 was generated by tandem insertions of the original Hand cardiac-specific enhancer ( Han and Olson , 2005 ) in the pGal4 vector . The wild type WDR5 cDNA was obtained from OriGene , which encodes the 334 a . a . protein with GenBank ID P61964 . The WDR5-K7Q cDNA was generated by introducing the K7Q mutation into the wild type WDR5 cDNA sequence using PCR . To generate UAS-WDR5 and UAS-WDR5-K7Q constructs , the above cDNAs were cloned into the pUASTattB vector and the transgenes were introduced into a fixed chromosomal docking site by germ line transformation . Pharate ( pre-eclosion ) adult flies were carefully removed from pupa cases and affixed by double-sided scotch tape to glass slides . A single injection of 100 nl of uranin solution ( 1 µg/µl in PBS ) was delivered into the abdomen using a Drummond Nanoject II auto-nanoliter injector with a glass capillary . Dye accumulation in the fly head was monitored over a 3 min interval using a Zeiss ApoTome . 2 microscope with a 5x Plan-Neofluar 0 . 16 air objective . Experiments were performed in triplicate ( 3 independent dye injection experiments ) . ImageJ software Version 1 . 49 was used for image processing and quantification . Fluorescence intensity in the fly head was measured at times 0 ( time of injection ) and 30 s post-injection . Increased fluorescence values represent fluorescence at t = 30 s minus t = 0 , and were expressed relative to head fluorescence of control flies not expressing an RNAi targeting transgene . RNA was isolated using Trizol Reagent ( Invitrogen , Carlsbad , CA ) from heart tissue dissected from 60 adult flies of the relevant genotype . RNA purity and concentration were determined using a Nanodrop-1000 ( Thermo Scientific , Wilmington , DE ) . Total RNA ( 1 μg ) was reverse transcribed using Superscript IV ( Invitrogen ) . SYBR Green based real-time qPCR ( Power Cyber Mastermix; Applied Biosystems , Carlsbad , CA ) was performed using a StepOne Plus ( Applied Biosystems ) . Gene-specific primer pairs were used . Quantitative values were determined using the 2-ΔΔCT method ( Livak and Schmittgen , 2001 ) , normalizing to Gapdh . Values are derived from three technical replicates of qRT-PCR experiments from pooled RNA . Fly preparation and imaging are described in more detail at Bio-protocol ( Zhu et al . , 2017 ) . Larvae and adult flies were dissected and fixed for 10 min in 4% paraformaldehyde in phosphate-buffered saline ( PBS ) . Alexa Fluor 555 Phalloidin was obtained from Thermo Fisher . Mouse anti-Pericardin antibody ( EC11 ) was used at 1:500 dilutions , followed by Cy3-conjugated secondary antibodies ( Jackson Lab ) . Confocal imaging was performed with a Zeiss ApoTome . 2 microscope using a 20× Plan-Apochromat 0 . 8 N . A . air objective . For quantitative comparisons of intensities , common settings were chosen to avoid oversaturation . ImageJ Software Version 1 . 49 was used for image processing . For quantitative comparisons of cardiac myofibrillar density , cardioblast cell numbers , and Pericardin deposition we analyzed six to ten control flies and six to ten flies of each experimental genotype . These features were quantitatively assessed in the same area of the same segment in each fly . Cardiac myofiber levels of Z-stack projections were carefully selected for analysis , avoiding the ventral muscle layer underlying the heart tube . Sample size determinations were based upon extensive previous experience in analyzing fly heart morphology . Within 24 hr of egg laying Drosophila larvae were transferred from 25°C to 29°C to boost UAS-transgene expression . Adult male flies were maintained at 29°C in groups of 15 or fewer; 50 flies in total were assayed per genotype . Statistical tests were performed using PAST . exe software ( http://folk . uio . no/ohammer/past/index . html ) unless otherwise noted . Sample errors are given as standard error of the mean ( s . e . m ) . Data were first tested for normality by using the Shapiro-Wilk test ( α = 0 . 05 ) . Normally distributed data were analyzed either by Student's t-test ( two groups ) and Bonferroni comparison to adjust p value or by a one-way analysis of variance followed by a Tukey-Kramer post-test for comparing multiple groups . Non-normal distributed data were analyzed by either a Mann-Whitney test ( two groups ) and Bonferroni comparison to adjust the p value or a Kruskal-Wallis H-test followed by a Dunn's test for comparisons between multiple groups . Statistical significance was defined as p<0 . 05 . | Around one in 100 children are born with heart defects caused by congenital heart disease . Studying the genetic sequences of people with congenital heart disease has revealed many genes that may play a role in causing the condition , but few of these findings have been confirmed experimentally in animal model systems . The fruit fly species Drosophila melanogaster is often used in genetic studies because it is a relatively simple organism . The insights gained from studying flies are often valuable for determining the direction of subsequent investigations in more complex animals – such as humans – that involve experiments that are more costly and less efficient . Zhu , Fu et al . have now used fruit flies to investigate the effects of 134 genes that have been suggested to contribute to congenital heart disease . The investigation used a method that rapidly allowed the activity of specific genes to be altered in the flies . The effects that these alterations had on many aspects of heart development , structure and activity were then measured . Of all the genes tested , 70 caused heart defects in the flies . Several of these genes help to modify the structure of proteins called histones; these modifications play important roles in heart cell formation and growth . Further tests showed that the effects of specific genetic errors that had been identified in people with congenital heart disease could be reliably reproduced in the flies . This may allow individual cases of congenital heart disease to be replicated and studied closely in the lab , helping to create treatments that are personalised to each patient . Studying congenital heart disease in flies provides a fast and simple first step in understanding the roles that different genes play in the disease . Moving forward , precise gene editing techniques could be used to generate flies to examine the role of each of the genetic mutations that occur in individual patients . Ultimately , when gene editing techniques are ready to be used in humans , this could lead to cures for congenital heart disease at the DNA level , so that these mutations won’t be passed on to the next generation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2017 | High throughput in vivo functional validation of candidate congenital heart disease genes in Drosophila |
High-speed atomic force microscopy was employed to observe structural changes in actin filaments induced by cofilin binding . Consistent with previous electron and fluorescence microscopic studies , cofilin formed clusters along actin filaments , where the filaments were 2-nm thicker and the helical pitch was ∼25% shorter , compared to control filaments . Interestingly , the shortened helical pitch was propagated to the neighboring bare zone on the pointed-end side of the cluster , while the pitch on the barbed-end side was similar to the control . Thus , cofilin clusters induce distinctively asymmetric conformational changes in filaments . Consistent with the idea that cofilin favors actin structures with a shorter helical pitch , cofilin clusters grew unidirectionally toward the pointed-end of the filament . Severing was often observed near the boundaries between bare zones and clusters , but not necessarily at the boundaries .
Actin filaments are involved in a variety of important functions in eukaryotic cells , including muscle contraction , amoeboid movement , cytokinesis , intracellular transport , and transcriptional regulation within the nucleus . These diverse functions depend on the interaction between actin and specific actin binding proteins ( ABPs ) , and it is generally assumed that specific biochemical signaling is involved in the spatial and temporal regulation of each actin–ABP interaction . During migration of amoeboid cells , for instance , cofilin plays essential roles in continuous extension of lamellipodia by severing actin filaments to promote filament depolymerization or to initiate polymerization from new barbed ends ( reviewed by Bravo-Cordero et al . , 2013 ) . Three independent biochemical mechanisms are known to inhibit cofilin activity; phosphorylation of Ser3 , sequestration to phosphatidylinositol 4 , 5-bisphosphate ( PIP2 ) in plasma membrane , and lower pH ( Bernstein and Bamburg , 2010; Bravo-Cordero et al . , 2013 ) . These inhibitory mechanisms are implicated in the regulation of cofilin activity during cell migration , since experimental unleashing of inactive cofilin has been shown to initiate cofilin-dependent processes ( Bernstein and Bamburg , 2010; Bravo-Cordero et al . , 2013 ) . Critically speaking , however , these data do demonstrate that active cofilin is required , but there is little experimental evidence that localized activation of cofilin is necessary for proper cofilin functions in cell migration . Indeed , overexpression of constitutively active S3A cofilin did not inhibit motility ( Endo et al . , 2003; Popow-Wozniak et al . , 2012 ) . Thus , biochemical signaling is not sufficient to explain how cofilin activities are properly regulated spatially and temporally in cells . On the other hand , many ABPs alter the atomic structure of actin subunits within filaments , and in certain cases these conformational changes are cooperative . For instance , the pioneering work of Oosawa and his colleagues demonstrated that the increase in fluorescence intensity of a fluorophore on actin saturates when only one molecule of the myosin motor domain is added for each 10 actin subunits within filaments ( Oosawa et al . , 1973 ) . Similar myosin-induced cooperative conformational changes have been detected in various other assays ( Tawada , 1969; Fujime and Ishiwata , 1971; Loscalzo et al . , 1975; Miki et al . , 1982; Prochniewicz et al . , 2010 ) . In addition , dense binding of cofilin shortens the helical pitch of actin filaments by 25% ( McGough et al . , 1997; Galkin et al . , 2001; Sharma et al . , 2011 ) , and time-resolved phosphorescence anisotropy ( Prochniewicz et al . , 2005 ) and differential scanning calorimetry ( Dedova et al . , 2004; Bobkov et al . , 2006 ) showed that one molecule of bound cofilin changes the structure of ∼100 actin subunits within a filament . Moreover , binding of cofilin to actin filaments is cooperative ( Hawkins et al . , 1993; Hayden et al . , 1993; McGough et al . , 1997; De La Cruz , 2005; Hayakawa et al . , 2014 ) . This implies that cooperative conformational changes induced by an ABP are propagated to neighboring actin subunits , increasing their affinity for that , or another , ABP . This could provide a novel mechanism by which actin filaments change their function by regulating their affinities for various ABPs ( Tokuraku et al . , 2009; Michelot and Drubin , 2011; Schoenenberger et al . , 2011; Uyeda et al . , 2011; Galkin et al . , 2012; Romet-Lemonne and Jegou , 2013 ) . However , the currently available information on structural changes to actin filaments is limited to high-resolution static images ( electron microscopy ) , low-resolution dynamic changes ( fluorescence microscopy ) , and bulk biochemical analyses . There is little information available in the molecular mechanism that mediates the propagation of structural changes to neighboring actin subunits . Atomic force microscope ( AFM ) is unique in that it enables detailed structural analysis of wet protein samples ( Müller and Dufrêne , 2008 ) , and recent dramatic improvements in scanning speed now enables real time imaging of conformational changes in protein samples with high-spatial resolution ( Ando et al . , 2013 ) . This high-speed AFM ( HS-AFM ) has been used to visualize molecular movements such as the stepping motion of myosin V along actin filaments ( Kodera et al . , 2010 ) , rotary catalysis of F1 ATPase without a rotor ( Uchihashi et al . , 2011 ) , and light-induced conformational changes in bacteriorhodopsin ( Shibata et al . , 2010 ) . Here , we used HS-AFM to directly visualize conformational changes in actin filaments induced by cofilin binding . We found that conformational changes within cofilin clusters unidirectionally propagate to the neighboring bare actin zone in a cooperative manner and that the growth of the cofilin cluster follows this unidirectional cooperative conformational change .
For AFM , actin filaments must be immobilized on the stage , yet they must have the freedom of movement to bind cofilin and exhibit the resultant changes in the helical twist that accompany cofilin binding . We therefore formed a bilayer of positively charged lipid on the surface of freshly peeled mica ( Yamamoto et al . , 2010 ) fixed to the observation stage . A solution of actin filaments was then placed on the supported lipid bilayer and HS-AFM was performed . Right-handed double helical filaments were clearly visualized ( Figure 1 ) , as in earlier reports ( Weisenhorn et al . , 1990; Schmitz et al . , 2010 ) . 10 . 7554/eLife . 04806 . 003Figure 1 . HS-AFM observation of control and cofilin-bound actin filaments . ( A ) Control actin filaments without cofilin . ( B ) Actin filaments fully bound with cofilin over an extended distance . ( C ) Paracrystals of actin filaments . Bars: 25 nm , Z-scale: 0–12 nm . ( D and E ) Histograms of peak heights ( D ) and lengths of half helical pitches ( E ) in control actin filaments and cofilin-decorated actin segments . N was between 1722 and 2536 . ( F ) Half helical pitches of actin paracrystals ( N = 1009 ) . Solid lines show Gaussian fittings with confidence intervals of 99 . 73% . For comparison , the dark line in ( F ) shows the Gaussian fitting of control actin filaments . Measurements were made in F buffer containing 1 mM ATP ( A ) , 1 mM ATP , and 75 nM cofilin ( B ) or 1 mM ATP and 30 mM MgCl2 ( C ) . Student's t-test comparing control and cofilin-decorated actin segments showed that the differences in peak heights and half helical pitches are statistically significant at p ≤ 0 . 00001 . The mean of the half helical pitches of control actin filaments and paracrystals did not differ significantly . Models of control actin filaments and cofilin-decorated actin filaments with two different orientations on substrates are shown in Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 00310 . 7554/eLife . 04806 . 004Figure 1—figure supplement 1 . Models of control actin filaments and cofilin-decorated filaments on a flat substrate . ( A and B ) show predicted peak heights of control actin filaments ( PDB ID 3G37; Murakami et al . ( 2010 ) ) , and ( C and D ) show fully cofilin-decorated filaments ( PDB ID 3J0S; Galkin et al . ( 2011 ) ) . These images were generated using Chimera software . When the true crossover point coincides with the position of an actin subunit ( A ) or a bound cofilin molecule ( C ) , the position of the peak is equal to that of the crossover point , and its height is the tallest of the various possible orientations of the filament on the substrate . When the true crossover point is at the middle of two actin subunits ( B ) or two bound cofilin molecules ( D ) , the peak height is the shortest of the possible orientations . In this latter case , the observed peak position is away from the true crossover point by as much as ∼2 . 5 nm . Depending on the orientation of the filament , therefore , the distance between the observed peak position and the true crossover point fluctuates between 0 and 2 . 5 nm . Random distribution of distances between the true crossover points and the highest points of the closest actin subunits would result in ±1 . 7 nm of error in the estimation of the positions of the true crossover points . If the peak positions at both ends of a half helix have this much error , the half helical pitch would have ±2 . 3 nm ( ±1 . 7 nm × √2 ) error . This is somewhat smaller than the ±3 . 0 nm standard deviation of the half helical pitches of actin filament paracrystals , and the difference should derive from small structural fluctuations in the paracrystals and/or measurement error from various other sources . In either case , it is safe to use the 3 . 0 nm standard deviation as the upper limit of the combined measurement errors and to estimate the lower limit of the structural fluctuations of control and cofilin-bound filaments . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 004 We approximated the positions of the crossover points , where the two strands of actin filament are aligned vertically , using the position of the highest point ( peak ) in each half helix identified in the AFM images . Quantitative analysis indicated that the height of those peaks , or the thickness of the filaments , was 8 . 6 ± 0 . 8 nm ( average ± SD ) , and the spacing between the peaks , or half helical pitch , was 36 . 8 ± 4 . 3 nm ( Figure 1D , E ) . These values , which are consistent with previously reported structural data ( Hanson and Lowy , 1963; Hanson , 1973; Egelman et al . , 1982 ) , will hereafter be referred to as normal height and normal half helical pitch , respectively . The distribution of half helical pitches ranged from ∼26 nm to ∼45 nm and was well fit by a normal distribution with a standard deviation of 4 . 3 nm . This distribution should reflect natural variation in half helical pitch , as was proposed on the basis of a similar distribution of half helical pitches in electron micrographs of negatively stained specimens ( Hanson , 1967; Egelman et al . , 1982 ) and also measurement errors . To assess the overall magnitude of the measurement errors , we prepared paracrystals of actin filaments in the presence of 30 mM MgCl2 ( Figure 1C ) ( Hanson , 1973 ) . The distribution of the half helical pitches of the paracrystals measured under the same conditions was narrower than that of free filaments , with a mean of 36 . 5 nm and standard deviation of 3 . 0 nm ( Figure 1F ) . In a hypothetical case where the paracrystals do not undergo spontaneous conformational changes , this distribution is the upper limit of the measurement errors in our system . Furthermore , assuming normal distributions of the true half helical pitches and measurement errors , the true standard deviation of the half helical pitches of control filaments was calculated to be 3 . 1 nm or larger , by subtracting the variance of the actin paracrystals from that of control filaments . This indicated that the structure of the actin filaments does indeed vary with changes in helical twist . When 75 nM cofilin was added to actin filaments immobilized on the lipid surface , cofilin gradually bound to the actin filaments , and sections of filaments bound with cofilin molecules were easily identified , as the bound filaments appeared thicker and the peaks were taller ( Figure 1B ) . Peak heights within long cofilin clusters ( longer than eight consecutive half helices ) showed a single distribution of 10 . 6 nm ± 1 . 0 nm , approximately 2 nm taller than the control filaments ( Figure 1D ) . This is consistent with electron microscopic analysis ( Galkin et al . , 2011 ) , which showed that filaments fully decorated with cofilin are ∼2 . 3-nm thicker than control filaments due to the presence of cofilin molecules ( Figure 1—figure supplement 1 ) . Moreover , the height distribution was fit well by a single normal distribution , which suggests that within those long cofilin clusters , both actin strands are homogeneously bound with cofilin molecules . Cofilin binding shortened the half helical pitch by 27% to 26 . 9 nm ( Figure 1E ) , which is again consistent with earlier electron microscopic ( McGough et al . , 1997; Galkin et al . , 2001 ) and AFM analyses ( Sharma et al . , 2011 ) . The standard deviation of the distribution of half helical pitches of the fully cofilin-decorated filaments was 3 . 8 nm , and subtracting the maximum possible measurement error yielded a standard deviation of 2 . 3 nm or larger . This value is smaller than that of the control filaments ( 3 . 1 nm ) and suggests that cofilin binding restricts torsional movements of the filaments within cofilin clusters . We next analyzed shorter cofilin clusters along otherwise apparently bare filaments ( Figure 2 ) so that both ends of the cluster were visible within the imaged area . Intriguingly , we noticed that the half helical pitch in the apparently bare section on one side of the cofilin cluster was short , while the pitch on the other side of the cluster was nearly normal . 10 . 7554/eLife . 04806 . 005Figure 2 . Asymmetric structure of bare actin zones neighboring a cofilin cluster . ( A and B ) HS-AFM image of a short cofilin cluster transiently associating with two S1 molecules ( yellow arrowheads ) , which persisted for ∼1 s , enabling identification of the filament polarity ( B ) . Measurements were made in F buffer containing 20 nM S1 , 75 nM cofilin , 1 mM ADP , and 0 . 1 mM ATP . Bar: 25 nm; Z-scale: 0–12 nm . ( C and D ) Time-dependent changes in the heights of the indicated peaks ( white arrowheads ) and half helical pitches between the indicated peaks . ( E and F ) Histograms of the lengths of the half helical pitches of bare actin segments immediately neighboring cofilin clusters . Filaments were incubated in F buffer containing 1 mM ADP and 0 . 1 mM ATP for 5 min before observation ( E ) as in ( A and B ) or were incubated in F buffer containing 1 mM ADP for 30 min before observation ( F ) . Pitches of the half helices of the first immediate neighbor on each side of cofilin clusters were measured . These values , together with those for the second neighbors , are summarized in Table 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 005 The observation buffer was then changed to F buffer containing 0 . 1 mM ATP , 1 mM ADP , hexokinase , glucose , 20 nM subfragment-1 ( S1 ) of skeletal myosin II , and 75–100 nM cofilin , so that the polarity of the actin filaments could be identified from the characteristic tilted binding angle seen when S1 transiently binds the filament ( Huxley , 1963 ) . This analysis revealed that the half helical pitch of the bare zone on the pointed-end side of the cofilin cluster was short , while that on the barbed-end side was slightly longer than the normal pitch ( Figure 2 and Table 1 ) . Student's t-test indicated that the differences in the mean half helical pitches between the first neighbor bare zone on either side of the cluster and the control filaments are statistically significant ( p < 0 . 00001 and p < 0 . 001 for pointed-end and barbed-end side , respectively ) . Pitches of the neighboring half helices second from the cofilin clusters were nearly normal ( Table 1 ) . 10 . 7554/eLife . 04806 . 006Table 1 . Peak heights and lengths of half helical pitches in bare actin segments neighboring cofilin clustersDOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 0061 mM ADP + 0 . 1 mM ATP1 mM ADPPeak height ( nm ) Half helix ( nm ) Peak height ( nm ) Half helix ( nm ) First neighbor on the P-end side9 . 2 ± 1 . 028 . 8 ± 4 . 58 . 9 ± 1 . 028 . 7 ± 4 . 7Second neighbor on the P-end side9 . 1 ± 1 . 036 . 5 ± 4 . 18 . 8 ± 1 . 136 . 9 ± 4 . 9First neighbor on the B-end side9 . 0 ± 0 . 937 . 3 ± 4 . 68 . 7 ± 0 . 938 . 4 ± 4 . 3Second neighbor on the B-end side9 . 2 ± 0 . 635 . 4 ± 3 . 98 . 7 ± 0 . 836 . 7 ± 4 . 2Actin filaments were incubated in F buffer containing 1 mM ADP and 0 . 1 mM ATP for 5 min or in F buffer containing 1 mM ADP for 30 min prior to the addition of cofilin . Filaments under the latter condition were shorter than those under the former condition and were apparently in the process of spontaneous depolymerization . Each mean and SD were calculated from 423 to 446 data . The results summarized above demonstrate that cofilin clusters induce distinctively asymmetric conformational changes in bare zones immediately neighboring the clusters , but it is still uncertain whether a stretch of many bound cofilin molecules , as in clusters , is necessary to induce such conformational changes . To answer that question , we needed to visualize individual cofilin molecules bound to actin filaments , and measure the half helical pitch of the filament around those bound molecules . This has not been possible because cofilin molecules are too small to image individually using electron microscopy or AFM . We therefore engineered a fusion protein in which the N-terminal half of the rod domain of α-actinin was attached to the C-terminus of cofilin . In electron micrographs of negatively stained specimens ( Figure 3A ) , as well as in HS-AFM images ( Video 1 and Figure 3—figure supplement 4 ) , we were able to see clusters of cofilin-rod with a half helical pitch ∼25% shorter than the bare zones , which is similar to control cofilin molecules bound to actin filaments ( Figure 3C ) . 10 . 7554/eLife . 04806 . 007Figure 3 . Actin filaments with bound cofilin or cofilin-rod fusion protein . ( A–C ) are electron micrographs of negatively stained samples , and ( D ) is a HS-AFM image of a sample similar to that shown in ( B ) . ( A ) Actin filaments bound with cofilin-rod . Arrowheads show crossover points in clusters of cofilin-rod . The rod portions of the fusion proteins are not readily visible , which may be due to alignment of the rods along the cofilin clusters . Severing activity and stoichiometric binding of cofilin-rod to actin filaments were confirmed by HS-AFM ( Video 1 ) and co-sedimentation assays ( Figure 3—figure supplement 3 ) , respectively . ( B ) Cofilin-rod molecules sparsely bound to actin filaments , identified by the rod-like structures ( black arrowheads ) . ( C ) Actin filaments with bound cofilin molecules ( without rod fusion ) . Arrowheads show crossover points in clusters . Actin filaments and cofilin or cofilin-rod were mixed at a 2:1 ( A and C ) or 1:1 ( B ) molar ratio in F buffer containing 1 mM ATP . Bars: 25 nm . ( D ) HS-AFM image of an actin filament and an apparently singly bound cofilin-rod molecule ( blue arrowhead in the upper image ) near P2 ( white arrowhead ) . Conditions: F buffer containing 1 mM ATP and 75 nM cofilin-rod ( without His-tag ) . Bar: 25 nm . See Video 2 . ( E ) shows heights of the three peaks and ( F ) shows spacing between them . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 00710 . 7554/eLife . 04806 . 008Figure 3—figure supplement 1 . Co-sedimentation of cofilin ( with or without His-tag ) with actin filaments . ( A and B ) are SDS-PAGE of supernatant and pellet fractions after ultracentrifugation of mixtures of various concentrations of cofilin with ( A ) or without ( B ) His-tag and 1 µM actin filaments , respectively . Acrylamide concentration of the gel was 12% . His-tag was removed by incubation with 1/4 ( wt/wt ) of His-tagged TEV protease overnight at 5°C and then passed through a Ni-column in the presence of 10 mM imidazole . Lanes 1 to 4 in ( C ) show His-TEV protease , mixture of His-cofilin and His-TEV after reaction over night , His-cofilin , and the reaction mixture that flowed through a Ni-NTA column ( i . e . , cofilin without His-tag ) . Co-sedimentation experiments were performed as follows . Frozen aliquots of cofilin ( ±His-tag ) and rabbit skeletal actin were thawed on ice for 1 hr and were then clarified by ultracentrifugation at 80 , 000 rpm for 5 min at 5°C . The supernatant was collected , and protein concentrations were measured using Advanced Protein Assay reagent ( Cytoskeleton ) . Actin filaments were prepared by polymerization of G-actin ( 20 µM ) in F buffer containing 1 mM ATP on ice for 30 min . The filaments formed were diluted to 1 µM and gently mixed with the cofilin ( ±His-tag ) of various concentrations ( i . e . , 0 . 1 , 0 . 5 , 1 , and 2 µM ) in F buffer containing 1 mM ATP . These protein mixtures were further incubated for 3–5 min at room temperature , followed by ultracentrifugation at 80 , 000 rpm for 5 min at 25°C . Supernatant and pellet fractions were separately run on a SDS-polyacrylamide gel and stained with Coomassie Blue . Control experiments were performed in the same manner using either 1 µM actin filament only or 2 µM cofilin or its derivative only . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 00810 . 7554/eLife . 04806 . 009Figure 3—figure supplement 2 . Actin binding curves of cofilin and cofilin without His tag . Band intensities of gels shown in Figure 3—figure supplement 1 were quantified , and fractions of actin-bound with cofilin ( [cofilin]ppt/[actin]ppt ) were shown as a function of free cofilin concentration ( [cofilin]sup ) . Mean ± SD are shown ( N = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 00910 . 7554/eLife . 04806 . 010Figure 3—figure supplement 3 . Co-sedimentation of cofilin-rod with ( + ) and without ( − ) His-tag . ( A and B ) SDS-PAGE analysis of binding of cofilin-rod with and without His-tag to actin filaments , respectively . Experiments were carried out in F buffer containing 1 mM ATP , and protein concentrations used are shown above each gel . Cleavage of His-tag by TEV protease was performed as described in the legend to Figure 3—figure supplement 1 . Acrylamide concentration of the gel was 10% . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 01010 . 7554/eLife . 04806 . 011Figure 3—figure supplement 4 . Representative still images from Video 1 , demonstrating cluster formation and severing function of cofilin-rod without His-tag . Red arrowheads: severing points in end half helices in cofilin clusters; white arrowheads: cofilin-rod clusters . Z-scale: 0–12 nm . Scale bar: 25 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 01110 . 7554/eLife . 04806 . 012Video 1 . Cluster formation and severing functions of cofilin-rod . Conditions: F buffer containing 1 mM ATP , and 300 nM cofilin-rod without His-tag , imaging rate: 2 frames/s , and playing rate: 5 frames/s . White arrowheads indicate clusters of cofilin-rod , and red and blue arrowheads show severing points inside a cluster and in a bare half helix immediately neighboring a cluster , respectively . Z-scale was 0–12 nm . Related to Figure 3—figure supplement 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 012 Rod-like structures were often observed to stick out from apparently bare sections of the filaments ( Figure 3B ) . These are most likely the rod portions of the cofilin-rod molecules bound sparsely along seemingly bare sections of the filaments , because their length was ∼10 nm , which is expected for half an α-actinin rod ( Yan et al . , 1993; Anson et al . , 1996 ) , and because similar rod-like structures were rarely observed in control images of cofilin added to actin filaments ( Figure 3C ) . We therefore estimated the helical pitches on both sides of single rod-like structures sticking out from the filaments by measuring the filament length that encompassed three actin subunits along one strand on either side of the bound molecule . In addition , because we did not know the polarity of each filament , we compared the helical pitches on both sides by dividing the longer helical pitch by the shorter one . This yielded a value of 1 . 02 ± 0 . 02 ( average ± SD , n = 9 ) , which was much smaller than the 1 . 30 ± 0 . 18 ( n = 4 ) measured for cofilin clusters in electron micrographs or 1 . 37 calculated from the AFM data ( Figure 1E ) and was comparable to the value of randomly selected regions along control actin filaments ( 1 . 04 ± 0 . 05 , n = 18 ) . Furthermore , the helical pitches around the apparently singly bound cofilin-rod molecules ( 36 . 0 ± 1 . 0 nm , n = 5 ) did not significantly differ from the control helical pitch ( 36 . 3 ± 1 . 0 nm , n = 8 ) measured in electron micrographs . In HS-AFM observations , we were able to observe four cases of an apparently single cofilin-rod molecule binding transiently to an actin filament ( Figure 3D and Video 2 ) . Consistent with the electron microscopic observation described above , we were unable to detect significant changes in peak heights and half helical pitches , such as those observed around cofilin clusters , around those singly bound cofilin-rod molecules ( Figure 3E , F ) . 10 . 7554/eLife . 04806 . 013Video 2 . Sparse binding of individual cofilin-rod molecules to actin filaments . Conditions: F buffer containing 1 mM ATP and 75 nM cofilin-rod without His-tag , imaging rate: 4 frames/s , and playing rate: 5 frames/s . Transient binding of cofilin-rod to and dissociation from an actin filament was followed for approximately 90 . 5 s , shown by the presence and absence of a blue arrowhead . No severing was observed in all four similar cases of successful imaging of sparse binding of cofilin-rod to actin filaments . Z-scale was 0–12 nm . Related to Figure 3D . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 013 These results suggest that a single cofilin molecule cannot induce conformational changes , either symmetric or asymmetric , that involve detectable changes in helical pitch when it binds to bare sections of the filament . Real time AFM imaging taken at 1 . 5 or 2 frames/s enabled us to follow the growth of individual cofilin clusters along actin filaments . In the case illustrated in Figure 4A and Video 3 , observed in F buffer containing 1 mM ADP and 0 . 1 mM ATP , P4 was already tall when imaging started ( ∼10 . 5 nm , average of t = 0–5 s ) , while P3 had an intermediate height ( ∼9 . 5 nm ) , and P1 and P2 were normal ( ∼8 . 5 nm ) ( Figure 4A , middle ) . We interpreted this to mean that the initial cofilin cluster extended beyond P4 but ended near P3 . P3 gradually became taller and plateaued beyond ∼25 s . Thereafter , P2 started to rise at ∼15 s and plateaued at ∼63 s , whereas P1 started to rise at ∼60 s . The distance between P2 and P3 rapidly shortened to ∼27 nm between 0 and 15 s , before P2 started to rise , while the distance between P1 and P2 shortened between 30 and 40 s , before P1 started to rise ( Figure 4A , bottom ) . This sequence of events implies gradual growth of this cofilin cluster into the neighboring bare zone on the pointed-end side where the helical pitch was shortened . 10 . 7554/eLife . 04806 . 014Figure 4 . Growth of cofilin clusters along actin filaments . Growth of cofilin clusters along actin filaments in F buffer containing 1 mM ADP and 0 . 1 mM ATP ( A ) , along actin filaments carrying ADP , prepared by incubating filaments in F buffer containing 1 mM ADP , hexokinase , and glucose at room temperature for 30 min ( B ) , and along actin filaments carrying ADP and Pi , prepared by incubating filaments in 1 mM ADP and 10 mM Pi for 10 min at room temperature ( C ) . The concentrations of cofilin were 75 nM ( A and B ) or 900 nM ( C ) and those of S1 were 20 nM ( A and B ) or 150 nM ( C ) . Each panel consists of four sequential snapshots ( top ) , a figure showing the heights of the indicated peaks ( middle ) , and the half helical pitches between the indicated peaks ( bottom ) . Yellow arrowheads show the transient association of S1 . Note that P1 in ( A ) and P3 in ( B ) rose in two substeps . Bars: 25 nm; Z-scale: 0–12 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 01410 . 7554/eLife . 04806 . 015Video 3 . Growth of a cofilin cluster toward the pointed end of a filament in F buffer containing 1 mM ADP , 0 . 1 mM ATP , 20 nM S1 , and 75 nM cofilin . Imaged at 2 frames/s and played at 5 frames/s . White arrowheads show growth of the cofilin cluster , and yellow and magenta arrowheads show binding of S1 . Magenta arrowheads indicate S1 molecules whose binding angle could not be determined , either for geometric reasons ( i . e . , binding on the upper face of the filament ) or because the binding was too short-lived . Z-scale was 0–12 nm . For magnifications and polarity of the analyzed filaments , refer to Figure 4 in the main text . Related to Figure 4A . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 015 Cofilin binds preferentially to actin subunits carrying ADP , compared to those carrying ATP or ADP and Pi ( Carlier et al . , 1997; Blanchoin and Pollard , 1999 ) . Since actin filaments in F buffer containing ATP should have ATP–actin caps near the barbed ends , and since subunits on the pointed-end side tend to carry ADP only due to ATP hydrolysis and Pi release ( Carlier , 1990 ) , the directional growth of cofilin clusters toward the pointed-end might reflect the asymmetric distribution of actin subunits with different nucleotides . To test this possibility , we prepared two different types of actin filaments with homogenous nucleotide states along the lengths . In the first case , filaments polymerized in buffer containing 1 mM ATP were incubated for 30 min in buffer containing 1 mM ADP , which is much longer than the 350 s required to hydrolyze ATP and release the resultant Pi from half of the polymerizing ATP–actin molecules ( Melki et al . , 1996 ) , so that most of the actin subunits should carry ADP only . In the second case , filaments polymerized in buffer containing 1 mM ATP were incubated for 10 min in buffer containing 1 mM ADP and 10 mM Pi . Under this condition , most of the actin subunits should carry ADP and Pi , considering a Kd of 1 . 5 mM for Pi ( Carlier and Pantaloni , 1988 ) . HS-AFM observations after the addition of 20 nM S1 and 75 nM cofilin or 150 nM S1 and 900 nM cofilin demonstrated that , in both cases , the growth of cofilin clusters was primarily to the pointed-end direction ( Figure 4B and Video 4 , and Figure 4C and Video 5 ) . Results of a large number of observations are compiled in Figure 5 . 10 . 7554/eLife . 04806 . 016Video 4 . Growth of a cofilin cluster toward the pointed end of a filament in F buffer containing 1 mM ADP , 20 nM S1 , and 75 nM cofilin . Imaged at 2 frames/s and played at 5 frames/s . For color codes of arrowheads , see the legend to Video 3 . Z-scale was 0–12 nm . For magnifications and polarity of the analyzed filaments , refer to Figure 4 in the main text . Related to Figure 4B . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 01610 . 7554/eLife . 04806 . 017Video 5 . Growth of a cofilin cluster toward the pointed end of a filament in F buffer containing 1 mM ADP , 10 mM Pi , 150 nM S1 , and 900 nM cofilin ( without His-tag ) . For color codes of arrowheads , see the legend to Video 3 . For magnifications and polarity of the analyzed filaments , refer to Figure 4 in the main text . Under this condition , binding of S1 was so short-lived that the tilted binding was not obvious in some cases ( magenta ) . The S1 molecule indicated by a double magenta arrowhead appears to tilt in the direction opposite to other S1 molecules indicated by yellow arrowhead , and we speculate that this is because this S1 molecule was not stably bound to the filament when imaged . Imaged at 2 frames/s and played at 3 frames/s . Z-scale was 0–12 nm . Related to Figure 4C . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 01710 . 7554/eLife . 04806 . 018Figure 5 . Directional preference of the growth of cofilin clusters . The growth of cofilin clusters was observed under three buffer conditions: in the presence of 1 mM ADP and 0 . 1 mM ATP ( +ADP +ATP ) ; 1 mM ADP ( +ADP ) and 1 mM ADP and 10 mM Pi ( +ADP +Pi ) , as in Figure 4 . Growth of a cluster by one-half helix was counted as one growth event . The total number of observed growth events was 37 , 46 , and 188 for each condition . We speculate that at least some of the cluster growth events in the barbed-end direction were actually growth in the preferred direction from invisibly small clusters on the barbed side of a visible cluster . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 018 The rate of formation and growth of cofilin clusters depended on the nucleotide state of actin subunits as well as the concentration of cofilin . In most experiments we used 75 nM cofilin since we were able to observe de novo formation of clusters and tractable growth , after mixing and settling of the system . When 1 µM cofilin was added and mixed , we were unable to find bare zones of actin filaments . When actin filaments were incubated with 1 mM ADP and 10 mM Pi , much higher concentrations of cofilin ( e . g . , 900 nM ) was needed to induce cofilin clusters . The growth rates of cofilin clusters were within tractable range even in the presence of 1 µM cofilin , when 10 mM Pi was present ( Figure 4C ) . These results are consistent with the weaker affinity of cofilin for actin subunits carrying ADP and Pi ( Carlier et al . , 1997; Blanchoin and Pollard , 1999 ) . Cofilin is an actin filament severing protein ( reviewed by Pollard ( 2000 ) ) , and although our experiments were performed at pH 6 . 8 , which should suppress severing activity ( Yonezawa et al . , 1985; Hawkins et al . , 1993; Pavlov et al . , 2006 ) , we observed frequent severing of the filaments near or within short cofilin clusters ( Figure 6A–C , Videos 6–10 and Figure 6—figure supplements 1 and 2 ) . Because a previous study by Adrianantoandro and Pollard ( 2006 ) found that severing activity is highest at 10 nM of human cofilin and that this activity sharply declines above and below 10 nM , we observed severing events at several different cofilin concentrations , including 10 nM . In the presence of 10 or 20 nM cofilin , severing was very infrequent ( Video 11 and Figure 6—figure supplement 3 ) . Severing was often observed between 40 ( Video 12 and Figure 6—figure supplement 4 ) and 200 nM cofilin . In the presence of 650 nM cofilin , many of the filaments were fully decorated with cofilin when observations started , and severing was observed exclusively near the ends of the long clusters ( Video 12 and Figure 6—figure supplement 4 ) . When the cofilin concentration was increased to 1 µM , all the filaments were fully decorated with cofilin , and severing was infrequent along those fully decorated filaments ( Video 12 ) . Thus , we also observed that cofilin's severing activity was highest in the medium concentration range ( i . e . , 40–200 nM ) , but this concentration is at least several-fold higher than that reported by Adrianantoandro and Pollard ( 2006 ) . These two studies used the same pair of proteins ( i . e . , skeletal muscle actin and human cofilin ) , and we can only speculate that this discrepancy is due to differences in experimental conditions . 10 . 7554/eLife . 04806 . 019Figure 6 . Severing of actin filaments near cofilin clusters . ( A–C ) Typical cases of filament severing ( red arrowheads ) within or near cofilin clusters ( white arrowheads ) . The observation buffers were F-buffer containing 1 mM ADP ( A and C ) and 1 mM ATP ( B ) . Concentration of cofilin was 40 nM ( A and C ) and 75 nM . The first break in ( A ) was inside a cluster , while the second was at or near the junction between a bare zone and a cluster . Red , blue , and green arrowheads show severing points within cofilin clusters , in bare zone close to cofilin clusters , and in bare zones more than half a helix away from a cofilin cluster , while white arrowheads show cofilin clusters . Bars: 25 nm; Z-scale: 0–12 nm . See Videos 6–8 . ( D ) Classification of severing sites into four categories: ( 1 ) in ‘far’ bare zone ( between a tall and a short black arrow or between two short black arrows , indicated by green bars in ( E ) ) ; ( 2 ) in bare zone half helices immediately neighboring a cofilin cluster ( between a tall black and a tall orange arrow , indicated by blue bars in ( E ) ) ; ( 3 ) in ‘end’ cofilin cluster half helices immediately neighboring bare zones ( between a tall and a short orange arrow , indicated by red bars in ( E ) ) ; and ( 4 ) in ‘inner’ cofilin cluster half helices ( between two short orange arrows , indicated by a red bar in ( E ) ) . Comparison of the last two categories demonstrates that severing within cofilin clusters occurs preferentially near the ends . Note , however , that this comparison does not necessarily show a quantitative difference in the susceptibility to severing between end and inner half helices , since the number of end and inner helices examined are not the same . ( E ) A schematic summary of the proposed distributions of bound cofilin molecules ( red spheres ) , segments of normal ( yellow ) and shortened ( orange ) helical pitch , and normal ( black arrows ) and tall ( orange arrows ) crossover points . Free cofilin molecules tend to bind to the supertwisted bare zone on the pointed-end side of the cluster ( gray arrows ) , driving the growth of the cluster in the pointed-end direction . This is most certainly an oversimplification , ignoring a number of complex issues , some of which are discussed in the main text . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 01910 . 7554/eLife . 04806 . 020Figure 6—figure supplement 1 . Representative still images from Video 9 , demonstrating severing of actin filaments by cofilin with His-tag . Red , blue , and green arrowheads indicate severing in half helices in cofilin clusters , in a bare half helix immediately neighboring a cofilin cluster , and in bare zones more than half a helix away from cofilin clusters , respectively . White arrowheads: cofilin clusters . Numbers ( 1–10 ) refer to the sequence number in the video . Z-scale: 0–12 nm . Scale bar: 25 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 02010 . 7554/eLife . 04806 . 021Figure 6—figure supplement 2 . Representative still images from Video 10 , showing severing of actin filaments by cofilin without His-tag . Red , blue , and green arrowheads indicate severing in half helices in cofilin clusters , in a bare half helix immediately neighboring a cofilin cluster , and in bare zones more than half a helix away from cofilin clusters , respectively . White arrowheads: cofilin clusters . Numbers ( 1–7 ) refer to the sequence number in the video . Z-scale: 0–12 nm . Scale bar: 25 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 02110 . 7554/eLife . 04806 . 022Figure 6—figure supplement 3 . Representative still images from Video 11 , showing severing of actin filaments in the presence and absence of low concentration of cofilin . Blue arrowheads: severing in a bare half helix immediately neighboring a cofilin cluster; green arrowheads: severing in bare zones more than half a helix away from cofilin clusters; white arrowheads: cofilin clusters . Z-scale: 0–12 nm . Scale bar: 25 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 02210 . 7554/eLife . 04806 . 023Figure 6—figure supplement 4 . Representative still images from Video 12 , showing severing in actin filaments decorated with high concentrations of cofilin . Red arrowheads: severing in half helices in cofilin clusters; blue arrowheads: severing in a bare half helix immediately neighboring a cofilin cluster; white arrowheads: cofilin clusters . Z-scale: 0–12 nm . Scale bar: 25 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 02310 . 7554/eLife . 04806 . 024Video 6 . Severing of actin filaments in a cofilin cluster . Conditions: F buffer containing 1 mM ADP and 40 nM cofilin . Severing of actin filaments occurred within a cofilin cluster ( at 53 s ) and then at or near the boundary between a bare zone and another cofilin cluster ( at 64 . 5 s ) . Imaged at 2 frames/s and played at 5 frames/s . Red , blue , and green arrowheads indicate severing in half helices in cofilin clusters , in a bare half helix immediately neighboring a cofilin cluster , and in bare zones more than half a helix away from cofilin clusters , respectively . White arrowheads: cofilin clusters . Z-scale was 0–12 nm . Related to Figure 6A . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 02410 . 7554/eLife . 04806 . 025Video 7 . Severing of actin filaments at or near a boundary between a bare zone and a cofilin cluster . Conditions: F buffer containing 1 mM ATP and 75 nM cofilin . Severing occurred at 42 s . Imaged at 2 frames/s and played at 5 frames/s . For color codes of the arrowheads , see the legend to Video 6 . Z-scale was 0–12 nm . Related to Figure 6B . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 02510 . 7554/eLife . 04806 . 026Video 8 . Severing of actin filaments in a bare zone more than one half helix away from a cofilin cluster . Conditions: F buffer containing 1 mM ADP and 40 nM cofilin . Severing occurred at 28 s . Imaged at 2 frames/s and played at 5 frames/s . For color codes of the arrowheads , see the legend to Video 6 . Z-scale was 0–12 nm . Related to Figure 6C . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 02610 . 7554/eLife . 04806 . 027Video 9 . Severing of actin filaments by cofilin . To show more general view of severing events , in addition to the small number of representative cases shown in Videos 6–8 , 10 different image sequences from different experiments were merged . Sequence numbers are shown in the first 10 frames of each sequence . For color codes of the arrowheads , see the legend to Video 6 . Conditions: F buffer containing 1 mM ATP ( sequences 1–6 ) , 1 mM ADP ( sequence 7–8 ) or 1 mM ATP + 10 mM Pi ( sequence 9–10 ) . The concentration of cofilin shown in this video was 75 nM , except in sequences 9 and 10 , in which it was 300 and 150 nM , respectively . Z-scale was 0–12 nm . Related to Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 02710 . 7554/eLife . 04806 . 028Video 10 . Severing of actin filaments by cofilin without His-tag . In this video , seven different image sequences from different filaments and experiments were merged . Conditions: F buffer containing 1 mM ATP and 75 nM cofilin , except in sequence 3 in which cofilin concentration was 150 nM . For color codes of the arrowheads , see the legend to Video 6 . Z-scale was 0–12 nm . Related to Figure 6—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 02810 . 7554/eLife . 04806 . 029Video 11 . Severing of actin filaments in the absence or presence of low concentrations of cofilin . Four independent image sequences are merged . Conditions: F buffer containing 1 mM ATP and 0 , 10 or 40 nM cofilin without His-tag . In this video , three data sets which represent three cases of the absence or presence of cofilin are sequentially shown and indicated before each sequence begins as ( i ) Control: Without Cofilin , ( ii ) 10 nM Cofilin ( two different image sequences ) , and ( iii ) 40 nM Cofilin . Note that severing of actin filaments was not observed not only in the absence but also in the presence of 10 nM cofilin . In the presence of 40 nM cofilin , severing was infrequently observed . For color codes of the arrowheads , see the legend to Video 6 . Images were taken at 0 . 5 frames/s , except in the presence of 40 nM cofilin they were recorded at 0 . 25 frames/s , and the video is played at 5 frames/s . Z-scale was 0–12 nm . Related to Figure 6—figure supplement 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 02910 . 7554/eLife . 04806 . 030Video 12 . Actin filaments decorated with high concentrations of cofilin . The first three quarter of this video were taken in F buffer containing 1 mM ATP and 1 µM cofilin , and the last one quarter was taken in the presence of 650 nM cofilin . In the presence of 1 µM cofilin , filaments were fully decorated along the length , and no severing was observed . In the presence of 650 nM cofilin , there were some bare zones , and severing occurred near the boundary of the bare zone and the cofilin clusters , regardless of the size of cofilin clusters . For color codes of the arrowheads , see the legend to Video 6 . Images were taken at 2 frames/s and played at 5 frames/s . Z-scale was 0–12 nm . Related to Figure 6—figure supplement 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 04806 . 030 We next quantitatively analyzed the severing positions of the filaments relative to the short cofilin clusters in the presence of 40–200 nM cofilin ( Figure 6D ) . Approximately 60% of severing events occurred inside the cofilin clusters , mostly within half helices neighboring bare zones . Severing also occurred in half helices in bare zones neighboring a cofilin cluster , which accounted for approximately 40% of severing events . Overall , ∼80% of the severing events occurred within one-half of a helix on either side of the boundary between a bare zone and a cofilin cluster . Severing in bare zones far from cofilin clusters was very rare . In the presence of 40 nM cofilin , the lowest concentration we used for quantitative severing assay , we observed a total of 22 cases of severing events in 31 actin filaments . Among those , 18 cases occurred in half helices immediately neighboring the boundary between a bare zone and a cofilin cluster , even though cluster formation was relatively rare at this concentration of cofilin . Three cases occurred in ‘far’ bare zones more than half a helix away from the boundary , and one occurred in an ‘inner’ cofilin cluster more than half a helix away from the boundary ( Video 11 ) . Based on these observations , we conclude that severing under the present experimental condition preferentially occurs near the boundary between a bare zone and a cofilin cluster , but not necessarily at the boundary . Considering the difference in spatial resolution , this conclusion is also consistent with that made by Suarez et al . ( 2011 ) , who used fluorescence microscopy and concluded that severing occurs at the boundary between cofilin clusters and bare zones .
The short helical pitch in the cofilin clusters was propagated to the immediately neighboring bare zone on the pointed-end side of the cluster . Since our current analysis method only measures heights and distances between peaks , we were unable to examine if the supertwisted structure returns to a normal state abruptly or gradually . Also , we were unable to determine the positions of the ends of cofilin clusters precisely above the spatial resolution of half helices , making it difficult to estimate accurately how far the supertwisting conformational changes propagate into neighboring bare zones . Nonetheless , the fact that the pitches of the neighboring half helices second from the cofilin clusters were nearly normal suggests that the effect does not propagate longer than one-half of a helix . The neighboring bare zone on the barbed-end side of the cluster had slightly longer helical pitch than the control . Although a t-test indicated that the untwisting was statistically significant , we are not certain if it was an active process . If interactions between actin filaments and the lipid surface produced resistive force against local rotation of the filament , supertwisting in the cluster would induce passive , compensating untwisting of the nearby helix . If that is the case , the fact that actin filaments are not necessarily free to rotate in vivo suggests similar passive untwisting may occur in vivo as well . In any case , it is noteworthy that cofilin clusters induced distinctly asymmetric cooperative conformational changes in neighboring bare zones . Propagation of the supertwisted conformation from cofilin clusters to neighbor bare zones had been suggested by image analysis of electron micrographs of cofilin–actin complexes ( Galkin et al . , 2001 ) . Our observation that the cofilin clusters grew in the pointed-end direction supports the conclusion of Galkin et al . ( 2001 ) that cofilin preferentially binds to the supertwisted segments of actin subunits . This tendency of cofilin clusters to grow unidirectionally was unaffected when most of the filament subunits carried ADP or ADP+Pi , demonstrating that the directional growth of cofilin clusters does not depend on a gradient of different nucleotide states on actin subunits within each filament . Instead , it presumably depends on asymmetric cofilin-induced conformational changes in the actin filaments , which is an intrinsic property of the polar structure of the actin filaments . The rise of peaks and shortening of half helical pitches that accompanied the growth of cofilin clusters were sometimes rapid , but at other times slow and gradual ( Figure 4 ) . If we assume that the rise was due to simple addition of cofilin molecules at the two binding sites closest to the crossover point , and if the cofilin clusters grow in perfect synchrony along two strands of the double helix , then the crossover points should rise abruptly in one step . If , on the other hand , there is a time-lag between cofilin bindings at the two critical binding sites , due to delayed growth along one strand , for example , the rise would occur in two smaller discrete substeps . Indeed , in many cases , the rise of peaks appeared to occur in two discrete substeps ( Figure 4 ) , supporting this possibility . In other cases , however , the peaks rose gradually over the course of 40 s ( e . g . , P2 in Figure 4A ) , although we have no mechanistic explanation for those events . In certain cases the next peak ( P1 in the case shown in Figure 4A ) only rose when the previous peak ( P2 ) had reached a plateau , whereas in other cases the new peak ( P2 ) started to rise while the previous one ( P3 ) was still rising . This latter case is more consistent with the possibility that cofilin clusters do not necessarily grow in concert along the two strands , and the delay may sometimes exceed one-half of a helix . Another uncertainty is whether the new cofilin molecule always binds to the vacant binding site immediately neighboring the cluster . If the unidirectional cluster growth to the pointed-end is driven by the shortened helical pitch on the pointed-end side of the cluster , the new molecule is likely to bind any of the vacant binding sites in the supertwisted , apparently , bare zone on the pointed-end side of the cluster . This view is consistent with the result of recent high-resolution single molecule fluorescence microscopic assays that showed that free cofilin molecules prefer to bind within 65 nm of an already bound cofilin molecule , but not necessarily to the site immediately neighboring the already bound molecule ( Hayakawa et al . , 2014 ) . We thus speculate that there are vacant binding sites behind the advancing front of cofilin clusters , which are eventually filled by other cofilin molecules to form tight clusters , and that the advances are not necessarily in concert between the two strands of the double helix . In the presence of cofilin , actin filaments are frequently severed at or near the boundary between cofilin clusters and bare zones , at least when examined at the spatial resolution of fluorescence microscopy ( Suarez et al . , 2011 ) . That observation is consistent with the idea that severing preferentially occurs at sites of structural discontinuity , such as the boundary between supertwisted and normal helical pitches ( Michelot et al . , 2007; De La Cruz , 2009 ) . An alternative view is that cofilin binding weakens longitudinal contacts between actin subunits within cofilin clusters and also in neighboring bare zones , but because bound cofilin bridges two actin subunits , strengthening interactions between them , severing occurs in nearby supertwisted bare zones ( Galkin et al . , 2001; Bobkov et al . , 2002 ) . Roughly 40% of the severing events in our AFM observations occurred at the boundary between a cluster and a bare zone or within the neighboring bare zone , which may be explained by these hypotheses . However , more than half of the severing events occurred within the cofilin clusters , albeit close to the cluster ends . One plausible explanation is that the scattered unoccupied cofilin binding sites near the ends of the cluster , discussed above , may be easy to break due to the lack of bridging cofilin molecules . Alternatively , if growth of cofilin clusters on one of the two filament strands lags the growth on the other , conformational stress may develop between the two strands , leading to breaks in the filament near the ends of cofilin clusters . Previous biochemical ( Andrianantoandro and Pollard , 2006 ) and simulation studies ( De La Cruz , 2005; De La Cruz , 2009 ) suggested that one or a few bound cofilin molecules are sufficient to sever actin filaments . However , our AFM observations are more consistent with the view that efficient severing requires a cofilin cluster longer than one-half of a helix , since most of the severing events were observed within or very near cofilin clusters that were recognizable in our AFM images . The need of contiguously bound cofilin molecules for efficient severing is consistent with our earlier mutant analysis . Within filaments in solution , D11Q mutant actin subunits rapidly exchange bound ADP for ATP in solution . Consequently , most subunits within D11Q actin filaments have ATP bound , reducing their affinity for cofilin and protecting them from severing . Interestingly , cofilin was able to bind to copolymers of D11Q and wild-type actins , but severing was inefficient ( Umeki et al . , 2012 ) , suggesting that contiguous clusters of cofilin-bound actin subunits are necessary for efficient severing activity . A related question is how many bound cofilin molecules are required to induce cooperative supertwisting conformational changes in actin filaments ? Answering this question is technically difficult when the imaging power is not sufficient to see individual bound cofilin molecules , but our observations using a cofilin-rod fusion protein showed that one is not enough . Consistent with this view , previous simulation studies suggested that two cofilin molecules bound close to one another along an actin filament serve as a nucleus to initiate cooperative binding of cofilin to form a cluster ( Ressad et al . , 1998; Blanchoin and Pollard , 1999 ) . Future high-resolution HS-AFM studies using cofilin fused with a rod or some other structural marker to make it visible with AFM will directly test those hypotheses . Even if a singly bound cofilin molecule cannot induce supertwisting of the helix or initiate cluster growth , it does not exclude the possibility that single bound cofilin molecules induce subtler , perhaps longer range , cooperative conformational changes , such as those indirectly detected through biophysical measurements ( Dedova et al . , 2004; Prochniewicz et al . , 2005; Bobkov et al . , 2006 ) . Apparently cofilin is able to induce at least two distinct types of cooperative conformational changes: one that requires cluster formation , involves ∼25% supertwisting of the helix , and propagates over half a helix toward the pointed-end of the filament; and a second type that singly bound cofilin molecules can induce which involves relatively subtle conformational changes , and propagates much longer . Simulation studies by De La Cruz and Sept ( 2010 ) suggest that there are two distinct states of cofilin–actin complexes , which may be correlated with the two types of cooperative conformational changes we propose here . Cooperativity in the binding of cofilin to actin filaments could have multiple physiological implications . Generally speaking , cooperativity would amplify small changes in the input ( concentration of active cofilin ) to a larger difference in output ( cluster formation and severing ) . In addition , the cellular concentration of cofilin is lower than that of polymerized actin , and cooperativity would be a useful means of disrupting selected filaments under those conditions ( Pollard et al . , 2000 ) . Second , we propose that the propagation of cofilin-induced conformational changes into neighboring cofilin-unbound zones of actin filaments would give cofilin an advantage in competition with other ABPs , once a small cofilin cluster is established as a foothold . For example , cofilin is implicated in severing and depolymerization of aged actin filaments in lamellipodia . However , those actin filaments are often bound with tropomyosin ( Gunning et al . , 2008 ) , which inhibits binding of cofilin and protects the filaments from cofilin's severing and depolymerizing activities ( Bernstein and Bamburg , 1982; Ono and Ono , 2002 ) . If cofilin forms a small cluster at a vacant site on an actin filament that is otherwise decorated and protected by tropomyosin , the cluster would induce , or apply conformational stress to induce , supertwisting on the neighboring tropomyosin-bound segment on the pointed-end side . This would accelerate dissociation of tropomyosin , resulting in faster growth of the cofilin cluster than when cluster growth needed to wait for spontaneous dissociation of neighboring tropomyosin molecules . Furthermore , that cluster growth is directed only to the pointed-end may be beneficial in selective disassembly of aged actin filaments . Within cells , formin remains bound to the barbed-end when it catalyzes filament elongation . This causes relative rotation between the formin molecule and the filament ( Mizuno et al . , 2011 ) . Thus , if the formin molecule and the filament are not free to rotate , the filament will be untwisted , and if cofilin has a lower affinity for untwisted actin filaments , rapidly polymerizing actin filaments in a formin-dependent manner will be protected from severing by cofilin ( Mizuno and Watanabe , 2012 ) . It should be noted , however , that formin bound to the barbed-end of an actin filament also allosterically changes the structure of the filament ( Bugyi et al . , 2006 ) , which may interfere with the interactions of the filament with cofilin , independent of mechanically forced untwisting of the filament helix . Sharma et al . ( 2012 ) discovered that the untwisted conformation of actin filaments induced by a drebrin N-terminal fragment propagates to neighboring bare zones; however , their findings seem to indicate that this propagation is in both directions from the drebrin clusters , though the authors did not address that point . It thus appears that there are multiple forms of ABP-induced cooperative conformational changes to actin filaments that propagate into bare zones , which implies there are also multiple physiological functions for such cooperative conformational changes to actin filaments .
cDNA encoding human cofilin 1 was amplified from a human cDNA library using PCR with primers 5′-ggtaccatggcctccggtgt and 5′-tctagacaaaggcttgccctcca . After confirmation of its sequence , the amplified DNA fragment was subcloned into pColdI expression vector ( Takara Bio , Otsu , Japan ) at the XbaI and KpnI sites . The pColdI vector had been modified to contain a TEV cleavage site between the His tag and the multi-cloning sites , so that the amino acid sequence near the N-terminus was MNHKVHHHHHHIEGRHMENLYFQGTMASGVAVS… ( italics indicate the TEV cleavage site ) . The cofilin-rod fusion gene was constructed by cloning the cDNA encoding the first half of the Dictyostelium α-actinin rod downstream of the cofilin gene in pColdITEV . The amino acid sequence at the junction of the two proteins was …SAVISLEGKPLEQTKSDYLKRA… , and the C terminal sequence was …QKIEDSLVSR ( italics show extra amino acid residues derived from recognition sites for restriction enzymes ) . The first half of the Dictyostelium α-actinin rod , corresponding to amino acid residues 265–505 of the parent molecule , forms a monomeric12-nm long rod-like structure ( Yan et al . , 1993 ) and has been used as an artificial lever arm of myosin motors ( Anson et al . , 1996 ) . The proteins were expressed in Escherichia coli , purified using Ni-NTA affinity chromatography , and dialyzed against a buffer containing 10 mM HEPES , pH 7 . 4 , 50 mM KCl , 0 . 1 mM DTT , and 0 . 01% NaN3 overnight at 4°C . After concentrating with a centrifugal concentrator ( Amicon Ultra 4 ) , aliquots were snap-frozen in liquid nitrogen and stored at −80°C . Unless otherwise stated , the experiments used those His-tagged cofilin or cofilin-rod proteins . However , we repeated some key experiments after removing the His-tag by treatments with TEV protease and obtained qualitatively similar results . These experiments included asymmetric conformational changes of actin filaments on either side of a cofilin cluster , unidirectional growth of cofilin clusters in the pointed-end direction , and frequent severing of filaments near the boundary between a bare zone and a cofilin cluster . Cofilin with and without His-tag bound to actin filaments at a 1:1 molar ratio with a similar affinity ( Figure 3—figure supplements 1 and 2 ) . Cofilin-rod with and without His-tag also did not sediment on its own , then bound to actin filaments with affinities similar to cofilin without the rod fusion ( Figure 3—figure supplement 3 ) . Rabbit skeletal muscle actin and chymotryptic subfragment-1 , S1 , were purified as described previously ( Spudich and Watt , 1971; Margossian and Lowey , 1982 ) and stored in liquid nitrogen . Some experiments used G-actin that was further purified by gel filtration column chromatography , yielding identical results . Before use , an aliquot of frozen stock was thawed for 1 hr on ice and clarified by ultracentrifugation at 80 , 000 rpm for 5 min at 5°C . Protein concentration was then measured using an Advanced Protein Assay ( Cytoskeleton , Denver , CO ) , using calibrated skeletal actin as the standard . We used a laboratory built high-speed atomic force microscope ( HS-AFM ) as described previously ( Ando et al . , 2013 ) . HS-AFM imaging was carried out in the tapping mode with small cantilevers ( BL-AC10DS-A2 , Olympus , Tokyo , Japan ) whose spring constant , resonant frequency in water , and quality factor in water were ∼0 . 10 N/m , ∼400 kHz , and ∼2 , respectively . The probe tip was grown on the original tip end of a cantilever through electron beam deposition and was further sharpened using a radio frequency plasma etcher ( PE-2000 , South Bay Technology , Redondo Beach , CA ) under an argon gas atmosphere ( typically at 180 mTorr and 15 W for 3 min ) . During HS-AFM imaging , the free-oscillation peak-to-peak amplitude of the cantilever ( A0 ) was set to ∼2 nm , and the feedback amplitude set point was set at more than 0 . 9A0 . Details of the method for HS-AFM imaging are described elsewhere ( Uchihashi et al . , 2012 ) . We prepared small unilamellar vesicles ( SUVs ) and mica-supported lipid bilayer as described previously ( Yamamoto et al . , 2010; Uchihashi et al . , 2012 ) . The typical lipid composition was 1 , 2-dipamitoyl-sn-glycero-3-phosphocholine ( DPPC ) and 1 , 2-dipalmitoyl-3-trimethylammonium-propane ( DPTAP ) at a weight ratio of 9:1 . The lipids were purchased from Avanti Polar Lipids ( Alabaster , AL ) . SUVs were dispersed in Milli-Q water at 2 mg/ml and stocked at −20°C . Before use , the SUVs were diluted in 5 mM MgCl2 to 0 . 5 mg/ml and sonicated with a bath sonicator ( AUC-06L , AS ONE , Osaka , Japan ) for 1 min . An aliquot of the sonicated SUVs was deposited on the surface of freshly cleaved mica , which had been glued onto a sample stage beforehand , and incubated for more than 3 hr at room temperature ( 24–26°C ) in a humid sealed container to avoid surface drying . Up to 10 sample stages were prepared simultaneously and stored in the sealed container . Before deposition of actin filaments , the surface of the sample stage was rinsed with a large amount of Milli-Q water ( ∼20 µl × five times ) to remove excess SUVs and lipid bilayers . Actin filaments were then deposited onto the lipid bilayer using one of the following methods . G-actin ( 5–10 µM ) was polymerized in F buffer ( 40 mM KCl , 20 mM PIPES–KOH , pH 6 . 8 , 1 mM MgCl2 , 0 . 5 mM EGTA , 0 . 5 mM DTT ) containing 1 mM ATP for 30 min on ice . The resultant actin filaments were diluted to 0 . 5–1 . 0 μM in F buffer containing 1 mM ATP . Water on the lipid bilayer on a sample stage was replaced with 1–2 µl of F buffer containing 1 mM ATP , to which 2 µl of the diluted actin solution was added . After 5–10 min of incubation at room temperature , unattached actin filaments were removed by exchanging the solution with 50 µl of one of the four different F buffer-based observation buffers each containing ( i ) 1 mM ATP , ( ii ) 0 . 1 mM ATP and 1 mM ADP , ( iii ) 1 mM ADP , 5 U/ml hexokinase , and 10 mM glucose , and ( iv ) 1 mM ADP and 10 mM Pi . In the case ( iii ) , incubation was continued for 30 min to ensure that all actin subunits within the filaments carried ADP . Alternatively , G-actin ( 20 µM ) was polymerized in F buffer containing 1 mM ATP and 30 mM MgCl2 for 1 hr at room temperature . After introduction of this solution to the sample stage and 10 min of incubation at room temperature , unattached actin filaments were removed by gently exchanging the solution with F buffer containing 1 mM ATP and 30 mM MgCl2 . Finally , the sample stage was mounted to the z-scanner of a HS-AFM apparatus and immersed in a liquid cell containing the same observation buffer used in the last step , and HS-AFM imaging was performed . To follow binding of cofilin to actin filaments , 6 µl of cofilin diluted in the observation buffer was injected into the observation cell during AFM imaging . In some experiments , S1 was added in the observation buffer in the concentration of 20 nM ( in cases [i] , [ii] , and [iii] ) or 150 nM ( in case [iv] ) to identify the polarity of actin filaments . HS-AFM images were viewed and analyzed using the laboratory built software , Kodec4 . 4 . 7 . 39 . In brief , a low-pass filter to remove spike noise and a flattening filter to make the xy-plane flat were applied to individual images . The position and height of the peak within each half helix were determined semi-automatically using the following steps . First , the most probable highest point near a crossover point was selected manually . Second , the actual highest point was automatically determined by searching a 5 × 5 pixel area ( typically 7 . 5 × 7 . 5 nm2 ) around the selected point . Third , the peak position was refined based on a center of mass calculation using information on the heights and positions within the 5 × 5 pixel area around the selected point , after which the refined peak position and height were used to represent the peak of the half helix . The Kodec4 . 4 . 7 . 39 for HS-AFM image viewing and analysis software is coded in Visual C# ( Visual Studio 2010 , Microsoft , USA ) and is available as Source code 1 . All filters and subroutines for image analysis used in the present study are included in the software . We confirmed the compatibility between the software and computers operated with Windows 7 or 8 . Installer of the software , Kodec4_Setup . msi , is available in the subfolder of ‘Kodec 4 . 4 . 7 . 39\Setup\Release’ . This software should be cited as: Sakashita M , M Imai , N Kodera , D Maruyama , H Watanabe , Y Moriguchi , and T Ando . 2013 . Kodec4 . 4 . 7 . 39 . Actin filaments were prepared by polymerization of G-actin ( 20 µM ) in F buffer containing 40 mM KCl , 20 mM PIPES , pH 6 . 8 , 1 mM MgCl2 , 0 . 5 mM EGTA , 0 . 5 mM DTT , and 1 mM ATP for 30 min on ice . Cofilin or cofilin-rod was added to 1 µM actin filaments in 50 µl of F buffer at a molar ratio of 1:2 or 1:1 , and the solutions were mixed by gentle pipetting , after which the mixture was incubated for 3–5 min at room temperature . This mixture was then added to F buffer containing sodium phosphate ( pH 6 . 8 ) so that the concentration of actin was 0 . 5–1 µM and that of Pi was 5 mM , and a drop of this solution was immediately placed on a copper grid . The samples were fixed and negatively stained using a solution containing 1% uranyl acetate and 20 µg/ml bacitracin ( Katayama , 1989 ) . The fixed samples were dried under an incandescent lamp to form films over the holes of the grid . Electron microscopic data were then acquired using a Hitachi H-7650 transmission electron microscope . | Actin is one of the most abundant proteins found inside all eukaryotic cells including plant , animal , and fungal cells . This protein is involved in a wide range of biological processes that are essential for an organism's survival . Actin proteins form long filaments that help cells to maintain their shape and also provide the force required for cells to divide and/or move . Actin filaments are helical in shape and are made up of many actin subunits joined together . Actin filaments are changeable structures that continuously grow and shrink as new actin subunits are added to or removed from the ends of the filaments . One end of an actin filament grows faster than the other; the fast-growing end is known as the barbed-end , while the slow-growing end is referred to as the pointed-end . Over 100 other proteins are known to bind to and work with actin to regulate its roles in cells and how it forms into filaments . Cofilin is one such protein that binds to and forms clusters on actin filaments and it can also sever actin filaments . Severing an actin filament can encourage the filament to disassemble , or it can help produce new barbed ends that can then grow into new filaments . Previous work had suggested that cofilin severs actin filaments at the junction between regions on the filament that are coated with cofilin and those that are not . It was also known that cofilin binding to a filament causes the filament to change shape , and that the shape change is propagated to neighboring sections of the filaments not coated with cofilin . However , the details of where cofilin binds and how changes in shape are propagated along an actin filament were not known . Furthermore , the findings of these previous studies were largely based on examining still images of actin filaments , which are unlike the constantly changing filaments of living cells . Ngo , Kodera et al . have now analyzed what happens when cofilin binds to and forms clusters along actin filaments using a recently developed imaging technique called high-speed atomic force microscopy . This technique can be used to directly visualize individual proteins in action . Consistent with previous findings , Ngo , Kodera et al . observed that filaments coated with cofilin are thicker than those filaments without cofilin; and that cofilin binding also substantially reduces the helical twist of the filament . Ngo , Kodera et al . also found that these changes in shape are propagated along the filament but in only one direction—towards the pointed-end . Moreover , cofilin clusters also only grew towards the pointed-end of the actin filament—and the filaments were often severed near , but not exactly at , the junctions between cofilin-coated and uncoated regions . Such one-directional changes in shape of the actin filaments presumably help to regulate how other actin binding proteins can interact with the filament and consequently regulate the roles of the filaments themselves . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | Cofilin-induced unidirectional cooperative conformational changes in actin filaments revealed by high-speed atomic force microscopy |
Large social insect colonies exhibit a remarkable ability for recognizing group members via colony-specific cuticular pheromonal signatures . Previous work suggested that in some ant species , colony-specific pheromonal profiles are generated through a mechanism involving the transfer and homogenization of cuticular hydrocarbons ( CHCs ) across members of the colony . However , how colony-specific chemical profiles are generated in other social insect clades remains mostly unknown . Here we show that in the honey bee ( Apis mellifera ) , the colony-specific CHC profile completes its maturation in foragers via a sequence of stereotypic age-dependent quantitative and qualitative chemical transitions , which are driven by environmentally-sensitive intrinsic biosynthetic pathways . Therefore , the CHC profiles of individual honey bees are not likely produced through homogenization and transfer mechanisms , but instead mature in association with age-dependent division of labor . Furthermore , non-nestmate rejection behaviors seem to be contextually restricted to behavioral interactions between entering foragers and guards at the hive entrance .
The ability to recognize ‘self’ plays an important role in regulating diverse processes across biological organizational levels ( Tsutsui , 2004 ) . Analogous to the acquired immunity system , which depends on self-recognition at the cellular and molecular levels ( Boehm , 2006 ) , adaptive organismal social interactions often depend on the recognition of kin and/or group-members to increase cooperation or to suppress inbreeding ( Hamilton , 1964a; Hamilton , 1964b; Pusey and Wolf , 1996; Trivers , 1971; West et al . , 2007; Wilkinson , 1988 ) . One remarkable example of organismal recognition of ‘self’ comes from colonies of social insects , which depend on a robust non-nestmate discrimination system ( more commonly called ‘nestmate recognition’ ) to prevent the loss of expensive resources to non-nestmates , and to maintain overall colony integrity ( Hefetz , 2007; van Zweden and D’Ettorre , 2010 ) . As in other self-recognition systems , theoretical models suggest that nestmate recognition in social insect colonies depends on the ability of individual colony members to reliably match colony-specific phenotypic cues , or ‘labels’ , carried by other colony members , to stored neural ‘templates’ ( Buckle and Greenberg , 1981; Errard , 1994; Gamboa et al . , 1986; Getz , 1982; Hölldobler and Michener , 1980; Lacy and Sherman , 1983; Reeve , 1989; Tsutsui , 2004; van Zweden and D’Ettorre , 2010 ) . In some social insect species , the cues used in recognizing individual members of the colony have been reported to be visual ( Baracchi et al . , 2015 ) , but in most cases are thought to be chemical ( van Zweden and D’Ettorre , 2010 ) . Cuticular hydrocarbons ( CHCs ) , which evolved to function as hydrophobic , anti-desiccant barriers in terrestrial arthropods , have been co-opted to also function as pheromones in diverse insect communication systems , including nestmate recognition in social insect species ( Chung and Carroll , 2015; van Zweden and D’Ettorre , 2010 ) . Whether the overall profile , or more specific components of it , represent the actual nestmate recognition cue remains unknown . However , previous studies have indicated that variations in the relative amounts of each compound in the CHC profile across individuals from different colonies are likely sufficient for the chemical recognition of nest membership ( van Zweden and D’Ettorre , 2010 ) . Nevertheless , how large groups of hundreds to thousands of individuals coordinate the production and recognition of a robust colony-specific chemical cue remains unknown for most species . Because members of social insect colonies are often genetically related , it was initially assumed that the production of similar colony-specific pheromones by individual colony members is intrinsically driven by shared allelic variants ( Crozier and Dix , 1979; Getz , 1982; Getz , 1981 ) . However , empirical studies revealed that , surprisingly , in many social insect species colony and social environmental factors play the most dominant role in defining colony-specific cues , and can often mask genetic relatedness ( Breed et al . , 1988; Downs and Ratnieks , 1999; Heinze et al . , 1996; Lahav et al . , 2001; Liang and Silverman , 2000; Singer and Espelie , 1996; Stuart , 1988 ) . Although these colony ‘environmental’ factors remain unknown for most social insect species , it has been suggested that contributions from nest building materials ( Breed et al . , 1988; Couvillon et al . , 2007; D'ettorre et al . , 2006; Espelie et al . , 1990; Singer and Espelie , 1996 ) , the queen ( Carlin and Hölldobler , 1988; Carlin and Holldobler , 1987; Carlin and Holldobler , 1986; Carlin and Hölldobler , 1983 ) , and diet ( Buczkowski et al . , 2005; Buczkowski and Silverman , 2006; Liang and Silverman , 2000; Richard et al . , 2004; Richard et al . , 2007 ) could , at least in part , provide unique chemical components to the chemical signature shared by colony members . Consequently , empirical and theoretical studies suggested that individual colony members acquire their colony-specific chemical signature largely through a homogenization process involving the exchange of relevant chemicals , including CHCs , through interactions between colony members or contact with nest building materials , often referred to as the ‘Gestalt’ model ( Crozier and Dix , 1979 ) . Empirical evidence in support of this model has been reported for a few ant species , which are known to transfer mixed blends of CHCs between individuals through trophallaxis and grooming via the action of the postpharyngeal gland ( PPG ) ( Boulay et al . , 2000; Lenoir et al . , 2001; Meskali et al . , 1995; Soroker et al . , 1994; Soroker et al . , 1995b; van Zweden et al . , 2010 ) . However , other studies suggest that such CHC homogenization processes might not fully represent how colony-specific chemical cues develop in all social insect species . For example , some ant species do not display robust trophallaxis behaviors , the main mode of chemical transfer across colony members ( Soroker et al . , 1994; Soroker et al . , 1995b ) , and in others , the CHC profiles of individual colony members are likely modulated by genetic relatedness ( Teseo et al . , 2014 ) , age ( Cuvillier-Hot et al . , 2001; Teseo et al . , 2014 ) , and/or task ( Martin and Drijfhout , 2009; Sturgis and Gordon , 2013; Wagner et al . , 2001; Wagner et al . , 1998 ) . Together , these data suggest that the regulation of chemical cues in different species is more variable and complex than initially hypothesized ( Esponda et al . , 2015; Newey , 2011; Sturgis and Gordon , 2012 ) , and remains unknown for most social insect species . Consequently , here we investigated the development of CHC profiles and nestmate recognition cues in the European honey bee , Apis mellifera , a species of economic importance and one of the best studied social insect species . Numerous previous studies have demonstrated that honey bees exhibit a robust nestmate recognition system that is based on the chemical recognition of pheromones ( van Zweden and D’Ettorre , 2010 ) . Analyses of CHC profiles showed that newly emerged honey bee workers express significantly lower amounts of total CHCs and lower overall CHC chemical diversity in comparison to older foragers , which are expected to elicit the strongest nestmate recognition response from guards at the entrance to the hive ( Breed et al . , 2004; Kather et al . , 2011 ) . Additionally , other studies have suggested that honey bee nestmate recognition cues might be derived from various environmental sources ( Downs and Ratnieks , 1999 ) , and hive building materials such as the honeycomb wax ( Breed , 1998; Breed et al . , 1988; Couvillon et al . , 2007; D'ettorre et al . , 2006 ) . Based on these studies , it has been hypothesized that , similar to some ant species , the CHC profile of newly eclosed workers represents a ‘blank slate’ ( Breed et al . , 2004; Lenoir et al . , 1999 ) , and that nestmate recognition cues are subsequently acquired by individual workers primarily through the homogenization and transfer of chemicals via direct social interactions and intermediate environmental factors ( Breed et al . , 2015 ) . Furthermore , it has recently been proposed that the cephalic salivary gland of honey bee workers is functionally analogous to the PPG in ants , and could be involved in the homogenization and transfer of the CHCs between colony members ( Martin et al . , 2018 ) . However , when and how honey bee chemical nestmate recognition cues mature , and whether CHC homogenization mechanisms play a role in this process have not been directly investigated . Here , we provide empirical evidence that the maturation of the CHC profile of individual honey bee workers is primarily regulated by innate developmental processes associated with age-dependent behavioral tasks and modulated by the social colony environment , and that mature colony-specific recognition cues are primarily associated with the foraging task . Specifically , we find that individual workers exhibit stereotypic quantitative and qualitative changes in their CHC profile as they transition from in-hive tasks to foraging outside , that these changes are associated with innate transcriptional changes in CHC biosynthetic pathway genes , and that only forager honey bees are behaviorally rejected from the entrance of an unrelated hive . Together , our findings suggest that not all members of honey bee colonies display a uniform cuticular chemical profile via the direct acquisition of CHC mixes . Instead , our data indicate that CHC profiles , and likely nestmate recognition cues , in honey bees are more likely a product of a genetically-determined developmental program that is modulated by colony-specific factors .
Given that newly emerged honey bees have lower amounts of total CHCs , and exhibit less chemical diversity compared to older bees ( Breed et al . , 2004 ) , we initially sought to determine the age at which the CHC profile of individual honey bee workers matures . To achieve this goal , we analyzed the CHC profiles of individual workers from a single age-cohort that was reintroduced back into its source colony and then collected at different ages . This analysis revealed that the total amount of CHCs increases between one-day post-reintroduction and 14 days post-reintroduction and then remains stable ( Figure 1A , Kruskal-Wallis , H = 9 . 21 , df = 3 , p=0 . 026 , FDR pairwise contrasts: Day 1 vs . Day 7 p=0 . 11 , Day 1 vs . Day 14 p=0 . 036 , Day 1 vs . Day 21 p=0 . 04 , Day 7 vs . Day 14 = 0 . 613 , Day 7 vs . Day 21 p=0 . 691 , Day 14 vs . Day 21 p=0 . 79; Figure 1—figure supplement 1A , ANOVA , F ( 3 , 28 ) = 6 . 40 , p=0 . 002 , FDR pairwise contrasts: Day 1 vs . Day 7 p=0 . 036 , Day 1 vs . Day 14 p=0 . 001 , Day 1 vs . Day 18 p=0 . 007 , Day 7 vs . Day 14 = 0 . 412 , Day 7 vs . Day 18 p=0 . 993 , Day 14 vs . Day 18 p=0 . 305 ) . Additionally , individual compounds vary in total amount across bees of different ages ( Figure 1C , D , Figure 1—figure supplement 1C , D , Table 1 , Table 2 ) . Independently of the age-related quantitative changes , we also found that the CHC profiles of workers exhibit age-related qualitative changes in the overall CHC chemical composition ( Figure 1B , Permutation MANOVA , F ( 1 , 31 ) = 22 . 86 , R2 = 0 . 43 , p<0 . 001 , FDR pairwise contrasts: Day 1 vs . Day 7 p=0 . 002 , Day 1 vs . Day 14 p=0 . 002 , Day 1 vs . Day 21 p=0 . 002 , Day 7 vs . Day 14 = 0 . 017 , Day 7 vs . Day 21 p=0 . 002 , Day 14 vs . Day 21 p=0 . 31; Figure 1—figure supplement 1B , Permutation MANOVA , F ( 3 , 28 ) = 2 . 35 , R2 = 0 . 22 , p=0 . 038 , FDR pairwise contrasts: Day 1 vs . Day 7 p=0 . 024 , Day 1 vs . Day 14 p=0 . 011 , Day 1 vs . Day 18 p=0 . 018 , Day 7 vs . Day 14 = 0 . 406 , Day 7 vs . Day 18 p=0 . 212 , Day 14 vs . Day 18 p=0 . 524 ) , as well as in the relative amounts of individual CHCs ( Figure 1E , F , Figure 1—figure supplement 1E , F , Table 3 , Table 4 ) . These data confirm that not all members of a honey bee colony share a common CHC profile ( Kather et al . , 2011 ) , and suggest that age-dependent processes might be playing an important role in the regulation of both the quantitative and qualitative dimensions of the cuticular chemical profiles of individual honey bee workers . Honey bee workers exhibit age-related division of labor , which is characterized by a stereotypic sequence of in-hive behavioral tasks such as nursing and food handling , followed by the final transition to foraging outside the colony at about three weeks of age ( Robinson , 1992; Smith et al . , 2008; Søvik et al . , 2015 ) . Consequently , under natural colony settings , it is impossible to separate the possible independent impacts of ‘age’ and ‘task’ on the expression of forager-specific CHC profiles . Therefore , we next analyzed the CHC profiles of individual nurse and forager bees from single-cohort-colonies ( SCC ) , a well-established experimental approach to uncouple behavioral maturation from chronological age ( Ben-Shahar et al . , 2004; Ben-Shahar et al . , 2002; Greenberg et al . , 2012; Robinson et al . , 1989; Whitfield et al . , 2003 ) . Because these artificial colonies are initially comprised of a single age-cohort of day-old bees , a small proportion of these young workers will accelerate their behavioral maturation to become precocious foragers that are the same age as typical nurses ( ~7 days old ) ( Ben-Shahar et al . , 2002; Greenberg et al . , 2012; Huang and Robinson , 1992 ) . The comparison of the CHC profiles of typical young nurses and precocious foragers of identical age revealed a significant effect of task on the CHC profile of individual workers ( Figure 2A , Permutation MANOVA , F ( 1 , 15 ) = 13 . 79 , R2 = 0 . 50 , p<0 . 001 ) . Similarly , we observed a significant effect of task on the CHC profiles of individual ‘over-aged’ nurses and typical-aged foragers at three weeks of age ( Figure 2B , Permutation MANOVA , F ( 1 , 15 ) = 45 . 41 , R2 = 0 . 76 , p<0 . 001 ) . In contrast , task and age had no effect on total CHC amount ( Figure 2C , Two-way ANOVA , age: F ( 1 , 28 ) = 0 . 55 , p=0 . 46 , task: F ( 1 , 28 ) = 0 . 37 , p=0 . 55 , age*task: F ( 1 , 28 ) = 5 . 37 , p=0 . 03 ) . Together , these data suggest that processes associated with the behavioral maturation of honey bee workers , not chronological age , are primarily responsible for the observed forager versus nurse CHC profiles of individual honey bee workers . Previous studies in Harvester ants suggested that exposure to the environment outside the nest is sufficient to induce stereotypical changes in the CHC profiles of individual social insects ( Wagner et al . , 2001 ) . Therefore , we next asked whether spending time outside the hive is sufficient to induce the observed forager-specific CHC profile by comparing the CHC profiles between ‘undertakers’ , nurses , and foragers from typical colonies . ‘Undertakers’ are a small group of highly specialized older pre-foraging workers ( 2–3 weeks of age ) , which are responsible for removing dead bees by carrying them outside and away from the colony ( Robinson , 1992; Smith et al . , 2008; Søvik et al . , 2015; Trumbo et al . , 1997 ) . Therefore , because undertakers and foragers perform their respective tasks outside the hive , while nurses and other younger , pre-foraging bees rarely do , we reasoned that if outdoor exposure defines the distinct forager-specific CHC profile then the CHC profiles of undertakers should be more similar to foragers than to nurses . However , we found that the CHC profiles of undertakers are markedly different from those of foragers , and are more similar to those of nurses ( Figure 2D , Permutation MANOVA , F ( 2 , 23 ) =12 . 60 , R2 = 0 . 55 , p<0 . 001 , FDR pairwise contrasts: undertaker vs . forager p=0 . 003 , undertaker vs . nurse p=0 . 176 , forager vs . nurse p=0 . 003 ) . These data suggest that some outdoor exposure is not sufficient to drive forager-specific CHC profiles . We next asked whether the CHC profiles of foragers are a direct consequence of their behavioral state by using ‘big back colonies’ ( Ben-Shahar et al . , 2000; Withers et al . , 1995 ) , which allowed us to compare active foragers to bees of a similar age and behavioral state that are unable to forage outside ( see Materials and methods ) . We found that the overall CHC profiles of ‘big-back’ bees were different from those of their actively foraging sisters ( Figure 2E , Permutation MANOVA , F ( 1 , 14 ) =5 . 91 , R2 = 0 . 313 , p<0 . 001 ) . These data suggest that the physiological transition to foraging behaviors is not the sole factor that defines forager-specific CHC profiles , and that it could be modulated by additional factors associated with the act of foraging itself and/or extended exposure to various outdoor environmental factors . However , the fact that foraging nestmates express very similar CHC profiles , which are markedly different from those of non-nestmate foragers sharing a similar foraging environment ( Figure 2—figure supplement 1 , Permutation MANOVA , F ( 1 , 15 ) = 12 . 5 , R2 = 0 . 47 , p<0 . 001 ) suggests that forager-specific CHC profiles are not simply defined by the foraging environment . Additionally , to test whether extended exposure to outdoor environmental factors induces predictable changes in CHCs , we compared the relative amounts of individual compounds between forager bees and in-hive bees across our various experiments . We did not find a single compound that varied between foragers and in-hive bees in a consistent manner across our experiments ( e . g . always increases or always decreases in association with foraging activity ) ( Table 5 ) , indicating that CHCs do not change in a stereotypic manner in association with extended outdoor exposure , as they do in Harvester ants ( Wagner et al . , 2001 ) . Nevertheless , to further examine whether forager-specific CHC profiles were solely environmentally determined , we also analyzed the CHC profiles of typical-age foragers that were forced to revert to a nursing state ( Robinson et al . , 1992 ) . However , we did not find any differences between the CHC profiles of reverted nurses and active foragers ( Figure 2F ) . These data suggest that once foragers acquire their signature CHC profile , it remains stable independent of the task they perform and despite the typical short CHC half-life in insects ( Kent et al . , 2007 ) . Together , these data suggest that forager-specific CHC profiles are derived from a combination of factors associated with an innate behavioral maturation process , as well as being physically engaged in foraging activity . Previous work indicates that guard bees will accept foraging-age nestmates and reject foraging-age non-nestmates , independent of genetic relatedness ( Downs and Ratnieks , 1999 ) . This suggests that factors associated with the hive environment play a dominant role in specifying the colony-specific chemical signatures used for nestmate recognition . Yet , our data also indicate that CHC profile development in individual workers is a developmentally-regulated process that is closely associated with the age-dependent division of labor among workers . To address this potential conundrum , we next asked whether the effects of task and colony environment on the development of CHC profiles of individual workers are independent by using a reciprocal cross-fostering strategy . To achieve our goal , we introduced cohorts of newly eclosed bees from two different typical colonies back into their source colony , as well as a reciprocal foster colony , and then recollected marked workers from both cohorts in each reciprocal colony at different ages . CHC analyses revealed that through Day 14 , the CHC profiles of bees were more similar to the profiles of their same-aged non-nestmate sisters than those of unrelated nestmates of similar age ( Figure 3A , Two-way Permutation MANOVA , foster colony ( environment ) : F ( 1 , 31 ) = 2 . 19 , R2 = 0 . 06 , p=0 . 06 , source colony ( genetics ) : F ( 1 , 31 ) = 5 . 94 , R2 = 0 . 16 , p<0 . 001 , foster colony*source colony: F ( 1 , 31 ) =0 . 46 , R2 = 0 . 01 , p=0 . 82; Figure 3B , Two-way Permutation MANOVA , foster colony: F ( 1 , 31 ) = 1 . 13 , R2 = 0 . 03 , p=0 . 33 , source colony: F ( 1 , 31 ) = 3 . 18 , R2 = 0 . 09 , p=0 . 02 , foster colony*source colony: F ( 1 , 31 ) = 1 . 78 , R2 = 0 . 05 , p=0 . 15; sample size assessment depicted in Figure 3—figure supplement 1 indicates sample size is adequate ) . In contrast , once workers shift to foraging activity , we found that the CHC profiles of fostered bees are different from the profiles of both foraging sisters raised in the source colony and unrelated host foragers of similar age ( Figure 3C , Two-way Permutation MANOVA , foster colony: F ( 1 , 31 ) = 4 . 04 , R2 = 0 . 10 , p=0 . 02 , source colony: F ( 1 , 31 ) = 7 . 65 , R2 = 0 . 19 , p=0 . 001 , foster colony*source colony: F ( 1 , 31 ) =0 . 48 , R2 = 0 . 01 , p=0 . 67 ) . Together , these data suggest that genetic variations , or other long-term effects associated with the source colony , play an important role in defining the CHC profiles of individuals during the early phases of the age-dependent behavioral development of worker honey bees . However , by the time bees start foraging , the mature CHC profile of individual workers is defined by an interaction between factors associated with both the source and foster colonies ( Figure 3C ) . Homogenization models for the development of colony-specific nestmate recognition cues predict that cue specificity is acquired by individuals via physical contact with other colony members and/or environmental sources of hydrocarbons ( Breed et al . , 2015; Crozier and Dix , 1979; Lenoir et al . , 2001; Meskali et al . , 1995; Soroker et al . , 1994; Soroker et al . , 1995b; van Zweden et al . , 2010 ) . However , because our data indicate that the maturation of the CHC profile of individual honey bees is actually regulated in association with the stereotypic age-dependent division of labor in this species , we next hypothesized that the CHC profiles of worker honey bees develop , at least in part , via an intrinsic age-dependent regulation of the CHC biosynthetic pathways in the pheromone producing oenocytes ( Chung and Carroll , 2015; Falcón et al . , 2014; Makki et al . , 2014; Yew and Chung , 2015 ) . Thus , we next examined whether age and/or task are associated with the mRNA expression levels of genes that encode elongases and desaturases , the primary CHC diversity producing classes of enzymes in the CHC biosynthesis pathway ( Chung and Carroll , 2015 ) . To identify candidate genes for our analyses , we first used a bioinformatic approach to identify all putative members of both protein families in the honey bee genome ( Table 6 ) . Subsequently , we used real-time quantitative RT-PCR to compare mRNA levels of each candidate gene in dissected abdominal cuticles from bees of different ages raised in their source colony ( sisters of the bees analyzed for Figure 1- Figure 1—figure supplement 1 ) , as well as foraging sister bees raised in either their source colony or an unrelated foster-colony . Our analyses revealed that the expression levels of at least one elongase and two desaturase genes are associated with either age or colony environment ( Figure 4 and Table 7 ) . Thus , our data suggest that individual worker honey bees regulate CHC expression through an innate age-dependent developmental process that is further modulated by other factors such as task and the social environment . Because previous studies have indicated that nestmate recognition in honey bee colonies is likely driven by components of the CHCs profile ( van Zweden and D’Ettorre , 2010 ) , and our discovery that the CHC profiles of individual workers seem to mature in association with the well-described age-dependent division of labor in this species ( Robinson , 1992; Smith et al . , 2008; Søvik et al . , 2015 ) , we next hypothesized that , in honey bees , nestmate recognition cues themselves mature in association with age-dependent division of labor , and reach maturation during foraging . To test this hypothesis , we investigated the behavioral responses of guard bees to related and unrelated focal bees of different ages ( Day 1 , Day 7 , Day 14 , and foragers on Day 21 ) . At each test colony , the behavioral responses of guards to random related and unrelated returning foragers of unknown age were used as the benchmark for the baseline level of nestmate recognition behavior . Behavioral observations revealed that bees are accepted at the entrance of their own colony , regardless of age ( Figure 5A , Pearson’s Chi-Squared , Day1: χ2 = 49 . 05 , df = 2 , p<0 . 001 , Day 7: χ2 = 19 . 07 , df = 2 , p<0 . 001 , Day 14: χ2 = 44 . 89 , df = 2 , p<0 . 001 , Day 21: χ2 = 28 . 32 , df = 2 , p<0 . 001 ) . In contrast , at the entrance to an unrelated colony , bees were accepted on Days 7 and 14 , but rejected as foragers ( Day 21 ) ( Figure 5B , Day1: χ2 = 11 . 61 , df = 2 , p=0 . 003 , Day 7: χ2 = 15 . 51 , df = 2 , p<0 . 001 , Day 14: χ2 = 11 . 91 , df = 2 , p=0 . 002 , Day 21: χ2 = 7 . 35 , df = 2 , p=0 . 04 ) . These data support the hypothesis that nestmate recognition cues in honey bee colonies mature in association with age-dependent division of labor , and suggest that nestmate recognition is specific to behavioral interactions between guards and foragers at the entrance to the hive . Surprisingly , we also observed that while young Day 1 bees are accepted by related guards , they are often rejected by unrelated guards ( Figure 5B ) . This finding contradicts the broadly accepted ‘blank slate’ hypothesis , which predicts that because day-old bees are devoid of any defining chemical signatures , they should be always accepted by guards independent of relatedness ( Breed et al . , 2004 ) . While we do not yet know which specific components of the CHC profile of young bees , if any , might have triggered a rejection by unrelated guards in our colonies , one plausible interpretation of these data is that the observed response of guards to unrelated Day 1 bees is an artifactual experimental outcome of a forced behavioral interaction between two bee groups , which in colonies with a typical demography , do not normally interact in the context of the hive entrance .
The ability of colonies of social insects to reliably recognize group membership is one of the remarkable adaptations that enabled their immense ecological success . Yet , the molecular and physiological mechanisms that underlie this complex trait remain unknown for most species . In the well-studied honey bee , previous studies suggested that the chemical cues that drive nestmate recognition are absent in newly eclosed bees , and subsequently develop primarily through the homogenization and transfer of chemicals between colony members via direct interactions such as allogrooming and trophallaxis , and indirect interactions such as physical contact with wax and other nest materials ( Breed et al . , 2015; Breed et al . , 2004 ) . However , the data we present here suggest that the overall development of individual CHC profiles of honey bee workers primarily depends on an innate developmental process that is associated with the stereotypic age-dependent division of labor in this species , and that colony-specific cues are likely only carried by foragers . Therefore , we posit that it is unlikely that CHC profiles in honey bees develop through homogenization and transfer mechanisms between nestmates and hive materials . Furthermore , given the established implicated role of CHCs in nestmate recognition ( van Zweden and D’Ettorre , 2010 ) , we additionally posit that CHC homogenization mechanisms are unlikely to play a key role in the production of colony-specific cues in honey bees . A major line of investigation in understanding nestmate recognition of social insects has been to determine how colony-specific cues are determined . Cue specificity has historically been proposed to be determined by mechanisms under genetic control or acquired from the environment ( Crozier and Dix , 1979 ) . Although our studies do not directly address the mechanism by which cue specificity is determined in honey bees , data from cross-fostering experiments suggest that cue development and specificity are defined by interactions between factors derived from the colony-of-origin of individual workers and the actual hive environment they develop in . Therefore , our data suggest that CHC profiles of honey bee workers develop via a biphasic process that is governed , at least in part , by the intrinsic physiology of individual workers , the specific behavioral tasks they are engaged in , and the hive environment they age in . In phase one , similar to other social insect species ( Soroker et al . , 1995a ) , the total CHC amount builds up , possibly to increase the resistance of workers to desiccation while still inside the protective hive environment ( Chung and Carroll , 2015 ) . In phase two , the total amount of CHCs remains constant but the relative abundances of individual components shift in association with the age-dependent behavioral maturation of workers , at least in part , via the transcriptional regulation of CHC biosynthetic enzymes . Which specific components of the honey bee CHC profile represent the nestmate recognition cue remains unknown . Although it has been shown that CHCs are likely used for nestmate recognition in honey bees ( van Zweden and D’Ettorre , 2010 ) , it is unlikely that all components of the CHC profile contribute to this process ( Akino et al . , 2004; Dani et al . , 2005; Dani et al . , 2001; Martin et al . , 2008; Ruther et al . , 2002 ) . In fact , it has previously been shown that alkenes seem to play a more prominent role in nestmate recognition in the honey bee than alkanes ( Dani et al . , 2005 ) . Our data also indicate that although unrelated foragers raised in the same colony are equally accepted , their overall CHC profiles remain somewhat qualitatively different ( Figure 3C ) . These data provide two important insights . First , guards are not likely using the full CHC profile of individuals to determine group membership . Second , differences in the CHC profiles of co-fostered nestmate foragers of similar age that originated from different source colonies indicate that the chemical profiles of individual workers are not likely to be the product of a stochastic CHC homogenization and transfer between colony members . The observation that the mRNA expression levels of genes that encode CHC-biosynthesis enzymes vary in association with age and/or task further indicate that the primary mechanism for the dynamic regulation of the CHC profile of individual honey bee workers is directly associated with the well-established age-dependent division of labor in the honey bee ( Robinson , 1992; Smith et al . , 2008; Søvik et al . , 2015 ) . Although these data do not directly exclude the possibility that some particular CHCs are transferred across colony members , they do indicate that the overall observed qualitative age- and task-dependent changes in the CHC profiles of individual workers are affected by intrinsic molecular dynamics of the CHC synthesis pathway . However , our studies also importantly show that genetically-related bees that age in different colonies exhibit qualitatively different CHC profiles and CHC biosynthesis gene expression levels , which suggests that the CHC synthesis process is also plastic and could be modulated by factors associated with the hive/social environment . We were initially surprised by our observation that Day one bees are accepted at the entrance to their source colony but rejected by guards at the entrance of an unrelated colony since previous studies hypothesized that the lower amounts of total CHCs in young bees represent a ‘blank slate’ in terms of the nestmate recognition cue because these bees are readily ‘accepted’ when introduced into unrelated colonies ( Breed et al . , 2004 ) . In fact , this phenomenon was exploited here to introduce cohorts of bees to foster colonies , typically by placing the new bees on the top frames of experimental hives . This apparent conundrum highlights an important , yet often underappreciated , aspect of the nestmate recognition system in honey bees and other social insect species , which is that the ‘rejection’ behavior by guards is highly contextual . Conceptually analogous to other biological systems responsible for the detection of ‘self’ versus ‘non-self’ ( e . g . , the acquired immunity system in vertebrates ) , behaviors associated with nestmate recognition are restricted to interactions between guards and incoming bees at the entrance to the hive ( Couvillon et al . , 2013 ) . Therefore , we speculate that because nestmate recognition is spatially restricted to specific behavioral interactions between entering bees and guards at the entrance , the commonly observed ‘acceptance’ of day old bees outside the specific context of the hive entrance actually represents the lack of behavioral ‘rejection’ rather than a true self-recognition-dependent ‘acceptance’ . Consequently , the observation that Day one bees are rejected at the entrance of an unrelated colony suggests that nestmate recognition of young bees either depends on components of the CHC profile that are already present in Day one bees , non-CHC chemical cues , or an altogether different sensory modality . Alternatively , because newly eclosed bees usually perform cell cleaning behaviors at the interior of the hive , and therefore do not typically interact with guards at the hive entrance ( Robinson , 1992; Smith et al . , 2008; Søvik et al . , 2015 ) , differences in rejection of Day one bees between these two colonies might represent an experimental artifact resulting from differences in tolerance to the forced behavioral interaction between two bee groups that normally do not interact . Additionally , it has previously been shown that observed levels of guarding behaviors in honey bees are plastic , and could fluctuate in response to various environmental factors such as seasonal weather patterns , overall colony size , food availability , and ‘robbing’ pressures from other colonies or predators ( Downs and Ratnieks , 2000 ) . Likewise , more extreme forms of plasticity in nestmate recognition systems have been reported in other social species . For example , some social insects can switch between using visual or chemosensory modalities for nestmate recognition under different circumstances ( Baracchi et al . , 2015 ) . Together , it seems that instead of being driven by simple binary decisions , nestmate recognition systems in the honey bee and other social insect species depend on a plastic recognition of ‘friends’ versus ‘foes’ as part of a broader group-level optimization of colony fitness . In conclusion , we propose that nestmate recognition cue production and acquisition in honey bees are not likely to be primarily driven by CHC homogenization and transfer mechanisms as previously described in some ant species ( Boulay et al . , 2000; Lenoir et al . , 2001; Meskali et al . , 1995; Soroker et al . , 1994; Soroker et al . , 1995b; van Zweden et al . , 2010 ) . Instead , we propose a new model for the regulation of nestmate recognition in honey bee colonies , which stipulates that unknown factors associated with the hive environment play a direct or indirect role in defining the developmental kinetics and specificity of nestmate recognition cues by modulating the cellular and molecular processes that are responsible for pheromone synthesis . Thus , it is plausible that the colony/social environment drives the intrinsic development of similar pheromone profiles by individual colony members , which in typical honey bee hives , is associated with the physiological processes that drive age-dependent division of labor . If true , this model could resolve previous seemingly contradictory data which suggested that honey bee CHC profiles are defined by genetic ( Page et al . , 1991 ) versus environmental ( Downs and Ratnieks , 1999 ) factors , as well as open the door for comparative mechanistic studies of how complex social traits evolve and function in different social insect clades .
Honey bee ( Apis mellifera ) colonies were reared and managed using standard beekeeping techniques across two locations near St . Louis , MO: Tyson Research center ( 38° 31’N , 90° 33’W ) and a residential home . For all experiments that included collections of bees at specific ages , capped brood frames were taken from a colony and placed in a humidified 32°C incubator . Once eclosed , about 1000 bees ( <24 hr old ) were marked with a spot of paint ( Testors , Vernon Hills , IL , USA ) on their thorax , and then reintroduced into either their source or a foster colony , depending upon the experiment . For collections of bees at specific ages , marked bees were collected from internal frames of the colony one day post reintroduction ( Day 1 ) , seven days post reintroduction ( Day 7 ) , 14 days ( Day 14 ) post reintroduction , and as returning foragers , identified by pollen loads on their hind legs or having a distended abdomen due to nectar loads , between 18 and 21 days post reintroduction . Bees used for chemical and molecular analyses were placed in individual 1 . 7 mL microtubes and immediately placed on dry ice . All samples were kept at −80°C until further analysis . Single-cohort colonies ( SCC ) were established as previously reported ( Ben-Shahar et al . , 2004; Ben-Shahar et al . , 2002; Greenberg et al . , 2012; Robinson et al . , 1989; Whitfield et al . , 2003 ) . In short , about 1000 newly eclosed bees ( <24 hr old ) were placed in a small wooden nucleus hive-box with a young , unrelated mated queen , one honey frame from their source colony , an empty comb frame , and three new frames with wax covered plastic foundation . Bees were collected as typical-aged nurses and precocious foragers one week after introduction , and as over-aged nurses and typical-aged foragers at three weeks after introduction . Bee samples were collected and stored as above . To induce ‘undertaking’ behavior , about 1000 dead bees were placed into the top of two different colonies , and the first 20 bees that were observed removing dead bees from the colony were collected from the entrance . Returning foragers and in-hive nurses of unknown ages were also collected from each colony at the same time . Samples were stored and processed as described above . Big-back colonies were established as previously described ( Ben-Shahar et al . , 2000; Withers et al . , 1995 ) . In short , bees were introduced in two cohorts to a 5-frame hive box containing three empty comb frames , two brood frames , and a new queen . In the first cohort , 200 day-old bees were collected as described above and marked on the thorax with paint . Half of these bees were marked with a plastic tag attached to the thorax ( ~3 mm diameter , ~1 mm thick; ‘big-back’ bees ) . Day-old bees in the other cohort were collected and introduced 4 days later as described above to increase the proportion of precocious foragers in the first group . The entrance to the colony was blocked by a piece of Plexiglas with holes in it that prevented ‘big-back’ bees from leaving the hive , but allowed paint marked bees to leave . Bees were collected at 7 days of age: returning foragers were collected as described above , and ‘big-back’ bees were collected as they were attempting to leave the hive via the holes in the plastic . Reversion colonies were made by collecting ~1000 foragers from a single source colony by vacuuming them directly into a sealed 5-frame hive box containing two brood frames , one honey frame , and two empty comb frames . The hive was sealed and moved to a new location ~30 miles away from the source colony , and a new queen was added that night . The hive was sealed for 3 days , and then was opened to allow normal foraging activity to resume . During this time , in the absence of nurses , some foragers reverted back to nursing behaviors ( Robinson et al . , 1992 ) . Actively foraging bees were collected at the hive entrance as described above and reverted nurses were collected from internal frames as described above . 1000 day-old bees from two independent source colonies were collected and marked as above . Half of the bees in each marked cohort were randomly reintroduced to both their own source colony and the reciprocal foster colony . Subsequently , marked bees of defined age were recollected from internal frames of each colony as described above . Every day over a three-week period , newly eclosed bees ( <24 hr old ) from a single source colony were collected as described above , uniquely color-marked , and then reintroduced into their source colony . Subsequently , on each experimental day , bees from the following groups were collected , placed in individual 15 mL plastic tubes ( Corning , Corning , NY , USA ) , and chilled on wet ice in an ice cooler up to 10 min before the assay in order to limit heat related stress: bees of the focal age ( identified by color of mark ) , returning nectar foragers ( denoted by distended abdomen and lack of pollen ) of unknown age from the source colony , and returning nectar foragers of unknown age from an unrelated colony . All foragers , which served as behavioral controls , were painted the same color as the experimental bees just after collection . Tubes were numbered in a randomized order and blinded to the experimenter conducting the behavioral assays . Fifteen bees per group were prepared for each colony each experimental day . Behavioral assays were conducted simultaneously at two colonies ( source and unrelated ) by two researchers , as well as recorded using digital video cameras . As described previously ( D'ettorre et al . , 2006; Downs and Ratnieks , 2000 ) , acceptance at the colony entrance was used as a proxy for nestmate recognition by placing individual bees on a modified entrance platform and recording the behavioral reactions of guard bees for ~5 min . Bees were considered ‘Rejected’ if they were bit , stung and/or dragged by at least one guard bee ( Video 1 ) . Bees were considered ‘Accepted’ if they were approached by guards , antennated and/or licked and then left alone ( not bit ) , if they immediately entered the colony and were not removed by other bees , or if they remained on the platform and did not receive aggression ( Video 2 ) . After 5 min , focal bees that remained on the platform outside the colony were removed before the next assay . All behaviors were scored in real time , and videos were retained as back-up . All behavioral assays were conducted during a period of 10 days , between 12 and 4pm , with two days focusing on each age of experimental bee ( N = 20–30 bees per group ) . CHCs were extracted from whole bees by placing individual bees into 6 mL glass vials fitted with 16 mm PTFE/silica septa screw caps ( Agilent Crosslab , Santa Clara , CA , USA ) . Bee CHCs were extracted in 500 µL hexane containing 10 ng/μl of octadecane ( C18 ) and 10 ng/μl of hexacosane ( C26 ) ( Millipore Sigma , St . Louis , MO ) , which served as injection standards . To achieve efficient extraction , each vial was gently agitated by vortexing ( Fisher Scientific , Waltham , MA , USA ) for 2 min at minimum speed . Extracts were immediately transferred to new 2 mL glass vials fitted with 9 mm PTFE lined caps ( Agilent Crosslab , Santa Clara , CA , USA ) . In cases where experiments involved forager honey bees , all bees ( including non-foragers ) had their hind legs removed prior to extraction , in order to ensure removal of pollen . 100 µL of each extract was transferred to a new 2 mL glass vial and stored at −20°C for further analysis; the remaining 400 µL was stored at −80°C as back-up . Representative pooled samples of foragers and nurses of known age were first analyzed by combined gas chromatography/mass spectrometry ( GC/MS ) for compound identification . Samples were run from 1500 ( 3 min hold ) to 3000 at 50/min . Compounds were identified by their fragmentation pattern as compared to synthetic compounds . For profile characterizations of individual bees , samples were analyzed using an Agilent 7890A gas chromatograph system with a flame ionization detector ( GC/FID ) and PTV injector ( cool-on-column mode ) , and outfitted with a DB-1 20 m x 0 . 18 mm Agilent 121–1022 fused silica capillary column ( Agilent Technologies , IncSanta Clara , CA , USA ) . Sample volumes of 1 . 0 μl were injected onto the column . Helium was the carrier gas and applied at a constant flow rate of 1 ml/min . Analysis of the extract was carried out with a column temperature profile that began at 50C ( held for 1 min ) and was ramped at 36 . 6 °C/min to 150C and then at 5 C/min to 280C , where it was held for 10 min . The injector and FID temperatures were programmed to 280C and 300C , respectively . Agilent OpenLAB CDS ( EZChrom Edition ) software was used to calculate the retention time and total area of each peak . Data were normalized to known quantity ( ng ) of internal standard hexacosane and all ng data are listed in source data . Members of the highly conserved desaturase and elongase gene families were identified in the honey bee genome by using the protein BLAST search tool ( https://blast . ncbi . nlm . nih . gov/Blast . cgi ) with annotated Drosophila melanogaster amino acid sequences ( https://flybase . org ) of elongase and desaturase genes known to play a role in CHC biosynthesis ( Chung and Carroll , 2015 ) . Initial homologs in the honey bee genome were chosen by picking the top match ( highest total score and query cover , lowest E value ) for each D . melanogaster gene , and possible paralogs of these putative genes were identified by subsequently using the NCBI protein BLAST tool ( RRID:SCR_004870 ) with these genes’ amino acid sequences . Many of these putative elongase and desaturase genes have previously been identified as possible CHC biosynthesis pathway genes in the honey bee ( Falcón et al . , 2014 ) . E-values from the BLAST scans of the honey bee genome by using three canonical Drosophila melanogaster CHC biosynthesis genes , EloF ( elongase subfamily ) , Elo68α ( elongase subfamily ) , and desat1 ( fatty acid desaturase subfamily ) , are listed in Table 6 . To measure mRNA levels of individual genes , the cuticles from the abdomens of four bees per group were dissected out , and total RNA was extracted using the Trizol Reagent ( Life Technologies , Grand Island , NY , USA ) . SuperScript II ( Life Technologies , Grand Island , NY , USA ) reverse transcriptase was used to generate cDNA templates from 500 ng of total RNA per sample by using random hexamers . A Bio-Rad ( Hercules , CA , USA ) CFX Connect Real-Time PCR Detection System and Bio-Rad iTaq Universal SYBR Green Supermix were subsequently used for estimating relative differences in mRNA levels across samples ( N = 4 per group , run in triplicate technical replications ) . Expression levels of the EIF3-S8 , a housekeeping gene that has previously been used as a reference gene in honey bee studies of gene expression by us and others ( Alaux et al . , 2009; Fischer and Grozinger , 2008; Greenberg et al . , 2012; Mao et al . , 2015; Richard et al . , 2008 ) , was used as a loading control . To further ensure that the reported expression data for the experimental genes are robust , we first confirmed that the raw EIF3-S8 Ct values per total RNA used in the individual RT reactions were not affected by any of the studied groups included in our current study ( Kruskal-Wallis , H = 3 . 299 , df = 4 , p=0 . 5091 ) . Ct data is listed in Figure 4—source data 1 . The specific RT-PCR primers for each gene-specific assay are listed in Table 6 . All CHC analyses included a set of 19 peaks that represent well-established honey bee CHCs , identified by comparing GC traces to published data ( Kather et al . , 2011 ) . For the comparisons of total CHCs across groups ( as in Figure 1A ) , total ng of all identified CHCs in each bee were analyzed using ANOVA followed by Tukey’s HSD in R 3 . 3 . 2 ( Team , 2016 ) . For the remainder of the datasets , the relative proportion of each compound in each sample was calculated and then used in further statistical analysis . For each dataset , a permutation MANOVA was run using the ADONIS function in the vegan package of R ( RRID:SCR_011950 ) with Bray-Curtis dissimilarity measures ( Oksanen et al . , 2017 ) . Pairwise comparisons with FDR p-value correction were subsequently run on experiments where more than two groups were compared . Data were visualized using non-metric multidimensional scaling ( metaMDS function in the vegan package of R ( RRID:SCR_011950 ) ( Oksanen et al . , 2017 ) ) using Bray-Curtis dissimilarity , and either 2 or three dimensions in order to minimize stress to <0 . 1 . For Table 1 , Table 2 , Table 3 , and Table 4 an ANOVA followed by Tukey’s HSD post-hoc comparison , or Kruskal-Wallis followed by Dunn’s Test with FDR adjustment was performed using total ng ( Table 1 and Table 2 ) or proportions ( Table 3 and Table 4 of each compound across bees of the four time point collections . For cross-fostering studies , power was assessed by performing pseudo multivariate dissimilarity-based standard error , a method for assessing sample-size adequacy in multivariate data , as described in and using code from Anderson and Santana-Garcon ( 2015 ) . For behavioral data , the proportion of bees accepted by guard honey bees was calculated for each experimental group at each colony at each day of age . A Pearson’s chi-square was run for each day of age at each colony with subsequent pairwise comparisons . For qPCR data , relative expression levels were calculated as previously described ( Greenberg et al . , 2012; Hill et al . , 2017; Zheng et al . , 2014 ) , using eIF3-S8 as a loading control . Fold-expression data were generated by using the 2-ΔΔCT method ( Livak and Schmittgen , 2001 ) and designating a single individual from the ‘Day 1’ group ( Figure 5 ) as a calibrator . For statistical analyses , the 2-ΔΔCT scores were compared within each gene across bees of different groups using an ANOVA followed by Tukey’s HSD post-hoc comparison , or Kruskal-Wallis followed by Dunn’s Test with FDR adjustment . Overall test p-values were then adjusted using FDR correction to account for 16 independent comparisons ( Benjamini and Hochberg , 1995 ) . | Honey bees are social insects that live in large groups called colonies , within structures known as hives . The young adult bees stay within the hive to build nests and care for the young , while the older bees leave the hive to forage for food . Honey bees store food and other valuable resources in their hives , so they are often targeted by predators , parasites and ‘robber’ bees from other colonies . Therefore , it is important for bees to determine whether individuals trying to enter the nest are group members or intruders . While it is known that social insects use blends of waxy chemicals called cuticular hydrocarbons to identify group members at the entrance to the colony , it is not clear how members of the same colony acquire a similar blend of cuticular hydrocarbons . Some previous work suggested that in some ant species ( which are also social insects ) , colony members exchange cuticular hydrocarbons with each other so that all members of the colony are covered with a similar blend of chemicals . However , it was not known whether honey bees also share cuticular hydrocarbons between colony members in order to identify members of a hive . Vernier et al . used chemical , molecular and behavioral approaches to study the cuticular hydrocarbons found on honey bees . The results show that , rather than exchanging chemicals with other members of their colony , individual bees make their own blends of cuticular hydrocarbons . As a bee ages it makes different blends of cuticular hydrocarbons , and by the time it starts to leave the hive to forage it makes a blend that is specific to the colony it belongs to . The production of this final blend is influenced by the environment within the hive . Thus , the findings of Vernier et al . indicate that honey bees guarding the entrance to a hive can only identify non-colony-member forager bees as intruders , rather than any non-colony-member bee that happens upon the hive entrance . Honey bees play an essential role in pollinating many crop plants so understanding how these insects maintain their social groups may help to improve agriculture in the future . Furthermore , this work may aid our understanding of how other social insects interact in a variety of biological situations . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology"
] | 2019 | The cuticular hydrocarbon profiles of honey bee workers develop via a socially-modulated innate process |
Caspase-3 carries out the executioner phase of apoptosis , however under special circumstances , cells can survive its activity . To document systematically where and when cells survive caspase-3 activation in vivo , we designed a system , CasExpress , which drives fluorescent protein expression , transiently or permanently , in cells that survive caspase-3 activation in Drosophila . We discovered widespread survival of caspase-3 activity . Distinct spatial and temporal patterns emerged in different tissues . Some cells activated caspase-3 during their normal development in every cell and in every animal without evidence of apoptosis . In other tissues , such as the brain , expression was sporadic both temporally and spatially and overlapped with periods of apoptosis . In adults , reporter expression was evident in a large fraction of cells in most tissues of every animal; however the precise patterns varied . Inhibition of caspase activity in wing discs reduced wing size demonstrating functional significance . The implications of these patterns are discussed .
The cell death program known as apoptosis was originally described as a series of morphological changes that cells undergo as they die ( Tenev et al . , 2005 ) . The reproducibility of the sequence suggested an underlying molecular program , and a conserved set of enzymes , the caspases , emerged as key regulators and executioners of apoptosis ( Martin and Green , 1995; Jacobson and Evan , 1994; Thornberry , 1998 ) . While caspase activation is frequently a terminal event resulting in swift cellular demise ( Chang et al . , 2002 ) , cell survival following caspase activation has been described ( e . g . , [Florentin and Arama , 2012; Kuranaga and Miura , 2007; Kumar , 2004; Meinander et al . , 2012] ) . In some cells and tissues , caspases promote localized or partial destruction of the cell without actually killing it ( Arama et al . , 2003; Huh et al . , 2004; Connolly et al . , 2014 ) . A variety of primary cells and cell lines can survive caspase activation following a lethal dose of an apoptotic stimulus , as long as it is transient and thus sublethal in time ( Tang et al . , 2012 ) . This reversal of late stage apoptosis has been named anastasis ( Greek for 'rising to life' ) . Cell survival following caspase activation in response to a sublethal dose of irradiation has also been reported ( Florentin and Arama , 2012; Liu et al . , 2015; Ichim et al . , 2015 ) . Such survival following caspase activation has the potential for both beneficial and harmful effects . It may limit permanent damage to the heart following transient ischemia ( Kenis et al . , 2010 ) ; however it can also be oncogenic ( Tang et al . , 2012; Liu et al . , 2015; Ichim et al . , 2015 ) , and could in principle allow tumor cells to escape chemotherapy . Apoptosis is a critical feature of normal development in multicellular organisms ( Miura , 2012; Denton and Kumar , 2015; Vaux and Korsmeyer , 1999 ) . Studies in model organisms such as worms and flies have made important contributions to unraveling the underlying mechanisms ( Connolly et al . , 2014; Denton and Kumar , 2015; Orme and Meier , 2009; Steller , 1995 ) . It is unknown whether cells ever recover from the brink of apoptotic cell death during development . The observations that cultured cells and adult cardiac myocytes recover from transient insults that cause caspase-3 activation raised the question as to how widespread cell survival following caspase activation might be in vivo , whether this ever occurs during normal development , and if so what function it might serve . Identification of cells that survive transient caspase activation is challenging because they bear no known distinguishing characteristic . Therefore we developed a genetic system to mark and manipulate cells that survive caspase activation in Drosophila ( Figure 1 ) . Using these CasExpress transgenic flies , we discovered that the majority of cells in the adult derive from cells that survive caspase activation during normal development . We observed distinct categories of CasExpress activation . For example , in some organs , every cell activated the sensor over an extended period of development without evidence of apoptosis or morphological remodeling , suggesting a function for caspase-3 unrelated to cellular destruction . In other tissues , activation was sporadic in temporal and spatial pattern , suggesting a stochastic process . In these tissues , the precise patterns differed from animal to animal , and occurred in regions that normally exhibit apoptosis . These observations suggest that some cells recover from the brink of apoptotic cell death and undergo developmental anastasis . We propose that these different patterns represent distinct functions of executioner caspases during normal development . 10 . 7554/eLife . 10936 . 003Figure 1 . Widespread CasExpress activation in adult tissues . ( A ) A schematic of CasExpress and G-trace . ( B ) A schematic showing the sequence of the DQVD caspase cleavage site in CasExpress and the point mutation in the DQVA control . ( C–L ) Confocal micrographs showing overlays of DAPI , RFP and GFP from CasExpress/G-Trace flies . ( D’–L’ ) GFP channel only . ( D”–L” ) RFP channel only . Arrows in D–D’’ indicate examples of GFP+ progenitor cells , and arrowheads point to examples of GFP- progenitor cells . Dotted lines in F–F’’ mark the boundary between midgut and hindgut . Scale bars in C and I-L are 100 μm; scale bars in D–H are 25 μm . ( M ) A schematic summarizing the general pattern of GFP and RFP expression in adult . Although GFP expression was present in all body wall muscle , only part is shown in green for simplicity and presentation clarity . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 003
In order to detect and follow the fates of cells that survive caspase activation , we designed a caspase-inducible Gal4 transcription factor ( Figure 1A ) . To keep Gal4 inactive in the absence of caspase activity , we tethered it to the plasma membrane by fusing it to mCD8 ( mouse cluster of differentiation 8 ) . To render the protein caspase-inducible , we inserted the caspase-3-binding and cleavage domain from the Drosophila Inhibitor of Apoptosis Protein 1 ( DIAP1 ) ( Ditzel et al . , 2003 ) in between CD8 and Gal4 . As a negative control we created a second transgene with a DQVD to DQVA amino acid substitution in the caspase cleavage site ( Figure 1B ) in order to render it caspase insensitive , hereafter the 'DQVA control . ' To allow for detection of caspase activation in as many cell types as possible , the fusion protein was expressed under the control of the ubiquitin ( ubi ) enhancer/promoter . We characterized the expression and activity of transgenic flies bearing a site-directed insertion into the attP40 landing site , selected for its ability to allow relatively uniform , moderate levels of expression in a variety of tissues ( Markstein et al . , 2008 ) . We also generated an insertion into a random site for comparison . We named this system CasExpress for its ability to drive expression of downstream genes and proteins under the control of caspase-3 activity . To detect caspase activity , we crossed the sensor and control to G-Trace ( Evans et al . , 2009 ) a fly line that expresses two fluorescent protein targets , under the control of Gal4-responsive UAS ( upstream activating sequences ) . G-Trace flies contain three transgenes , all on the second chromosome: UAS-RFP , UAS-FLP , which encodes a yeast recombinase enzyme , and a ubi-FRT-STOP-FRT-GFP cassette where FRT stands for FLP Recombination Target sequence . Crossing the mCD8-DQVD-Gal4 sensor to G-Trace should lead to permanent GFP expression in any cell that survives transient caspase activation and in all of its progeny , in contrast to other caspase activity reporters ( Bardet et al . , 2008 ) . We expected the caspase-activated Gal4 protein to be short-lived because we had observed rapid degradation of other caspase reporters ( Tang et al . , 2012 ) , so we anticipated RFP would be transient and limited to the cells that activated caspase-3 but not their progeny . We first examined adult tissues where , to our surprise , we found widespread GFP expression ( Figure 1C–L ) . In the intestine for example , GFP was evident in the most anterior structure , the proventriculus ( Figure 1C ) , although little RFP was evident there , suggesting that caspase had been active earlier during development . In the midgut both RFP and GFP appeared in a partially overlapping pattern ( Figure 1C , D–D” ) . Large nuclei corresponding to differentiated epithelial cells expressed both RFP and GFP suggesting ongoing caspase activation , whereas a subset of small progenitor cells expressed GFP but not RFP ( Figure 1D–D” arrows ) . Visceral muscle and hindgut showed a mixture of GFP+/RFP- cells as well as some GFP+/RFP+ cells ( Figure 1E–F” ) . The adult eye and antenna exhibited widespread nuclear GFP but only infrequent RFP ( Figure 1G–H” ) , suggesting that caspase had been activated earlier in development either in a large fraction of cells , or in precursors that gave rise to a large fraction of adult cells; however little activation of caspase appeared to be ongoing in the adult . In the adult central brain and nerve cord , a minority of cells expressed GFP and/or RFP ( Figure 1I–J” ) . In the optic lobe , many but not all cells expressed GFP and/or RFP ( Figure 1I–I” ) , whereas in body wall muscle , nearly every cell expressed GFP ( Figure 1K–K” ) . In the female reproductive system , every cell of the oviduct was GFP+/RFP- in every animal , whereas the majority of germline and somatic cells in egg chambers lacked FP expression ( Figure 1L ) . Figure 1M summarizes these findings schematically . Recently a similar strategy detected similarly widespread adult expression ( Tang et al . , 2015 ) . The adult expression suggested that CasExpress was activated during development . To document when caspase activation first appeared , we examined embryonic and larval stages . In Drosophila embryos , the only tissue that activated CasExpress robustly was the salivary gland beginning at stage 12 ( Figure 2A–A” ) . Salivary gland expression was not detected in the DQVA control , demonstrating that this was not due to leaky or background expression from the G-Trace transgenes or random breakdown of the fusion protein that might separate Gal4 from the transmembrane domain . We also confirmed that the DQVD sensor and DQVA control showed similar patterns and levels of fusion protein expression at the cell surface detected with anti-mCD8 antibody staining throughout the embryo and in most tissues and stages of development ( Figure 2—figure supplement 1 ) . In the embryo RFP was also detected in some randomly distributed cells , likely corresponding to a subset of cells that normally undergo apoptosis ( Figure 2A and A” ) ; little if any GFP was detected in those cells , presumably because dying cells were not active enough to transcribe and translate FLP , undergo DNA recombination , and then transcribe and translate GFP to detectable levels . While RFP was detected in the salivary gland beginning in stage 12 , GFP expression became evident later ( Figure 2B–B” ) , confirming that these FPs exhibit different timing of activation . 10 . 7554/eLife . 10936 . 004Figure 2 . CasExpress activation in embryos and larvae . ( A–B” ) RFP and GFP expression in Drosophila embryos ( A–A” ) stage 12 , ( B–B” ) stage 17 . ( C–C” ) Acridine orange detection of apoptotic cells in stage 13 embryos of the indicated genotypes . ( D–D” ) 1st instar larva , ( E–E” ) 2nd instar larva and ( F–F” ) 3rd instar larva . ( G ) A schematic summarizing of GFP and RFP expression in above stages . Red represents RFP expression . Green represents GFP . Yellow/Orange indicates either a mixture of GFP positive and RFP positive cell populations or the presence of cells expressing both . Scale bars represent: 100 μm ( A–C ) ; 200 μm ( D ) ; 400 μm ( E ) ; and 600 μm ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 00410 . 7554/eLife . 10936 . 005Figure 2—figure supplement 1 . Expression of CasExpress DQVD and the DQVA control . Anti mCD8 staining of the DQVD CasExpress sensor ( A , C , E , G , I , K ) or DQVA caspase-insensitive control ( B , D , F , H , J , L ) in stage 10 embryos ( A , B ) , and 3rd instar larval central nervous system ( C , D ) , wing disc ( E , F ) , eye-antennal disc ( G , H ) and leg disc ( I , J ) Stage 10 egg chambers of the adult ovary ( K , L ) . Scale bars in A–D and K , L are 100 μm , in G–J are 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 005 Although the DQVD and DQVA proteins contained the caspase binding sequence from DIAP1 ( Tenev et al . , 2005 ) , which in principle could function as a dominant-negative inhibitor of caspase activity if expressed at high enough levels , the flies expressing the sensor were viable and fertile and showed no discernible morphological defects . The modest expression level and membrane localization presumably prevented any dominant negative effect . Moreover , there was no decrease in the number , or change in distribution , of apoptotic cells in DQVD and DQVA embryos compared to w1118 embryos ( Figure 2C–C” ) . Both RFP and GFP continued to be expressed throughout embryonic and larval development ( Figure 2D–F” ) . The temporal appearance of RFP and GFP in embryonic and larval life are indicated schematically in Figure 2G . During larval development CasExpress activation appeared over time in many cell types and tissues including all imaginal discs , oenocytes , and in subsets of neurons ( Figure 2E and Figure 3 ) . Tissues from flies carrying G-Trace in the absence of the caspase sensor or in combination with the DQVA caspase-insensitive control exhibited infrequent FP expression in small clones in a minority of animals ( Figure 3—figure supplement 1 ) . The frequency and patterns were very similar regardless of the presence or absence of the DQVA control transgene ( Figure 3—figure supplement 1 ) , suggesting that this minor background was due to leaky , Gal-4-independent FLP expression from the UAS-FLP transgene . In contrast , expression in the presence of the DQVD caspase-sensitive construct was present in every animal ( Figure 3—figure supplement 1 ) , and in large fractions of cells ( Figure 3 ) . 10 . 7554/eLife . 10936 . 006Figure 3 . CasExpress activation in larval tissues . ( A–F ) Confocal micrographs showing overlays of DAPI , RFP and GFP expression in the indicated tissues of wandering 3rd instar larvae . ( A’–E’ ) GFP only . ( A”–E” ) RFP only . The brackets in D mark the eye and antenna parts of the disc , in E mark the position of optic lobe , central brain and ventral nerve cord , and in F mark the different regions of the gut . Scale bars in A and F are 200 μm , in B and C are 50 μm , in D and E are 100 μm . ( G ) A schematic summarizing of GFP and RFP expression in larvae . There is little GFP/RFP expression in trachea or muscles , which are not included in diagram . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 00610 . 7554/eLife . 10936 . 007Figure 3—figure supplement 1 . Comparison of GFP expression in the CasExpress DQVD sensor , the DQVA control or G-trace alone . The percentage of wing , leg , haltere and eye-antennal discs containing GFP-expressing cells for the G-Trace alone ( blue bars ) , G-Trace together with the DQVA caspase-insensitive control ( orange bars ) , or G-Trace together with the DQVD caspase sensor ( gray bars ) are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 00710 . 7554/eLife . 10936 . 008Figure 3—figure supplement 2 . Diverse CasExpress activation patterns in larval CNS . ( A–F ) Six different examples of RFP ( red ) and GFP ( green ) CNS expression patterns in late 3rd instar larvae . DAPI is shown in blue . ( A’–F’ ) GFP only . ( A”–F” ) RFP only . All scale bars are 100 μm . Note that patterns differ from animal to animal and are not bilaterally symmetric . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 00810 . 7554/eLife . 10936 . 009Figure 3—figure supplement 3 . Diverse CasExpress patterns in leg and wing discs . Six different examples of RFP and GFP expression patterns in leg ( A–F ) and wing ( G–L ) discs of late 3rd instar larvae . ( A’–L’ ) GFP only . ( A”–L” ) RFP only . All scale bars are 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 00910 . 7554/eLife . 10936 . 010Figure 3—figure supplement 4 . CasExpress patterns in larval intestines . ( A–E ) Five different examples of RFP and GFP expression patterns in larval intestines . DAPI is shown in blue . All scale bars are 400 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 010 Different tissues exhibited distinct temporal and spatial patterns . For example oenocytes exhibited RFP and GFP expression in virtually every cell and in every animal ( Figure 3A–A” ) . In contrast , in imaginal discs fewer cells expressed RFP as compared to GFP ( Figure 3B–D” ) . Although every disc from every animal exhibited expression , the precise patterns varied ( Figure 3—figure supplements 2–4 ) . In the developing central nervous system ( CNS ) the patterns were not bilaterally symmetric ( Figure 3—figure supplement 2 ) . In the imaginal discs the patterns , particularly of RFP , varied from animal to animal and did not appear to correspond to known developmental patterns of known signaling pathways or cell types ( Figure 3—figure supplement 3 ) . Tissues that showed little or no activation of the sensor during normal development up through the third instar included somatic muscles , trachea , and the ventral nerve cord ( Figure 3E–E” ) . Although most of the nervous system showed little sensor activation , a consistently large fraction ( 50–80% ) of cells in the developing optic lobes were GFP-positive ( Figure 3E–E” ) . In the larval intestine , partially overlapping GFP and RFP expression patterns were observed ( Figure 3F ) , and while the overall regional patterns were conserved from one animal to the other , the details varied ( Figure 3—figure supplement 4 ) . The third instar larval patterns are summarized schematically in Figure 3G . The unexpectedly widespread activation of CasExpress raised the question as to its caspase-dependence . The sensor inserted into the attP40 site and the random insertion demonstrated similar patterns . The absence of expression in the DQVA control demonstrated that a proteolytic cleavage at the aspartic acid was likely necessary . To address the possibility that a protease other than caspase activated CasExpress , we crossed the sensor and G-Trace into a homozygous dronc mutant background . Dronc encodes the upstream apoptotic caspase in Drosophila ( equivalent to caspase-9 in mammals , ( Meier et al . , 2000; Hawkins et al . , 2000 ) and its activity is necessary for activation of both fly executioner caspase molecules Drice and Dcp-1 ( Florentin and Arama , 2012; Song et al . , 1997; Fraser et al . , 1997; Fraser and Evan , 1997; DeVorkin et al . , 2014; Muro et al . , 2006 ) . Although Dronc mutants are homozygous lethal , they survive to the third instar larval stage allowing us to assess CasExpress at that stage . As expected , the homozygous dronc mutant background eliminated caspase activity detected with an antibody against cleaved and activated Dcp-1 ( c-Dcp-1 ) ( Figure 4A–F’ ) . The dronc mutant also eliminated virtually all RFP and GFP expression in imaginal discs from CasExpress ( Figure 4A–F ) . 10 . 7554/eLife . 10936 . 011Figure 4 . Caspase-dependence of CasExpress . ( A–F ) Confocal micrographs showing overlays of DAPI , RFP and GFP expression in third-instar larval eye-antennal disc ( A–B ) , wing disc ( C–D ) , and leg disc ( E–F ) . CasExpress and G-trace were crossed into heterozygous ( A , C , E ) or dronc homozygous ( B , D , F ) dronc mutants . ( A’–E’ ) Cleaved Dcp-1 staining of corresponding discs . Scale bars are 50 μm . ( G–H ) RFP expression in eye-antennal discs of late third-instar larvae with CasExpress and G-trace with ( H ) or without ( G ) GMR-p35 . The dashed line encircles the region where p35 is expressed . ( I ) Quantification of of RFP: DAPI area . Error bars show standard error of the mean , and *** indicates p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 01110 . 7554/eLife . 10936 . 012Figure 4—figure supplement 1 . Anti-cleaved caspase-3 ( red ) and DAPI ( blue ) staining of stage 14 embryo ( A ) and high magnification of a salivary gland ( B ) . ( A'–B' ) are cleaved caspase-3 only . Scale bar in A is 20 μm , in B is 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 01210 . 7554/eLife . 10936 . 013Figure 4—figure supplement 2 . GFP and RFP expression of CasExpress in a wild type eye-antennal disc ( A–A” ) , or one carrying the GMR-p35 transgene ( B–B” ) . Scale bars are 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 01310 . 7554/eLife . 10936 . 014Figure 4—figure supplement 3 . Loss of dredd does not change CasExpress patterns . ( A–F ) Confocal micrographs showing overlays of DAPI , RFP and GFP expression in third-instar larval CNS ( A–B ) , larval intestines ( C–D ) , and adult intestines ( E–F ) . CasExpress and G-trace were crossed into heterozygous ( A , C , E ) or homozygous ( B , D , F ) dredd mutants . ( A'–B' ) GFP only , ( A''–B'' ) RFP only . Scale bars in A and B are 100μm , in C and D are 800μm , in E and F are 400μmDOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 014 Homozygous dronc mutant embryos retained RFP and GFP expression in the salivary gland , possibly due to the perdurance of maternal caspase expression . To confirm the presence of cleaved caspase in embryonic salivary gland cells , which has not been previously reported , we stained CasExpress embryos with an antibody against cleaved caspase-3 ( Figure 4—figure supplement 1 ) . Despite the absence of other apoptotic markers in these cells , salivary glands did label with this antibody , suggesting a non-apoptotic function for caspase in this tissue . The Baculovirus p35 protein inhibits both executioner caspases DrIce and Dcp-1 , but not Dronc ( Meier et al . , 2000 ) . Therefore we crossed GMR-p35 , which is a transgene that expresses p35 in the eye imaginal disc posterior to the morphogenetic furrow ( Hay et al . , 1994 ) , into the CasExpress/G-Trace flies . GMR-p35 significantly reduced the number of RFP+ cells in the posterior eye disc compared to the control ( Figure 4 G-I ) , whereas no change in RFP was observed in the antennal disc , which served as an additional internal control . The few remaining RFP+ cells in the posterior eye disc likely were cells that had activated Gal4 prior to the onset of expression of the GMR promoter . GFP expression was still evident , indicating that caspase activation preceded expression of p35 from the GMR enhancer/promoter in those cells ( Figure 4—figure supplement 2 ) . One known non-apoptotic role for caspase activity is in the innate immune response . Specifically the upstream caspase Dredd activates NFkB signaling and expression of anti-microbial peptides ( Meinander et al . , 2012; Leulier et al . , 2000 ) . The gut is known to have a highly active innate immune response . Therefore to determine whether the CasExpress activity we detected in the gut was due to the immune response , we crossed CasExpress into dredd mutant animals . However we detected no difference in the GFP or RFP expression level or pattern between dredd mutants and heterozygous wild type siblings , in any tissue examined ( Figure 4—figure supplement 3 ) . To address when the CasExpress was activated in various tissues , we silenced CasExpress during most of development , by crossing in the temperature-sensitive ( ts ) version of Gal80 ( Gal80ts ) , which represses expression from UAS transgenes even in the presence of Gal4 . When flies carrying CasExpress , Gal80ts , and G-TRACE were grown at 18°C , GFP was completely repressed , and even the infrequent , random clones due to leaky expression of UAS-FLP was suppressed ( Figure 5—figure supplement 1 ) . The large percentage of GFP+ cells in late third instar larval tissues raised the question as to whether caspase was activated in a significant fraction of cells at one particular stage in development , or alternatively whether caspase was activated sporadically in time . To address this question , we grew Gal80ts/CasExpress/G-Trace flies at 18°C and then shifted them to 29°C for 24 hr either at the first instar ( Figure 5 , upper panels ) , the second instar ( Figure 5 , middle panels ) or the mid third instar ( Figure 5 , lower panels ) . We then returned them to 18°C and dissected them at the late third instar larval stage . Rather than all the GFP+ cells arising at one particular stage , sporadic expression was observed regardless of when the temperature shift occurred . This was true in wing ( Figure 5A–C ) , leg ( Figure 5D–F ) and eye-antennal discs ( Figure 5G–I ) . Cells that activated the sensor later produced smaller patches of cells , as expected if the patches represent clonal descendants of a single event . However we cannot rule out the possibility that separated cells that activate the sensor , coalesced into patches based on differential adhesion . In contrast to the discs , caspase activation in the brain was limited to late larval stages ( Figure 5J–L ) . 10 . 7554/eLife . 10936 . 015Figure 5 . Timing of caspase activation in larval tissues . Larvae with CasExpress , G-trace and Gal80ts were grown at 18°C ( blue in the timeline bars on the left ) for 2d ( A , D , G , J ) , 6d ( B , E , H , K ) , or 12d ( C , F , I , L ) , shifted to 29°C ( red in the timeline bars ) for 1d , then kept at 18°C until late third instar . Induction of GFP expression occurs in wing discs ( A–C ) , leg discs ( D–F ) and eye-antennal discs ( G–I ) throughout the larval stage; whereas few cells in brain ( J–L ) survive caspase activation before third instar . Scale bars are 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 01510 . 7554/eLife . 10936 . 016Figure 5—figure supplement 1 . GFP , RFP and DAPI fluorescence in imaginal discs from flies carrying CasExpress , Gal80ts , and G-TRACE raised at 18°C . GFP and RFP are completely repressed . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 016 Apoptosis plays a particularly important role in the nervous system , and cleaved caspase is detected throughout larval CNS ( CNS ) development , both in w1118 and in DQVD sensor flies ( Figure 6—figure supplement 1 ) . Therefore we characterized the temporal and spatial activation of CasExpress in this tissue in more detail . We combined CasExpress , G-Trace and Gal80ts . Flies kept at 18°C to silence CasExpress throughout development exhibited virtually no expression of RFP or GFP . However if we shifted them to 29°C for 12 hr at the early ( Figure 6A ) , middle ( Figure 6B ) , or late ( Figure 6C , D ) third instar and dissected them near the end of larval development , we observed GFP expression in seemingly random locations . Similar numbers of GFP-expressing cells appeared regardless of precise developmental stage . Figure 6E–H shows the patterns observed in 10 different animals , each in a different color , demonstrating the variability . The patterns were clearly not bilaterally symmetrical . We conclude that cells survive caspase activation sporadically during CNS development . This pattern seems more consistent with that expected for developmental anastasis , that is recovery from the brink of apoptotic cell death , rather than a precise role for caspase in the development of a specific cell type . 10 . 7554/eLife . 10936 . 017Figure 6 . Timing of CasExpress activation in larval CNS . Larvae with CasExpress G-Trace and Gal80ts were grown in 18°C ( blue bars in the middle ) until 3rd instar and shifted to 29°C for 12 hr . Then larvae were kept at 18°C for 72 hr ( A , E ) , 48 hr ( B , F ) , 24 hr ( C , G ) or 12 hr ( D , H ) until they reached late 3rd instar . ( A–D ) Four examples of GFP expression patterns in the larval CNS presented in Rainbow RGB , which shows different levels of GFP intensity in different colors . ( E–H ) Z-projections of different samples were slightly transformed and fit into the diagram of brain . The positions of GFP positive cells for each sample are indicated with different colors . Scale bars are 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 01710 . 7554/eLife . 10936 . 018Figure 6—figure supplement 1 . Caspase activity during larval CNS development . ( A , B , C ) Anti-cleaved Dcp1 antibody staining ( green ) detects caspase activity during larval brain development in w1118 ( A ) CasExpress DQVD ( B ) and DQVA control ( C ) . ( A–C ) ( A'–C' ) are Z-projection of images generated by maximum intensity algorithm , ( A''–C'' ) ( A'''–C''' ) are single slices of each image . ( A'–C' ) ( A'''–C''' ) are cleaved Dcp1 only . ( D ) Quantification of the percentage of cDcp-1-expressing cells . Scale bars are 100μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 018 Metamorphosis requires remodeling of some tissues and wholesale destruction and rebuilding of others ( Baehrecke , 2002; Yu and Schuldiner , 2014 ) . Apoptosis contributes substantially to these processes . To address how much of the adult expression arose during metamorphosis in each tissue , we crossed in the Gal80ts repressor and grew flies at 18°C to prevent the induction of CasExpress . We then shifted the flies to 29°C , the non-permissive temperature for Gal80ts , to allow induction only during specific time windows corresponding to larval , pupal , or adult stages respectively ( Figure 7A ) . Distinct patterns were observed in different tissues . In the antenna ( Figure 7B–D” ) and brain ( Figure 7E–G” ) , CasExpress was activated during larval and pupal stages but virtually none was detected in adulthood . Activation during the pupal period could be responsible for remodeling of connections during metamorphosis and was not unexpected , however the more extensive activity during the larval period suggests an additional function for caspase in earlier nervous system development . In midgut enterocytes , some activation occurred during pupal life but more appeared in the adult ( Figure 7H–I” ) , possibly related to the biology of midgut enterocytes which face damage and undergo rapid turnover in adults even under normal physiological conditions . In visceral muscle surrounding the midgut ( Figure 7J–K” ) and in the hindgut ( Figure 7L–M” ) activation was limited to pupal stages , consistent with a role for caspase in metamorphosis of this tissue . 10 . 7554/eLife . 10936 . 019Figure 7 . Developmental timing of caspase activation in adult tissues . ( A ) A schematic of the timing of the temperature shifts ( blue: 18°C , red: 29°C ) during the growth of flies with CasExpress , G-trace , and Gal80ts . ( B–M ) GFP and RFP expression in antenna ( B–D ) , brain ( E–G ) , midgut ( H–I ) , visceral muscle surrounding midgut ( J–K ) , and hindgut ( L–M ) in flies with CasExpress , G-trace , and Gal80ts that grown at the condition indicated in the panels . Panels marked with prime showed separated channels of the left . Arrows in H–I’’ point to some examples of GFP+ progenitor cells . Dotted lines in L–M’’ mark the boundary between midgut and hindgut . Arrows in L and M point to the hindgut proliferation zone . Scale bars in B–D , and H–M are 25 μm . Scale bars in E–G are 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 019 When CasExpress induction was allowed only during the pupal stage , some progenitor cells in the midgut ( Figure 7I–I’’ ) , visceral muscle surrounding it ( Figure 7K–K’’ ) , and the whole hindgut including the proliferation zone , which contains progenitor cells ( Figure 7M–M’’ ) , showed GFP expression . Thus caspase was activated during metamorphosis and some cells survived . This is intriguing because during metamorphosis the larval gut degenerates and the adult gut is reconstituted by progenitor cells ( Micchelli , 2012 ) . We did not detect RFP in progenitor cells at any stage that we analyzed , and we only detected GFP in progenitors when CasExpress was allowed to be active during the pupal stage . Therefore caspase is likely activated for a brief period during pupal life . The progenitor cells , like the rest of the animal , are exposed to apoptotic stimuli such as systemic ecdysone ( Jiang et al . , 1997 ) , yet they survive to reconstitute the adult gut . They might survive either because they are particularly resistant to caspase activity and apoptosis , as is postulated for stem/progenitor cells generally . Alternatively caspases may actually promote their proliferation or maintenance as has been described for some mammalian progenitors ( Li et al . , 2010; Yoneyama et al . , 2014 ) ; or some progenitor cells may undergo anastasis and recover from the brink of apoptotic cell death . Although we cannot currently distinguish definitively between these possibilities , the observation that a subset of progenitor cells activates CasExpress might indicate that some cells resist the apoptotic stimulus prior to activation of caspase-3 whereas others experience caspase activity and recover from it . The observation that pupal visceral muscle cells exhibit RFP and GFP in nearly every cell suggests prolonged caspase activation . Together these observations demonstrate that survival of caspase activation occurs in distinct spatial and temporal patterns for different cell types and tissues , possibly due to differing epigenetic states , developmental mechanisms , and/or physiological functions ( see Discussion ) . We wondered if the observed caspase activity was functionally significant . The homozygous Drice mutant is lethal , as are dronc mutants . The interpretation is that these mutations prevent apoptosis , and that apoptosis is essential . Our results suggest an additional possibility , which is that non-apoptotic caspase activity may be important during normal development . To test this , we crossed two different UAS-p35 transgenes to the rotund-Gal4 ( rn-Gal4 ) line , which expresses in the pouch region of the wing imaginal disc , the region that gives rise to the adult wing blade . We then evaluated the morphology and size of the adult wing for defects in growth and/or patterning . Although the wings appeared normally patterned , they showed a small ( 10% ) but reproducible and significant reduction in area ( Figure 8A–D ) , demonstrating the functional importance of caspase activity in this tissue . We repeated this experiment using engrailed-Gal4 , which drives expression only in the posterior compartment of the wing , and compared the area of the posterior compartment to that of the anterior compartment as an internal control . Again , inhibition of apoptosis by expression of p35 caused a small but significant reduction in size . If the only function of caspase were to promote apoptosis , inhibition of caspase should result in excess cells , and therefore a larger size . The observation of a smaller wing suggests a different function for caspase in this tissue ( see Discussion ) . 10 . 7554/eLife . 10936 . 020Figure 8 . Inhibition of caspase activity reduces wing size . ( A–C ) Representative wings from progeny of rn-Gal4 crossed to ( A ) control w1118 , ( B ) UAS-p35 on chromosome II , or ( C ) UAS-p35 on the X chromosome . ( D ) Quantification of wing area in arbitrary units . ( E ) Schematic showing the regions used for area measurement in wings with or without p35 expressed under en-Gal4 . In the anterior compartment , we measured the area anterior to L3 vein , which is highlighted in blue and marked as A ( L3 ) . In posterior compartment , we measured the area posterior to L4 , which is highlighted in purple and marked as P ( L4 ) . ( F ) Quantification of the ratio between P ( L4 ) and A ( L3 ) in wings from progeny of en-Gal4 crossed to w1118 , UAS-p35 on the second chromosome , and UAS-p35 on the X chromosome . Error bars show standard error of the mean , and **** indicates p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 10936 . 020
When extra apoptosis is artificially induced in Drosophila imaginal disc cells , it stimulates surviving cells to proliferate ( Fan and Bergmann , 2008 ) . Dying cells secrete growth factors to facilitate the survival and proliferation of their neighbors in the process known as compensatory cell proliferation ( Xing et al . , 2015; Kuranaga et al . , 2011 ) . The marking system that we report here demonstrates cell autonomous survival of caspase activation . Both autonomous and non-autonomous survival and proliferation may cooperate to promote recovery of tissues from insults that kill some but not all cells . A role for caspases in injury repair and tissue regeneration has been demonstrated in Hydra , Xenopus , planaria , newts and in mouse liver ( reviewed in [Connolly et al . , 2014] ) , indicating that this is a well-conserved and general phenomenon . Our observation that the majority of cells in the adult fly descend from cells that survive caspase activation at some point suggests that , in addition to the well-documented compensatory proliferation in response to injury , there may autonomous compensatory proliferation in cells that survive caspase-3 activation during normal development . The idea is that some cells die and need to be replaced so the cells that survive proliferate . Such an autonomous increase in proliferation might explain the abundance of GFP-expressing cells in the adult . It could also explain the otherwise paradoxical result that inhibition of executioner caspase activity in the wing imaginal disc by p35 reduced wing area in the adult . If inhibiting caspases only blocked apoptosis , one would expect the tissue to contain excess cells and to be either larger , abnormally patterned , or both . In contrast we observed a small decrease in wing area , consistent with the idea that inhibiting caspase activity might also inhibit compensatory cell proliferation during normal development . An earlier study ( de la Cova et al . , 2004 ) showed that inhibiting apoptosis in the wing disc led to variability in the size of the disc later in development; however this study did not address the ultimate effect on the size of the adult wing . It will be interesting in the future to examine CasExpress in models of injury , repair and regeneration to determine if cell autonomous compensatory proliferation occurs in those settings as well . Two papers document examples of cell recovery from apoptosis during C . elegans development ( Reddien et al . , 2001; Hoeppner et al . , 2001 ) . When phagocytosis is impaired , a fraction of cells that normally die are able to reverse the morphological signs of apoptosis , which are caused by caspase-3 activity . These cells not only survive , they differentiate . One interpretation of these findings is that phagocytosis normally occurs so early in the death process that it prevents anastasis . However development in C . elegans is far more stereotyped than it is in most organisms . In C . elegans the fate of every single cell is precisely determined . However in organisms with greater numbers of cells , cell survival or death is not thought to be a predetermined cell fate; rather there is a selection process in which cells compete ( Moreno and Rhiner , 2014; Merino et al . , 2015; Vincent et al . , 2013; de Beco et al . , 2012 ) . Our results indicate that many cells in adult flies derive from cells that survive caspase activity at some point during their development . An alternative interpretation of the C . elegans studies is that the ability to survive and recover even after caspase-3 activation is a fundamental and ancient cellular property that evolved early and still exists in a latent form , even in an animal that does not normally need it . Even in C . elegans , the precise moment of engulfment is not predetermined; and it is not always the same cell that consumes the dying cell . Therefore in organisms with larger numbers of cells whose fates are far less predictable , it is unlikely that engulfment always occurs at a precise time point during the apoptotic process . The results presented here demonstrate that it is not rare for cells to survive caspase-3 activation during normal Drosophila development , and such cells make a major contribution to normal adult tissues .
The following transgenic and mutant strains were used: The CasExpress biosensor ( pattB-Ubi-CasExpress-DQVD ) and caspase-insensitive control ( pattB-Ubi-CasExpress-DQVA ) were newly generated as follows . First , a backbone pattB-synaptobrevin-7-QFBDAD-hsp70 ( gift from Christopher J . Potter lab ) was linearized with restriction enzymes AatII and BamHI . The poly-ubiquitin promoter was cloned by PCR from pUWR ( Addgene , Cambridge , MA ) , and the product was inserted to backbone by In-Fusion Cloning Kit ( Clontech Laboratories , Mountain View , CA ) . The product , which was verified by sequencing was named pattB-Ubi . Second , pattB-Ubi was linearized with restriction enzymes NdeI and PstI as a backbone . An insert consisting of the sequence of MCD8 , DIAP1 ( residues 2–147 ) and Gal4 in 5’ to 3’ order and two 15 bp sequences overlap with backbone on both 3’ and 5’ end was generated by PCR and In-Fusion cloning . Residues 21 and 22 , immediately following the DQVD cleavage site in DIAP1 , were mutated from sequence NN to GV , in order to protect the cleaved product from possible N-end rule degradation . Third , the insert and backbone were ligated using the In-Fusion kit . A product verified by sequencing was named pattB-Ubi-CasExpress-DQVDGV . Finally , nucleotide 59 of DIAP1 sequence in pattB-Ubi-CasExpress was mutated to change amino acid 20 from D to A by single point mutagenesis . A product verified by sequencing was named pattB-Ubi-CasExpress-DQVAGV . CasExpress and control plasmids were sent to BestGene Inc . , inserted to Perrimon strain P{CaryP}attP40 through a PhiC31 integrase mediated transgenesis . Random insertions of CasExpress-DQVDNN and CasExpress-DQVANN were also generated . DroncI29 was a gift from Kenneth D . Irvine . The following strains were obtained from the Bloomington Stock Center: G-Trace ( Bloomington #28280 ) ; tub-Gal80ts ( Bloomington #7018 ) ; GMR-p35 ( Bloomington #5774 ) ; UAS-p35 BH1 and BH2 ( Bloomington #5072 and 5073 ) . All lines and crosses were kept at 25°C except where otherwise indicated . Larval intestines , oenocytes ( together with surrounding cuticle ) and adult muscles , brains , eyes , ovaries , oviducts , uteri , tissues were dissected in PBS . For adult ventral nerve cords , whole thoraxes were used for fixation . For larval tissues , the anterior 1/3 part of larvae was cut off and turned inside out , all tissues remained attached to cuticle during fixation . Tissues were fixed in 4% paraformaldehyde in PBS at room temperature for 10 min ( larval cuticles with CNS and imaginal discs ) 30 min ( adult thoraces ) . Other tissues were fixed for 15 min . After fixation , adult ventral nerve cords were dissected from adult thoraces . The samples were then washed with PBS/0 . 3% Triton X-100 ( PBSt ) for 3 x 10 min and blocked with 5% goat serum for 30 min . Fluorescence of RFP and GFP were detected directly without antibody staining . Mouse anti-mCD8 ( Santa Cruz , Dallas , TX , #51735 , 1:50 ) , and rabbit anti-Cleaved Dcp-1 ( Asp216 ) ( Cell Signaling , Danvers , MA , #9578 , 1:100 ) were incubated with dissected tissues overnight at 4°C , followed by 3 x 10 min PBSt washing and secondary antibody incubation for 2 hr at room temperature . Samples then were washed twice for 15 min each with PBSt and incubated for 15 min with 10 ng/ml Hoechst 33 , 342 in PBSt . After Hoechst staining , larval CNS and imaginal discs were dissected away from cuticle . All samples were mounted in Vectashield mounting media ( Vector Laboratories , Burlingame , CA , H-1000 ) . A Zeiss AxioZoom microscope was used for imaging whole larvae . A Zeiss LSM 780 confocal microscope was used for the rest of images . Embryo collections , fixation and acridine orange ( Sigma-Aldrich , St . Louis , MO , A6014 ) staining of embryos are as described ( McCall and Peterson , 2004 ) Images were processed with Fiji . A threshold for each channel of interest ( COI , e . g . RFP , GFP ) was set by auto-threshold ( method: Default , Dark ) . For anti-cDcp1 the threshold was set using MaxEntropy . Area above threshold was measured as S[COI] . The area of DNA was measure in the same manner as S[DNA] . A ratio of S[COI]/S[DNA] was then calculated . To test suppression of CasExpress by the caspae inhibitor p35 , larvae with the genotype GMR-p35; CasExpress/G-Trace; TubGal80ts were raised in 18°C until early third instar . Larvae were then transferred to 29°C and incubated for 48 hr . Antennal-eye discs were then dissected , fixed and stained with Hoechst 33342 , followed by a Z-stack imaging on LSM780 microscope ( Objective: 20x Zeiss plan-apochromat dry , 0 . 8 NA; step-size: 1 . 46 µm ) . Images were process with Fiji . A Z-projection of each image was generated by maximum intensity algorithm . An ROI was drawn to define the boundary of the antennal disc . A threshold for the RFP channel was determined by auto-threshold ( method: Default , Dark ) . Another ROI was then drawn to define the boundary of region of the eye disc posterior to the morphogenetic furrow . Threshold of RFP channel of original image without Z-projection was set as ( 0 . 5a , b ) . Area of RFP above threshold was measured for each layer and summed . Area of DNA above threshold determined by auto-threshold ( method: Default , Dark ) was also measured for each layer and summed . The ratio of summed areas of RFP and DNA was then calculated . Crosses were maintained at 25°C . The progeny of the desired genotypes were collected and dehydrated in 100% ethanol . The wings were then mounted in Canada balsam ( Gary’s magic mountant , Sigma ) and photographed using a Zeiss AxioZoom microscope . Wing sizes were quantified using ImageJ software . Statistical significance was determined using the unpaired two-tailed t test for two-sample comparison or one-way ANOVA for multiple samples analysis , with p<0 . 05 set as the threshold for significance . The Tukey test was used to derive adjusted P values for multiple comparisons . In figures , *** indicates p<0 . 001 , and **** indicates p<0 . 0001 . Error bars show standard error of the mean . | Every day , individual cells in our body actively decide whether to live or die . There are enzymes called executioner caspases that help cells to die in a carefully controlled process called apoptosis . Although the activation of executioner caspases generally leads to apoptosis , there are some circumstances in which cells are able to survive . Fruit flies are often used in research as models of how animals grow and develop . Ding , Sun et al . set out to find out more about the circumstances in which cells manage to survive caspase activation in fruit flies . The experiments used a new method that results in cells that survive caspase activity producing a fluorescent marker protein . This allowed Ding , Sun et al . to track when and where these events occurred in the flies . Few cells in fruit fly embryos survive the activation of executioner caspase . However , in later stages of development , more and more cells survive this process . Cells in different parts of the body responded differently . For some types of cells , every cell seemed to survive caspase activity with no evidence of apoptosis . In other tissues like the central brain , in which a few cells normally choose to die , some cells occasionally managed to survive the activation of caspases . This rescue from the brink of death was more common than Ding , Sun et al . had anticipated . The next step will be to uncover the molecular mechanisms that enable the cells to survive caspase activity . This knowledge may help us to develop treatments that can promote the survival of useful cells like heart muscle cells and brain cells , or trigger the death of cancer cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2016 | CasExpress reveals widespread and diverse patterns of cell survival of caspase-3 activation during development in vivo |
A number of TP53-MDM2 inhibitors are currently under investigation as therapeutic agents in a variety of clinical trials in patients with TP53 wild type tumors . Not all wild type TP53 tumors are sensitive to such inhibitors . In an attempt to improve selection of patients with TP53 wild type tumors , an mRNA expression signature based on 13 TP53 transcriptional target genes was recently developed ( Jeay et al . 2015 ) . Careful reanalysis of TP53 status in the study validation data set of cancer cell lines considered to be TP53 wild type detected TP53 inactivating alterations in 23% of cell lines . The subsequent reanalysis of the remaining TP53 wild type cell lines clearly demonstrated that unfortunately the 13-gene signature cannot predict response to TP53-MDM2 inhibitor in TP53 wild type tumors .
A number of TP53-MDM2 inhibitors are currently under investigation as therapeutic agents in a variety of clinical trials across multiple tumor types . Mechanistically , only tumors with wild-type ( WT ) TP53 can potentially be sensitive to TP53-MDM2 inhibitors as confirmed in part by sensitivity of WT MEFs cells and by the loss of sensitivity in TP53 knockout MEFs ( Efeyan et al . , 2007 ) . Therefore , clinical trials of TP53-MDM2 inhibitors only include patients with WT TP53 tumors . Based on pre-clinical work , it is clear that not all WT TP53 tumors are sensitive to TP53-MDM2 inhibitors . Multiple attempts have been made to try to predict sensitivity to TP53-MDM2 inhibitors in WT TP53 tumors . Unfortunately , despite these efforts , there is currently no clinically validated and FDA-approved assay to identify WT TP53 tumors most likely to respond to TP53-MDM2 inhibitors . Recently , Jeay et al . , ( 2015 ) attempted to find a messenger RNA ( mRNA ) predictive expression signature to selective TP53-MDM2 inhibitor NVP-CGM097 using a panel of cell lines from the Cancer Cell Line Encyclopedia ( CCLE ) ( Barretina et al . , 2012 ) with corresponding genetic and genomic datasets . As a result of this work , Jeay et al . , ( 2015 ) described the mRNA signature based on 13 TP53 transcriptional target genes . The signature was generated using TP53-MDM2 inhibitor sensitive versus insensitive cell lines without regard to the TP53 status . As a critical part of the validation work , Jeay et al . , ( 2015 ) used an independent set of 52 cancer cell lines that were considered to be TP53 WT . Since the signature was generated without considering TP53 status and the fact that TP53-MDM2 inhibitors can only be effective in WT TP53 tumors , the signature is likely to represent a proxy for TP53 status . Therefore , it would not be expected to enhance the ability to predict sensitivity to TP53-MDM2 inhibitors in TP53 WT tumors , so the reported predictive ability of the Jeay et al . , ( 2015 ) signature in the set of 52 cancer cell lines considered by authors as TP53 WT is surprising .
One potential explanation for the reported predictive ability of the Jeay et al . , ( 2015 ) signature in a validation set of 52 cancer cell lines that were considered TP53 WT is the possibility that some of these cell lines have TP53 inactivating alterations that were missed during cell lines selection . TP53 could be inactivated by a variety of mechanisms including inactivating mutations , DNA loss and loss of mRNA expression . The CCLE provides sequencing , copy number and mRNA expression data , enabling careful examination of TP53 status in the set of 52 cancer cell lines used for validation by Jeay et al . , ( 2015 ) . Careful examination of TP53 status using publicly available CCLE mutation calls , copy number and mRNA expression ( described in Materials and methods ) identified 12 out of 52 cancer cell lines containing inactivating TP53 alterations , which are summarized in Table 1 . 10 . 7554/eLife . 10279 . 003Table 1 . List of 12 cell lines with inactivated TP53 in the validation set of 52 cancer cell lines considered to be TP53 wild-type by Jeay et al . , ( 2015 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10279 . 003Cell line nameTP53 inactivating mutation ( s ) Alternative reads/reference readsTP53 mRNA ( MAS5-150 201746_at ) TP53 CN ratioJeay et al . 13-gene signature predictionNVP-CGM097 sensitivityKASUMI-1p . R248Q52/02650 . 54insensitiveinsensitiveCOLO-818p . C135R34/02571 . 14insensitiveinsensitiveIGR-37p . C229fs110/1190 . 59insensitiveinsensitiveHCC202p . T284fs35/4140 . 8insensitiveinsensitiveEFM-192Ap . F270fs7/1100 . 74insensitiveinsensitiveNCI-H1568p . H179R89/12020 . 82insensitiveinsensitiveCOLO-783p . P27L38/03041 . 05sensitiveinsensitiveGA-10p . I232N , p . P152L94/50 , 52/764930 . 81insensitiveinsensitiveVMRC-RCWp . I332_splice192/68631 . 65insensitiveinsensitiveJHH-5p . PPQH190del107/412721 . 03insensitiveinsensitiveHDLM-210 . 94insensitiveinsensitiveRERF-LC-KJ251 . 3insensitiveinsensitive As can be seen in Table 1 , the majority of 12 cell lines have inactivating TP53 point mutations , three cell lines with TP53 frame shift mutations exhibit loss of TP53 mRNA expression likely due to nonsense-mediated mRNA decay , two other cell lines also have loss of TP53 mRNA expression . ( Gene expression and Copy Number ( CN ) cutoffs are defined in Materials and methods ) . Importantly , since only TP53 WT tumors have a chance of being sensitive to TP53-MDM2 inhibitors , all 12 cell lines are insensitive to NVP-CGM097 . In order to re-evaluate the performance of the signature in TP53 WT settings , the 12 cancer cell lines with inactivated TP53 listed in Table 1 have been removed from the Jeay et al . , ( 2015 ) validation list of cell lines , resulting in set of 40 likely WT cancer cell lines listed in Supplementary file 1A with information on sensitivity to NVP-CGM097 and Jeay et al . , ( 2015 ) 13-gene signature prediction . Results of reevaluation of signature performance are listed in Table 2 . 10 . 7554/eLife . 10279 . 004Table 2 . Performance of Jeay et al . , ( 2015 ) 13-gene signature prediction in validation set of 40 likely TP53 wild-type cancer cell lines . DOI: http://dx . doi . org/10 . 7554/eLife . 10279 . 004Performance measureCell sensitivity defined by NVP-CGM097Sensitivity89% ( 24/27 ) Specificity15% ( 2/13 ) {DAN-G removal 8% ( 1/12 ) *}PPV68 . 6% ( 24/35 ) NPV40% ( 2/5 ) {DAN-G removal 25% ( 1/4 ) *}Response rate67 . 5% ( 27/40 ) * DAN-G has TP53 mRNA expression of 33 ( MAS5-150 201746_at ) indicating the probable loss of TP53 mRNA . ( Stringent TP53 mRNA expression cutoff is set at 32 ( MAS5-150 201746_at ) to indicate loss of TP53 mRNA ) . NPV - negative predicted value; PPV - positive predicted value . As can be seen from Table 2 , the 13-gene signature positive predicted value for NVP-CGM097 does not noticeably differ from the response rate to the inhibitor . Also the specificity and negative predicted value ( NPV ) are low and it is likely that the actual specificity and NPV are even lower considering that DAN-G cell line has TP53 mRNA expression just above cutoff for TP53 mRNA loss . ( NVP-CFC218 is another TP53-MDM2 inhibitor used by Jeay et al . , ( 2015 ) that is structurally and biochemically very similar to NVP-CGM097 and , as can be seen from Supplementary file 1B , signature has the same pattern of performance for NVP-CFC218 as for NVP-CGM097 ) . Figure 1 provides a visual overview of data used for evaluating the 13-gene signature performance for NVP-CGM097 in the validation set of 40 likely TP53 WT cancer cell lines . Figure 1 clearly illustrates that the Jeay et al . , ( 2015 ) 13-gene signature cannot differentiate between sensitive and insensitive cell lines . Keeping in mind that DAN-G cell line may have TP53 mRNA loss , there is only one cell line , HCC-95 , that is correctly predicted to be insensitive . The presented reanalysis of signature performance in this article strongly suggests that 13-gene signature is a proxy for TP53 status . In such case , one can put forward the hypothesis that HCC-95 may harbor an inactivating alteration ( s ) that has been missed . If this is the case , the Jeay et al . , ( 2015 ) 13-gene signature has zero specificity and zero NPV in the validation set of TP53 WT cancer cell lines . 10 . 7554/eLife . 10279 . 005Figure 1 . Cell lines sensitivity to NVP-CGM097 in validation set of 40 likely TP53 WT cell lines . DOI: http://dx . doi . org/10 . 7554/eLife . 10279 . 005 In clinical sequencing , special care is taken to make sure sufficient coverage is obtained across all target regions in order to reliably detect point mutations , insertions/deletions , fusions and copy number aberrations ( Frampton et al . , 2013 ) . Often the additional step of manual review of sequencing analysis results is added to the workflow to detect false positive/negative calls due to particular sequence composition or computational pipeline artifacts . In preclinical sequencing , such detailed workflow is often too expensive to obtain . In the CCLE , RainDance technology ( Mazutis et al . , 2009 ) was used to fill some blind spots in the hybrid capture process , but such a process does not necessarily mitigate all problematic regions . This may explain the potentially missed TP53 inactivating alteration ( s ) in HCC-95 . In addition to providing sensitivity to NVP-CGM097 in the validation set , Jeay et al . , ( 2015 ) also provided sensitivity data to NVP-CFC218 in the set of 356 CCLE cancer cell lines . In this set , Jeay et al . , ( 2015 ) reported the presence of 113 cell lines with WT TP53 with response rate of 38% to NVP-CFC218 . Based on the detection of 12 cell lines with inactivated TP53 in the validation set of 52 cell lines considered to be TP53 WT by Jeay et al . , ( 2015 ) , a careful examination of TP53 status in the set of 113 cell lines considered to be TP53 WT resulted in identification of 29 cancer cell lines containing inactivating TP53 alterations , which are summarized in Supplementary file 1C . Importantly , since only TP53 WT tumors have a chance of being sensitive to TP53-MDM2 inhibitors , all 29 cell lines are insensitive to NVP-CFC218 . Sensitivity to NVP-CFC218 in the remaining 84 likely TP53 WT cell lines is summarized in Supplementary file 1D . Analysis of this data indicates the response rate of 51% ( 43/84 ) . Based on the current analysis , it is clear that the Jeay et al . , ( 2015 ) 13-gene signature is a proxy for TP53 status . It has a good , but of course not perfect , ability to detect cell lines with inactivated TP53 and this ability could be useful in some of the preclinical work . For example , it could be useful to look at the cell lines that are insensitive to TP53-MDM2 inhibitor ( s ) and also predicted by Jeay et al . , ( 2015 ) 13-gene signature to be insensitive , but not annotated as TP53 inactivated; it is likely that significant fraction of such cell lines harbor undetected TP53 inactivating alterations . In summary , it is clear that Jeay et al . , ( 2015 ) 13-gene signature unfortunately cannot predict response to TP53-MDM2 inhibitor in TP53 WT tumors . Therefore the ability to predict sensitivity to TP53-MDM2 inhibitors in WT TP53 tumors is still out of reach . The development of such prediction capacity would be clinically beneficial and also may provide valuable insights into the understanding of some of important areas of cancer biology .
NVP-CGM097 and NVP-CFC218 pharmacologic cell line profiling data , 13-gene signature predictions have been obtained from ( Jeay et al . , 2015 ) . Affymetrix U133Plus2 mRNA expression , Affymetrix SNP 6 . 0 data , OncoMap mutation calls ( MacConaill et al . , 2009 ) , exome sequencing data ( Hodges et al . , 2007 ) have been obtained from CCLE website ( http://www . broadinstitute . org/ccle/home ) . TP53 mutation calls have been obtained from the following two files: CCLE_hybrid_capture1650_hg19_NoCommonSNPs_NoNeutralVariants_CDS_2012 . 05 . 07 . maf ( 22-May-2012 ) and 1650_HC_plus_RD_muts . maf . annotated ( 24-Nov-2014 ) , both files are available for download from CCLE website . Genomic characterization section in Supplementary methods ( Barretina et al . , 2012 ) provides a detailed description of sequencing data generation and variant calling pipeline . Supplementary file 1E provides COSMIC information on TP53 mutations in cell lines listed in the Table 1 and also includes COSMIC sample ID for each of the cell lines . Copy number ( CN ) ratio is the ratio of signal intensity in a tumor sample versus normal reference samples normalized to total DNA quantity; thus a CN ratio of 1 corresponds to a diploid locus . CN ratio <0 . 6 indicates ‘allelic loss’ . CN ratio <0 . 25 indicates ‘bi-allelic loss’ . All mRNA expression values are MAS5 normalized , with a 2% trimmed mean of 150 ( Hubbell , Liu , and Mei 2002 ) . TP53 Affymetrix ( 201746_at ) mRNA MAS5-150 normalized expression values below 32 are considered to be indicative of TP53 ‘mRNA loss’ . | Damaged cells in the human body can develop into tumors if left unchecked . TP53 ( also called p53 ) is a protein that normally helps to repair or eliminate these damaged cells and prevent tumors from forming . About half of all cancerous tumors have mutations that prevent TP53 from working . In tumors with normal TP53 ( called TP53 wild type tumors ) , another protein that acts to keep TP53 in check is often overly active . This overactive protein ( called MDM2 ) prevents TP53 from suppressing tumor development . Many scientists are developing anticancer drugs called TP53-MDM2 inhibitors to target the potentially overactive protein in TP53 wild type tumors , and importantly only a tumor with working TP53 would have a chance of responding to this kind of inhibitor . Earlier in 2015 , a team of researchers at the Novartis Institutes for BioMedical Research reported the results of a screen of hundreds of cancer cell lines that investigated which ones were sensitive to TP53-MDM2 inhibitors . Using mix of TP53 mutant and TP53 wild type cancer cell lines , the Novartis team identified a set of 13 genes that were highly expressed in cell lines that were sensitive to one of these inhibitors . This 13-gene signature was then suggested as a way to identify which cancer patients with TP53 wild type tumors would be the most likely to benefit from treatment with TP53-MDM2 inhibitors . However , now Dmitriy Sonkin has reanalyzed the validation set of TP53 wild type cancer cell lines used by the Norvartis team and found that many of them had been mistakenly identified as TP53 wild type . That is to say around a quarter of the cell lines thought to have normal TP53 actually had mutations in the gene for TP53 . Sonkin then repeated the analysis using only those cell lines that were from TP53 wild type tumors . This revealed that the 13-gene signature cannot predict how cancer cells from a TP53 wild type tumor will respond to a TP53-MDM2 inhibitor . Further work would be beneficial in order to find an accurate test to determine which cancer patients will benefit the most from treatment with TP53-MDM2 inhibitors . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"computational",
"and",
"systems",
"biology",
"cancer",
"biology"
] | 2015 | Expression signature based on TP53 target genes doesn't predict response to TP53-MDM2 inhibitor in wild type TP53 tumors |
Tau pathology first appears in the transentorhinal and anterolateral entorhinal cortex ( alEC ) in the aging brain . The transition to Alzheimer’s disease ( AD ) is hypothesized to involve amyloid-β ( Aβ ) facilitated tau spread through neural connections . We contrasted functional connectivity ( FC ) of alEC and posteromedial EC ( pmEC ) , subregions of EC that differ in functional specialization and cortical connectivity , with the hypothesis that alEC-connected cortex would show greater tau deposition than pmEC-connected cortex . We used resting state fMRI to measure FC , and PET to measure tau and Aβ in cognitively normal older adults . Tau preferentially deposited in alEC-connected cortex compared to pmEC-connected or non-connected cortex , and stronger connectivity was associated with increased tau deposition . FC-tau relationships were present regardless of Aβ , although strengthened with Aβ . These results provide an explanation for the anatomic specificity of neocortical tau deposition in the aging brain and reveal relationships between normal aging and the evolution of AD .
Alzheimer’s disease ( AD ) is characterized by amyloid-β ( Aβ ) plaques and hyperphosphorylated forms of the tau protein as neurofibrillary tangles ( NFTs ) ( Braak and Braak , 1991 ) . Both of these aggregated proteins can be found in the brains of cognitively normal older adults ( OA ) , suggesting that they reflect the 10–25 year incubation period for AD ( Jack et al . , 2010 ) . Using positron emission tomography ( PET ) and radiotracers that target Aβ and tau , we can now investigate these pathologies in vivo ( Schöll et al . , 2019 ) , and examine how the deposition of these proteins in the aging brain may lead to AD . Neither human PET studies nor autopsy data have pointed to a precise single focus where Aβ deposition begins , rather suggesting that this pathology appears multifocally and soon encompasses the majority of association cortex ( Braak and Braak , 1991; Palmqvist et al . , 2017; Thal et al . , 2002; Whittington et al . , 2018 ) . In contrast , neuropathological studies show that cortical tau deposition begins focally in the entorhinal cortex ( EC ) , specifically in the transentorhinal cortex , i . e . the transition between lateral portions of the EC and perirhinal cortex ( Braak and Braak , 1985; Braak and Braak , 1991; Kaufman et al . , 2018 ) . Strikingly , tau pathology is found in the EC of the majority of OA , including those without concurrent Aβ pathology ( Braak and Braak , 1997; Maass et al . , 2017 ) . The mechanisms that cause tau to spread out of the EC and to other cortical regions are key to understanding , and ultimately preventing , the development of AD . Postmortem studies mapping tau deposition have generated inferences about how tau spreads through the brain , beginning in EC , and then progressing in a stereotyped spatiotemporal pattern first to temporal and limbic regions and then widely throughout association cortex ( Braak and Braak , 1991 ) . Both cross-sectional , and more recently , longitudinal PET studies have largely supported these ‘Braak Stages’ ( Cho et al . , 2016; Harrison et al . , 2019; Maass et al . , 2017; Schöll et al . , 2016; Schwarz et al . , 2016 ) . Cellular and molecular data indicate that tau can spread trans-synaptically through axonal projections , driven by neuronal activity , and inducing pathology in downstream neurons ( de Calignon et al . , 2012; Pooler et al . , 2013; Wu et al . , 2016; Yamada et al . , 2014 ) . Both PET and molecular data indicate that this phenomenon is at least partly accelerated by Aβ ( Hurtado et al . , 2010; Khan et al . , 2014; Pooler et al . , 2015; Schöll et al . , 2016; Vemuri et al . , 2017 ) . This stereotypical pattern of tau distribution in conjunction with the molecular mechanisms of tau spread strongly suggest that tau progresses through the brain along neural pathways . As cortical tau pathology most likely originates in the EC , the patterns of tau spread may be mapped by tracing the connectivity of EC in large-scale neural networks . While the major projection of EC is to the hippocampus , it is also reciprocally connected with limbic and association cortex ( Van Hoesen , 1982; Swanson and Köhler , 1986; Witter et al . , 1989 ) . The human EC contains two anatomically and functionally distinct subregions , the anterolateral EC ( alEC ) and the posteromedial EC ( pmEC ) , which are the human homologues of the lateral and medial entorhinal areas described in rodents ( Maass et al . , 2015; Navarro Schröder et al . , 2015 ) . The alEC is more strongly connected to anterior temporal regions including perirhinal cortex and is involved in processing object-related memory , while the pmEC is more strongly connected to posterior medial regions and is involved in processing spatial memory ( Kerr et al . , 2007; Ranganath and Ritchey , 2012; Schultz et al . , 2012; Reagh and Yassa , 2014; Berron et al . , 2018 ) . The alEC is particularly vulnerable to the effects of aging ( Olsen et al . , 2017; Reagh et al . , 2018 ) and preclinical AD ( Khan et al . , 2014 ) , which has been proposed to be due to its early susceptibility to tau pathology . In contrast , the pmEC seems to be largely spared from early tau pathology ( Braak and Braak , 1985 ) and age- or preclinical AD-related vulnerability ( Khan et al . , 2014; Olsen et al . , 2017; Reagh et al . , 2018 ) , and may not become affected until later in the disease process . This natural dissociation of connectivity and tau deposition allows us to test the hypothesis that functional connectivity ( FC ) patterns of the earliest tau deposition region , that is the alEC , is a better predictor of subsequent cortical tau deposition than the pmEC . The goal of this study was to examine whether tau spreads out of the EC through neuronal pathways in humans by contrasting whether FC networks derived from the alEC and pmEC were differentially associated with patterns of cortical tau deposition in OA . We generated distinct FC networks using seeds in each EC subregion as well as the entire EC with resting state fMRI in healthy young adults ( YA ) . We chose to model FC in YA because of concerns that the networks generated in OA may have been modified by tau pathology ( Schultz et al . , 2017 ) , and thus the YA networks are more likely to reflect healthy adult network structure . These FC networks were then used to examine patterns of tau deposition in the brain , measured with PET in cognitively normal OA . Our main hypothesis was that because tau originates in transentorhinal or lateral portions of the EC , cortical regions functionally connected to alEC should demonstrate more tau pathology than those connected to pmEC or in non-connected cortical regions . Furthermore , we hypothesized that the amount of tau deposition in a region would be proportional to the strength of alEC connectivity to that region , and that relationships between FC and tau would be strengthened in the presence of Aβ .
Fifty-five YA ( aged 20–35 ) and 123 cognitively normal OA ( aged 60+ ) from the Berkeley Aging Cohort Study ( BACS ) were included in the study . All YA participants underwent structural and resting state functional 3T MRI . All OA participants received tau-PET with 18F-Flortaucipir ( FTP ) , Aβ-PET with 11C-Pittsburgh Compound-B ( PiB ) , and a standard cognitive assessment . A subset of OA ( n = 87 ) also completed the same 3T MRI protocol as the YA participants , and their resting state fMRI data was used for supplemental FC analyses . Demographic information for each sample is presented in Table 1 . We investigated the resting state FC of three different entorhinal seeds . To investigate the full extent of the EC , including the transentorhinal region , we used a structural entorhinal seed derived from the FreeSurfer segmentation of each participant’s native space MRI ( Figure 1a ) . To investigate EC subregions , we used template space alEC and pmEC seeds defined in a previous study ( Maass et al . , 2015 ) ( Figure 1b ) . We performed seed-to-voxel FC analyses using the CONN Toolbox ( Whitfield-Gabrieli and Nieto-Castanon , 2012 ) . First-level models were performed with semi-partial correlations , focusing on unilateral seeds and within-hemisphere FC to more accurately approximate EC neural pathways ( see Materials and methods ) . Second-level results were obtained with one-sample t-tests controlling for age and sex , and thresholded at both the voxel ( p<0 . 001 ) and cluster level ( p<0 . 05 , FDR correction ) . Group level patterns of FC derived from the YA sample are depicted in Figure 1c–f . The EC ( Figure 1c ) was functionally connected to other medial temporal lobe structures such as the hippocampus , amygdala , and temporal pole , and lateral temporal lobe structures such as the middle and inferior temporal gyrus . FC was also found with regions such as the angular gyrus , posterior cingulate , and medial frontal cortex . Patterns of alEC FC ( Figure 1d ) were largely limited to anterior temporal regions , such as the medial temporal lobe and inferior temporal gyrus , but also found in angular gyrus . Patterns of pmEC FC ( Figure 1e ) included posterior medial regions such as the parahippocampal gyrus , posterior cingulate , and precuneus , and was also found in the medial orbitofrontal cortex . FC patterns for the alEC and pmEC seeds were distinct , showing minimal spatial overlap ( Figure 1f ) . We repeated these analyses using the data from the 87 OA with fMRI ( Figure 1—figure supplement 1 ) . While the patterns were generally similar , the alEC connectivity was somewhat expanded into typical regions of pmEC connectivity , and pmEC connectivity was reduced . The alEC and pmEC connectivity also spatially overlapped to a greater extent . Importantly , all subsequent analyses shown relating FC to tau utilized the YA FC data . However , we conducted parallel analyses using the OA FC to confirm and extend our findings , which are briefly described at the end of each section and provided in full as supplemental information . To test the hypothesis that tau deposition paralleled patterns of entorhinal FC , we extracted and binarized each YA FC map , removing each seed region from their respective FC mask to derive non-EC FC masks . For the EC seed , we compared tau deposition within regions of FC to cortical gray matter regions that did not demonstrate significant FC ( ‘outside cortical regions’ ) . For the alEC and pmEC subregions , we compared tau deposition between each FC mask and to outside cortical regions not included within either the alEC or pmEC FC masks . We quantified tau deposition in OA as the proportion of suprathreshold FTP voxels ( >1 . 4 SUVR ) within a region , which has been demonstrated as a reliable marker of AD-related tau pathology ( Maass et al . , 2017 ) and is not influenced by region size . We explored effects of Aβ by classifying each OA participant as Aβ- or Aβ+ based upon their global PiB DVR . To compare tau deposition between EC FC regions and outside cortical regions , we performed a repeated measures ANCOVA . We contrasted tau deposition by including region ( EC FC vs . outside cortical regions ) as a within subjects factor , Aβ status as a between subjects factor , and age and sex as covariates . We found a significant main effect of region ( F ( 1 ) =119 . 30 , p<0 . 001 ) ; Figure 2a ) . Post-hoc paired t-tests indicated higher tau deposition within regions of EC FC compared to outside cortical regions ( t ( 121 ) =10 . 01 , p<0 . 001 ) . We further found a significant region by Aβ status interaction ( F ( 1 ) =10 . 38 , p=0 . 002 ) , such that Aβ+ participants had a greater mean difference in tau deposition in regions of EC FC compared to outside cortical regions than did Aβ- participants ( t ( 73 . 57 ) = 3 . 03 , p=0 . 003 ) . We next tested the hypothesis that alEC FC would be a better predictor of tau deposition than pmEC FC or outside cortical regions . We performed a repeated measures ANCOVA with region ( alEC FC vs . pmEC FC vs . outside cortical regions ) as a within subjects factor , Aβ status as a between subjects factor , and age and sex as covariates . We again found a significant effect of region ( F ( 1 . 65 ) = 43 . 88 , p<0 . 001 ) . Post-hoc paired t-tests indicated higher tau deposition within regions of alEC FC compared to pmEC FC ( t ( 121 ) =3 . 42 , p=0 . 001 ) , as well as for both alEC FC ( t ( 121 ) =6 . 97 , p<0 . 001 ) and pmEC FC ( t ( 121 ) =6 . 66 , p<0 . 001 ) compared to outside cortical regions . We also found a region by Aβ status interaction ( F ( 1 . 65 ) = 6 . 09 , p=0 . 005 ) , which was driven by Aβ+ participants having a greater mean difference in tau deposition in regions of alEC FC compared to outside cortical regions than did Aβ- participants ( t ( 59 . 53 ) = 2 . 64 , p=0 . 01 ) , while this difference in alEC FC compared to pmEC FC regions was trending ( t ( 56 . 64 ) = 1 . 90 , p=0 . 06 ) . We repeated these analyses using the OA FC masks and results were consistent with the YA results , except the region by Aβ status interaction in the subregion analysis was reduced to a trend ( p=0 . 07 ) ( Figure 2—figure supplement 1; Supplementary file 1 ) . Based on these results , both entorhinal FC and the presence of Aβ are related to the cortical distribution of tau . We next sought to further probe how the amount of cortical Aβ and EC tau influenced tau deposition at the FC targets of EC . We constructed ‘FC specific’ measures of tau deposition defined as the mean difference per participant between suprathreshold FTP voxels in 1 ) each of the three FC masks compared to outside cortical regions and 2 ) alEC FC compared to pmEC FC regions . We first assessed the relationship between FC specific tau and continuous levels of Aβ across the cortex with a measure of global PiB DVR . As global PiB DVR increased , FC specific tau deposition increased within regions of EC FC ( r = 0 . 37 , p<0 . 001 ) , alEC FC ( r = 0 . 47 , p<0 . 001 ) , and pmEC FC ( r = 0 . 20 , p=0 . 03 ) , as well as within alEC FC compared to pmEC FC ( r = 0 . 41 , p<0 . 001 ) . We repeated this analysis within only Aβ+ participants to ensure that associations were not driven by a floor effect of global PiB DVR . All associations remained significant ( p’s < 0 . 05 ) except for pmEC FC-specific tau deposition ( p=0 . 26; Supplementary file 2 ) . This finding indicated a strong relationship between continuous Aβ levels and proportionally greater tau deposition within EC and alEC FC targets that occurred across the Aβ+ spectrum . Although Aβ is diffusely distributed in cortex , we next examined whether its location in the cortical connectivity targets of EC was important in driving tau spread . We quantified the mean PiB DVR within each FC mask for each participant . Across participants , PiB DVR was significantly higher within regions of pmEC FC compared to alEC FC ( paired samples t-test , t ( 121 ) =26 . 61 , p<0 . 001 ) and to EC FC ( t ( 121 ) =18 . 92 , p<0 . 001 ) . However , associations between FC specific tau deposition and PiB DVR within the FC masks were of similar strength to that of global PiB DVR ( Supplementary file 2 ) , and therefore did not offer more precise information about tau deposition in these FC regions . PiB DVR within the FC masks was highly correlated with global PiB DVR ( r’s > 0 . 99 ) , which may explain the similarity of the findings . Again , all results remained significant when analyzing only the Aβ+ participants , except for pmEC FC specific tau deposition ( Supplementary file 2 ) . Finally , we sought to determine whether higher levels of EC tau were associated with proportionally greater tau deposition within its FC targets , with the hypothesis that more tau would be available to spread from the EC to downstream regions . We therefore examined the relationship between mean EC FTP SUVR and FC-specific tau deposition across subjects . As EC FTP SUVR increased , FC-specific tau deposition increased within regions of EC FC ( r = 0 . 62 , p<0 . 001 ) , alEC FC ( r = 0 . 46 , p<0 . 001 ) , and pmEC FC ( r = 0 . 30 , p=0 . 001 ) , as well as within alEC FC compared to pmEC FC ( r = 0 . 33 , p<0 . 001 ) . We repeated this analysis controlling for global FTP ( mean SUVR across the cortex ) to try to further isolate the effects of EC tau . All correlations remained significant ( p’s < 0 . 02 ) except for pmEC FC-specific tau deposition ( p=0 . 59; Supplementary file 2 ) , indicating that EC tau has a stronger relationship with tau deposition within regions of EC and alEC FC than within pmEC FC . We repeated the Aβ and EC tau analyses using the OA FC masks , and the overall pattern of results was similar ( Supplementary file 2 ) . We next investigated whether stronger average FC between an entorhinal seed and a region was associated with higher levels of tau deposition in that region . We subdivided each seed’s FC mask into regions of low , medium , and high FC based on the YA group-average FC strength ( beta value ) in each voxel using one-dimensional k-means clustering ( see Materials and methods ) . Results of this clustering are depicted in Figure 3a–c . We then calculated the proportion of suprathreshold FTP voxels within each FC strength region for each seed’s FC mask , as this tau measure was not influenced by the different FC strength region sizes . To test whether regions of stronger FC had a higher level of tau deposition , we performed a repeated measures ANCOVA within each seed’s FC mask separately . We contrasted tau deposition between FC strength regions ( low vs . medium vs . high FC ) as a within subjects factor , included Aβ status as a between subjects factor , and age and sex as covariates . We found significant main effects of FC strength for the EC ( F ( 1 . 14 ) = 73 . 29 , p<0 . 001 ) ; Figure 3d ) , alEC ( F ( 1 . 05 ) = 29 . 99 , p<0 . 001; Figure 3e ) , and pmEC ( F ( 1 . 22 ) = 33 . 55 , p<0 . 001; Figure 3f ) seeds . For the EC and alEC , stronger FC was associated with an increase in tau deposition in a stepwise fashion , with low < medium < high FC ( post-hoc paired t-tests , p’s < 0 . 001 ) . However , we found the inverse association for the pmEC , where stronger FC was associated with a decrease in tau deposition ( p’s < 0 . 05 ) . We additionally found a FC strength by Aβ status interaction for the EC ( F ( 1 . 14 ) = 15 . 16 , p<0 . 001 ) , alEC ( F ( 1 . 05 ) = 8 . 43 , p=0 . 004 ) , and pmEC ( F ( 1 . 22 ) = 4 . 46 , p=0 . 03 ) seeds . For the EC , the difference in tau deposition between all FC strength regions was greater in the Aβ+ compared to Aβ- participants ( independent samples t-tests , all p’s < 0 . 01 ) . For the alEC , this interaction was mainly driven by a greater tau deposition difference in the Aβ+ compared to Aβ- participants in high FC compared to medium and low FC ( p’s < 0 . 05 ) , while the medium to low FC comparison was trending ( p=0 . 051 ) . These interactions indicate that Aβ is specifically associated with more tau deposition within regions of stronger EC and alEC connectivity rather than regions of lower connectivity . However , for the pmEC , the interaction showed that while both the Aβ- and Aβ+ groups had the least amount of tau deposition in the high FC regions , the difference between low-high and medium-high was larger for the Aβ+ than the Aβ- subjects ( p’s < 0 . 05 ) , while there was no significant difference in the medium-low FC comparison across groups ( p>0 . 05 ) . This interaction indicates that Aβ is associated with more tau deposition within regions of low and medium pmEC connectivity rather than regions of higher pmEC connectivity . These analyses were repeated using the OA FC masks and provided largely consistent results , except that the interaction between pmEC FC strength and Aβ status was reduced to a trend ( p=0 . 09; Figure 3—figure supplement 1; Supplementary file 3 ) . We found that stronger FC with EC is related to higher cortical tau deposition , and that the amount of EC tau is related to proportionally greater tau deposition within its FC targets , but it was not clear whether the correlation between EC tau and cortical tau was directly related to the FC strength . To test this , we began by performing a voxelwise regression across all OA participants using EC FTP SUVR as the independent variable and cortical FTP SUVR within each voxel as the dependent measure , controlling for age and sex ( see Materials and methods ) . This analysis produced a group-level EC-cortical tau association map that shows voxels in the brain where tau is significantly positively correlated with the amount of tau in the EC across subjects ( Figure 4a ) . The strongest associations were in the medial temporal lobe , middle and inferior temporal gyrus , and posterior parietal lobe , while other significant regions included the retrosplenial cortex , precuneus , and frontal cortex . To determine whether stronger FC between the EC and a voxel was correlated with a stronger association between EC tau and that same voxel’s tau , we performed a voxelwise correlation between the group-average YA FC strength ( beta value ) of each voxel , and the group-average EC-cortical tau association strength ( beta value ) at each voxel . We found a significant positive correlation between EC FC strength and EC-cortical tau association strength ( r = 0 . 44 , rs = 0 . 29 , p’s < 0 . 001; Figure 4b ) , indicating that stronger FC from the EC to a voxel is related to more EC-associated tau in that voxel . We similarly found a positive correlation between alEC FC strength and EC-cortical tau association strength ( r = 0 . 29 , rs = 0 . 15 , p’s < 0 . 001; Figure 4c ) . While pmEC FC strength was also correlated with EC-cortical tau association strength , the relationship was very weak ( r = 0 . 08 , rs = 0 . 04 , p’s < 0 . 001; Figure 4d ) . These results replicated when using the OA group FC strengths ( Figure 4—figure supplement 1 ) .
In this study , we demonstrate that patterns of EC FC predict the spatial topography and level of cortical tau deposition in a sample of cognitively normal OA , where regions with high entorhinal connectivity have more tau . Moreover , connectivity of the alEC subregion more strongly predicted tau in connected cortical regions than pmEC connectivity . These associations were not dependent on the presence of Aβ , however , higher Aβ increased the strength of FC-tau relationships . Additionally , the relationship between FC and cortical tau was closely associated to the amount of tau within EC , suggesting that as tau in EC increases , it begins to spread to connected cortex . Together , these results support a model in which early tau in transentorhinal or lateral portions of the EC spreads to downstream cortical regions through functional connections , accelerated by the presence of Aβ . Using resting state fMRI from YA participants , we obtained healthy patterns of FC from three different entorhinal seeds . The full EC seed , which included the transentorhinal region , was functionally connected to regions that are known to receive structural projections from the EC , including temporal and limbic regions ( Van Hoesen et al . , 1975; Witter et al . , 1989 ) . This pattern agreed with previous work showing entorhinal FC to the default mode network ( Huijbers et al . , 2014 ) . The alEC and pmEC subregions demonstrated distinct FC patterns to anterior temporal and posterior medial regions , respectively , which parallel their unique structural connectivity ( Witter et al . , 1989 ) and functional involvement in the processing of object versus spatial information ( Ranganath and Ritchey , 2012; Reagh and Yassa , 2014 ) . These FC patterns were largely consistent with a previous study investigating cortical FC of the alEC and pmEC ( Navarro Schröder et al . , 2015 ) , and were similar to FC patterns of their closely associated regions: the perirhinal cortex and parahippocampal gyrus ( Kahn et al . , 2008; Libby et al . , 2012 ) . The concordance of our FC networks with anatomical literature and previous FC findings increased our confidence in relating these FC patterns to tau deposition , which was the primary goal of this study . Both the spatial extent and strength of EC and alEC FC was associated with the level of tau deposition in cognitively normal OA . The initial site of cortical tau deposition appears to be transentorhinal cortex and lateral EC ( Braak and Braak , 1985; Kaufman et al . , 2018 ) , which is consistent with other reports suggesting that alEC is particularly vulnerable to effects of age and preclinical AD ( Berron et al . , 2018; Khan et al . , 2014; Reagh et al . , 2018 ) . Our results indicate that alEC may serve as an epicenter of tau pathology , initiating propagation to distant areas of cortex in a topographically specific manner reflecting its connectivity . The idea that tau spreads trans-synaptically via large scale neural networks also has support from other laboratories which have indicated that nodes with higher FC show more tau or higher covariance of tau deposition ( Cope et al . , 2018; Franzmeier et al . , 2019 ) . Evidence for the trans-synaptic spread of tau in humans has also been demonstrated using diffusion tensor imaging ( Jacobs et al . , 2018 ) . Our findings expand upon this previous literature by indicating an anatomic specificity to the earliest stages of cortical tau deposition , and provide an explanation for why tau is preferentially deposited in the anterior temporal network in aging and AD ( Maass et al . , 2019 ) . Patterns of pmEC FC did not predict the location or amount of tau deposition as well as the EC or alEC , and its FC strength demonstrated an inverse relationship with levels of tau deposition . However , tau pathology does occur in the posterior medial network which are targets of pmEC connectivity , such as posteromedial cortex , although this is likely at a later stage than in alEC connected regions ( Braak and Braak , 1991; Cho et al . , 2016; Harrison et al . , 2019; Maass et al . , 2017; Maass et al . , 2019; Vemuri et al . , 2017 ) . Because the alEC has reciprocal axonal projections to cortex , while pmEC mostly lacks these efferent projections ( Witter et al . , 1989 ) , tau may indirectly spread from the alEC to posteromedial regions through multisynaptic projections . This is supported by a recent PET study in humans that suggests tau spreads from the hippocampus to the posterior parietal cortex via the cingulum bundle ( Jacobs et al . , 2018 ) . These data suggest that tau deposition in the posterior medial system is a later stage , dependent on further downstream spread to the hippocampus . This is consistent with longitudinal inferences drawn from cross-sectional autopsy-based Braak staging ( Braak and Braak , 1991 ) . The association between entorhinal FC and tau deposition did not require the presence of Aβ , as we observed associations within our Aβ- sample . This finding is consistent with another recent human neuroimaging study that observed relationships between whole-brain FC and tau covariance strength in OA without Aβ ( Franzmeier et al . , 2019 ) , as well as animal data indicating that tau spreads trans-synaptically without Aβ ( de Calignon et al . , 2012; Wu et al . , 2016 ) . The finding that tau may spread out of the entorhinal cortex to connected cortical regions in the absence of Aβ is important to better understand the results of clinical treatments of AD . Recent Aβ-reducing therapies have failed to prevent AD from worsening ( Egan et al . , 2018; Salloway et al . , 2014 ) . This may be because removing Aβ does not prevent the spread of tau , which is more strongly related to neurodegeneration and cognitive impairment ( Jagust , 2018 ) . However , while Aβ may not be necessary for tau to spread out of the EC , it may still accelerate this spread , as the tau-FC associations we observed were stronger in Aβ+ participants and increased continuously with Aβ levels . While this finding is consistent with other reports in the animal literature ( Hurtado et al . , 2010; Pooler et al . , 2015 ) , stronger tau-FC relationships with increasing Aβ have not been previously demonstrated in humans using neuroimaging . Because local interactions between Aβ and tau in the EC are unlikely in a cognitively normal OA sample due to their separate spatial topographies ( Lockhart et al . , 2017; Sepulcre et al . , 2016 ) , we investigated whether the amount of Aβ at the cortical targets of the EC explained FC-related tau deposition . However , Aβ within the FC targets was not a better predictor of FC-related tau deposition than global levels of Aβ , although these two Aβ measures were highly correlated . It is possible that tau-Aβ interactions might be better explained by soluble oligomeric forms of Aβ not measured by PET , or by other unknown molecular interactions beyond the scope of this study . We also found a strong relationship between the amount of EC tau and the association between FC and cortical tau . Increased tau deposition within the EC was associated with proportionally greater tau deposition within its cortical FC targets , particularly within alEC FC . Additionally , higher EC and alEC FC strength with a cortical target was correlated with a stronger association between EC tau and tau in that cortical target , indicating that the association between EC tau and cortical tau was related to the FC strength between them . Together , these findings suggest that as tau accumulates and increases within EC , it begins to spread to functionally connected cortex in a pattern consistent with its connectivity strength . The alEC is particularly vulnerable to the effects of aging , largely reflecting its early role in the tau pathological cascade . In OA , structural and functional alterations in the alEC seem to be related to both age and AD-related pathology , while the pmEC does not show these same changes ( Berron et al . , 2018; Olsen et al . , 2017; Reagh et al . , 2018 ) . As the alEC is one of the most interconnected regions of the brain ( Bota et al . , 2015; Swanson and Köhler , 1986 ) , alEC tau pathology has the potential to spread to and disrupt widespread cortical regions . Therefore , characterizing the pattern of alEC connectivity and its relationship to tau deposition is a critical factor in understanding and potentially predicting the trajectory of tau spread . While the majority of OA have tau deposition within the EC ( Braak and Braak , 1997; Maass et al . , 2017; Schöll et al . , 2016 ) , the spread of tau to alEC-connected cortex may be one of the first signs of AD or a related neurodegenerative disease such as primary age related tauopathy ( PART ) . Our finding that Aβ is associated with more tau deposition in alEC-connected cortex indicates that the earliest stage of AD , characterized by Aβ-facilitated tau spread , requires a background of normal aging , which partly explains the poorly understood relationship between normal aging and AD . Because tau and Aβ both affect FC ( Schultz et al . , 2017 ) , we chose to use YA FC to establish normal connectivity patterns and to avoid any tau-induced changes that may be late-life alterations that have little impact on long-term patterns of tau spread . It is important to note , however , that the results relating the OA FC data to tau deposition were highly similar . To our knowledge , patterns of alEC and pmEC FC have not previously been described in OA . It is interesting to note that alEC connectivity in OA expanded to encompass regions of posteromedial and prefrontal cortex that are normally associated with the posterior medial scene processing system , while pmEC FC was reduced in scope . This could be due to age-related dedifferentiation of the networks ( Goh , 2011; Maass et al . , 2019 ) , or to atrophy causing more overlap between the seeds and thus increased similarity of the networks . Future examination of age and pathology related changes in these networks using more precise native-space subregions would be useful for better understanding cognitive aging . There are several limitations of our study . First , we acknowledge that measurements of FC based upon fMRI data are indirect measures of neural activity . However , the BOLD signal is reliably associated with neural activity ( Fox and Raichle , 2007 ) , and resting state fMRI has been successfully used to measure FC in previous related studies ( Cope et al . , 2018; Franzmeier et al . , 2019 ) . We were unable to distinguish the directionality of FC , and thus our results may also represent activity directed to our seed regions rather than solely from projections . This lack of directionality also prevents us from determining whether the patterns of tau deposition we observed reflect anterograde or retrograde spread of tau out of the EC , which has important implications for tracking the progression of tau spread . Future studies should aim to differentiate these processes . Additionally , because FC measures may also reflect multisynaptic connections ( Fox and Raichle , 2007 ) , and tau can progress across multisynaptic connections ( Wu et al . , 2016 ) , our results may include the spread of tau from other downstream regions . Further , we note that the resolution of our 3T fMRI data was relatively low compared to previous high-resolution fMRI studies defining functional EC subregions and their connectivity ( Maass et al . , 2015; Navarro Schröder et al . , 2015 ) . While we found anatomically distinct connectivity patterns of the entorhinal subregions resembling previous results ( Navarro Schröder et al . , 2015 ) , there was some minimal overlap between the networks . Although the patterns of our networks still explained vulnerability to tau deposition , the use of higher resolution fMRI may reveal more precise networks and therefore even better predictions for tau spread . Due to the resolution of PET , we were unable to distinguish tau deposition in the EC subregions to confirm that alEC was affected sooner than pmEC; however , neuropathological data and the vulnerability of alEC compared to pmEC strongly suggests this occurrence ( Braak and Braak , 1991; Khan et al . , 2014; Reagh et al . , 2018 ) . Additionally , the 18F-Flortaucipir tracer exhibits known off-target binding to the choroid plexus that prevents accurate measurements of tau deposition within the hippocampus ( Baker et al . , 2017 ) . Thus , we were not able investigate specific associations between hippocampal tau and entorhinal FC . Finally , the cross-sectional nature of these associations limits our interpretation of causality . While it is possible that tau deposition increases FC ( rather than the reverse ) , this seems unlikely because of our use of FC data from YA , who do not have appreciable tau deposition . In addition , considerable evidence suggests that tau reduces neuronal activity and FC strength ( Busche et al . , 2019; Schultz et al . , 2017 ) . Thus , our data appear to support the first stage of a bidirectional relationship between tau and FC , where stronger FC initially guides tau spread , and tau deposition later reduces FC . Our findings provide an explanation for the topographically specific patterns of neuronal vulnerability to tau deposition , suggesting this pattern reflects tau spread from the transentorhinal or lateral EC to functionally connected cortex . The observation that this process is not dependent on , although facilitated by , Aβ suggests that AD may develop from processes of normal aging . Efforts to develop effective treatments for AD are moving towards earlier and earlier detection , and away from Aβ-lowering therapies which have failed in clinical trials . The ability to detect the very earliest spread of tau may be crucial in selecting individuals for tau-directed therapies before symptoms of cognitive decline appear . In order to do this , the biology of how and where tau spreads needs to be better understood .
Fifty-five young adults ( YA ) ages 20–35 were included for analysis . YA participants received structural and functional 3T MRI . YA were recruited for the study through flyers posted on the UC Berkeley campus and by word of mouth . Inclusion criteria included no major health problems , no current or recent history of psychiatric illness , no history of neurological disorders or traumatic brain injury , no substance abuse problems , no depression or antipsychotic medications , and fluency in English . 123 cognitively normal OA from the Berkeley Aging Cohort Study ( BACS ) were included for analysis . All OA participants received tau-PET imaging with 18F-Flortaucipir ( FTP ) and Aβ-PET with 11C-Pittsburgh Compound B ( PiB ) , 1 . 5T structural MRI , and neuropsychological testing as part of normal BACS protocol . A subset of 87 OA also received structural and functional 3T MRI . Eligibility requirements for the BACS participants included age ≥60 years , cognitively normal status ( Mini-Mental State Examination score ≥25 and normal neuropsychological examination , defined as within 1 . 5 SDs of age , education , and sex adjusted norms ) ; no serious neurological , psychiatric , or medial illness; no major contraindications found on MRI or PET; and independent living in the community . This study was approved by the Institutional Review Boards of the University of California , Berkeley , and the Lawrence Berkeley National Laboratory . All participants provided written informed consent . We removed each seed region from their respective FC mask to ensure that FC results did not reflect autocorrelations , and due to our interest in assessing FC between the EC and non-EC cortical regions . To match the resolution of the fMRI data , we smoothed each seed with a Gaussian kernel of FWHM of 4 . 77 mm , calculated by accounting for both the original resolution of the data ( 2 . 6 mm ) and the extra spatial smoothing applied ( 4 mm ) . The smoothed seeds were then thresholded to include voxels that contained >15% influence from the seed , which ensured expanded coverage around the original seed region . These expanded seed regions were then removed from their respective FC mask . To examine associations between FC strength and tau deposition , we classified each seed’s FC mask into regions of low , medium , and high FC strength . To do this , we extracted the group-average beta value from each voxel within each final group level FC mask . The beta value at each voxel represents group-average Fisher’s z’ transformed correlation coefficient between the time series of the seed and that voxel . Because the beta values were non-normally distributed , we performed segmentation of the beta values into FC strength regions in a data driven manner rather than picking arbitrary cut-offs . We applied one-dimensional k-means clustering to the beta values within each FC mask , using the package ‘Ckmeans . 1d . dp’ ( Wang and Song , 2011 ) , implemented in R version 3 . 5 . 1 ( http:// www . r-project . org/ ) . Like traditional k-means clustering , one-dimensional k-means clustering classifies data into groups with minimum variability within each group , but across one dimension ( Wang and Song , 2011 ) . This analysis resulted in each beta value being classified into one of three groups , each representing either low , medium , or high beta values , and therefore FC strength . We then created a separate FC strength mask for the voxels contained within each FC strength group , and calculated the proportion of suprathreshold FTP voxels within each FC strength mask . We compared demographic information between the full sample of OA with PET and the subsample of OA who additionally received fMRI . Categorical variables were compared with Chi-squared tests , and continuous variables were compared with independent samples t-tests . Analyses were performed using SPSS version 25 . Repeated measures ANCOVAs were performed using SPSS . In cases where sphericity was violated , degrees of freedom and corresponding p-values were corrected using Greenhouse-Geisser estimates . Post-hoc analyses further investigating the main effects were performed using paired-samples t-tests , and investigation of Aβ status interactions were performed using independent-samples t-tests . We did not further explore any effects of age or sex , as they were entered into the model as covariates of no interest . ANCOVA main effects and interactions , as well as post-hoc t-tests , were considered significant at a p<0 . 05 threshold . We explored associations between EC tau and cortical tau using voxelwise regressions across all OA participants with SPM12 . The mean partial volume corrected FTP SUVR was extracted from separate left and right hemisphere EC ROIs for each OA participant . For each unilateral EC ROI , we entered mean FTP SUVR as the predictor , age and sex as covariates of no interest , and the voxelwise FTP images as the dependent variables . Results were thresholded at the voxel level ( p<0 . 001 uncorrected ) and at the cluster level ( cluster size >100 voxels ) . To be consistent with the FC results , we then extracted left hemisphere results for the left hemisphere EC ROI , and right hemisphere results for the right hemisphere EC ROI , and combined them to produce the final voxelwise regression t-map . We then investigated the voxelwise correlation between FC strength and the EC-cortical tau associations . We first took the results of the EC-cortical tau regression analysis , where the beta value at each voxel represented the strength of association between FTP in the EC and in that voxel , and masked it with the significant FC mask of each seed . We next extracted ( 1 ) the beta value within each FC mask , representing FC between the seed region and that voxel , and ( 2 ) the beta value within the masked EC-cortical tau map , representing the association between EC FTP and that voxel’s FTP . For each FC seed , we performed a correlation across all the beta values from the FC mask and the beta values from the masked EC-cortical tau regression with both Pearson’s and Spearman’s correlation to account for the distribution of the data . | The changes in the brain that cause Alzheimer's disease begin up to 25 years before the first symptoms appear . During this long incubation period , two proteins accumulate in brain tissue: amyloid-β and tau . Amyloid-β forms clumps known as plaques , while tau forms structures called tangles . But whereas amyloid plaques accumulate evenly throughout the brain , this is not the case for tau . Instead tau accumulates first within a region called the entorhinal cortex , which is important for memory . Findings in animals suggest that tau then spreads out of the entorhinal cortex to other brain regions through neural connections . The entorhinal cortex itself consists of two subregions , which each accumulate tau at different times . The anterolateral subregion ( or alEC for short ) develops tau first , followed by the posteromedial subregion ( pmEC ) . These two subregions process different types of memory and so have connections to different areas of the brain . Does tau therefore spread to brain regions connected to the alEC before it spreads to regions connected to the pmEC ? To test this prediction , Adams et al . scanned the brains of healthy young adults to map their brain connectivity patterns . Young adults were chosen because the aging process itself can alter this connectivity . The brains of healthy older adults , aged 60 or more , were then scanned to measure amyloid-β and tau . None of the older adults had cognitive symptoms of Alzheimer's disease . Despite this , many showed deposits of amyloid-β and tau in their brains . As predicted , alEC-connected regions contained more tau than pmEC-connected regions . Indeed , the stronger the connection between a brain region and the alEC , the more tau that region contained . These relationships occurred in older adults with and without amyloid-β in their brains . However , they were stronger in the individuals with amyloid-β . This adds to evidence suggesting that amyloid-β promotes the spread of tau . Future experiments should measure how tau spreads within an individual's network of connections over time . In the long run , researchers may even find that therapies that stop tau from spreading out of the alEC could help prevent Alzheimer's disease from taking hold . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2019 | Cortical tau deposition follows patterns of entorhinal functional connectivity in aging |
Different types of neurons in the retina are organized vertically into layers and horizontally in a mosaic pattern that helps ensure proper neural network formation and information processing throughout the visual field . The vertebrate Dscams ( DSCAM and DSCAML1 ) are cell adhesion molecules that support the development of this organization by promoting self-avoidance at the level of cell types , promoting normal developmental cell death , and directing vertical neurite stratification . To understand the molecular interactions required for these activities , we tested the functional significance of the interaction between the C-terminus of the Dscams and multi-PDZ domain-containing scaffolding proteins in mouse . We hypothesized that this PDZ-interacting domain would mediate a subset of the Dscams’ functions . Instead , we found that in the absence of these interactions , some cell types developed almost normally , while others resembled complete loss of function . Thus , we show differential dependence on this domain for Dscams’ functions in different cell types .
The vertebrate retina provides an advantageous model to study how specific neuronal cell types organize themselves during development to form functional circuits . The ~100 neuronal cell types of the retina vertically organize into layers , with light-transducing photoreceptors in the outermost cellular layer ( ONL , or outer nuclear layer ) and retinal ganglion cells ( RGCs ) – responsible for the sole output from the retina – in the innermost layer ( RGL , or retinal ganglion cell layer ) . Between these , both functionally and physically , are the interneuron cell types ( horizontal , amacrine , and bipolar cells ) in the inner nuclear layer ( INL ) , which process and relay visual information from photoreceptors to RGCs . These three cellular layers are separated by two synaptic plexiform layers , where the neurites from these cell types mingle and form synaptic connections with specific partners in specific substrata . Additionally , many cell types are horizontally spaced in a mosaic pattern , such that there is a low probability of finding two neurons of the same subtype ( i . e . , homotypic ) in close proximity . This pattern ensures that the information processing provided by each subtype is distributed across the retina ( Masland , 2012 ) . Establishing this pattern requires cells to be able to recognize other cells of the same type and avoid them , while also stably interacting with appropriate synaptic partners ( Garrett and Burgess , 2011 ) . Critical parts of this recognition code include Down syndrome cell adhesion molecule ( DSCAM ) and the highly similar Dscam-like1 ( DSCAML1 ) , collectively referred to as Dscams . Dscam and Dscaml1 encode homophilic members of the Ig-superfamily of cell adhesion molecules , and are expressed in non-overlapping neuronal subtypes in the retina ( Agarwala et al . , 2001; Fuerst et al . , 2009; Yamagata and Sanes , 2008 ) . The Dscams promote self-avoidance at the cell type level: When either gene is mutated , the cell types that normally express the Dscam lose their mosaic spacing and often form clusters ( Fuerst et al . , 2009 , 2008 ) . The neurites fail to evenly cover their receptive fields , and instead form fascicles with neighboring homotypic cells . This clustering and fasciculation is , with few exceptions , homotypic – cells of one subtype rarely cluster with the cells of another subtype . Self-avoidance requires homophilic Dscam interactions between cells , demonstrated in mosaic experiments where neurons lacking Dscam fasciculate with homotypic neurons with intact Dscam ( Fuerst et al . , 2012 ) . This self-avoidance function is consistent with studies in Drosophila , which have four Dscam genes . Most notably , Dscam1 promotes self-avoidance at the individual cell level by using alternative splicing to produce 19 , 008 distinctly homophilic isoforms , allowing each neuron to recognize and avoid 'self' while still interacting with 'non-self' during processes like dendrite arborization and axon branching ( reviewed in [Zipursky and Grueber , 2013] ) . Dscams are also required for normal developmental cell death . In the mutant mice , there is an overabundance of each affected cell type , resulting in a severe expansion of the retina through the cellular and plexiform layers . The extent of cell number expansion varies with cell type . Some cell types are expanded beyond even that seen in mutants for the pro-apoptotic Bax gene , while others are more modestly expanded compared to Bax mutants ( Fuerst et al . , 2009 , 2008; Keeley et al . , 2012 ) . Dscams also contribute to the vertical organization of the retina . In chick , Dscams label-specific sublaminae of the IPL and can instruct neurite targeting to these layers ( Yamagata and Sanes , 2008 ) . DSCAM protein localization is punctate throughout the IPL in mouse , and is not confined to specific sublaminae ( de Andrade et al . , 2014 ) . Despite this , some neuronal types do have disorganized neurite stratification in Dscam mutants , although the disorganization varies with genetic background ( Fuerst et al . , 2010 ) . There are also indications that the synaptic connections that form do not mature normally . For instance , Dscaml1 is expressed both in rod bipolar cells and AII amacrine cells , which connect at dyad ribbon synapses in the IPL . These synapses can still be found in Dscaml1 mutants , but are morphologically abnormal , with indistinct , detached presynaptic ribbons , and functionally abnormal , with much slower decay of the synaptic current ( Fuerst et al . , 2009 ) . The mechanisms by which the Dscams mediate their developmental functions are unknown . One attractive hypothesis is that different functions , including cell death , self-avoidance , and synapse maturation , are mediated by different molecular interactions in the cytosol . Signaling molecules have been found in complex with Dscam1 in Drosophila ( e . g . , Dock/Pak , Ableson , tubulin binding cofactor D [Okumura et al . , 2015; Schmucker et al . , 2000; Sterne et al . , 2015] ) and DSCAM in vertebrates ( e . g . , PAK1 , FAK , Fyn [Purohit et al . , 2012] ) . Both DSCAM and DSCAML1 also have canonical PDZ-interacting motifs at their C-termini by which they interact with scaffolding proteins in the MAGI ( membrane-associated guanylate kinase with inverted orientation ) and PSD95 families ( Yamagata and Sanes , 2010 ) . Because this motif is common to both Dscams , we chose to test the functional significance of these C-terminal interactions , with the initial hypothesis that this interaction would be required for a specific subset of Dscams’ functions . We engineered mouse mutations in which the sequences encoding the final 10 amino acids of DSCAM and DSCAML1 , which interact with PDZ domain-containing proteins , were replaced with epitope tags . Contrary to our initial hypothesis , we found that rather than distinguishing phenotypes based on specific molecular mechanisms , these mutations distinguished cell types . Some cell types were essentially like controls in cell number , spacing , and stratification , whereas others were nearly as disrupted as null mutants . Still other cell types were intermediate in severity . Together , these results demonstrate that different cell types have different dependencies on the PDZ-interacting C-termini of Dscams for function , indicating multiple intracellular molecular mechanisms are involved .
The molecular interactions through which DSCAM could function may involve extracellular interactions , intracellular interactions with other membrane proteins , or initiation of intracellular signaling pathways . We reasoned that a tractable first step in dissecting these possibilities would be to disrupt the C-terminus of DSCAM . Both DSCAM and DSCAML1 have canonical PDZ-interacting domains at their C-termini , and have been shown to interact with PSD-95 and MAGI family members ( Yamagata and Sanes , 2010 ) ( Figure 1—figure supplement 1 ) . To assess the functional relevance of DSCAM’s PDZ-binding motif , we generated a targeted allele of Dscam , in which the sequence encoding the C-terminal 10 amino acids was replaced with sequence encoding a Myc epitope tag ( DscamΔC , see Figure 1A and Materials and methods ) . Deletion of these final 10 amino acids disrupts the canonical binding to the hydrophobic pocket of PDZ domains ( Doyle et al . , 1996 ) , and this DSCAM-∆C mutation markedly reduced MAGI-3 association in co-immunoprecipitation experiments with the DSCAM intracellular domain ( ICD ) when co-transfected in HEK293T cells ( Figure 1B ) . The residual interaction may be an artifact of overexpression , or may reflect interactions between the ICD of DSCAM and MAGI3 that are not dependent on the canonical C-terminal PDZ-interacting domain , but nonetheless , the affinity is greatly reduced and given results described below for DSCAML1-△C , we have successfully interfered with the interactions between Dscams and PDZ-domain-containing proteins using this strategy . Unlike null mutants , which do not survive on a C57BL/6 background , DscamΔC/ΔC mice survived without hydrocephaly or any of the overt phenotypes observed in null animals ( Amano et al . , 2009; Fuerst et al . , 2010 , 2008 ) . 10 . 7554/eLife . 16144 . 003Figure 1 . The C-terminus of DSCAM is not required for protein stability or localization . ( A ) In the Dscam∆C allele , the sequence encoding the final ten amino acids was replaced with a Myc tag by homologous recombination . ( B ) Western blots of protein immunoprecipitated from HEK293T cells co-transfected with MAGI-3 and V5-tagged DSCAM intracellular domain ( ICD ) or V5-tagged DSCAM-∆C ICD ( ∆C ) demonstrates that the ∆C mutation disrupts the PDZ-binding of DSCAM’s C-terminus . ( C ) Western blots of DSCAM protein immunoprecipitated from neonatal brains showed no change in the size or amount of DSCAM in Dscam∆C/∆C mutants . The antibody specificity is confirmed by the lack of signal from Dscam-/- brains . ( D–F ) Immunofluorescent labeling for DSCAM in vertically sectioned retinas from 3-week old mice demonstrated that the protein is found in a normal , punctate localization in the synaptic plexiform layers in Dscam+/+ ( D ) and Dscam∆C/∆C ( E ) mice , consistent with earlier reports for wild type DSCAM ( de Andrade et al . , 2014 ) . No staining above background was found in Dscam-/- retinas ( F ) , demonstrating the specificity of the DSCAM antibody . ( G–I ) Hematoxylin and eosin staining of adult retinas shows that , compared to controls ( G ) , Dscam-/- retinas ( I ) are severely expanded and disorganized . Dscam∆C/∆C retinas ( H ) have modest expansion , but not the extensive disorganization found in the null mutant . Scale bar is 100 µm . See also Figure 1—figure supplement 1 and Figure 1—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 00310 . 7554/eLife . 16144 . 004Figure 1—figure supplement 1 . DSCAM’s C-terminus interacts with PDZ domains . ( A , B ) Both MAGI-2 and MAGI-3 were found to interact with DSCAM by yeast two-hybrid . The C-terminal 20 amino acids of DSCAM ( PDZ ) were used as bait and the Gal4 binding domain alone ( GBD ) was used as a negative control . Successful interaction is revealed by the expression of β-galactosidase on LacZ ( A , note colony color ) and expression of HIS3 promoting survival on plates without histidine ( B , -His ) . When transfected into HEK293T cells , both DSCAM ( C ) and DSCAM-∆C ( D ) protein localized to the cell surface as revealed by live cell staining with an antibody raised against the entire extracellular domain . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 00410 . 7554/eLife . 16144 . 005Figure 1—figure supplement 2 . DSCAM protein localization is grossly unchanged in Dscam∆C/∆C retinas . ( A–F ) Confocal images of P9 retinas immunolabeled for DSCAM and melanopsin show normal colocalization between the proteins in control ( A–C ) Dscam∆C/∆C animals ( D–F ) . The colocalization is quantified in G . N = 4 retinas per genotype . *p<0 . 05 . ( H–L ) Confocal images of P7 retinas immunolabeled for DSCAM collected under standardized microscope settings from Dscam+/+ ( H ) , Dscam-/+ ( I ) , Dscam∆C/∆C ( J ) , and Dscam-/- ( K ) retinas show the relative staining intensities in each mouse . ( L ) Intensities were measured along a 10 µm line adjacent and perpendicular to the INL , a region that includes S1 . Fluorescence intensity in Dscam∆C/∆C retinas was not reduced . Box plots represent the median , first and third quartile , range , and outliers . N = 6 retinas each for Dscam+/+ , Dscam-/+ , and Dscam∆C/∆C genotypes and 2 retinas for Dscam-/- . *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 005 Consistent with the milder phenotype of DscamΔC/ΔC mice , the protein appears to be stable and properly localized . When transfected into HEK293T cells , DSCAM-∆C protein was produced and targeted to the membrane at levels similar to that of full length DSCAM ( Figure 1—figure supplement 1 ) . To test if DSCAM-∆C protein was stable in vivo , we immunoprecipitated protein from neonatal brains using an antibody against DSCAM , and performed Western blots with a second anti-DSCAM antibody . The relative abundance and size of DSCAM protein was indistinguishable between Dscam+/+ and DscamΔC/ΔC samples , while negative control Dscam-/- samples were devoid of DSCAM protein , as expected ( Figure 1C ) . In the retina , DSCAM acquires a punctate localization in the synaptic plexiform layers ( de Andrade et al . , 2014 ) . At three weeks of age , there was no obvious mislocalization or reduction in labeling in cryosections from DscamΔC/ΔC retinas labeled for DSCAM by immunofluorescence ( Figure 1D–F , and Figure 1—figure supplement 2 ) . In contrast to spontaneous Dscam-/- mutants , which display severely disrupted retinal histology , with expanded and grossly disorganized inner nuclear , inner plexiform , and retinal ganglion cell layers ( Fuerst et al . , 2008 ) ( Figure 1G , I ) , DscamΔC/ΔC retinas showed some expansion and disorganization , but a milder phenotype than in Dscam-/- animals ( Figure 1H ) . Thus , DscamΔC/ΔC mice produce and localize DSCAM normally , and have a less severe gross histological defect . This intermediate phenotype could be the result of milder , partial-loss-of-function phenotypes in all Dscam-expressing cell types , or it could be that now only some cell types display the Dscam phenotype , whereas others are normal . The latter appears to be the case , based on our results below . We analyzed the spacing and density of dopaminergic amacrine cells ( DA cells , tyrosine hydroxylase-positive , Figure 2A–C ) and bNOS-positive amacrine cells ( Figure 2H–J ) , both of which form homotypic clusters and increase in number in Dscam-/- mutants ( Fuerst et al . , 2008 ) . We measured cell spacing with three distinct tests: density recovery profiling ( DRP ) , Voronoi tessellation domain analysis , and nearest neighbor analysis . Each of these tests provides a measure of spacing independent of overall density . DRP plots the density of cells at binned distances from each individual cell . When cells are randomly spaced , the density within each bin is equal to the overall density . When cells are mosaically spaced , there is an 'exclusion zone' where bins close to each reference cell have a lower density than the overall field . If cells are clustered then the near bins have a higher density than the overall field ( Rodieck , 1991 ) . In addition to providing a visual representation of cell spacing , DRP also calculates a 'packing factor , ' which we used for statistical comparison . A perfectly ordered array of cells has a packing factor of 1 , while an array of cells with no exclusion zone has a packing factor of 0 ( Rodieck , 1991 ) . In Voronoi tessellation domain analysis , each point in the image is assigned to the domain of the nearest cell , creating a tessellation ( e . g . , Figure 2—figure supplement 1 ) . In images of mosaically spaced cells , the domains are relatively uniform in size , while images of more irregularly spaced cells display a greater variance of domain areas . Calculating the ratio of the variance to the mean of these areas controls for overall cell density ( Khiripet et al . , 2012 ) . Finally , nearest neighbor analysis measures the distance from each cell to its nearest neighbor . The nearest neighbor regularity index ( NNRI ) is calculated for each image by dividing the mean nearest neighbor distance by the standard deviation . To control for cell density , the measured NNRI is divided by the NNRI from a randomly generated array of an equal number of points ( Keeley and Reese , 2014; Rodieck , 1991 ) . p-values from all pairwise comparisons are in Supplementary file 1 . 10 . 7554/eLife . 16144 . 006Figure 2 . DSCAM-mediated self-avoidance requires C-terminal interactions in only some amacrine cell types . ( A–C ) Dopaminergic amacrine cells ( stained for tyrosine hydroxylase , TH ) are non-randomly spaced in wild type retinas ( A ) , but lose mosaic spacing and form neurite fascicles in two-week old Dscam∆C/∆C ( B ) and Dscam-/- ( C ) retinas . ( D ) In both mutants , there was a significant increase in cell density . Spacing was quantified by DRP analysis; relative cell densities normalized to the overall density at increasing distances from reference cells are plotted in ( E ) . By Voronoi ( F ) and nearest neighbor ( G ) analyses spacing in Dscam∆C/∆C retinas was not significantly different than in Dscam-/- animals . ( H–J ) Conversely , bNOS-positive amacrine cells were not visibly different between controls ( H ) and Dscam∆C/∆C retinas ( I ) despite clear fasciculation and loss of mosaic spacing in Dscam-/- mice ( J ) . Dscam∆C/∆C values were intermediate between control and Dscam∆C/∆C in cell density ( K ) , DRP ( L ) , Voronoi ( M ) , and nearest neighbor ( N ) analyses , but differences from control were not statistically significant . Means ± s . e . m . are represented in D–E , K–L . Box plots represent the median , first and third quartile , range , and outliers . N = 4–8 retinas per genotype . *p<0 . 05; **p<0 . 01; ***p<0 . 001; n . s . is not significant by Tukey post-hoc test between indicated genotypes or compared to controls . Scale bar is 100 µm . Representative Voronoi domains are in Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 00610 . 7554/eLife . 16144 . 007Figure 2—figure supplement 1 . Examples of Voronoi tessellation domains in Dscam mutants . Representative Voronoi tessellation domains of TH-positive DA cells ( A–C ) and bNOS-positive amacrine cells ( D–F ) show the differential effect of the C-terminal deletion . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 007 Unexpectedly , when these tests were applied to DA and bNOS-positive cells , we found a differential dependence on the PDZ-binding domain . DA cells had a significant increase in cell density and a loss of mosaic spacing in Dscam∆C/∆C mice ( Figure 2D–G , Figure 2—figure supplement 1 ) . By all measures , DA cell organization in Dscam∆C/∆C retinas was indistinguishable from that of Dscam-/- mice . Conversely , bNOS-positive amacrine cells in Dscam∆C/∆C mice showed little , if any , disruption in organization . bNOS-positive amacrine cells maintained a clear exclusion zone in Dscam∆C/∆C mice , and were not significantly different from wildtype controls by any spacing measure , nor did their overall cell density differ from controls ( Figure 2K–N , Figure 2—figure supplement 1 ) . To ask if this differential dependence on the PDZ-interacting C-terminus extended to other cell types , we analyzed two populations of RGCs: intrinsically photosensitive retinal ganglion cells ( ipRGCs ) , and cells labeled in the Cdh3-GFP GENSAT transgenic line ( Osterhout et al . , 2011 ) . ipRGCs include as many as five distinct populations which can be differentiated by morphology , but only three of these subtypes – M1 , M2 , and M3 – are labeled by an antibody to melanopsin postnatally ( Schmidt et al . , 2011 ) . M1 cells stratify their dendrites to the outermost lamina in the OFF region of the inner plexiform retina and stain the brightest for melanopsin . M2 cells stratify in the ON layer much closer to the RGC cell bodies . M3 cells are bistratified , but are much more rare than the M1 or M2 population . ipRGCs are among the most clustered and fasciculated cell types in Dscam-/- mutants; both M1 and M2 cells are severely clustered with tight dendritic fascicles ( Figure 3A , C , and [Fuerst et al . , 2009] ) . In Dscam∆C/∆C mutants , cell bodies were clearly clustered , albeit not as severely as in Dscam-/- mice ( Figure 3B , E–G , Figure 3—figure supplement 1 ) . Cell density appeared to be modestly increased compared to controls , but the difference was not statistically significant ( Figure 3D ) . Thus , in general , ipRGCs were intermediately affected . The Cdh3-GFP RGC cell number was increased in Dscam-/- mutants , and the cell bodies aggregated into clusters ( Figure 3H , J ) . In contrast , cell number was not significantly increased in DscamΔC/ΔC retinas , and the cell body clustering observed in Dscam-/- mutants was not observed ( Figure 3I , K–N , Figure 3—figure supplement 1 ) . Thus , in both amacrine and ganglion cells , some cell types require the PDZ-interacting C-terminus of DSCAM for its function , whereas in other cell types , this domain is largely dispensable . Furthermore , the functions of DSCAM in promoting cell death and self-avoidance change in parallel and are not separated by the perturbation of this intracellular interaction . 10 . 7554/eLife . 16144 . 008Figure 3 . RGCs also display differential dependence on DSCAM C-terminal interactions for self-avoidance . ( A–C ) Melanopsin-positive intrinsically photoresponsive retinal ganglion cells are found in a mosaic pattern in wild type retinas ( A ) , but at two weeks of age , ipRGC cell bodies in Dscam∆C/∆C ( B ) retinas are pulled into clusters similar to those in Dscam-/- ( C ) retinas . ( D ) Overall ipRGC density was not significantly increased in Dscam∆C/∆C retinas . ( E ) By DRP , cell body clustering was intermediate between Dscam+/+ and Dscam-/- retinas . Voronoi ( F ) and nearest neighbor ( G ) analyses also revealed a clear intermediate defect . Values from Dscam∆C/∆C retinas were significantly different from both control and Dscam-/- mutants . ( H–J ) Cdh3-GFP RGCs are mosaically spaced in control retinas ( H ) , and this spacing is not perturbed in the Dscam∆C/∆C retinas ( I ) , but these cells form clusters in Dscam-/- animals ( J ) , indicating interactions mediated by DSCAM’s C-terminus are dispensable to prevent these cells from clustering . ( K ) Cdh3-GFP RGC overall cell density was significantly increased in Dscam-/- retinas , but not in Dscam∆C/∆C mutants . ( L ) A clear exclusion zone is detectable by DRP analysis in Dscam+/+ and Dscam∆C/∆C retinas . This exclusion zone is lost in Dscam-/- animals , where cell density is increased adjacent to any given cell , indicative of clustering . Similarly , Voronoi ( M ) and nearest neighbor ( N ) analyses describe a clear spacing defect in Dscam-/- but not in Dscam∆C/∆C retinas . Means ± s . e . m . are represented in D–E , K–L . Box plots represent the median , first and third quartile , range , and outliers . N = 6 retinas per genotype . *p<0 . 05; **p<0 . 01; ***p<0 . 001; n . s . is not significant by Tukey post-hoc test between indicated genotypes or compared to controls . Scale bar is 250 µm . Representative Voronoi domains are in Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 00810 . 7554/eLife . 16144 . 009Figure 3—figure supplement 1 . Examples of Voronoi tessellation domains in Dscam mutants . Representative Voronoi tessellation domains of melanopsin-positive ipRGCs ( A–C ) and Cdh3-GFP RGCs ( D–F ) show the differential effect of the C-terminal deletion . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 009 One possible explanation for this differential dependence could be that DSCAM-∆C protein is fully functional , but selectively unstable in some cell types , such as DA cells and ipRGCs , which are affected by the C-terminal deletion , but maintained at normal levels in other cell types , such as bNOS-positive amacrine cells , which are largely unaffected by the C-terminal deletion . We used two methods to assess selective instability . First , we quantified the colocalization between DSCAM and melanopsin to see if there was a reduction in DSCAM specifically in ipRGCs . We did not find any difference in this colocalization between Dscam+/+ and Dscam∆C/∆C retinas ( Figure 1—figure supplement 2 ) . Second , we quantified the fluorescence intensity of Dscam staining in cryosections along a 10 µm line perpendicular to the INL directly adjacent to this cellular layer . This region includes the S1 lamina in which DA cells stratify their neurites . No reduced fluorescence intensity was detectable in Dscam∆C/∆C retinas , whereas we could detect a trend towards reduced labeling in heterozygous Dscam+/- images relative to Dscam+/+ controls , ( Figure 1—figure supplement 1E–I ) . Thus , neither immunofluorescence nor immunoprecipitation ( Figure 1B ) provide any evidence supporting protein instability . It remains possible , however , that a subtle difference in cell-type specific protein stability or localization beyond the resolution of these experiments could contribute to the range of phenotypes we have observed . Similarly , the tag could be interfering with function in some cell types; however , we do not believe this to be the case as we performed similar experiments with DSCAML1 using a different tag and obtained similar results , as described below . Dscam and Dscaml1 serve similar functions in the retina , but in different cell types based on their non-overlapping expression patterns . Dscaml1 is expressed in the cells that contribute to the rod circuit , including rod photoreceptors , rod bipolar cells ( RBCs ) , and AII amacrine cells ( Fuerst et al . , 2009 ) , as well as a population of previously undefined cells in the inner nuclear layer . To test if DSCAML1 functions , like DSCAM , require C-terminus-mediated PDZ-interactions in only some cell types , we created mice harboring a similar replacement of the final ten amino acids of DSCAML1 with an epitope tag ( HA ) using the same strategy as described for DSCAM ( Figure 4A ) . DSCAML1-∆C protein was stable and localized to the cell surface in transfected HEK293T cells . In a surface biotinylation assay , comparable proportions of DSCAML1-∆C and full-length DSCAML1 proteins were at the surface ( not shown ) . DSCAML1-∆C protein had the expected membrane topology , but did not co-immunoprecipitate with MAGI-3 ( Figure 4—figure supplement 1 ) . Like Dscam , Dscaml1∆C/∆C retinas were modestly expanded , but not so severely as in Dscaml1-/- mutants ( Figure 4B–D ) . 10 . 7554/eLife . 16144 . 010Figure 4 . DSCAML1-mediated self-avoidance requires C-terminal interactions in only some cell types . ( A ) Dscaml1∆C/∆Cmutant mice were generated by replacing the sequence encoding the final ten amino acids with an HA tag by homologous recombination . See also Figure 4—figure supplement 1 . ( B–D ) Hematoxylin and eosin staining of adult retinas shows that , compared to controls ( B ) , Dscaml1-/- retinas ( D ) are significantly expanded and disorganized . Dscaml1∆C/∆C retinas ( C ) have a more intermediate expansion without the extensive disorganization found in the null mutant . ( E–G ) AII amacrine cells ( Dab1-positive ) are organized in a mosaic pattern in two-week old control retinas ( E ) . This pattern is disrupted in Dscaml1∆C/∆C retinas ( F ) , but not as severely as in Dscaml1-/- retinas ( G ) , where the cells form clusters . ( H ) There was a significant increase in cell density in Dscaml1-/- but not in Dscaml1∆C/∆C animals . I ) DRP analysis revealed an intermediate effect in Dscaml1∆C/∆C retinas between the clear exclusion zone in control and clustering in Dscaml1-/- . AII amacrine spacing was slightly disrupted in Dscaml1∆C/∆C retinas by Voronoi analysis ( J ) but not by nearest neighbor analysis ( K ) . L-N ) Conversely , compared to controls ( L ) the disruption of VGLUT3-positive amacrine cell spacing in Dscaml1∆C/∆C ( M ) retinas was more similar to that in Dscaml1-/- ( N ) retinas . ( O ) VGLUT3-positive amacrine cell density was significantly increased in both Dscaml1∆C/∆C and Dscaml1-/- animals . ( P ) DRP analysis reveals the loss of exclusion zone in both mutants . ( Q , R ) Dscaml1∆C/∆C values were significantly different from controls in both Voronoi and nearest neighbor analyses . Means ± s . e . m . are represented in H–I , O–P . Box plots represent the median , first and third quartile , range , and outliers . N = 6–8 retinas per genotype . *p<0 . 05; ***p<0 . 001; n . s . is not significant by Tukey post-hoc test between indicated genotypes or compared to controls . Scale bars are 100 µm . Representative Voronoi domains are in Figure 4—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 01010 . 7554/eLife . 16144 . 011Figure 4—figure supplement 1 . DSCAML1-∆C has a normal membrane topology , but does not interact with MAGI-3 . Live staining of HEK293T cells transfected with full-length DSCAML1 with an N-terminal HA tag ( A ) and DSCAML1-∆C with a C-terminal HA tag ( B ) show that , for both proteins as expected , the N-terminus is presented to the extracellular space and the C-terminus to the cytoplasm . ( C ) Western blots of protein immunoprecipitated from HEK293T cells co-transfected with MAGI-3 and V5-tagged DSCAML1 intracellular domain ( ICD ) or V5-tagged DSCAML1-∆C ICD ( ∆C ) demonstrates that the ∆C mutation disrupts the PDZ-binding of DSCAML1’s C-terminus . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 01110 . 7554/eLife . 16144 . 012Figure 4—figure supplement 2 . Examples of Voronoi tessellation domains in Dscaml1 mutants . Representative Voronoi tessellation domains of AII amacrine cells ( A–C , Dab1 ) and VGLUT3-positive amacrine cells ( D–F ) show the differential effect of the C-terminal deletion . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 012 We assessed the density and spacing of two cell types: AII amacrine cells and VGLUT3-positive amacrine cells . Both cell types increase in number and lose mosaic spacing in Dscaml1-/- mutants ( Figure 4E , G , L , N and [Fuerst et al . , 2009] ) . AII amacrine cells in Dscaml1∆C/∆C mice were subtly disrupted , but more similar to the control condition than to that of Dscaml1-/- animals ( Figure 4F ) . There was no increase in cell density , and cells did not form clusters in Dscaml1∆C/∆C mice , as they did in Dscaml1-/- animals , as measured by DRP or nearest neighbor analysis ( Figure 4H , I , K ) . There was , however , an increased covariance of Voronoi tessellation domain areas compared to controls ( Figure 4J , Figure 4—figure supplement 2 ) indicative of irregular spacing . Conversely , VGLUT3-positive amacrine cells in Dscaml1∆C/∆C mice were more similar to Dscaml1-/- than to controls ( Figure 4L–N ) . Cell number was significantly increased , and DRP , Voronoi , and nearest neighbor analyses all indicated a significant loss of mosaic spacing ( Figure 4O–R , Figure 4—figure supplement 2 ) . Thus , as with DSCAM , some cell types have a greater dependence on DSCAML1 C-terminus than do others . Increased cell number from a lack of cell death can disrupt mosaic spacing in some cell types , although not as severely as Dscam-/- mutation ( Keeley et al . , 2012 ) . To ask how much the increased cell number in ∆C mutants contributed to abnormal spacing , we compared the cell types affected in Dscam∆C/∆C ( ipRGCs and DA cells ) and Dscaml1∆C/∆C ( VGLUT3-positive amacrine cells ) to Bax-/- mutants . In all three cases , cell density was , if anything , higher in Bax-/- retinas than in ∆C mutants ( Figure 5A , E , I ) , although this was only significant for VGLUT3 cells . ipRGCs were more severely clustered in Dscam∆C/∆C than in Bax-/- animals , as measured by DRP packing factor ( Figure 5B ) and Voronoi domain analysis ( Figure 5C ) , but were not significantly different by nearest neighbor analysis ( Figure 5D ) . DA cell spacing was similarly disrupted in Bax-/-mutants as in Dscam∆C/∆C ( Figure 5F–H ) , while VGLUT3-positive cells in Dscaml1∆C/∆C were significantly more disrupted by Voronoi and nearest neighbor analyses , but not by DRP packing factor , despite a significantly lower cell density than Bax-/- retinas ( Figure 5I–L ) . Therefore , cell spacing of ipRGCs and VGLUT3-positive amacrine cells was more severely disrupted in ∆C mutants than could be explained by increased cell number alone . 10 . 7554/eLife . 16144 . 013Figure 5 . Increased cell density is not sufficient to explain spacing defects . ( A ) ipRGC cell density was similar in Dscam∆C/∆C and Bax-/- retinas . Despite this , clustering was more severe in Dscam∆C/∆C as measured by DRP ( B ) and Voronoi ( C ) , but not by nearest neighbor analysis ( D ) . ( E ) Likewise , DA cell density was similar in Dscam∆C/∆C and Bax-/- retinas . However , DA cell spacing was not significantly different between Dscam∆C/∆C and Bax-/- mutants ( F–H ) . ( I ) VGLUT3-positive amacrine cell density was significantly higher in Bax-/- than in Dscaml1∆C/∆C retinas . Mosaic spacing was more disrupted in Dscaml1∆C/∆C as measured by Voronoi domain analysis ( K ) and nearest neighbor ( L ) but not by DRP ( J ) . Means ± s . e . m . are represented in A–B , E–F , I–J . Box plots represent the median , first and third quartile , range , and outliers . *p<0 . 05; **p<0 . 01; ***p<0 . 001; n . s . is not significant by student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 013 The exception to this observation was DA cells , in which case the increased density in Bax-/- retinas disrupted cell body mosaics to a similar extent as loss of the C-terminus of DSCAM ( Figure 5F–H ) . However , we observed that the DA neurites were not severely fasciculated in Bax-/- mutants ( Figure 6D ) . To separately evaluate neurite fasciculation , we developed a method to quantify fasciculation . In this technique , observers blind to genotype were asked to choose which of two randomly presented images was less fasciculated without regard to cell number . A score was generated for each image based on iterative head-to-head matchups by an Elo algorithm ( Elo , 1978 ) , which uses these win-loss matchups to efficiently sort the images ( Elo score , see Materials and methods ) . Images that the observers consistently deemed less fasciculated received high scores , while those deemed more fasciculated received low scores . The mean score per retina was used to compare across genotypes by a Wilcoxon rank sum test with the Benjamini and Hochberg correction . For DA cells , there was significantly more severe fasciculation in Dscam∆C/∆C and Dscam-/- retinas than in controls or Bax-/- retinas ( Figure 6A–D , M ) . 10 . 7554/eLife . 16144 . 014Figure 6 . Neurite fasciculation is separable from density-dependent cell body clustering . ( A ) DA cell neurites evenly fill their receptive fields in control mice , but form fascicles in Dscam∆C/∆C ( B ) and Dscam-/- ( C ) animals . ( D ) DA fascicles are rarely observed in Bax-/- mutants . ( E ) M2 ipRGC dendrites imaged in the ON region of the IPL are evenly distributed in wild type mice and largely remain so in Dscam∆C/∆C ( F ) and Bax-/- ( H ) mutants , while severe fasciculation is evident in Dscam-/- retinas ( G ) . I ) In the OFF strata of the IPL , M1 ipRGC dendrites are diffusely organized . There is modest fasciculation in Dscam∆C/∆C ( J ) and Bax-/- ( L ) mice , while fasciculation in Dscam-/- retinas ( K ) is much more severe . ( M–O ) Elo ranking of fasciculation severity between genotypes demonstrates that DA neurites ( M ) are clearly fasciculated in Dscam∆C/∆C and Dscam-/- retinas , but not in Bax-/- , which had loss of mosaic cell body spacing . Conversely , ipRGCs in Dscam∆C/∆C did not have significantly more fasciculation than Bax-/- either in the ON ( N ) or OFF ( O ) layers , despite having a more severe cell body clustering . Box plots represent the median , first and third quartile , range , and outliers . *p<0 . 05; **p<0 . 01; ***p<0 . 001; n . s . is not significant by Wilcoxon rank sum test . Scale bar is 250 µm . See also Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 01410 . 7554/eLife . 16144 . 015Figure 6—figure supplement 1 . Fasciculation of Cdh3-GFP RGCs and bNOS-positive amacrine cells . Fasciculation of Cdh3-GFP RGC dendrites ( A–C ) and bNOS-positive neurites ( D–G ) were compared between genotypes at two weeks of age . ( H ) Elo ranking of fasciculation severity demonstrates that Cdh3-GFP RGC dendrites are clearly fasciculated in Dscam-/- retinas , but not in Dscam∆C/∆C or control retinas . ( I ) Similarly , bNOS-positive neurites were fasciculated in Dscam-/- retinas , but not in Dscam∆C/∆C or controls , and there was only mild fasciculation in Bax-/- animals . Box plots represent the median , first and third quartile , range , and outliers . *p<0 . 05; **p<0 . 01; n . s . is not significant by Wilcoxon rank sum test . Scale bar is 250 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 015 We next applied this technique for grading dendrite fasciculation to ipRGCs by collapsing individual z-stacks of images into distinct projections through the ON portion ( Figure 6E–H ) and through the OFF portion ( Figure 6I–L ) of the IPL . Here , clear differences between Dscam∆C/∆C and Dscam-/- retinas were visible . In the ON layer ( predominantly M2 cells ) dendritic labeling was denser directly above clustered cell bodies , but spread out to evenly cover the field in Dscam∆C/∆C images ( Figure 6F ) . The Elo score could not distinguish Dscam∆C/∆C retinas from control or from Bax-/- retinas , but Dscam-/- fasciculation was significantly more severe than in all other genotypes ( Figure 6G , N ) . In the OFF layer ( predominantly M1 cells ) there were smaller and fewer fascicles in Dscam∆C/∆C than in null retinas ( Figure 6J–K ) . Indeed , by Elo score Dscam∆C/∆C retinas were intermediately fasciculated between control and Dscam-/- but were indistinguishable from Bax-/- mutants ( Figure 6O ) . We also assessed the fasciculation of Cdh3-GFP RGC dendrites and found tight fasciculation in Dscam-/- mutants , but no discernable difference between DscamΔC/ΔC and control retinas ( Figure 6—figure supplement 1 ) . Similarly , bNOS-positive amacrine cells had mild fasciculation , if any , in DscamΔC/ΔC and Bax-/- mutants ( Figure 6—figure supplement 1 ) . Thus , increased cell density can contribute to abnormal spacing and dendrite fasciculation in some cell types as previously reported ( Keeley et al . , 2012 ) , but is not sufficient to explain the self-avoidance deficits and especially the fasciculation of processes found Dscam-/- retinas , and to varying degrees , in ∆C mutant retinas . Dscams promote normal neurite stratification in some cell types , although this is sensitive to genetic background . bNOS-positive amacrine cells are largely disorganized in Dscam mutants on a C3H background , but not in mixed C57BL/6 – BALBc strains ( Fuerst et al . , 2010 ) . This has not been assessed in null mutants on a clean C57BL/6 background , as these mice die at birth ( Amano et al . , 2009 ) . We inspected bNOS-positive amacrine stratification in Dscam∆C/∆C retinas , which are on a C57BL/6 background , and found no abnormalities ( Figure 7A–B ) . DA cell and M1 ipRGCs co-stratify correctly in the OFF region of the IPL in Dscam-/- mutants ( Fuerst et al . , 2009 ) , and both correctly targeted in Dscam∆C/∆C mutants as well ( Figure 7C–D ) . DA and M1 ipRGCs provide the only clear example we have seen of co-fasciculation in Dscam-/-mutants ( Fuerst et al . , 2009 ) . In Dscam∆C/∆C mutants , DA cells are severely fasciculated ( Figure 6M ) , while M1 ipRGCs more mildly so ( Figure 6O ) . Interestingly , there is a clear co-fasciculation between these cell types in Dscam∆C/∆C mutants ( Figure 7E ) , suggesting that the M1 fasciculation may be influenced by the DA cells . 10 . 7554/eLife . 16144 . 016Figure 7 . Laminar specificity in ∆C mutants . Neurite stratification in the IPL was analyzed in immunolabeled cryosections . ( A , B ) bNOS-positive amacrine cells stratified normally in Dscam∆C/∆C retinas , as did ipRGCs and DA cells ( C , D ) , which co-stratify in the OFF region adjacent to the INL . ( E ) Imaged en face , DA neurites co-fasciculated with ipRGC dendrites in Dscam∆C/∆C mutants ( arrowheads ) , as we have previously found in Dscam-/- retinas ( Fuerst et al . , 2009 ) . In Dscaml1∆C/∆C mutants , AII amacrine cells ( F , G ) and rod bipolar cells ( H , I ) terminate their processes normally . ( J–M ) TEM analysis revealed that Dscaml1∆C/∆C retinas contained structurally normal dyad synapses between rod bipolar cells and AII/A17 amacrine cells with distinct ribbons ( arrows ) . ( K ) Dscaml1-/- RBC dyad synapses are characterized by excess in synaptic vesicle number and indistinct ribbons . Four retinas analyzed by TEM per genotype , > 10 synapses inspected per retina . ( N , O ) VGLUT3-positive amacrine cells misprojected beyond the ON ChAT layer in Dscaml1∆C/∆C mutants . ( P , Q ) These ectopic neurites became associated with AII amacrine terminals adjacent to the retinal ganglion cell layer ( arrowheads ) . This association was observable at 3 weeks of age ( P ) and persisted through adulthood ( Q , 18 months of age ) . ( R ) These misprojections were quantified by imaging whole-mount retinas stained for VGLUT3 and Dab1 en face and calculating the percent of area occupied by VGLUT3 in projections through Dab1-positive AII amacrine terminals . Means ± s . e . m . at three threshold levels are represented in R . n = 6–8 retinas per genotype . Scale bar is 20 µm in A–O , 110 µm in E , 10 µm in P , Q , and 500 nm in J–M . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 016 In Dscaml1-/- mutants , stratification and synaptic pairing between rod bipolar cells and AII amacrine cells is preserved ( Fuerst et al . , 2009 ) . We found that to be the case in Dscaml1∆C/∆C mice as well: both cell types project to the ON region of the IPL adjacent to the RGL ( Figure 7F–I ) where they connect at dyad synapses . In Dscaml1-/- mice , these synapses displayed features reminiscent of immature synapses , including malformed ribbons and an overabundance of synaptic vesicles ( Figure 7K and [Fuerst et al . , 2009] ) . We inspected the dyad synapses in Dscaml1∆C/∆C retinas by transmission electron microscopy , and failed to detect these synaptic phenotype reminiscent of the Dscaml1-/- mice ( Figure 7L–M ) , indicating that DSCAML1 promotes synapse maturation independent of C-terminal interactions . We determined VGLUT3-positive amacrine cells depend on DSCAML1 C-terminal interactions for self-avoidance and normal developmental cell death ( Figure 4 ) . This population stratifies its neurites to the ON-OFF region of the IPL between the ChAT-positive laminae in wild type mice ( Figure 7N ) . We found that in Dscaml1∆C/∆C mutants , some of these neurites projected past the ON ChAT band into the region where AII amacrines and RBCs target ( Figure 7O ) . Indeed , much of the VGLUT3 labeling below the ON ChAT band was directly adjacent to Dab1-positive AII amacrine terminals ( Figure 7P ) , an association that persisted through adulthood ( 18 months , Figure 7Q ) . We quantified this misprojection in whole mount retinas stained for VGLUT3 and Dab1 imaged en face . We made projections through the Dab1-positive terminals blind to the VGLUT3 channel and to genotype , then quantified the area occupied by VGLUT3 labeling at three threshold levels . VGLUT3 occupied a significantly greater percentage of area in Dscaml1∆C/∆C animals than in controls ( 2-way ANOVA , p<0 . 0001 , Figure 7R ) . Thus , in ∆C mutants , we found cell types displaying defects in all three categories of Dscam function – cell death , self-avoidance , and neurite stratification – as well as cell types with relatively little dysfunction .
We tested the functional significance of the Dscams’ PDZ-interacting domains in vivoby replacing the C-terminal ten amino acids with epitope tags ( Dscam∆C and Dscaml1∆C ) . The Dscams have similar functions in the different cell types in which they are expressed: ( 1 ) They promote developmental cell death , ( 2 ) they promote self-avoidance at the cell type level both between cell bodies and neurites , preventing cell body clustering and fasciculation of processes , and ( 3 ) they can promote the laminar specificity of neurite stratification and synapse maturation ( Fuerst et al . , 2009 , 2008 ) . We hypothesized that interactions with PDZ domain-containing proteins via the C-termini would mediate a subset of these functions . Instead , we found that a subset of cell types required the PDZ interaction for all of these functions; while in other cell types these processes could proceed relatively normally without the Dscams’ C-termini . The phenotypes found in ∆C mutants compared to null mutants are summarized by cell type in Table 1 . 10 . 7554/eLife . 16144 . 017Table 1 . Phenotypes by cell type in ∆C mutants . The phenotypes assessed in Dscam∆C/∆C and Dscaml1∆C/∆C are summarized . n . a . is not applicable , n . s . is not shown . DOI: http://dx . doi . org/10 . 7554/eLife . 16144 . 017Cell typeCell spacingCell densityFasciculationNeurite stratificationDAnearly nullnearly nullnearly nullnormalbNOSnearly wild typenearly wild typenearly wild typenormalCdh3-GFPnearly wild typenearly wild typenearly wild typenormal ( n . s . ) ipRGCIntermediatenearly wild typeIntermediatenormalVGLUT3nearly nullnearly nulln . a . misprojectionAIIIntermediatenearly wild typen . a . normal Phrased genetically , the Dscam-/- and Dscaml1-/- alleles represent a complete loss of function . The functions we have defined , promoting developmental cell death , promoting self-avoidance in both processes and cell bodies to enable mosaic spacing and coverage factor , and promoting laminar specificity and synapse maturation in at least some cell types and genetic backgrounds , are based on the loss of function mutant phenotype . The deletion of the PDZ-binding C-termini of DSCAM and DSCAML1 had several possible outcomes in this context . It could have resulted in an unstable or mis-trafficked protein , causing a complete loss of function . Our data do not support this , as protein is found at normal levels by both immunoprecipitation and immunohistochemistry , is on the cell surface by live cell staining and surface biotinylation , and is in the correct topological orientation , although more subtle or specific defects may still have been missed by these analyses . The PDZ-interacting C-termini of DSCAM and DSCAML1 could also have been totally superfluous for its function . Consistent with previous results ( Yamagata and Sanes , 2010 ) , this also is not the case . A partial loss of function could have resulted in a subset of the phenotypically defined functions being present , and others being lost . Given the interaction with synaptic scaffolding proteins , a phenotype of impaired synapse maturation with normal developmental cell death and self-avoidance was a reasonable anticipated outcome . Interestingly , we instead found a partial loss of function at the level of cell types . Some cell types appear to almost completely require the C-termini , whereas in other cell types they are almost completely dispensable . In yet other cell types , the phenotype was intermediate , between wild type and the null , but all phenotypically defined functions changed roughly in parallel . An additional caveat is that the epitope tags used to replace the PDZ-binding motifs may contribute to these phenotypes . While this is formally possible , we believe that it is unlikely to be the case , as we used two different epitope tags ( Myc and HA ) with two different genes and found a similar range of phenotypes between different cell types . Thus , our main conclusion from these studies is that the requirement for a PDZ-interacting C-terminus in DSCAM or DSCAML1 is variable and depends on the cell type . How cell adhesion molecules function to prevent adhesion and promote self-avoidance remains and interesting question . The other well described mediators of self-avoidance function by generating thousands of distinctly homophilic recognition units ( Zipursky and Grueber , 2013 ) . Dscam1 in Drosophila uses three banks of alternatively spliced exons to produce 19 , 008 isoforms with distinct extracellular domains ( Schmucker et al . , 2000 ) . Each of these three exons encode regions of the protein essential for homophilic binding , resulting in 19 , 008 potential molecules that preferentially recognize other copies of the same isoform ( Sawaya et al . , 2008; Wojtowicz et al . , 2004 , 2007; Wu et al . , 2012 ) . By biased stochastic exon choice , a given neuron expresses an estimated 10–50 isoforms , most of which will differ from neighboring neurons ( Miura et al . , 2013; Neves et al . , 2004; Zhan et al . , 2004 ) . This gives each neuron a distinct fingerprint allowing it to recognize and avoid 'self' through repulsion while still interacting with its neighbors , a process called self/non-self discrimination ( Hughes et al . , 2007; Matthews et al . , 2007; Soba et al . , 2007 ) . As might be expected with such a mechanism in which each cell is uniquely identified , vast isoform diversity is required for Dscam1 to function normally ( Hattori et al . , 2009 , 2007 ) . Similarly , the vertebrate gamma protocadherin cluster ( Pcdhg ) can also produce diverse recognition units . There are only 22 Pcdhg isoforms , specified by alternative promoter choice within the gene cluster instead of alternative splicing , but the resulting proteins form cis multimers with trans homophilic binding specificity at the multimer level ( Schreiner and Weiner , 2010; Thu et al . , 2014; Wu and Maniatis , 1999 ) . This incorporation of isoforms into multimers can generate thousands of distinctly homophilic interactors . Intriguingly , the multimer compositions are regulated not only by promoter choice , but also by the relative expression levels of each isoform . PCDHG promotes self-avoidance in cerebellar Purkinje cells and retinal starburst amacrine cells ( SACs ) ( Lefebvre et al . , 2012 ) . Here again , diversity is required for normal function . When only one isoform was expressed in SACs , self-avoidance at the single cell level was preserved , but interactions between neighboring SACs were aberrant , resulting in circuit level dysfunction ( Kostadinov and Sanes , 2015 ) . Both Dscam1 and Pcdhg promote self-avoidance by conferring self/non-self discrimination . Without extensive isoform diversity , vertebrate DSCAMs are not able to provide this fine level of self-recognition . Our conclusion that DSCAMs function through at least two molecular mechanisms , one that requires a C-terminal PDZ-interacting motif , and one that does not , fits a model in which DSCAMs interact with cell-type-specific adhesion mechanisms to serve their function . This is most easily discussed for self-avoidance , where the deletion of DSCAM or DSCAML1 results in the clumping and fasciculation of homotypic cells – phenotypes that could be described as excessive adhesion . Under this model , the DSCAMs serve a generic role , and the specificity and complexity of the system is conferred by the cell-type-specific adhesion mechanisms . Thus , in cell types in which the PDZ-interacting C-terminus is required , the predominant cell adhesion system may also interact with PDZ-scaffolded complexes . These systems could include Ig-superfamily CAMs such as L1-CAM or NRCAM , or possibly neurexins or neuroligins . Alternatively , the CAMs could interact with PDZ domains indirectly through mediator proteins . In cell types that do not require the C-terminus of DSCAM or DSCAML1 , the predominant cell adhesion systems may not involve PDZ interactions , such as cadherins or protocadherins . In cell types with intermediate phenotypes , a mix of PDZ-dependent and –independent mechanisms may be involved . This is not surprising , as each cell type is expected to express more than one class of cell adhesion molecule . The C-termini of both DSCAM and DSCAML1 interact with at least six different multi-PDZ domain-containing proteins ( Yamagata and Sanes , 2010 ) and ( Figure 1—figure supplement 1 ) . de Andrade and colleagues analyzed the co-localization of DSCAM with seven different multi-PDZ proteins , including MAGI-2 and MAGI-3 , and found little co-localization in the adult IPL . Interestingly , there was increased co-localization during development , indicating that these interactions are dynamic and transient ( de Andrade et al . , 2014 ) . None of the analyzed PDZ proteins were clearly localized to specific laminae , but were punctate throughout the IPL . Thus , DSCAM and DSCAML1 could interact with different PDZ proteins in different cell types throughout development . As mentioned , we have thus far been generally unable to genetically separate functions such as developmental cell death and self-avoidance . Cell types with any phenotype in ∆C mutants had multiple phenotypes , which changed roughly in parallel . For example , DA cells were more numerous with disrupted spacing and neurite fasciculation , while bNOS-positive amacrine cells and Cdh3-GFP RGCs both had a relatively normal density , spacing , and neurite arborization ( Figures 2 , 3 ) . In ipRGCs , cell density was variably increased and did not reach statistical significance , and cell spacing and dendrite fasciculation were both intermediate in severity between controls and null mutants ( Figure 3 ) . This covariation could suggest that there is one primary function and the other phenotypes are secondary to this . Keeley and colleagues tested if cell death was this primary function in DA cells and ipRGCs , and found that while increasing cell density by inhibiting developmental cell death could contribute to spacing defects , it was not sufficient to explain the loss of self-avoidance in Dscam-/- mutants ( Keeley et al . , 2012 ) . We confirmed those findings here , and extended them in three ways . ( 1 ) We added a third cell type , VGLUT3-positive amacrine cells , and found that their spacing was more disrupted in Dscaml1∆C/∆C mutants than Bax-/- mutants despite having a lower cell density ( Figure 5I–L ) . ( 2 ) In ipRGCs , cell clustering was more severe in Dscam∆C/∆Cmutants than in Bax-/- mutants despite comparable cell densities ( Figure 5A–D ) . ipRGC density in Dscam-/- mutants is much higher than in Bax-/- retinas , complicating the correlation of density and spacing defects . ( 3 ) We quantified neurite fasciculation independently from the cell body spacing , finding that cell density can contribute to fasciculation in ipRGCs , but does not significantly drive fasciculation in DA cells or bNOS cells ( Figures 6 , Figure 6—figure supplement 1 ) . Thus , promotion of cell death is not the sole primary function of the Dscams . This is consistent with our previous finding that ipRGC clustering and fasciculation occurs in Dscam-/-mutants even when cell density is severely reduced by Pou4f2 double mutation ( Fuerst et al . , 2012 ) . As we have not been able to induce clustering and fasciculation without an increase in cell density , it remains possible that the self-avoidance function of Dscams is primary and the clustering and fascicle formation is protective against cell death . It is also possible that the process through which the Dscams prevent fasciculation also promotes proper neurite stratification . We have now identified two examples of aberrant interaction between different cell types . DA cells co-fasciculate with M1 ipRGCs in Dscam∆C/∆C retinas ( Figure 7E ) and VGLUT3-positive amacrine cells become directly adjacent to AII amacrine cells in Dscaml1∆C/∆C mutants ( Figure 7P–Q ) . As DA and M1 ipRGCs normally co-stratify , this interaction does not result in any misprojection . VGLUT3-positve amacrine neurites , however , are normally confined in the ON-OFF strata between the ChAT layers , not in the ON region proximal to the RGL where AII projections reside . Improper interactions with AII amacrine cells could provide a mechanism for VGLUT3 misprojection ( Figure 7O–Q ) . In summary , DSCAM and DSCAML1 in different cell types function through different molecular mechanisms . We hypothesize that this mechanistic diversity is based on the diverse adhesion systems masked by the DSCAMs . Determining how DSCAMs interact with other systems to mask their adhesivity for self-avoidance , and whether this is the primary function and developmental cell death and synapse maturation are secondary , will require additional studies . However , what these results clearly demonstrate is that the molecular mechanisms through which DSCAMs function will need to be defined on a cell type by cell type basis , and that a single molecular complex or pathway will not account for DSCAMs’ function ( s ) in all cell types .
All animals were housed in the research animal facility at The Jackson Laboratory under standard housing conditions with a 12:12 light dark cycle and food and water ad libitum . All procedures using animals were performed in accordance with The Guide for the Care and Use of Laboratory Animals and were reviewed and approved by the Animal Care and Use Committee of The Jackson Laboratory . Previously described mouse strains include: Dscam-/- = B6 . CBy-Dscamdel17/Rwb , RRID:IMSR_JAX:008000 , described in ( Fuerst et al . , 2008 ) ; Cdh3-GFP , RRID:MMRRC_000236-UNC , courtesy of Dr . Andrew Huberman ( Osterhout et al . , 2011 ) ; Bax-/- = B6 . 129X1-Baxtm1Sjk/J , RRID:IMSR_JAX:002994 , described in ( Knudson et al . , 1995 ) ; Dscaml1-/- = Dscaml1GT/GT , RRID:MGI:4417834 , described in ( Fuerst et al . , 2009 ) . Dscaml1-/- mice were genotyped with a previously unreported primer set . A common forward primer ( ATGCCACTGTGCCTGGCTGTT ) was used with a reverse primer specific to either wild type sequence ( CCCAGCAGTTGAGTGCCCTGG ) or mutant sequence ( TATCCACAACCAACGCACCCAAGC ) . To disrupt the PDZ-interacting C-terminus of DSCAM , we replaced the sequence encoding the C-terminal ten amino acids with a Myc epitope tag sequence through standard knock-in techniques . The targeting vector was generated using bacterial recombineering of the BAC BMQ-206A7 ( from 129S7/SvEv ) . We made a recombineering targeting cassette by PCR of a loxP-flanked Neomycin cassette . The forward primer had the following sequence: GTGCAGAGCTGGGACAGGCAGCTAAAATGAGCAGCTCCCAAGAGTCACTGCTGGACTCCCGGGGCCATTTGAAAGGAAACGAACAAAAGCTGATCTCTGAGGAAGATCTGTAAcggcgcgcctagtcgacttc . This primer was composed of the 60 bases from 90 to 30 bases upstream of the Dscam stop codon in exon 33 ( underlined ) , Myc coding sequence with a stop codon ( bold ) , and the 5’ end of a loxP-flanked Neomycin cassette ( lowercase ) . The reverse primer had the following sequence: CGGAATTCAGTAAAAAAAAGGTAGCTTTGATTGGCTCGTTTAAATTGTATTTACAACCGCTGTCCATCAGGTGCCATGTGgcttagtttaaactcgagcc , composed of the 60 bases immediately after the stop codon in exon 33 ( underlined ) and the 3’ end of a loxP-flanked Neomycin cassette ( lowercase ) . The BAC was recombineered in SW102 cells as in ( Warming et al . , 2005 ) . The resulting BAC was digested with PmlI and XbaI to create a 9 . 2 kb fragment which was subcloned into pSL1180 at EcoRV and XbaI sites . The vector acquired a point mutation resulting in a G to E substitution two residues before the Myc tag . The 9 . 2 kb targeting vector in pSL1180 was linearized with NotI and electroporated into the C57BL/6N ES cell line JM8 . 96 neomycin-resistant clones were screened by long range PCR . Of five positive clones , two were expanded and injected into albino B6 ( Cg ) -Tyrc-2J/J blastocysts . Chimeric mice were bred to albino B6 ( Cg ) -Tyrc-2J/J mice to detect germline transmission . To remove the loxP flanked Neomycin cassette , mice harboring the mutation were crossed to a line expressing Cre under the CMV promoter - B6 . C-Tg ( CMV-cre ) 1Cgn/J – then backcrossed to C57BL/6J to segregate from Cre . Mice were genotyped by PCR using a primer pair spanning the retained loxP site: CCTCCACCTCTTCCACGCGAGAAG and AGTAGTCTTTGCGCTGTCTGTGG . Dscaml1∆C mice were generated in parallel by the same techniques . The BAC RP23-342M16 was recombineered to replace the 30 bases preceding the stop codon in exon 33 with an HA tag and a loxP-flanked Neomycin cassette . The following primers were used: AGGGACTCACTACTCGAAATGAGCACCCCAGGGGTAGGGCGTTCTCAGAAACAGGGGGCTTACCCATACGATGTTCCAGATTACGCTTAAcggcgcgcctagtcgacttcg and TGGTGTGCGGGGGCTGGAGGCGCAGAGGTCCCAGTGTGGAGCCCTTCTCCATTTGTCGGCgcttagtttaaactcgagcc , corresponding to exon 33 of Dscaml1 ( underlined ) , the HA tag ( bold ) , and the loxP-flanked Neomycin cassette ( lowercase ) . The recombineered BAC was digested with BmtI , and the liberated 7 . 7 kb fragment was subcloned into the pSL1180 vector with a diphtheria toxin negative selection cassette . The resulting targeting vector was linearized by NotI digestion and electroporated into the albino C57BL/6 ES cell line J-A18 ( B6 ( Cg ) -Tyrc-2J/J ) . Clones with confirmed homologous recombination were microinjected into C57BL/6 blastocysts to create chimeric mice . Sperm from chimeric males was genotyped to confirm the presence of the engineered allele , and then used for in vitro fertilization . As with Dscam∆C , the resulting mice were crossed with B6 . C-Tg ( CMV-cre ) 1Cgn/J to remove the Neomycin cassette , then backcrossed to C57BL/6 to segregate from Cre . Mice were genotyped by PCR using a primer pair spanning the retained loxP site: CCTCCATGAGGAACCTGACTCG and CATGACTGGGGATTTCTTTTTGAC . While the epitope tags were useful for analyzing protein in transfected cells , neither allowed for effective labeling in vivo . This was unfortunate , but not uncommon for a single-copy small epitope tag . A yeast two-hybrid screen was performed using a Clontech Matchmaker kit as per the manufacturer’s instructions ( Takara , Mountain View , CA ) . AH109 yeast transformed with a plasmid encoding the C-terminal 20 amino acids of DSCAM fused to the Gal4 binding domain were mated to Y187 yeast pre-transformed with a mouse cDNA library fused to the Gal4 activation domain . Successful interaction resulted in Gal4-driven expression of HIS3 , screened for by survival on plates without histidine and containing 3-AT; ADE2 , screened for by survival on plates without adenine; and LacZ , screened for by activity of the β-galactosidase enzyme . Positive interactors , including MAGI-2 and MAGI-3 , were verified by immunoprecipitation and Western blot using the Gal4 fusion constructs . pCAG-Dscam was used to express full length DSCAM ( Schramm et al . , 2012 ) and pCMV-Dscaml1 to express full length DSCAML1 ( [Yamagata and Sanes , 2008] , Addgene 18738 , Cambridge , MA ) . A Dscam∆C expression construct mimicking the knock-in mutation was made by PCR from pCAG-Dscam using the primers GGGGACAAGTTTGTACAAAAAAGCAGGCTGGACCATGTGGATACTGGCTCTCTCC and GGGGACCACTTTGTACAAGAAAGCTGGGTGTTACAGATCTTCCTCAGAGATCAGCTTTTGTTCGTTTCCTTTCAAATGGCCCCGGGAG . The PCR product was cloned into the Gateway cloning pDONR201 vector by BP reaction ( Thermo Fisher , Waltham , MA ) then into pDEST47 by LR reaction ( Thermo Fisher ) for expression . The Dscaml1∆C expression vector was made through identical steps starting with pCMV-Dscaml1 using primers GGGGACAAGTTTGTACAAAAAAGCAGGCTGGACCATGTGGCTGGTAACTTTCCTCCTG and GGGGACCACTTTGTACAAGAAAGCTGGGTGTTAAGCGTAATCTGGAACATCGTATGGGTAAGCCCCCTGTTTCTGAGAACG . Constructs to express V5-tagged intracellular domains ( ICD ) were made by PCR from pCAG-Dscam and pCMV-Dscaml1 . Both Dscam ICD constructs were made using the forward primer AATTAAGAATTCATGGGTAAGCCTATCCCTAACCCTCTCCTCGGTCTCGATTCTACGAGGAGACGGCGAGAGCAGAGGC where the V5 tag is underlined and the beginning of the ICD is italicized . The wild type ICD construct was made with the following reverse primer: TTATTCTAGATTATACCAAGGTGTAAGATTTTGC while the Dscam∆C ICD construct was made with the following reverse primer: TTATTCTAGATTACAGATCCTCTTCTGAGATGAGTTTTTGTTCGTTTCCTTTCAAATGGCCCC . These PCR products were ligated into the pEYFP-N1 vector ( Clontech ) , with the YFP sequence removed , digested at EcoRI ( sticky ) and NotI ( blunted ) sites . The Dscaml1 ICD constructs were likewise produced with a common forward primer encoding the V5 tag: AATTAAGAATTCATGGGTAAGCCTATCCCTAACCCTCTCCTCGGTCTCGATTCTACGCGAAAGAAGAGGAAGGAGAAGAGGC . Wild type ICD was made with reverse primer: AAATGCGGCCGCCTACACCAGGGTGTAGGATTTGG and Dscaml1∆C with reverse primer TATTGCGGCCGCTTAAGCGTAATCTGGAACATCGTATGGGTAAGCCCCCTGTTTCTGAGAACGC . PCR products were ligated into pEYFP-N1 at EcoRI and NotI sites . The MAGI-3 expression construct was generated by PCR from cDNA from the neonatal brain using the following primers: CACCATGTCGAAGACGTTGAAGAAG and TCACAGCTGTTTGTCAGCCATG . The PCR product was cloned into pENTR/D-TOPO by TOPO cloning reaction , then into pDEST47 by the LR cloning reaction . HEK293T cells were obtained from ATCC ( Manassas , VA , CRL-11268 , RRID:CVCL_1926 , lot 62312975 ) where they were tested free of mycoplasma and their identity was verified by STR analysis . Cells were transfected with Lipofectamine 3000 ( Thermo Fisher ) according to the manufacturer's protocol . Immunoprecipitation from whole P0 brain was performed according to standard protocols using mouse anti-DSCAM ( 1:25 , R&D Systems , Minneapolis , MN , clone 36661 , RRID:AB_2095452 ) and protein G magnetic beads ( Dynabeads , Thermo Fisher ) . Western blots were performed according to standard procedures using goat anti-DSCAM primary antibody ( 1:1000 , R&D Systems , RRID:AB_2230818 ) . For in vitrostudies , immunoprecpitation was performed 48 hr after transfection with an agarose-conjugated anti-V5 tag antibody ( ABCAM , Cambridge , MA , AB1229 , RRID:AB_308681 ) and Western blots with mouse anti-V5 tag ( 1:2000 , Pierce E10/V4RR , RRID:AB_10977225 ) and rabbit anti-MAGI-3 ( 1:1000 , Sigma , St . Louis , MO , RRID:AB_2619643 ) Tissue preparation and immunofluorescence staining were performed as described previously ( de Andrade et al . , 2014; Fuerst et al . , 2009 ) . Whole retinas were isolated and fixed in 4% paraformaldehyde for 4–8 hr . Retinas were stained free-floating in 2 . 5% BSA with 0 . 5% Triton-x-100 in the indicated antibodies for 48–72 hr at 4°C . After washing off unbound primary antibodies , secondary antibodies were applied in the same solution overnight at 4°C . For sectioning , lenses were removed from enucleated eyes . Eyecups were fixed , then cryopreserved in 30% sucrose and frozen in Tissue-Tek OCT ( Sakura , Torrance , CA ) . Cryosections were cut at 12 µm and immunostained on the slide . Primary antibodies were applied overnight in blocking solution at 4°C , and secondary antibodies for one hour at room temperature . For DSCAM staining , eyecups were only fixed for 50 min on ice . Longer fixation time reduced staining efficiency . For live staining of HEK293T cells , antibodies were diluted in PBS with 2 . 5% BSA and applied to cells on ice for 1 hr . Cells were then rinsed in PBS , fixed for 10 min at room temperature in 4% PFA , and counterstained as described above for cryosections . Histological analysis using H&E was performed using standard staining protocols with a Leica automated slide stainer . Tissue was prepared as above but with paraffin embedding . Sections were cut at 4 µm on a Leica microtome for staining . The following antibodies were used at the indicated dilutions: rabbit anti-GFP ( 1:500 , Millipore , Darmstadt , Germany , RRID:AB_91337 ) , mouse anti-DSCAM ( 1:50 , R&D Systems clone 36661 , RRID:AB_2095452 ) , sheep anti-tyrosine hydroxylase ( 1:500 , Millipore , RRID:AB_11213126 ) , rabbit anti-bNOS ( 1:500 , Sigma RRID:AB_260796 ) , rabbit anti-melanopsin ( 1:10 , 000 gift of Dr . Ignacio Provencio at the University of Virginia or Advanced Targeting Systems , San Diego , CA , RRID:AB_1266795 ) , rabbit anti-Dab1 ( 1:500 , Millipore , RRID:AB_2261451 ) , guinea-pig anti-VGLUT3 ( 1:10 , 000 , Millipore , RRID:AB_2187832 ) , rabbit anti-PKCa ( 1:1000 , Sigma RRID:AB_477345 ) , goat anti-ChAT ( 1:400 , Millipore , RRID:AB_2079751 ) , rabbit anti-HA tag ( 1:250 , Sigma , RRID:AB_260070 ) . All secondary antibodies were alexa-fluor conjugates ( 1:500 , Thermo Fisher ) . | Neurons in a part of the eye called the retina detect light and convert it into electrical signals that are sent to the brain . Different types of neurons in the retina are arranged vertically into layers and horizontally in a mosaic pattern so that two neurons of the same type are not next to each other . To establish this highly organized pattern , neurons in the developing retina must be able to recognize other neurons of the same type and avoid moving towards them – a process referred to as self-avoidance . A group of proteins called the Dscams are found on the surface of neurons and play key roles in positioning them in the retina . Dscams promote self-avoidance , help to establish connections between certain neurons and kill any excess neurons that are not needed . However , the mechanisms by which Dscams serve these three roles were not known . Scaffolding proteins in the cell interior interact with Dscams to hold them in place on the cell surface . Garrett et al . investigated whether Dscams need to interact with the scaffolding proteins in order to carry out any of their activities . The experiments used mice that had been genetically engineered to produce mutant Dscam proteins that cannot bind to the scaffolding proteins . Garrett et al . hypothesized that this would affect the activities of Dscams in all of the different types of neurons in the retina . However , the experiments show that the mutant Dscam proteins had different effects on the neurons . Some types of neurons developed normally , while others experienced disruptions in all three of the processes that Dscams are normally involved in . Some other neurons were affected to a moderate extent . This indicates that Dscams use different mechanisms in different types of neurons to carry out the same activities . The next step is to find out what other proteins Dscams need to interact with in different types of neurons . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Replacing the PDZ-interacting C-termini of DSCAM and DSCAML1 with epitope tags causes different phenotypic severity in different cell populations |
LINE-1 ( L1 ) retrotransposons represent approximately one sixth of the human genome , but only the human-specific L1HS-Ta subfamily acts as an endogenous mutagen in modern humans , reshaping both somatic and germline genomes . Due to their high levels of sequence identity and the existence of many polymorphic insertions absent from the reference genome , the transcriptional activation of individual genomic L1HS-Ta copies remains poorly understood . Here we comprehensively mapped fixed and polymorphic L1HS-Ta copies in 12 commonly-used somatic cell lines , and identified transcriptional and epigenetic signatures allowing the unambiguous identification of active L1HS-Ta copies in their genomic context . Strikingly , only a very restricted subset of L1HS-Ta loci - some being polymorphic among individuals - significantly contributes to the bulk of L1 expression , and these loci are differentially regulated among distinct cell lines . Thus , our data support a local model of L1 transcriptional activation in somatic cells , governed by individual- , locus- , and cell-type-specific determinants .
At least half of our DNA is derived from repeated and dispersed sequences called retrotransposons , a class of mobile genetic elements which proliferate via an RNA-mediated copy-and-paste mechanism termed retrotransposition ( see [Burns and Boeke , 2012; Hancks and Kazazian , 2012; Richardson et al . , 2015] for recent reviews ) . Since most copies have accumulated mutations or are truncated , they are unable to initiate new cycles of retrotransposition and could be considered as molecular fossils ( although in some cases they might nevertheless still be transcriptionally active [Macia et al . , 2011] ) . In contrast , the youngest and human-specific L1 subfamily , 'transcribed L1 , subset a' or L1HS-Ta , continues to retrotranspose and to accumulate in modern human genomes ( Boissinot et al . , 2000 ) . Hence , each individual has hundreds of additional copies not present in the reference genome , referred to as 'non-reference' L1HS-Ta , which contribute to our genetic diversity ( Xing et al . , 2009; Cordaux and Batzer , 2009; Ewing and Kazazian , 2010; Beck et al . , 2010; Huang et al . , 2010; Iskow et al . , 2010; Kidd et al . , 2010; Lupski , 2010; Ewing and Kazazian , 2011; Ray and Batzer , 2011; Stewart et al . , 2011; Mir et al . , 2015 ) . Recent advances in deep-sequencing technologies have also led to the discovery that L1HS-Ta are not only able to mobilize in the germline , in the early embryo and in embryonic stem cells - resulting in inheritable genetic variations ( van den Hurk et al . , 2007; Wissing et al . , 2012; Hancks and Kazazian , 2012; Macia et al . , 2014 ) - but can also retrotranspose in some somatic tissues such as brain ( Faulkner et al . , 2009; Coufal et al . , 2009; Baillie et al . , 2011; Evrony et al . , 2012; Erwin et al . , 2014; Richardson et al . , 2014; Upton et al . , 2015 ) , and in many epithelial cancers ( Miki et al . , 1992; Iskow et al . , 2010; Solyom et al . , 2012; Shukla et al . , 2013; Rodić and Burns , 2013; Pitkänen et al . , 2014; Helman et al . , 2014; Tubio et al . , 2014; Goodier , 2014; Ewing et al . , 2015; Rodić et al . , 2015; Doucet-O'Hare et al . , 2015; Paterson et al . , 2015 ) . The overall ability of L1 elements to retrotranspose presumably results from the balance between the activities of the L1 sequences themselves , and the effects of restricting cellular pathways . The first step required to initiate retrotransposition of a particular L1 instance is its transcriptional activation: this is primarily driven by an internal promoter located within the L1 5' UTR ( Swergold , 1990; Minakami et al . , 1992; Tchénio et al . , 2000; Athanikar et al . , 2004 ) , but can be repressed by CpG methylation ( Yoder et al . , 1997; Bourc'his and Bestor , 2004; Muotri et al . , 2010; Wissing et al . , 2012; Castro-Diaz et al . , 2014 ) . Production of L1 RNA transcripts is essential both for the translation of L1-encoded proteins , ORF1p and ORF2p , which are required for retrotransposition ( Moran et al . , 1996 ) , and to act as a template for reverse transcription itself ( Wei et al . , 2001 ) . After reverse transcription and genomic integration , the sequences of each L1 element can accumulate genetic alterations ( mutations , deletions , insertion of nested transposable elements ) , and these can alter the intrinsic integrity and biochemical activity of these copies . As a result , only a fraction of L1 elements are retrotransposition-competent , even when cloned in a plasmid and tested in cellular assays , with their expression driven by a strong constitutive promoter ( Brouha et al . , 2003; Beck et al . , 2010 ) . These so-called 'hot' L1 elements are highly enriched among the youngest L1HS-Ta insertions , which are polymorphic among individuals ( Beck et al . , 2010; Lupski , 2010; Beck et al . , 2011 ) . Finally , additional cellular pathways and restriction factors can limit L1 activities at multiple other stages of the L1 retrotransposition cycle ( see [Heras et al . , 2014; Richardson et al . , 2015; Pizarro and Cristofari , 2016] for reviews ) . Our understanding of L1 transcriptional activation , particularly in the context of different cell types , remains extremely limited . Indeed , studying this process is complicated by the extent of L1 insertional polymorphisms in individual genomes and the extreme level of sequence identity between the copies of the youngest ( and most-active ) L1HS-Ta subfamily and with older copies of retrotransposition-incompetent subfamilies . Theoretically , the high L1 activity observed in particular cell types could result from global unleashing of most L1HS-Ta copies . Alternatively , it could derive from a few deregulated L1HS-Ta instances . To resolve these competing models , we mapped the location of each L1HS-Ta element dispersed in the genome of a panel of normal and transformed human cells , identified a genomic signature for the transcriptionally active copies and investigated the contribution of each of them to the bulk of L1HS-Ta transcripts . We found that individual L1 instances exhibit both locus- and cell-type-specific activation , implying that L1 mutagenic activity originates from 'hot L1' inserted in permissive loci and suggesting an unforeseen new layer of cell-type specific regulation to control endogenous retrotransposons .
Human L1-derived RNA-transcripts and proteins , required for L1 retrotransposition , are detected in embryonic stem cells , in embryonal carcinoma cells , and other transformed cells or tumors , as well as in neuronal progenitor cells , but not in most primary cells , such as fibroblasts ( Faulkner et al . , 2009; Coufal et al . , 2009; Belancio et al . , 2010; Wissing et al . , 2012 ) . To study the polymorphism and expression of the L1HS-Ta subfamily at the level of individual genomic instances , we selected twelve widely used cell lines belonging to each of these different categories ( Supplementary file 1 ) . These included 10 cell lines which have been characterized in depth as part of the ENCODE project ( Bernstein et al . , 2012 ) , together with two others – the commonly used embryonic lung fibroblast line MRC-5 , and the embryonic carcinoma cell line 2102Ep , which is known to express high levels of endogenous L1HS-Ta ( Leibold et al . , 1990 ) . As a first estimate of L1 activity , we quantified and compared the endogenous levels of the L1-encoded ORF1p protein in distinct cell-lines . We detected ORF1p expression in whole cell extracts of half of the transformed cell lines ( Figure 1—figure supplement 1 ) , consistent with the proportion of human tumors expressing ORF1p ( Rodić et al . , 2014 ) and with previous work ( Belancio et al . , 2010 ) . As expected , no ORF1p was detected in primary fibroblasts . ORF1p associates with the L1 mRNA , and L1 ORF2p , to form a ribonucleoprotein particle ( RNP ) , which mediates the retrotransposition reaction ( Kulpa and Moran , 2006; Doucet et al . , 2010 ) . To ensure the highest sensitivity and to enrich for functional ORF1p ( at least able to bind RNA ) , we prepared L1 RNPs by sucrose cushion ultracentrifugation and probed ORF1p by immunoblot . We observed similar results as in whole cell extracts , except that ORF1p was faintly detected in two additional transformed cell lines ( HeLa S3 , Hep G2 , Figure 1a ) . In a complementary approach , we estimated the proportion of L1 transcripts originating from the L1HS-Ta subfamily by counting RNA-seq reads mapped on the L1HS consensus sequence , which encompass subfamily-diagnostic SNPs in the L1 3' UTR sequence ( ACA for L1HS-Ta , ACG for L1HS-PreTa and GAG for L1PA2 and older ) ( Boissinot et al . , 2000 ) . For this analysis , publicly available data from the hESC line H1 were also included ( no stranded polyA+ RNA-seq data were available for MRC-5 and HEK-293 cells ) . Consistent with L1 RNP quantification , MCF7 and 2102Ep cells exhibit the highest levels of L1HS-Ta RNA-seq tags . Most other transformed cells have intermediate levels , while HCT 116 and primary fibroblasts ( BJ , IMR-90 ) have extremely reduced L1HS-Ta levels ( Figure 1b ) . In agreement with previous studies on other hESC , H1 cells express relatively high levels of L1HS-Ta ( Garcia-Perez et al . , 2007; Macia et al . , 2011 ) . Altogether these data indicate that , in several - but not all - transformed cells and in hESC , L1HS-Ta retrotransposons can escape the epigenetic , transcriptional and post-transcriptional controls that usually limit their expression in most somatic cells ( Faulkner et al . , 2009 ) . 10 . 7554/eLife . 13926 . 003Figure 1 . Global expression of L1HS elements in a panel of human somatic cell lines . ( a ) ORF1p immunoblot analysis of L1 RNP accumulation in the indicated cell lines . Top , ORF1p immunoblot . Bottom , S6 Ribosomal Protein immunoblot as loading control . The quantity of RNP loaded is indicated at the bottom of the gel . ( b ) Global estimate of L1HS-Ta RNA levels obtained by counting RNA-seq reads mapping against the L1HS consensus and containing the Ta-specific ACA diagnostic signature , normalized by the total number of reads mapping in the human reference genome ( hg19 ) ( mean ± s . e . m . , n=2 except for MCF-7 where n=4 , and HCT 116 where n=1 ) . This analysis is based on stranded polyA+ RNA-seq data ( Supplementary file 1 ) . None were available for MRC-5 and HEK-293 cells , but data obtained from the hESC line H1 were included . See also Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 00310 . 7554/eLife . 13926 . 004Figure 1—figure supplement 1 . Analysis of L1 ORF1p expression in whole cell extracts of various cell lines by immunoblot . Top , ORF1p immunoblot . Bottom , Tubulin immunoblot as loading control . The quantity of whole cell extracts loaded is indicated at the bottom of the gel . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 004 We next asked whether the observed expression of L1HS-Ta in some cells results from the transcriptional activation of all or most L1HS-Ta genomic copies , or only of a few of them . As a first step , we mapped the genomic location of all L1HS-Ta elements in each cell line of our panel . To achieve this task , we adapted for deep sequencing an existing method termed ATLAS ( Badge et al . , 2003 ) . In brief , ATLAS-seq relies on the random mechanical fragmentation of the genomic DNA to ensure high-coverage , ligation of adapter sequences , suppression PCR-amplification of L1HS-Ta element junctions , and Ion Torrent sequencing using single-end 400 bp read chemistry ( Figure 2a–c and Figure 2—figure supplement 1a , see also Materials and methods section ) . A notable aspect of ATLAS-seq is that we combine amplification and mapping of L1 downstream flanking sequence as well as the L1 upstream flanking sequence of full length elements ( Figure 2a–c ) . This allows the unambiguous identification of full-length and potentially retrotransposition-competent genomic instances . In total , ATLAS-seq identified 7823 high-confidence L1HS-Ta insertions in the 12 cell lines analyzed , corresponding to 1633 distinct loci and including 358 full length elements ( 22% ) ( Supplementary files 2 and 3 ) . The human reference genome hg19 contains 485 L1HS-Ta insertions with a detectable 3' end . On average ( ± s . d . ) , each cell line contains 652 ( ±68 ) L1HS-Ta copies , including 178 ( ±12 ) full-length elements . Among them 393 ( ±10 ) are reference insertions , 179 ( ±18 ) are non-reference L1HS-Ta previously identified as L1 insertion polymorphisms and catalogued in euL1db ( Mir et al . , 2015 ) , and 80 ( ±60 ) are novel insertions ( Figure 2d ) . ATLAS-seq recovers 98% of the L1HS-Ta elements previously described as fixed in the human population ( see Materials and methods and Figure 2—figure supplement 1b ) , showing that this mapping approach is close to being comprehensive . To further validate the L1HS-Ta elements mapped by ATLAS-seq , we randomly selected and tested by PCR 72 non-reference insertions identified in HEK-293T cells with a broad range of supporting ATLAS-seq reads . Primers could be designed for 70/72 loci . We validated 66/70 of the tested L1HS-Ta , giving a true positive rate of 94% ( Supplementary file 4 ) . One fifth of the L1 loci are present in all tested cell lines and approximately 40% of them are present in only one of the cell lines . The remaining insertions show an intermediate level of polymorphism among the studied cell lines ( Figure 2—figure supplement 1c ) . Finally , each pair of distinct diploid normal fibroblast lines ( IMR-90 , BJ and MRC-5 in our panel ) differs at an average of 298 positions with regards to the presence or absence of a specific L1HS-Ta copy , in remarkable agreement with previous estimates ( Figure 2—figure supplement 1d ) ( Ewing and Kazazian , 2010 ) . Collectively , these data reinforce the notion that L1HS-Ta elements are highly polymorphic and contribute to the diversity of the human genome . 10 . 7554/eLife . 13926 . 005Figure 2 . The genetic landscape of L1HS-Ta insertional polymorphisms in 12 human somatic cell lines . ( a ) Principle of the ATLAS-seq procedure . The subsequent in silico steps are described in Figure 2—figure supplement 1a . ( b–c ) , Modified IGV genome browser views ( Thorvaldsdóttir et al . , 2013 ) of two non-reference polymorphic L1 instances detected in MCF7 cells ( b , full length L1 , note the two adjacent 5'- and 3'-ATLAS-seq peaks; c , truncated L1 ) . ( d ) L1HS-Ta insertions found in the various cells of the studied panel . See also Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 00510 . 7554/eLife . 13926 . 006Figure 2—figure supplement 1 . Fixed and polymorphic L1HS-Ta elements mapped by ATLAS-seq . ( a ) A scheme summarizing the principle of ATLAS-seq sequencing data analysis . ( b ) Barchart showing the discovery rate of fixed L1HS-Ta elements before ( empty bars ) and after ( plain blue bars ) implementation of in silico filters , in each of the analyzed cell lines . ( c ) Extent of L1HS-Ta insertional polymorphisms among the 12 cell lines analyzed by ATLAS-seq . ( d ) Venn-Diagram representing the number of common L1HS-Ta insertions , and the extent of L1-mediated structural variation , in three normal diploid cell lines ( fibroblasts ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 006 Full-length L1 elements contain internal sense and antisense promoters located in their 5'-untranslated region ( UTR ) . To gain insight into the pattern of L1 expression in human somatic cells at single copy resolution , we noted that the weak L1 polyadenylation signal in the 3' UTR allows a fraction of L1 transcripts to extend into the downstream genomic sequence ( Holmes et al . , 1994 ) . This property enables us to use these 3’ downstream transcripts as a measure of L1 transcriptional activity , thereby circumventing the difficulties associated with attempting to unambiguously determine the origin of transcript sequences derived from within the highly-identical L1 elements ( Figure 3a and Figure 3—figure supplement 1 ) . Similarly , the L1 antisense promoter activity generates antisense transcripts extending into the upstream genomic region flanking the L1 element ( Speek , 2001; Cruickshanks and Tufarelli , 2009; Rangwala et al . , 2009; Macia et al . , 2011; Denli et al . , 2015 ) . To identify such transcripts , we performed paired-end ( 2x150 bp ) and stranded poly ( A ) + RNA sequencing ( RNA-seq , Supplementary file 2 ) from the highly L1-expressing breast cancer cell line MCF7 ( Figure 1 ) . Then we used sense RNA-seq tags downstream of the L1 elements mapped by ATLAS-seq , or antisense RNA-seq tags upstream of them , as a proxy to monitor the sense and antisense promoter activities of each individual copy , respectively ( Figure 3a ) . 10 . 7554/eLife . 13926 . 007Figure 3 . Detection of transcriptionally active L1HS-Ta elements at individual copy resolution in MCF7 cells . ( a ) Theoretical scheme representing the outcome of RNA-seq and ChIP-seq read mapping at polymorphic L1 loci . The informative regions are highlighted in beige . ( b ) Genome browser views of reference ( left , TTC28 locus ) and non-reference ( right , NEDD4 locus ) L1 instances integrated with RNA-seq ( green ) and H3K4me3 ChIP-seq data ( blue ) . R1 and R2 , replicate #1 and #2 , respectively . ( c ) shRNA-mediated ORF1p knock-down . Top , immunoblot for ORF1p . Bottom , immunoblot for Actin , Tubulin and GAPDH as loading controls . R1 , R2 , and R3 are independent knock-down replicates performed in parallel and used subsequently for RNA-seq . Relative ORF1p levels normalized by the loading controls and scrambled shRNA controls are indicated between the two membranes . ( d ) Modified IGV genome browser views ( Thorvaldsdóttir et al . , 2013 ) of the TTC28 ( left ) and NEDD4 ( right ) L1 instances with RNA-seq data upon ORF1p shRNA-mediated knock-down . The informative L1 downstream region is highlighted in beige . Only one biological replicate out of three is shown for the sake of clarity . ( e ) Heat maps showing RNA-seq read accumulation 1 kb upstream and 1 kb downstream of each L1 copy . The downstream signal on the L1 strand ( left heat map ) is indicative of L1 sense promoter activity , while the upstream signal on the L1 antisense strand ( right heat map ) reflects L1 antisense promoter activity . L1 instances ( rows ) are sorted by decreasing L1 level of expression on the sense strand and the order is identical for the antisense strand . ( f ) Chromatin and transcription status around expressed ( blue , FPKM of downstream RNA-seq tag>0 . 05 ) and non-expressed L1HS-Ta instances ( pink ) . The indicated ChIP-seq and RNA-seq signals for each class of L1HS-Ta copies were aggregated and plotted centered around the position of the L1 insertion site . Note that the internal L1 region , when available ( reference L1 ) , is not included , but only its flanks . See also Figure 3—figure supplements 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 00710 . 7554/eLife . 13926 . 008Figure 3—figure supplement 1 . L1HS-Ta loci belongs to low mappability genomic regions . Genome browser view showing an L1HS-Ta element in the TTC28 locus . This insertion is evolutionary young ( human-specific , not present in other Primates ) . UCSC mappability tracks are shown in black and green . Numbers on the left refer to read length . Increasing read length resolve mappability issues in older repeats ( more divergent ) flanking the L1HS-Ta , but not the L1HS-Ta itself . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 00810 . 7554/eLife . 13926 . 009Figure 3—figure supplement 2 . Impact of shRNA-mediated ORF1 knockdown on RNA levels for each L1HS-Ta genomic instance . Heat maps showing for each L1HS-Ta instance ( row ) the Log2 fold change ( Log2FC ) of RNA-seq signals for each ORF1 shRNA versus a scrambled shRNA control ( n=3 ) . ( a ) MCF7 cells . ( b ) 2102Ep cells . The pink rows reflect L1HS-Ta copies for which no RNA-seq signal could be detected ( and therefore the ratio could not be calculated ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 009 This strategy enabled us to clearly identify individual , expressed copies of full-length L1HS-Ta , including both fixed and polymorphic instances , as exemplified by the two loci depicted in Figure 3b . The first one is a reference L1HS-Ta element integrated in the TTC28 gene ( 22q12 . 1 ) . Uniquely mapped RNA-seq tags are detected both in the body of the L1 sequence and in the immediate downstream genomic region . However , read mapping is performed against the reference human genome; thus , reads mapping within the L1 body could originate either from this particular reference L1 copy or from a non-reference L1 copy integrated somewhere else in the genome , and are therefore not informative . The second example is a non-reference L1HS-Ta element inserted in the NEDD4 gene ( 15q21 . 3 ) in opposite orientation . Again , a downstream RNA-seq peak immediately follows the insertion point . Interestingly , in both cases , we do not detect upstream antisense RNA-seq tags at the vicinity of these full-length L1HS-Ta copies , suggesting that sense and antisense L1 promoter activities - or the stability of their respective transcripts - are not necessarily coupled in a chromosomal environment , in agreement with previous results obtained from plasmid borne reporter assays ( Macia et al . , 2011 ) . Several lines of evidence confirmed that the downstream RNA-seq peaks emanate from these L1 copies and not from other overlapping genic or non-genic transcripts , or from distinct L1 copies carrying a 3' transduction ( see below ) . First , we observed H3K4me3 ChIP-seq peaks immediately upstream and adjacent to these L1 copies , a histone mark reflecting active or poised promoters ( Figure 3a–b ) ( Bernstein et al . , 2012 ) . Although the H3K4me3 signal is expected to be centered on the internal promoter within each L1 copy , a region that is either non-uniquely mappable or not included in the reference genome sequence , the ChIP-seq signal can be readily detected in the flanking genomic sequence . Second , we performed shRNA-mediated knockdown of all L1HS-Ta transcripts , targeting the ORF1 sequence ( Figure 3c ) . Two different ORF1 shRNAs greatly reduced the downstream RNA-seq signal when compared to a scrambled shRNA or to unmanipulated cells ( Figure 3d ) . Together , these data indicate that downstream RNA-seq tags originate from the same transcriptional unit as the considered L1 . Retrotransposition of L1 transcripts which include downstream genomic sequences can result in duplication of these sequences at the new insertion site , a phenomenon termed 3’-transduction ( Holmes et al . , 1994; Moran et al . , 1999 ) . Thus , it is possible , in principle , that downstream RNA-seq tags mapping to a particular L1HS-Ta copy could in fact emanate instead from a daughter copy with a 3' transduction , located elsewhere in the genome . However , the concomitant presence of an upstream H3K4me3 mark at many transcriptionally active L1 copies renders this situation very unlikely in most cases . To obtain a comprehensive view of the transcriptional activities of individual L1 copies , we applied this integrative approach to all the full-length L1HS-Ta elements identified by ATLAS-seq in MCF7 ( Figure 3e ) . Strikingly , only 5 L1HS-Ta copies show relatively high expression , and approximately 15 more copies exhibit low but detectable levels of expression . In addition , only 4 L1HS-Ta loci show evidence of L1 antisense promoter activity , including instances that are distinct from the few copies expressing high levels of L1 sense transcripts , suggesting uncoupling and differential regulation between these two L1 promoter activities . Consistent with the examples described at the TTC28 and NEDD4 loci ( Figure 3b ) , transcribed L1HS-Ta loci at the genome-wide level have a dual fingerprint with upstream active chromatin marks ( H3K4me3 and H3K27ac ) and Pol-II and downstream RNA-seq signal ( Figure 3f ) ; the latter being reduced upon shRNA-mediated L1 knockdown ( Supplementary file 2 and Figure 3—figure supplement 2a ) . The expression pattern of individual L1HS-Ta loci in standard growth conditions is highly reproducible , as revealed by the clustering of independent RNA-seq experiments obtained from the same cell line grown in two independent laboratory environments ( Figure 4a , MCF7_Cristofari and MCF7_ENCODE samples , Pearson correlation r=0 . 830 p<0 . 0001 ) , highlighting the non-random nature of this process and the robustness of our overall approach . 10 . 7554/eLife . 13926 . 010Figure 4 . Locus- and cell-type-specific reactivation of individual L1HS-Ta copies in normal and transformed cells . ( a ) Heat map displaying expression levels of each L1 instance in each of the analyzed cell lines . Expression level is defined as the number of RNA-seq fragments mapped in a 1 kb-window downstream of a particular L1 copy and on the same strand , normalized by the total amount of mapped fragment ( FPKM ) . Grey , absent polymorphic L1 copy . Most cell lines have at least two RNA-seq replicates ( R1 and R2 ) , which cluster based on their L1 expression profiles , showing their cell-line specificity . ( b ) The bulk of L1HS-Ta transcripts is produced by a limited number of loci . Scatter plot showing the number of L1 copies contributing to half of the total pool of L1HS-Ta transcripts . The y-axis represents the total L1 downstream tag FPKM count for each cell line . The x-axis represents the number of L1 loci contributing to half of this total FPKM . ( c ) Distribution of expressed and non-expressed L1 insertions in genic and non-genic regions . Bar chart indicating the fraction of L1 copies in genic ( dark grey ) and non-genic ( white ) regions with associated pie charts indicating the proportion of non-expressed L1 ( light blue ) and L1 expressed in at least 1 cell line of the panel ( dark blue ) . The distribution of expressed L1 insertions is not statistically different between genic and non-genic regions ( p=0 . 117 , binomial test ) . ( d ) Expression levels of genes associated with non-expressed or expressed L1 copies . Values of gene expression are considered independently for each cell line of the panel , and distributed whether the L1 insertion is expressed ( >0 . 05 FPKM ) or not ( ≤0 . 05 FPKM ) in each particular cell line . White oval shows the median; black box lower and upper limits indicate the 25th and 75th percentiles , respectively; whiskers extend to 1 . 5 times the interquartile range; violin shape represents density estimates of data and extend to extreme values ( out of scale range ) . Genes containing expressed L1s are more expressed than genes containing non-expressed L1s ( p<0 . 001 , Kolmogorov-Smirnov test ) . See also Figure 4—figure supplements 1 , 2 and 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 01010 . 7554/eLife . 13926 . 011Figure 4—figure supplement 1 . Heat maps for L1HS-Ta loci individual expression in various cell lines . Legend is identical to Figure 3e . MCF7 RNA-seq data are from the ENCODE Project ( MCF7_ENCODE ) and from this study ( MCF7_Cristofari ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 01110 . 7554/eLife . 13926 . 012Figure 4—figure supplement 2 . RT-PCR validation of individual L1 expression across several cell lines . PCR primers are anchored in the L1 internal sequence and in the flanking genomic region , respectively . Each RT-PCR included a control reaction without RT ( - ) to exclude possible genomic DNA contamination . Top , RT-PCR reactions . Bottom , PCR on genomic DNA using the same primers , showing polymorphic L1 copies among the various cell lines , and validating PCR conditions . RT , reverse transcriptase . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 01210 . 7554/eLife . 13926 . 013Figure 4—figure supplement 3 . Relationship of expressed L1HS-Ta copies with genes and ploidy . ( a ) Distribution of genic L1HS-Ta full-length copies regarding the orientation relative to the overlapping genes . Pie charts indicate the proportion of sense ( light blue ) and antisense ( dark blue ) L1 insertions relative to genes . In some instances , the L1 insertion overlaps two genes with both sense and antisense orientation ( labeled “both” in the pie charts , coral blue ) . L1 copies are more frequent in the antisense orientation with no statistical difference between non-expressed and expressed L1 copies ( p=0 . 259 , binomial test ) . ( b ) Distribution of L1 insertions regarding the presence or absence of copy number variation in the genomic region . Status of the genomic regions are considered independently for each cell line of the panel , when data are available , and distributed whether the enclosed L1 insertion is expressed ( >0 . 05 FPKM ) or not ( ≤0 . 05 FPKM ) in each particular cell line . Expressed L1 copies are not significantly enriched in amplified regions ( p=0 . 138 , Chi-square test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 013 To expand these observations to a range of cell-types we applied the same RNA-seq-based analysis strategy to the complete panel of cell lines ( except MRC-5 and HEK-293 cells for which stranded poly ( A ) + RNA-seq data were not available in public databases , Figure 4—figure supplement 1 ) . The identity of the individual L1HS-Ta loci which are expressed , as well as their levels of expression , varies considerably between cell types ( Figure 4a and Supplementary file 5 ) . Strikingly , many L1HS-Ta elements which are present at the same genomic location in distinct cell types are differentially expressed , indicating that L1HS-Ta re-activation in transformed cells may result from cell-type- and copy-specific regulation . As for MCF7 , L1HS-Ta expression profiles between biological replicates of all cell lines are similar; and global shRNA-mediated ORF1 knockdown in 2102Ep , followed by RNA-seq , again confirmed the L1-derived origin of downstream transcripts in this cell line ( Supplementary file 2 and Figure 3—figure supplement 2b ) . The total levels of RNA-seq tags 1 kb downstream of L1HS-Ta for a given cell line are highly correlated with the number of internal L1 reads containing the ACA diagnostic nucleotides ( Spearman r=0 . 8169 , p<0 . 0001 ) , reinforcing the idea that RNA-seq tags downstream of L1 can be used as a reliable proxy for L1 sense transcription . RT-PCR validation using primers anchored in the 3' UTR of L1 and in the downstream flanking genomic sequence confirmed the transcriptional state deduced from RNA-seq , for a selection of expressed and non-expressed L1HS-Ta elements ( Figure 4—figure supplement 2 ) . Interestingly , for most L1HS-Ta-expressing cells , only 5 to 15 individual copies contribute to the bulk of L1HS-Ta RNA , defined as the number of L1 copy contributing to half of the total FPKM count ( Figure 4b ) . As compared to MCF7 cells , the embryonal carcinoma cells 2102Ep accumulate comparable global levels of L1HS-Ta transcripts , but seem to have a higher number of permissive L1HS-Ta loci , each contributing to a smaller proportion of the total , although the number of active instances still represents a small fraction ( <10% ) of all L1HS-Ta copies in these cells . To test whether the expression of L1HS-Ta copies could be influenced by their genic environment , we first compared the proportion of expressed L1 insertions in genes as compared to non-expressed copies . Approximately 1/3 of all full-length L1HS-Ta copies were inserted in genes ( Figure 4c ) . Although the proportion of expressed L1s in genes was slightly higher than that of non-expressed copies , the difference was not significant . Then , we focused on the genic cohort of full-length L1HS-Ta . For the latter , we asked whether expressed vs . non-expressed elements were differentially oriented relative to the overlapping genes . Consistent with previous observations ( Szak et al . , 2002 ) , genic L1 copies are more often found in the antisense orientation ( Figure 4—figure supplement 3a ) . However , this proportion was not significantly different between expressed and non-expressed copies . Independently of their orientation relative to genes , we found that genes containing expressed L1 are often more expressed than those containing non-expressed L1 ( Figure 4d ) , suggesting that highly expressed gene loci might represent a favorable genomic environment for L1 reactivation . Finally , it is conceivable that expressed L1 could be located in larger chromosomal regions having undergone massive amplification . To address this possibility , we looked whether L1 insertions were located in genomic regions showing copy number variations ( CNVs ) . The majority of L1 copies were inserted in normal regions ( Figure 4—figure supplement 3b ) , whether they were expressed or not , and expressed L1 copies were not significantly enriched in amplified regions . Finally , we determined whether the top expressed L1HS-Ta elements identified as expressed in the panel of cell lines have the ability to achieve complete retrotransposition cycles and to generate new copies . To answer this question , we combined complementary strategies . First , we collated published data of retrotransposition assays in cultured cells obtained for different L1 instances ( Brouha et al . , 2003; Beck et al . , 2010 ) , and combined them with our own experimental results obtained with an additional newly-identified L1 copy following the same protocol ( Figure 5b ) . Second , we compared our set of highly-expressed L1HS-Ta copies with those which have been identified in earlier studies as mobilization-competent based on the detection of daughter copies with matching 3’-transduced sequences ( Tubio et al . , 2014 ) . Third , and most directly , we specifically searched for evidence of 3' transduction in our 3' ATLAS-seq data using split reads partly mapping downstream of two distinct L1HS-Ta copies ( see Materials and methods and Figure 5c ) . We found that 5 out of the 20 most highly-expressed L1HS-Ta copies across all cell lines fulfill at least two of these criteria strongly supporting their ability to retrotranspose and 6 additional ones could be identified as progenitors of other , daughter copies . The remaining nine elements could also be retrotransposition-competent , but have not been tested in cultured assays , nor could daughter L1 copies be unambiguously identified . Thus , in total , at least 11 out of the 20 most highly-expressed L1HS-Ta copies across all cell lines are retrotransposition-competent ( Figure 5a and Supplementary file 5 ) . 10 . 7554/eLife . 13926 . 014Figure 5 . Evidence of retrotransposition capability for selected L1HS-Ta copies . ( a ) Evidence of retrotransposition competence for the top 20 most expressed L1 copies across all cell lines analyzed . Cellular assays refer to retrotransposition cellular assays of plasmid-borne L1 instances , whose expression is driven by either the native L1 5’ UTR alone ( Brouha et al . , 2003 ) or supplemented by a strong CMV promoter ( [Beck et al . , 2010] and Figure 5b ) . These assays measure L1 intrinsic biochemical activity , independently of their actual expression in their genomic context . Three-prime transduction refers to the existence of progeny copies containing a 3' transduction , which can be traced back to the original locus and reflect a retrotransposition event . ( b ) Retrotransposition assay in cultured cells for MCF7 L1 copy EXP_ID_0447 ( NEDD4 locus ) . A full length transcribed L1HS-Ta copy present in the genome of MCF7 cells was subcloned by PCR in an expression vector containing a reporter gene to measure retrotransposition activity and generated four independent clones ( pVan610-1 to -4 ) . In transfected HeLa cells , de novo retrotransposition events of engineered L1 copies lead to the introduction of a functional genomic copy of the neomycin phosphotransferase gene , which expression confers resistance to G418 . Resistant foci were stained and counted to monitor retrotransposition activity compared to the positive ( pJM101/L1 . 3 , wild type L1HS-Ta ) and negative ( pJM105/L1 . 3 , mutant L1HS-Ta ) control conditions . The value of G418 resistant colonies obtained with the positive control was set to 100% . A picture of a representative well with stained colonies is displayed for illustrative purposes under each bar of the graph . The average value of three biological replicates is displayed with error bars corresponding to the standard deviation among the three biological replicates . ( c ) Detection of 3' transductions in ATLAS-seq data . This in silico screen identifies L1HS-Ta copy ( progeny element ) with ATLAS-seq clusters containing reads with non-aligning subsequences ( soft-clipped ) , which uniquely map downstream and adjacent to another full length L1HS locus ( progenitor element ) . The panel shows a genome browser view of such a 3' transduction , originating from a full length L1HS-Ta in the TTC28 gene ( 22q12 . 1 ) . The soft-clipped region of the reads is shown in color ( base code: T , red; A , green; C , blue; G; orange ) . As expected , the transduced region is flanked by 2 poly ( A ) tails ( poly ( T ) here since it is located on the reverse genomic strand ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 014 Some L1 elements with high retrotransposition activity ( ‘hot’ L1 ) belong to well-defined lineages with distinctive 3’ transductions . To evaluate the proportion of the highly expressed elements which belong to such lineages , we screened 3’ ATLAS-seq reads supporting each L1 insertion for sequence tags characteristic of the three most characterized lineages ( AC002980 , LRE3 and RP , see Methods ) ( Schwahn et al . , 1998; Brouha et al . , 2002; Myers et al . , 2002; Beck et al . , 2010; Macfarlane et al . , 2013 ) . We found only 1 insertion ( EXP_ID_0447 , in the NEDD4 gene ) among the 20 most highly expressed L1HS-Ta copies as deriving from the L1RP transduction family ( Supplementary file 5 ) . Nine additional copies were also part of one or another lineage , but were not expressed – or only moderately – in any of the cell lines analyzed . Thus , these findings suggest that the observed high level of expression of a small cohort of L1 insertions is not an intrinsic feature of any previously identified lineage .
Altogether our observations support a model ( Figure 6 ) where: ( i ) L1HS-Ta transcription is predominantly inactive in somatic human cells , including transformed cell lines; however ( ii ) a small number of L1HS-Ta copies can escape silencing , allowing their expression and transcript accumulation; ( iii ) the locus in which a particular L1HS copy integrates has a major influence on its ability to be subsequently reactivated; ( iv ) L1 instances at distinct genomic loci are subject to cell-type dependent activation ( potentially dependent on environmental or physiological signals ) . This model is consistent with previous observations made on few L1 instances suggesting that the transcriptional activity of the L1 promoter might be influenced by its immediate upstream genomic sequence ( Lavie et al . , 2004 ) . 10 . 7554/eLife . 13926 . 015Figure 6 . Schematic model showing the highly locus-specific and variable expression of L1HS-Ta elements among different somatic cell types and individuals . The colored boxes correspond to L1HS-Ta copies , some being polymorphic ( pink ) . The model is developed in the main text . DOI: http://dx . doi . org/10 . 7554/eLife . 13926 . 015 L1 retrotransposon expression has been proposed both as a potential biomarker of cancer prognosis and as the starting point for L1-mediated genome instability in tumors ( Piskareva et al . , 2011; Rodić and Burns , 2013; Rodić et al . , 2014 ) . Hence , understanding the means by which L1s can escape regulation in particular cancers is vital in order to improve their rational use as biomarkers or to predict their possible effects on disease progression . Strikingly , our results indicate that – in the cancer cell-lines studied – the general cellular regulation of most L1 instances is unimpaired , even in cells exhibiting high L1 activity . Thus , it appears that shut-off of global L1-regulatory pathways is not a prerequisite for L1 activation in cancer . Instead , we find that only a very limited set of L1 copies , at specific genomic loci , become activated in cancer cells . We note that the presence or absence of polymorphic L1s at these permissive loci , and their degree of transcriptional activation , may represent risk factors for particular cancer types , and more specific biomarkers than global L1 expression or methylation status . Furthermore , we provide here a unique resource consisting of near-complete maps of L1HS-Ta elements present in widely used normal and transformed model cell lines , for which the availability of genomic datasets is regularly increasing ( including several tier-1 & tier-2 cell-lines of the ENCODE project ) . These maps will be of broad utility in the future to address the impact and regulation of transposable elements in the human genome . Indeed , isolated RNA-seq and ChIP-seq signals resulting from the presence of non-reference L1 copies can only be correctly interpreted if such maps are available . Thus they can act as a platform to fully interpret and profit from the many expanding public datasets generated using the same cell lines . Recently , whole genome sequencing of human tumors has revealed recurrent retrotransposition events stemming from a handful of source elements ( Tubio et al . , 2014 ) , consistent with the notion that only a fraction of all full-length L1 elements is actually capable of retrotransposition ( Brouha et al . , 2003; Beck et al . , 2010 ) . Here , we have identified highly heterogeneous expression of individual L1HS-Ta copies , implicating L1 transcriptional activation as a key regulatory process which limits the mutagenic potential of L1 elements independently of their respective intrinsic biochemical activity . Therefore , we extend the concept of retrotransposition-competent L1 copies , previously described as 'hot L1s' ( Brouha et al . , 2003; Beck et al . , 2010 ) , to the transcriptional regulation of each individual locus and cell type . We conclude that L1-mediated mutagenesis results from the reactivation of a small subset of permissive loci , only a fraction of which contains retrotransposition-competent elements , combined with a favorable cellular environment ( i . e . diminished restriction factor and/or increased cofactor activities [Pizarro and Cristofari , 2016] ) . Several of these regulated loci are polymorphic with regards to the presence or absence of an L1 element among the human population , highlighting the role of genetic determinants in the global L1 mutagenic potential in a given individual . Overall , our data suggest that activation of L1 transcription in somatic cells is governed by individual- , locus- , and cell-type-specific determinants and provide a framework to study how distinct L1HS-Ta copies may be regulated by environmental , physiological and pathological triggers . Future work will determine the factors and cellular signaling pathways that contribute to the transcriptional reactivation of the different L1HS-Ta copies in somatic cells .
The cell lines used in this study are summarized in Supplementary file 1 . Cells were grown in the medium indicated in Supplementary file 1 , containing 4 . 5 g/L D-Glucose , 110 mg/L Sodium Pyruvate , and supplemented with 10% FBS , 100 U/mL penicillin and 100 µg/mL streptomycin . Growth medium was also supplemented with 862 mg/mL L-Alanyl-L-Glutamine ( Glutamax ) , or 2mM Glutamine ( HCT 116 and K-562 ) . The HEK-293 cell line is a clonal derivative of primary human embryonic kidney cells transformed with Adenovirus 5 ( Ad5 ) DNA ( Graham et al . , 1977 ) . The HEK-293T cell line ( also known as 293T or 293/tsA1609neo ) is a clonal derivative of HEK-293 cells stably transfected with a temperature sensitive SV40 Large T antigen allele and the neomycin resistance gene ( DuBridge et al . , 1987 ) . All above-mentioned cell cultures tested negative for mycoplasma infection using the MycoAlert Mycoplasma Detection Kit ( Lonza , Basel , Switzerland ) . Cell line authenticity was verified by multiplex STR analysis ( PowerPlex 21 PCR system [Promega , Madison , WI] , assays performed by Eurofins Genomics [Ebersberg , Germany] as a service provider ) and comparison with the DSMZ database ( https://www . dsmz . de/services/services-human-and-animal-cell-lines/online-str-analysis . html ) or with previously published profiles for H1 and 2102Ep cells ( Josephson et al . , 2007; Mallon et al . , 2014 ) . Oligonucleotides were synthesized by Sigma-Aldrich ( St Louis , MO ) , Eurogentec ( Liège , Belgium ) or Integrated DNA Technologies ( IDT , Coralville , IA ) . Those used for ATLAS-seq , shRNA constructs , PCR or RT-PCR assays are described in Supplementary file 2 and those used to validate ATLAS-seq results are described in Supplementary file 4 . For whole cell lysates , 3x106 cells were pelleted and lysed in 150 µL RIPA buffer ( 10 mM Tris-HCl pH 7 . 5 , 1 mM EDTA , 150 mM NaCl , 0 . 5% NP-40 , 0 . 5% sodium deoxycholate , 0 . 1% SDS , supplemented with complete Mini , EDTA-free Protease Inhibitor Cocktail tablets [Roche , Basel , Switzerland] ) by 30 cycles of sonication ( 30 s on/30 s off ) at 4°C with a Bioruptor sonicator ( Diagenode , Liège , Belgium ) . Protein concentration was determined by the BCA assay ( Interchim , Montluçon , France ) . L1 ribonucleoprotein particles ( RNP ) were enriched by ultracentrifugation on sucrose cushion as described previously ( Kulpa and Moran , 2006; Monot et al . , 2013; Viollet et al . , 2016 ) . Briefly , 107 cells were collected and washed with 1X phosphate-buffered saline ( PBS ) . Cell pellets were lysed with 1 mL of LEAP buffer ( 5 mM Tris-HCl pH 7 . 5 , 1 . 5 mM KCl , 2 . 5 mM MgCl2 , 1% sodium deoxycholate , 1% Triton X-100 , supplemented with complete Mini , EDTA-free Protease Inhibitor Cocktail tablets [Roche] ) overnight at 4°C on an end-over-end wheel . Cell lysates were loaded on top of a two-layered sucrose cushion ( 8 . 5% and 17% sucrose solution diluted in 80 mM NaCl , 5 mM MgCl2 , 20 mM Tris pH 7 . 5 and 1 mM DTT supplemented with complete , Mini , EDTA-free Protease Inhibitor Cocktail tablets ) . Samples were ultracentrifuged at 39 , 000 rpm in a Beckman SW 41 Ti rotor for 2 hr at 4°C and resuspended overnight in 50 µL of MilliQ water . Protein concentration was measured by fluorometry using the Qubit protein assay kit ( Life Technologies , Carlsbad , CA ) . Unless otherwise stated , 30 µg of whole cell lysates or 10 µg of RNP were resolved on a 4–12% gradient NuPAGE Bis-Tris gel ( Life Technologies ) and transferred with the semi-dry XCell II Blot module onto PVDF FL membrane ( Millipore , Billerica , MA ) . Membranes were incubated for 1 hr at room temperature in blocking solution ( Phosphate-buffered saline with 0 . 1% Tween 20 ( PBS-T ) , containing 5% ( w/v ) fat-free milk ) , and then overnight at 4°C with a primary antibody diluted in blocking solution . After 5 washes in PBS-T , the membranes were incubated for 1 hr at room temperature with the appropriate secondary antibody coupled to infrared fluorochromes diluted in Odyssey blocking buffer ( LI-COR Biosciences , Lincoln , NE ) and washed again 5 times in PBS-T and once in PBS for 5 min . The signal was detected and quantified with a dual-channel Odyssey infrared imaging system ( LI-COR Biosciences ) . When needed , membranes were stripped for 1 hr in LI-COR stripping buffer , washed 3 times in MilliQ water and were reprobed for loading control following the protocol described above . Primary antibodies were directed against hORF1p ( rabbit polyclonal antibody , serum SE-6798 , 1:5 , 000 ) ( Monot et al . , 2013 ) , S6 Ribosomal Protein ( RPS6 , rabbit monoclonal antibody , clone 5G10 , Cell Signaling Technology [Danvers , MA] , 1:2500 ) , β-Tubulin ( TUBB , mouse monoclonal antibody , clone BT7R , Pierce Biotechnology [Waltham , MA] , 1:5000 ) , β-Actin ( ACTB , mouse monoclonal antibody , clone 8H10D10 , Cell Signaling Technology , 1:5000 ) , and Glyceraldehyde-3-phosphate dehydrogenase ( GAPDH , mouse monoclonal antibody , clone GA1R , Pierce Biotechnology , 1:5000 ) . As secondary antibodies , we used IRDye 800CW Goat anti-Rabbit IgG , IRDye 800CW Goat anti-Mouse IgG , IRDye 680RD Goat anti-Rabbit IgG , or IRDye 680RD Goat anti-Mouse IgG ( all from LI-COR Biosciences , 1:10 , 000 ) . Whole cell total RNA was purified by a double TRI Reagent extraction following the manufacturer’s instructions ( Molecular Research Center , Cincinnati , OH ) and solubilized in 50 µL of milli-Q water . Subsequently , 20 µg of total RNA was treated with 2U of TURBO DNAse ( Life technologies ) for 20 min at 37°C followed by a 5 min incubation step at room temperature with DNase Inactivation Reagent . After centrifugation at 10 , 000 x g for 1 . 5 min , the supernatants containing the RNA samples were transferred into a new tube . RNA was quantified and quality-controlled by UV-spectroscopy ( NanoDrop 2000 ) and microfluidic electrophoresis ( Agilent 2100 Bioanalyzer ) , respectively . For RNA-seq samples an additional column clean up was performed ( RNeasy mini extraction kit , Qiagen , Hilden , Germany ) . Genomic DNA was extracted and prepared using QiaAmp DNA Blood mini kit ( Qiagen ) . The primers used in the ATLAS protocol ( Badge et al . , 2003 ) allow the specific amplification of the 3' flank of the L1HS-Ta subfamily ( 3'-ATLAS ) and of the 5' flank of the L1HS-Ta1d subfamily ( 5'-ATLAS ) . While representing less than 0 . 1% of human L1s , the L1HS-Ta and L1HS-Ta1d subfamilies are the youngest and most active elements ( Boissinot et al . , 2000 ) and are responsible for all but one described disease-producing insertions ( Chen et al . , 2005 ) . ATLAS-seq differs from the original ATLAS protocol in three main aspects: ( i ) Restriction enzyme digestion of the genomic DNA was replaced by mechanical fragmentation to reduce sequence bias and thus to increase the mappable part of the genome . DNA is end-repaired and A-tailed to permit ligation of the adapter to fragmented genomic DNA , before the suppression PCR step; ( ii ) Adapters have been modified to be compatible with A-tailed genomic DNA fragments and to include Ion Torrent primer sequences; ( iii ) Cloning and Sanger sequencing were replaced by Ion Torrent semiconductor sequencing . The full ATLAS-seq protocol and analysis scheme are detailed below . The analysis was performed using dedicated scripts starting from raw reads in fastq format . The different steps are detailed below . Cloning of L1 copy EXP_ID_0447 was obtained by nested PCR amplification from 50 ng of MCF7 genomic DNA using Phusion High-Fidelity DNA Polymerase ( Thermo Scientific , Waltham , MA ) in 50-µL reactions . The first PCR was performed with primers specific to the genomic DNA sequence flanking L1 copy EXP_ID_0447 ( LOU1652 and LOU1656 ) with the following program: 98°C for 3 min , 35 cycles [98°C for 10 s , 54°C for 15 s , 72°C for 3 . 5 min] , and 72°C for 10 min . A ~6 kb PCR product was resolved by 0 . 8% agarose gel electrophoresis , gel-purified using Wizard SV Gel and PCR Clean-Up System ( Promega ) , and resuspended in 50 µL of milli-Q water . The second PCR was performed using 1 µL of the purified PCR product with primers matching the L1HS 5’ UTR ( LOU1662 ) and the downstream flanking genomic DNA sequence of the L1 copy EXP_ID_0447 ( LOU1664 ) with the following program: 98°C for 3 min , 40 cycles [98°C for 10 s , 58°C for 15 s , 72°C for 4 min] , and 72°C for 10 min . A PCR product of ~6 kb fragment was gel-purified as described above . Both primers of the second PCR contained homology in their 5’ parts to the CMV promoter and to the retrotransposition reporter cassette contained in the L1 expression vector pAD135 ( Doucet et al . , 2010 ) , allowing SLiCE cloning , a method based on in vitro homologous recombination ( Zhang et al . , 2012 ) . Briefly , pAD135 was digested with NotI and BstZ17I to remove the engineered L1 . 3 element from the expression vector . SLiCE cloning to generate pVan609 plasmids was conducted in a 10-µL reaction by combining 100 ng of the digested and gel-purified vector fragment with the PCR product of L1 EXP_ID_0447 in a 1:3 molar ratio , 1X SLiCE buffer and 1 µL of PPY SLiCE extract for 1 hr at 37C ( Zhang et al . , 2012 ) . Sanger sequencing of positive clones confirmed the specific subcloning of the full-length L1 copy EXP_ID_0447 through the identification of the 3’ flanking region of genomic DNA . The flanking genomic DNA sequence was then removed by digestion of pVan609 clones with BstZ17I , an enzyme that cuts in the 3’ UTR of the subcloned L1 and just upstream of the retrotransposition reporter in the vector . This strategy allows , after re-ligation of the vector to itself to generate pVan610 clones , containing a full length L1 copy with an engineered 3’ UTR containing a retrotransposition reporter . Four independent pVan610 clones ( pVan610-1 to -4 ) were generated to rule out mutations that could have occurred through the second round of PCR amplification . L1 retrotransposition assay was conducted as described previously ( Moran et al . , 1996; Wei et al . , 2000 ) with minor modifications . Briefly , HeLa cells were plated in 6-well plates at 2x105 cells per well . The next day , cells were transfected with 1 µg of plasmid DNA and 3 µL of Lipofectamine 2000 ( Life Technologies ) diluted in 200 µL of Opti-MEM ( Life Technologies ) , and medium was replaced with fresh medium after ~5 hr . Two days after transfection , medium was supplemented with G418 ( Life Technologies ) at 1 mg/mL . After 10 days of selection , colonies were fixed and stained with 1 mL of a solution containing 30% methanol ( v/v ) , 10% acetic acid ( v/v ) and 0 . 2% Coomassie blue ( m/v ) . The colonies obtained in three independent biological replicates were manually counted to assess the retrotransposition efficiency of pVan610 clones compared to the positive control pJM101/L1 . 3 ( Sassaman et al . , 1997 ) and negative control pJM105/L1 . 3 ( RT mutant ) ( Wei et al . , 2001 ) . For each biological replicate ( four independent L1 clones , independently transfected three times each ) , two wells were used as internal technical replicates ( parallel transfections ) . Annealed complementary oligonucleotides with 3' overhangs ( see Supplementary file 2 for sequences ) were directly cloned into pLKO . 1 Puro lentiviral vectors ( Sigma-Aldrich , St Louis , MO ) between AgeI and EcoRI sites . Lentiviral particles were prepared by phosphate-calcium-mediated transfection of the vector and helper plasmids into HEK-293T cells . Briefly , one day before transfection , we plated 3 . 5x106 cells per 10 cm-dish . Cells were transfected with 8 . 6 µg of shRNA pLKO . 1 construct , 8 . 6 µg of pCMV-dR8 . 91 ( Addgene #2221 ) and 2 . 8 µg of pCMV-VSV-G ( Addgene #8454 ) . Growth medium was changed 8 hr post-transfection . Two days after transfection , cells supernatants from 5 dishes were collected , pooled , filtered , and concentrated by overnight centrifugation at 4 , 000 rpm . Pellets were resuspended in 350 µL of growth medium , aliquoted and stored at -80°C until use . Viral stocks were titrated by colony formation assay upon puromycin selection ( 8 days , 1 µg/mL ) using serial dilution of vector stocks . For RNA-seq experiments , we infected 500 , 000 cells in triplicates ( either MCF7 or 2102Ep cells ) at M . O . I 1 . Cells were collected for protein and RNA extraction after 1 week of puromycin selection ( 1 . 5 µg/mL ) . Whole cell total RNA was processed by Beckman Coulter Genomics according to the following steps . Poly ( A ) + RNA was isolated from the total RNA samples and fragmented with ultrasound ( 1 pulse of 15 s at 4°C ) . First-strand cDNA synthesis was primed with random hexamers . Then , the Illumina TruSeq sequencing adapters were ligated to the 5' and 3' ends of the cDNA . The cDNA was finally amplified with PCR using a proof reading enzyme and between 12 and 14 cycles . The TruSeq barcode sequences which are part of the 3' TruSeq sequencing adapters are included in Supplementary file 2 . The cDNA was purified using the Agencourt AMPure XP kit ( Beckman Coulter ) . For Illumina sequencing , the cDNA samples were pooled in approximately equimolar amounts . The cDNA pool was size fractionated in the size range of 350–550 bp on a preparative agarose gel and size range was verified by capillary electrophoresis ( Shimadzu MultiNA microchip electrophoresis system ) . Finally , 2x150 bp paired-end and strand-specific sequencing was performed with HiSeq SBS kit v4 on a HiSeq 2500 system ( Illumina , San Diego , CA ) . RNA-seq sequencing statistics are summarized in Supplementary file 2 . Raw ATLAS-seq and RNA-seq data were submitted to the ArrayExpress database ( www . ebi . ac . uk/arrayexpress ) under accession number E-MTAB-4676 and E-MTAB-3788 , respectively . RNA-seq data contain biological duplicates for untreated cells ( two independent RNA extractions ) , and triplicates for shRNA-treated cells ( three parallel lentivirus shRNA transductions , see above ) . The genomic locations of L1 insertions in the different cell lines are provided in Supplementary file 3 . Published RNA-seq data used in this work are described in Supplementary file 1 , and ChIP-seq data were obtained from the ENCODE Project: H3K9me3 ( ENCSR000EWQ ) , H3K4me3 ( ENCSR000DWJ ) , H3K27Ac ( ENCSR000EWR ) , H3K27me3 ( ENCSR000EWP ) , and Pol2 ( ENCSR000DMT ) . All RNA-seq and ChIP-seq datasets contain biological duplicates except HCT 116 RNA-seq ( single experiment ) . Copy number variation ( CNV ) data were obtained from the ENCODE Project: BJ ( ENCFF628TJX ) , H1 ( ENCFF228MUH ) , HeLaS3 ( ENCFF996ASY ) , HepG2 ( ENCFF074XLG ) , IMR90 ( ENCFF455ARY ) , K562 ( ENCFF486MJU ) and MCF7 ( ENCFF278UJF ) . RNA-seq raw reads were pre-processed to enhance their quality and to avoid a significant loss at the mapping stage . The quality of raw reads was verified for each lane per RNA-seq sample , using the FASTQC ( v0 . 11 . 2 ) tool ( http://www . bioinformatics . babraham . ac . uk/projects/fastqc/ ) . Reads were trimmed using sliding-window mode from Trimmomatic ( v0 . 32 ) ( Bolger et al . , 2014 ) . The minimum quality and read length allowed were 20 and 25 , respectively , as they showed the best mapping rates in several tests trying different values of these parameters . The rest of the Trimmomatic parameters were set as recommended for paired-end ( PE ) reads in the Trimmomatic tool manual . PE trimmed reads were aligned per lane to both the hg19 reference transcriptome and genome obtained from UCSC , using TopHat2 v2 . 0 . 13 ( Kim et al . , 2013 ) . For a better accuracy , the parameter --read-realign-edit-dist was set to 0 ( specifying alignment of all reads at each step , returning the best alignment for each read ) . The –r ( insert size ) and --mate-std-dev ( insert size standard deviation ) parameters were estimated using the tool CollectInsertSizeMetrics from Picardtools v1 . 128 ( http://broadinstitute . github . io/picard ) , after mapping a subset of the RNA-seq samples . The other mapping parameters were set to their default values . This section is related to Figure 1b . Reads were mapped against the consensus L1HS sequence from RepBase ( Bao et al . , 2015 ) using the end-to-end mode from bowtie2 ( v2 . 2 . 4 ) ( Langmead and Salzberg , 2012 ) . The total end-to-end mismatches ( specified by -score-min ) were -0 . 6 + 0 . 6*L , so -45 . 6 for 75 bp reads ( or no more than 7 high-quality mismatches ) or -90 . 6 for 150 bp reads ( or no more than 15 high-quality mismatches ) . These options allow two mismatches per seed , needed to allow the SNPs that distinguish each L1HS subfamilies . Reads that mapped to a diagnostic tri-nucleotide ( 5927–5929 ) within the L1 sequence were extracted using samtools ( Li et al . , 2009 ) , and the occurrences of each tri-nucleotide at this position was counted . These counts of tri-nucleotide-containing reads per library were normalized by the total number of reads mapped to the hg19 reference genome . Reads with an ACA diagnostic tri-nucleotide were counted as derived from L1HS-Ta elements . This section is related to Figure 3e and Figure 4—figure supplement 1 . Stranded coverage file ( wig files ) were generated from the alignments with the tool 'count' from igvtools ( v2 . 3 . 43 ) ( http://www . broadinstitute . org/igv ) . Only the strand of first mates and with a MAPQ of 20 were taken into account , ensuring that only one read per fragment was counted and only pairs uniquely mapped were retained for downstream analyses . The window size was set to 10 and the window function used was the median . Coverage wig files from all libraries were normalized according to the total number of mapped mates 1 of each library . After normalization , wig files were converted to bigWig files using the command wigToBigWig ( v4 ) from UCSC bioinformatics utilities . The bigWig files were used to identify transcriptional signals from the possibly full-length L1HS individual copies: RNA-seq signals upstream or downstream individual L1HS copies were used as a proxy to identify potential transcriptional signature of specific individual L1HS copies , as these genomic flanking sequences are more-often uniquely mappable . To obtain a global view of the transcription levels detectable at each LINE-1 locus , the computeMatrix command from deeptools ( v1 . 5 . 9 . 1 ) was employed ( Ramírez et al . , 2014 ) . For each RNA-seq strand and L1HS strand , the scores associated at the 5’ and 3’ of the L1HS copies in a 1 kb window size , upstream and downstream respectively , were computed separately and averaged between biological replicates . The window size was selected according to previous studies and to avoid overlapping of transcription signals from other genomic features . Annotated exons within the 1 kb window were masked . Heatmaps were produced with the R package heatmap . 2 , and lines were sorted according to the total transcription signal at the downstream flanking region . This section is related to Figure 3f . Mapped ChIP-seq data were retrieved from ENCODE . The replicates of each experiment were pooled together and read duplicates were removed from the merged bam files . The merged bam files were converted into wig files using 'count' from igvtools , with a window of 10 bp and the median function , and , then , converted into bigWig file with tool wigToBigWig from the UCSC tools . ChIP-seq signals were obtained using the ‘reference-point’ mode of computeMatrix from deeptools , specifying 1 kb upstream and 1 kb downstream of the TSS & TES , which were defined as the extremities of each L1HS copy , and with a window of 50bp and mean function . Using a threshold of 0 . 05 on the FPKM L1HS expression values , the L1HS copies were divided in two groups: expressed and non-expressed copies . The mean of each group of ChIP-seq signals was extracted and plotted . This section is related to Figure 3—figure supplement 2 . Separate annotation files ( . gtf ) were created for each cell line by appending the coordinates of 1 kb downstream of each L1HS copy to the hg19 annotation file . To count the number of reads associated with genomic features ( exons and flanking L1 sequences ) , HTSeq v0 . 6 . 1p1 was employed in the union mode , and a MAPQ threshold of 50 was used to take only unambiguously mapped reads into account ( Anders et al . , 2015 ) . Other parameters were assigned to their default value . In order to determine whether RNA-seq tags mapping within the 1 kb downstream of each L1HS actually reflect L1HS-derived transcripts , a differential expression analysis ( DEA ) was performed using HTSeq counts as inputs . Since reads mapping within the downstream 1 kb sequence flanking L1HS instances are often non-abundant , several DEA tools were tested for their ability to discriminate properly differential expression with low number of counts: from this test , DESeq2 ( Love et al . , 2014 ) gave the best performance ( data not shown ) and was used for the rest of the analysis . Data corresponding to each L1 shRNA was compared against control data corresponding to a scrambled shRNA , and default parameters were used for the entire analysis . This section is related to Figure 4a . The transcriptional activity of each L1-HSTa instance was estimated independently for each cell-line in which it is detected by the number of RNA-seq fragments which non-redundantly align to the downstream adjacent 1 kb of genomic sequence , and normalized to the number of non-redundantly aligned reads in the entire RNA-seq dataset ( FPKM ) . For this purpose , mates 1 stranded-specific counting was performed at 1 kb downstream L1HS copies with a dedicated script . Only mates with a MAPQ>= 20 were taken into account . The FPKM for each copy was calculated according to the total number of mapped mates 1 in each library . Clusters of cell-lines with similar patterns of detectable expression of distinct L1 copies were clustered using the R hclust function ( distance measure=’euclidean’ , clustering method=’ward’ ) . To identify groups of L1s which exhibit similar behaviour across cell-types , L1 expression was compared between cell-types using the reciprocal of the mean product of FPKMs as a distance measure . L1s with no detectable expression in any cell-type were excluded , and pairs of L1s which are not present in any common cell-types were assigned a neutral distance equal to the mean of all calculable pairwise distances . Clustering was performed using R hclust ( clustering method=’ward’ ) . First-strand complementary DNAs ( cDNA ) were obtained from the RNA samples using SuperScript III Reverse Transcriptase ( Life Technologies ) as per the manufacturer’s instructions . Briefly , 1 µg of DNase-treated RNA samples were combined in a 13-µL reaction including 1 µL of a 10 mM dNTP mix , 1 µL of a 50 µM RACE primer and milli-Q water . The samples were incubated for 5 min at 65C and placed on ice for 1 min . The reactions were then supplemented with a 7-µL cDNA Synthesis Mix containing 4 µL of 5X First-Strand buffer , 1 µL of 0 . 1 M DTT , 1 µL of 40 U/µL RNaseOUT and 1 µL of 200 U/µL of SuperScript III RT . For each sample , a control condition was prepared without reverse transcriptase . The tubes were incubated for 50 min at 55°C and then for 15 min at 70°C . PCR amplification of the cDNA was obtained using Platinum Taq DNA polymerase ( Life Technologies ) as per the manufacturer’s instructions . Briefly , 1 µL of cDNA sample was combined in a 50-µL reaction with 1X PCR Buffer , 1 . 5 mM MgCl2 , 1 U of Platinum Taq DNA polymerase , 0 . 2 mM dNTP , 0 . 2 µM of forward primer ( RB3PA1 [Badge et al . , 2003] ) , 0 . 2 µM of reverse primer ( specific to the 3’ flanking genomic DNA sequence of L1 copies , see Supplementary file 2 for details ) . Amplification was performed with an Applied Biosystems Veriti Thermal Cycler ( Life Technologies ) with the following program: one step at 94°C for 2 min , one step of 32 cycles at 94°C for 30 s , 56°C for 30 s , 72°C for 1 min , and one final step at 72°C for 5 min . The detection of GAPDH transcripts was performed similarly , with only 30 cycles of PCR amplification , using primers LOU0639 and LOU0576 . PCR products were resolved and visualized by 1 . 5% agarose gel electrophoresis in 0 . 5X TBE buffer and ethidium bromide staining ( 50 ng/mL , Life Technologies ) . This section is related to Figure 4c–d and Figure 4—figure supplement 3 . Raw RNA-seq data retrieved from ENCODE were processed as described above to obtain an FPKM value for each gene . We used the UCSC gene set ( hg19 version ) obtained from iGenomes ( https://support . illumina . com/sequencing/sequencing_software/igenome . html ) . When an L1 copy was inserted in a region corresponding to several overlapping genes oriented in the same direction , we considered the most expressed gene . Ploidy at each L1 insertion locus and for each cell line analyzed was determined using the copy number variation ( CNV ) tracks obtained from ENCODE ( see ‘Data’ section ) . They were deduced from Illumina Human 1M-Duo Infinium HD BeadChip assays and circular binary segmentation ( CBS ) , which segments the genome into amplified , normal , heterozygous deletion , homozygous deletion or uncovered regions . As expected , no L1 were inserted in region annotated as homozygous deletion . Uncovered regions were excluded from the analysis . The list of full-length L1 insertions provided in Supplementary file 5 was divided in two groups , expressed and non-expressed copies , using a threshold of 0 . 05 on the FPKM L1HS expression values . Due to CNV data availability , only L1 elements identified in BJ , H1 , HeLaS3 , HepG2 , IMR90 , K562 and MCF7 were considered for this analysis . | Retrotransposons , also known as jumping genes , have invaded the genomes of most living organisms . These fragments of DNA have the ability to move or copy themselves from one location of a chromosome to another . Depending on where they insert themselves , retrotransposons can modify the sequence of nearby genes , which can alter or even abolish their activity . Although these genetic modifications have contributed significantly to the evolution of different species , retrotransposons can also have detrimental effects; for example , by causing new cases of genetic diseases . Adult human cells have a number of mechanisms that work to keep the activity of retrotransposons at a very low level . However , in many types of cancers retrotransposons escape these defense mechanisms and ‘jump’ actively . This is thought to contribute to the development and spread of cancerous tumors . To understand how jumping genes are mobilized , a fundamental question must be answered: is the high jumping gene activity observed in some cell types a result of activating many copies of the retrotransposons , or only a few of them ? This question has been difficult to address because there are more than one hundred copies of retrotransposons that could potentially move in humans , many of which have not even been referenced in the human genome map . Furthermore , each copy is almost identical to another one , making it difficult to discriminate between them . Philippe et al . have now developed an approach that can map where individual retrotransposons are located in the genome of normal and cancerous cells and measure how active these jumping genes are . This revealed that only a very restricted number of them are active in any given cell type . Moreover , different subsets of jumping genes are active in different cell types , and their locations in the genome often do not overlap . Thus , whether jumping genes are activated depends on the cell type and their position in the genome . These results are in contrast to the prevalent view that retrotransposons are activated in a more widespread manner across the genome , at least in cancerous cells . Overall , Philippe et al . ’s findings pave the way towards characterizing the chromosome regions in which retrotransposons are frequently activated and understanding how they contribute to cancer and other diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"genetics",
"and",
"genomics"
] | 2016 | Activation of individual L1 retrotransposon instances is restricted to cell-type dependent permissive loci |
Several genes positively influence final leaf size in Arabidopsis when mutated or overexpressed . The connections between these growth regulators are still poorly understood although such knowledge would further contribute to understand the processes driving leaf growth . In this study , we performed a combinatorial screen with 13 transgenic Arabidopsis lines with an increased leaf size . We found that from 61 analyzed combinations , 39% showed an additional increase in leaf size and most resulted from a positive epistasis on growth . Similar to what is found in other organisms in which such an epistasis assay was performed , only few genes were highly connected in synergistic combinations as we observed a positive epistasis in the majority of the combinations with samba , BRI1OE or SAUR19OE . Furthermore , positive epistasis was found with combinations of genes with a similar mode of action , but also with genes which affect distinct processes , such as cell proliferation and cell expansion .
Since Bateson introduced the term epistasis to describe the phenomenon that some mutations seemed to be ‘stopping’ or ‘standing above’ the effect of other mutations ( Bateson , 1909 ) , it became clear that interactions between multiple genes influence many traits . Epistasis , or interaction between genes , therefore corresponds to any deviation from the expected phenotype , predicted by combining the effects of individual alleles or mutations ( Fisher , 1918; Phillips , 2008 ) . Only by identifying and understanding the nature of these underlying gene interactions , we will gain better insights in the regulation of complex traits and be able to dissect the architecture of biological networks . In the last decade , numerous studies on the effect of pairwise gene perturbations have been conducted , primarily in the budding yeast Saccharomyces cerevisiae , to systematically evaluate epistasis for several characteristics , such as fitness or synthetic lethality ( Tong et al . , 2004; Jasnos and Korona , 2007; St Onge et al . , 2007; Dixon et al . , 2009; Costanzo et al . , 2010 , 2011 ) . These genome-scale genetic interactions studies were facilitated by the availability of large collections of deletion strains and the development of automated platforms to analyze the phenotypes of double mutants ( Scherens and Goffeau , 2004 ) . Since the first large-scale genetic interaction study in yeast identified 4000 genetic interactions among 1000 genes when analyzing synthetic lethality in double deletion mutants ( Tong et al . , 2004 ) , the field advanced considerably . Currently , about 170 , 000 interactions are known among 5 . 4 million gene pairs screened to affect fitness ( Baryshnikova et al . , 2010; Costanzo et al . , 2010 ) . Interestingly , these studies have shown that the majority of the genes are infrequently connected in the genetic interaction network , while a small fraction of genes shows many interactions ( Baryshnikova et al . , 2010; Costanzo et al . , 2010 , 2011 ) . In higher organisms , large collections of mutants often do not exist and/or the generation of double mutants is much more labor-intensive and time-consuming . However , in the nematode Caenorhabditis elegans , in Drosophila cell cultures and in human cell lines , global analysis of genetic interactions have been performed by making use of RNA interference libraries to generate double mutants ( Lehner et al . , 2006; Byrne et al . , 2007; Barbie et al . , 2009; Horn et al . , 2011 ) . In C . elegans , systematic mapping of interactions between genes functioning in the signaling and the transcriptional networks that regulate development also revealed high connectivity of a small proportion of genes in the network , while most genes have few interactions ( Lehner et al . , 2006 ) . In plants , although large collections of mutants are available for some species such as Arabidopsis ( Alonso et al . , 2003; http://www . arabidopsis . org/ ) , large-scale epistasis studies on double mutants are experimentally and practically virtually impossible to achieve . On a smaller scale , newly identified mutants in Arabidopsis are crossed with known mutants with similar phenotypes or within the same biological process to test for allelic interaction or epistasis . For example , genetic interactions among late flowering Arabidopsis mutants have been studied by generating double mutants ( Koornneef et al . , 1998 ) . Further , genetic modifier screens are performed frequently through a random mutagenesis of individuals harboring one mutant gene to screen for second-site mutations that either enhance or suppress the primary phenotype . An example in relation to leaf size is the identification of enhancer mutations of da1-1 further increasing leaf and seed size ( Li et al . , 2008; Yao et al . , 2008; Xu and Li , 2011; Fang et al . , 2012 ) . While epistasis can easily be detected for qualitative traits , such as synthetic lethality , which are fairly straightforward to visually inspect , genetic interactions from quantitative traits , such as organ growth or gene expression , are more difficult to identify , especially in multicellular organisms ( Kroymann and Mitchell-Olds , 2005; Malmberg et al . , 2005; Xu and Jia , 2007; Chapman et al . , 2012; Steinhoff et al . , 2012; Huang et al . , 2014 ) . Estimating epistasis for quantitative categories of phenotypes implies calculating how much the phenotype of a double mutant deviates from an expected additive value based on the effect of the single mutations ( Fisher , 1918 ) , therefore requiring accurate measurements of the phenotype of the single and double mutants . Although enabling the identification of subtle interactions , these quantitative analyses of gene interactions are not easily amenable to large-scale studies of complex traits . One example of such a complex quantitative trait in higher plants is leaf growth . Leaves are essential to capture solar radiation and convert it into chemical energy by photosynthesis , therefore contributing to a large part of plant biomass production . As for most plant organs , their determinate growth pattern results in a relatively constant size within a fixed environment . Leaf growth is mediated by a cell proliferation phase followed by a cell expansion phase that initiates at the leaf top and proceeds basipetally ( Donnelly et al . , 1999; Andriankaja et al . , 2012 ) . At least five different parameters contribute to the final leaf size ( Gonzalez et al . , 2012 ) : the number of cells incorporated in leaf primordia; the rate of cell division; the developmental window of cell proliferation; the timing of meristemoid division; and the extent of cell expansion . Several genes have been described to , when downregulated or ectopically ( over ) expressed , increase the final leaf size in Arabidopsis ( Gonzalez et al . , 2009; Krizek , 2009; Breuninger and Lenhard , 2010 ) by affecting one or more processes governing leaf growth . Whereas much research has been done on single genes affecting leaf size , the interactions between these growth regulators remain unexplored . So far , only one case of positive epistasis in Arabidopsis leaf growth has been described when a dominant-negative point mutation in DA1 , encoding an ubiquitin receptor , is combined with the knock-out of the ENHANCER OF DA1 ( EOD ) 1/BIG BROTHER , coding for an E3 ubiquitin ligase ( Li et al . , 2008 ) . In this study , we performed a combinatorial screen of transgenic Arabidopsis plants producing larger leaves to identify positive epistatic effects on leaf growth . We aimed to gain further insight in the links between genes controlling growth and the mechanisms driving leaf development . We obtained binary combinations by crossing 13 transgenic lines with an increased leaf size and measured leaf and rosette area of the single and double transgenics . We found that the leaf area of 38% of all combinations was larger than the sum of those of the single mutants , resulting in positive epistatic effects , whereas 23% of the combinations were smaller , showing a negative epistatic effect .
To identify positive epistatic effects on leaf growth , we analyzed pairwise perturbations of 13 genes positively affecting final leaf size in a gain- or loss-of-function situation ( Table 1 ) by measuring the individual and total leaf area . We used lines in the Col-0 background , homozygous for a single-locus insertion of the transgene of interest and shown to have a positive effect on all rosette leaves or a subset of those ( Cho and Cosgrove , 2000; Gonzalez et al . , 2010; Spartz et al . , 2012 ) . This enhanced leaf growth can result from the perturbation of genes affecting cell division and/or cell expansion . The downregulation of SAMBA disturbs the early stage of leaf development , since larger meristems are formed resulting in larger leaves containing more cells ( Eloy et al . , 2012 ) . A point mutation in DA1 or the downregulation of its enhancer , EOD1 , leads to the production of larger leaves with more cells due to an extended cell proliferation phase ( Li et al . , 2008 ) . Similarly , in plants overexpressing ANGUSTIFOLIA3 ( AN3 ) , AINTEGUMENTA ( ANT ) , ARABIDOPSIS VACUOLAR-PYROPHOSPHATASE ( AVP1 ) , GROWTH-REGULATING FACTOR5 ( GRF5 ) under the control of the constitutive 35S promoter or BRASSINOSTEROID INSENSITIVE 1 ( BRI1 ) under the control of its own promoter , larger leaves containing more cells are formed because of an extension of the cell proliferation phase ( Wang et al . , 2001; Horiguchi et al . , 2005; Li et al . , 2005 ) . On the other hand , an increased cell proliferation at the edge of the leaf and a prolonged period of meristemoid division are observed when the miRNA JAW is overexpressed and the PEAPOD ( PPD ) genes are downregulated ( Palatnik et al . , 2003; White , 2006 ) . When GIBBERELLIN 20-OXIDASE 1 ( GA20OX1 ) is overexpressed , an increase in cell number and cell size leads to the formation of larger leaves ( Huang et al . , 1998; Gonzalez et al . , 2010 ) . Finally , in plants overexpressing EXPANSIN 10 ( EXP10 ) and SMALL AUXIN UP-REGULATED RNA 19 ( SAUR19 ) fused to a GFP tag , bigger leaves containing larger cells are produced ( Cho and Cosgrove , 2000; Spartz et al . , 2012 ) . Several of these leaf growth-promoting genes are involved in hormonal pathways , confirming the importance of plant hormones in the regulation of growth processes: BRI1 encodes a brassinosteroid receptor , GA20OX1 catalyzes rate-limiting steps in late gibberellic acid ( GA ) biosynthesis , ANT has been suggested to be involved in auxin signal transduction and both AVP1 and SAUR19 in auxin transport ( Huang et al . , 1998; Mizukami and Fischer , 2000; Wang et al . , 2001; Li et al . , 2005; Spartz et al . , 2012 ) . To obtain pairwise perturbations , our strategy was to cross the homozygous transgenic lines and to analyze the heterozygous progeny . We produced 102 heterozygous combinations , consisting of 78 paired combinations and 24 back-crosses with the wild type ( WT ) used as controls ( Figure 1—figure supplement 1 ) . Because the homozygous line can be used as pollen donor or receptor , care was taken that the crosses with the wild-type plants , producing the heterozygous control line , maintained the same directionality . For example , a cross between ami-ppd ( ♀ ) and SAUR19OE ( ♂ ) was compared to the offspring of the crosses ami-ppd ( ♀ ) X WT ( ♂ ) and WT ( ♀ ) X SAUR19OE ( ♂ ) . This approach standardizes for possible maternal effects ( Scott et al . , 1998 ) . Next , we checked the expression levels of the transgenes in the obtained heterozygous double mutants as well as in the heterozygous control lines . In the majority of the combinations , transgene expression levels were comparable with those of the heterozygous controls ( Figure 1—figure supplement 2 ) . In total , 61 combinations were used for further growth analysis . Sixteen plants per genotype were grown in three independent repeats and at 21 days after stratification ( DAS ) , the size of each individual leaf of the rosette was measured , resulting in 56 , 505 data-points , enabling us to estimate potential gene interactions for these quantitative traits ( Figure 1—figure supplement 3 ) . Leaf area ( LA ) of the paired combinations was compared to a theoretical , expected if non-interacting value ( EXPni ) , based on the size of the WT and both heterozygous controls . To estimate the EXPni , we applied an additive model on a multiplicative scale by transforming the data on log2 scale ( Koornneef et al . , 1998; Phillips , 2008; Horn et al . , 2011 ) :log2 ( LAEXPni ) =log2 ( LAheterozygous control 1 ) +log2 ( LAheterozygous control 2 ) −log2 ( LAwild type ) 10 . 7554/eLife . 02252 . 003Table 1 . Growth regulators and transgenics used for the binary combinationsDOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 003Gene nameGene symbolGene IDLine namePerturbationCellular process promotedReferenceANGUSTIFOLIA3AN3AT5G28640AN3OEOECell division ( Horiguchi et al . , 2005 ) AINTEGUMENTAANTAT4G37750ANTOEOECell division ( Mizukami and Fischer , 2000 ) ARABIDOPSIS V-PYROPHOSPHATASEAVP1AT1G15690AVP1OEOECell division ( Li et al . , 2005 ) BRASSINOSTEROID INSENSITIVE 1BRI1AT4G39400BRI1OEOECell division ( Wang et al . , 2001; Gonzalez et al . , 2010 ) DA1DA1AT1G19270da1-1LOFCell division ( Li et al . , 2008 ) ENHANCER OF DA1-1/BIG BROTHEREOD/BBAT3G63530eod1-2LOFCell division ( Li et al . , 2008 ) EXPANSIN 10EXP10AT1G26770EXP10OEOECell expansion ( Cho and Cosgrove , 2000 ) GIBBERELLIN 20 OXIDASE 1GA20OX1AT4G25420GA20OX1OEOECell division and expansion ( Huang et al . , 1998; Gonzalez et al . , 2010 ) GROWTH REGULATING FACTOR5GRF5AT3G13960GRF5OEOECell division ( Horiguchi et al . , 2005 ) miR-JAW/ miRNA 319miR-JAWAT4G23713jaw-DOECell division ( Palatnik et al . , 2003 ) PEAPODPPDAT4G14713 and AT4G14720ami-ppdLOFMeristemoid division ( White , 2006; Gonzalez et al . , 2010 ) SAMBASAMBAAT1G32310sambaLOFCell division and expansion ( Eloy et al . , 2012 ) SMALL AUXIN UP RNA 19SAUR19AT5G18010SAUR19OEOECell expansion ( Spartz et al . , 2012 ) OE: over-expression , LOF: loss of function . In order to identify combinations with synergistic or negative effects on leaf growth , we searched for significant leaf–genotype interactions ( FDR <0 . 05 ) . The significance of the difference between the EXPni and the observed value was determined using a mixed model ( ‘Materials and methods’ ) . This calculation and comparison was done for each combination ( Figure 1—figure supplement 4–64 ) . The LAs were analyzed using repeated measurements to take into account dependencies between the different leaves of the rosette . We also calculated the total rosette area , defined as the sum of all individual leaves . Similarly as for leaf area , a rosette EXPni was calculated . Among the 61 combinations analyzed , 23 pairwise crosses , almost 38% , were found to have a rosette size significantly exceeding the EXPni value ( FDR <0 . 05 , Figures 1 and 2 ) . In the strongest synergistic combinations , such as BRI1OE-eod1-2 , BRI1OE-EXP10OE , BRI1OE-SAUR19OE , GRF5OE-SAUR19OE , BRI1OE-da1-1 , ami-ppd-SAUR19OE and samba-eod1-2 ( at least 20% larger than the EXPni ) , the positive effect on size was observed for all rosette leaves . Remarkably , although out of the 13 genes that were selected for this screen only two are involved in increasing cell size ( EXP10OE and SAUR19OE ) , almost half of the synergistic combinations arose from combining cell proliferation-stimulating gene perturbations with these two cell expansion-promoting genes , particularly with SAUR19OE ( Figure 2; Table 1 ) . We also observed a positive epistasis in the majority of the combinations with samba , BRI1OE or SAUR19OE , suggesting that these growth regulators are more prone to lead to synergistic effects in binary combinations ( Figure 2—figure supplement 1A ) . 10 . 7554/eLife . 02252 . 004Figure 1 . Heat map representing the effect of the binary combinations for rosette and leaf area . The outer ring shows the percentage of the rosette size of the combinations compared to the WT ( C/W ) . In the middle rings , percentages of the observed sizes of the cotyledons ( L0 ) until leaf 6 ( L6 ) and the complete rosette are shown compared to the expected if non-interacting value ( EXPni ) . Significant differences to the rosette EXPni value ( FDR <0 . 05 ) allowed identifying synergistic interactions ( black line ) and negative interactions ( dashed line ) between two transgenic lines . The inner circle shows the color code with dark pink being the lowest and deep green being the highest value . Combinations that are at least 5% larger than each of their heterozygous controls are marked in bold . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 00410 . 7554/eLife . 02252 . 005Figure 1—figure supplement 1 . Overview of all heterozygous and homozygous combinations and their controls ( Col x mutant or mutant x Col ) obtained by crosses . Horizontally , the pollen donors are shown and vertically the pollen receptors . The samba mutant plants were mostly used as a pollen receptor because they produce little pollen . The heterozygous gene combinations are shown in light green and their controls in darker green . Reciprocal heterozygous crosses are shown in blue and the homozygous combinations are indicated with an ( * ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 00510 . 7554/eLife . 02252 . 006Figure 1—figure supplement 2 . Relative gene expression levels in the heterozygous binary combinations and their controls . Each graph represents the relative expression of a gene of interest in Col-0 , the heterozygous control and the heterozygous combinations . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 00610 . 7554/eLife . 02252 . 007Figure 1—figure supplement 3 . Phenotypic analysis workflow . 16 plants per genotype were grown in three independent repeats and at 21 DAS ( 1 ) , leaf series were made ( 2 ) to measure the individual leaf size ( 5 ) . Images of the leaf series were pre-processed ( 3–4 ) and the individual leaf area was measured ( 5 ) with ImageJ v1 . 45 ( NIH; http://rsb . info . nih . gov/ij/ ) . The data was analyzed using a mixed model and the effect size for the genotypes were calculated ( ‘Materials and methods’ ) . Finally , these estimates of the single control lines and the double transgenics were plotted in graphs ( 6 ) and compared to the wild-type and a calculated expected if additive value in heatmaps ( 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 00710 . 7554/eLife . 02252 . 008Figure 1—figure supplement 4 . Statistical output of the phenotypic data for the heterozygous combination AN3OE-ANTOE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 00810 . 7554/eLife . 02252 . 009Figure 1—figure supplement 5 . Statistical output of the phenotypic data for the heterozygous combination AN3OE-AVP1OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 00910 . 7554/eLife . 02252 . 010Figure 1—figure supplement 6 . Statistical output of the phenotypic data for the heterozygous combination AN3OE-BRIOE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 01010 . 7554/eLife . 02252 . 011Figure 1—figure supplement 7 . Statistical output of the phenotypic data for the heterozygous combination AN3OE-da1-1 . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1-L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 01110 . 7554/eLife . 02252 . 012Figure 1—figure supplement 8 . Statistical output of the phenotypic data for the heterozygous combination AN3OE-eod1-2 . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 01210 . 7554/eLife . 02252 . 013Figure 1—figure supplement 9 . Statistical output of the phenotypic data for the heterozygous combination AN3OE-EXP10OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 01310 . 7554/eLife . 02252 . 014Figure 1—figure supplement 10 . Statistical output of the phenotypic data for the heterozygous combination AN3OE-GA20OX1OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 01410 . 7554/eLife . 02252 . 015Figure 1—figure supplement 11 . Statistical output of the phenotypic data for the heterozygous combination AN3OE-GRF5OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 01510 . 7554/eLife . 02252 . 016Figure 1—figure supplement 12 . Statistical output of the phenotypic data for the heterozygous combination AN3OE-jaw-D . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 01610 . 7554/eLife . 02252 . 017Figure 1—figure supplement 13 . Statistical output of the phenotypic data for the heterozygous combination AN3OE-ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 01710 . 7554/eLife . 02252 . 018Figure 1—figure supplement 14 . Statistical output of the phenotypic data for the heterozygous combination AN3OE-SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 01810 . 7554/eLife . 02252 . 019Figure 1—figure supplement 15 . Statistical output of the phenotypic data for the heterozygous combination ANTOE-AVP1OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 01910 . 7554/eLife . 02252 . 020Figure 1—figure supplement 16 . Statistical output of the phenotypic data for the heterozygous combination ANTOE-BRI1OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 02010 . 7554/eLife . 02252 . 021Figure 1—figure supplement 17 . Statistical output of the phenotypic data for the heterozygous combination ANTOE-da1-1 . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 02110 . 7554/eLife . 02252 . 022Figure 1—figure supplement 18 . Statistical output of the phenotypic data for the heterozygous combination ANTOE-eod1-2 . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 02210 . 7554/eLife . 02252 . 023Figure 1—figure supplement 19 . Statistical output of the phenotypic data for the heterozygous combination ANTOE-EXP10OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 02310 . 7554/eLife . 02252 . 024Figure 1—figure supplement 20 . Statistical output of the phenotypic data for the heterozygous combination ANTOE-GRF5OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 02410 . 7554/eLife . 02252 . 025Figure 1—figure supplement 21 . Statistical output of the phenotypic data for the heterozygous combination ANTOE-ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 02510 . 7554/eLife . 02252 . 026Figure 1—figure supplement 22 . Statistical output of the phenotypic data for the heterozygous combination ANTOE-SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 02610 . 7554/eLife . 02252 . 027Figure 1—figure supplement 23 . Statistical output of the phenotypic data for the heterozygous combination AVP1OE-BRI1OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 02710 . 7554/eLife . 02252 . 028Figure 1—figure supplement 24 . Statistical output of the phenotypic data for the heterozygous combination AVP1OE-da1-1 . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 02810 . 7554/eLife . 02252 . 029Figure 1—figure supplement 25 . Statistical output of the phenotypic data for the heterozygous combination AVP1OE-eod1-2 . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 02910 . 7554/eLife . 02252 . 030Figure 1—figure supplement 26 . Statistical output of the phenotypic data for the heterozygous combination AVP1OE-EXP10OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 03010 . 7554/eLife . 02252 . 031Figure 1—figure supplement 27 . Statistical output of the phenotypic data for the heterozygous combination AVP1OE-ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 03110 . 7554/eLife . 02252 . 032Figure 1—figure supplement 28 . Statistical output of the phenotypic data for the heterozygous combination BRI1OE-da1-1 . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 03210 . 7554/eLife . 02252 . 033Figure 1—figure supplement 29 . Statistical output of the phenotypic data for the heterozygous combination BRI1OE-eod1-2 . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 03310 . 7554/eLife . 02252 . 034Figure 1—figure supplement 30 . Statistical output of the phenotypic data for the heterozygous combination BRI1OE-EXP10OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 03410 . 7554/eLife . 02252 . 035Figure 1—figure supplement 31 . Statistical output of the phenotypic data for the heterozygous combination BRI1OE-GA20OX1OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 03510 . 7554/eLife . 02252 . 036Figure 1—figure supplement 32 . Statistical output of the phenotypic data for the heterozygous combination BRI1OE-GRF5OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 03610 . 7554/eLife . 02252 . 037Figure 1—figure supplement 33 . Statistical output of the phenotypic data for the heterozygous combination BRI1OE-ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 03710 . 7554/eLife . 02252 . 038Figure 1—figure supplement 34 . Statistical output of the phenotypic data for the heterozygous combination BRI1OE-SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 03810 . 7554/eLife . 02252 . 039Figure 1—figure supplement 35 . Statistical output of the phenotypic data for the heterozygous combination da1-1-EXP10OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 03910 . 7554/eLife . 02252 . 040Figure 1—figure supplement 36 . Statistical output of the phenotypic data for the heterozygous combination da1-1-GA20OX1OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 04010 . 7554/eLife . 02252 . 041Figure 1—figure supplement 37 . Statistical output of the phenotypic data for the heterozygous combination da1-1-GRF5OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 04110 . 7554/eLife . 02252 . 042Figure 1—figure supplement 38 . Statistical output of the phenotypic data for the heterozygous combination da1-1-jaw-D . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 04210 . 7554/eLife . 02252 . 043Figure 1—figure supplement 39 . Statistical output of the phenotypic data for the heterozygous combination da1-1-ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 04310 . 7554/eLife . 02252 . 044Figure 1—figure supplement 40 . Statistical output of the phenotypic data for the heterozygous combination da1-1-SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 04410 . 7554/eLife . 02252 . 045Figure 1—figure supplement 41 . Statistical output of the phenotypic data for the heterozygous combination eod1-2-ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 04510 . 7554/eLife . 02252 . 046Figure 1—figure supplement 42 . Statistical output of the phenotypic data for the heterozygous combination EXP10OE-GRF5OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 04610 . 7554/eLife . 02252 . 047Figure 1—figure supplement 43 . Statistical output of the phenotypic data for the heterozygous combination EXP10OE-jaw-D . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 04710 . 7554/eLife . 02252 . 048Figure 1—figure supplement 44 . Statistical output of the phenotypic data for the heterozygous combination EXP10OE-ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 04810 . 7554/eLife . 02252 . 049Figure 1—figure supplement 45 . Statistical output of the phenotypic data for the heterozygous combination EXP10OE-SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 04910 . 7554/eLife . 02252 . 050Figure 1—figure supplement 46 . Statistical output of the phenotypic data for the heterozygous combination GA20OX1OE-GRF5OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 05010 . 7554/eLife . 02252 . 051Figure 1—figure supplement 47 . Statistical output of the phenotypic data for the heterozygous combination GA20OX1OE-jaw-D . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 05110 . 7554/eLife . 02252 . 052Figure 1—figure supplement 48 . Statistical output of the phenotypic data for the heterozygous combination GA20OX1OE-ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 05210 . 7554/eLife . 02252 . 053Figure 1—figure supplement 49 . Statistical output of the phenotypic data for the heterozygous combination GA20OX1OE-SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 05310 . 7554/eLife . 02252 . 054Figure 1—figure supplement 50 . Statistical output of the phenotypic data for the heterozygous combination GRF5OE-jaw-D . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 05410 . 7554/eLife . 02252 . 055Figure 1—figure supplement 51 . Statistical output of the phenotypic data for the heterozygous combination GRF5OE-ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 05510 . 7554/eLife . 02252 . 056Figure 1—figure supplement 52 . Statistical output of the phenotypic data for the heterozygous combination GRF5OE-SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 05610 . 7554/eLife . 02252 . 057Figure 1—figure supplement 53 . Statistical output of the phenotypic data for the heterozygous combination jaw-D-ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 05710 . 7554/eLife . 02252 . 058Figure 1—figure supplement 54 . Statistical output of the phenotypic data for the heterozygous combination jaw-D -SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 05810 . 7554/eLife . 02252 . 059Figure 1—figure supplement 55 . Statistical output of the phenotypic data for the heterozygous combination ami-ppd -SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 05910 . 7554/eLife . 02252 . 060Figure 1—figure supplement 56 . Statistical output of the phenotypic data for the heterozygous combination samba–AN3OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 06010 . 7554/eLife . 02252 . 061Figure 1—figure supplement 57 . Statistical output of the phenotypic data for the heterozygous combination samba -ANTOE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 06110 . 7554/eLife . 02252 . 062Figure 1—figure supplement 58 . Statistical output of the phenotypic data for the heterozygous combination samba -AVP1OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 06210 . 7554/eLife . 02252 . 063Figure 1—figure supplement 59 . Statistical output of the phenotypic data for the heterozygous combination samba -BRIOE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 06310 . 7554/eLife . 02252 . 064Figure 1—figure supplement 60 . Statistical output of the phenotypic data for the heterozygous combination samba -da1-1 . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 06410 . 7554/eLife . 02252 . 065Figure 1—figure supplement 61 . Statistical output of the phenotypic data for the heterozygous combination samba-eod1-2 . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 06510 . 7554/eLife . 02252 . 066Figure 1—figure supplement 62 . Statistical output of the phenotypic data for the heterozygous combination samba -EXP10OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 06610 . 7554/eLife . 02252 . 067Figure 1—figure supplement 63 . Statistical output of the phenotypic data for the heterozygous combination samba -ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 06710 . 7554/eLife . 02252 . 068Figure 1—figure supplement 64 . Statistical output of the phenotypic data for the heterozygous combination samba-SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 06810 . 7554/eLife . 02252 . 069Figure 2 . Network representing the combinations showing positive epistasis on total rosette area and leaf series of gene combinations with a large effect on leaf size . ( A ) The connections between two transgenics indicate the observation of a synergistic effect on rosette size . Two transgenics producing larger leaves resulting from an increased cell area are SAUR19OE and EXP10OE . ( B ) Both synergistic ( GRF5OE-SAUR19OE and ANTOE-SAUR19OE ) and additive combinations ( da1-1-GA20ox1OE and ANTOE-AVP1OE ) lead to plants strongly enlarged up to 39% compared to the WT . In order to flatten the leaves for area measurements , cuts were made in the blade . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 06910 . 7554/eLife . 02252 . 070Figure 2—figure supplement 1 . Occurrence of the growth-regulating genes in a ( A ) synergistic combination and ( B ) negative combinations . The values indicate for each gene the % of synergistic effect observed in all its combinations . For example , 83% of all combinations with eod1-2 synergistically enhance leaf growth . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 07010 . 7554/eLife . 02252 . 071Figure 2—figure supplement 2 . Phenotype of the homozygous combination da1-1-SAUR19OE . ( A ) Leaf series to illustrate the size increase compared to the WT . ( B ) Percentages of the observed sizes of the cotyledons ( L0 ) until leaf 6 ( L6 ) and the complete rosette ( R ) are shown compared to the WT . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 071 Of all binary crosses analyzed , 39 . 2% resulted in plants with a rosette size exceeding that of both heterozygous control lines and the WT ( Figure 1; Supplementary file 1 ) . Interestingly , 16 combinations resulted from a synergistic effect , while eight were the result of an additive effect . Among the largest plants , synergistic ( GRF5OE-SAUR19OE and ANTOE-SAUR19OE , 39% and 38% larger than the WT , respectively ) and additive ( da1-1-GA20ox1OE and ANTOE-AVP1OE , 38% and 36% larger than the WT , respectively ) effects could be found . In addition , we also found that 23% of the combinations led to the formation of smaller rosettes than expected . We observed that mainly combinations with jaw-D and ami-ppd led to cases of negative epistasis . The total rosette area of these combinations was similar or much smaller than that of WT plants , such as GRF5-jaw-D ( 46% smaller than the WT ) , with the exception of da1-1-ami-ppd , which was larger than the WT , but smaller than da1-1-Col-0 . ( Figure 1 ) . In conclusion , from this screen , we found that more than one third of the combinations showed positive epistasis on leaf growth , resulting from combining either two genes both stimulating cell proliferation , or either one gene enhancing cell proliferation and the other cell expansion . To strengthen the observed effects of pairwise perturbations and to further exclude that the observed phenotypes were influenced by maternal effects , we made reciprocal crosses of selected synergistic combinations ( SAUR19OE-ami-ppd , EXP10OE-BRI1OE , SAUR19OE-BRI1OE ) . We measured the leaf area at 21 DAS and could confirm the synergistic effects for all three combinations ( Figure 3A , Figure 3—figure supplement 1–3 ) . Next , we generated homozygous lines for two synergistic combinations , ami-ppd-SAUR19OE and samba-eod1-2 , and one additive combination , producing nevertheless a very large rosette , da1-1-SAUR19OE . Transgene expression levels in these homozygous lines were verified and found comparable to those in the homozygous single lines ( Figure 3—figure supplement 4 ) . We confirmed a synergistic effect on the rosette sizes in homozygous ami-ppd-SAUR19OE and samba-eod1-2 plants ( 24% and 8% larger than the rosette EXPni respectively ) ( Figure 3B , Figure 3—figure supplement 5 , 6 ) . The combination da1-1-SAUR19OE , which produced among the largest plants in the screen , but did not enhance leaf size synergistically , was also found to be particularly large when homozygous , since its rosette size was 61% larger than that of the WT ( Figure 3B , Figure 3—figure supplement 7 , Figure 2—figure supplement 2 ) . From these experiments we could confirm the observed positive epistatic effects in a selected set of double mutants from our screen of heterozygous combinations in a reciprocal direction and/or homozygous status . 10 . 7554/eLife . 02252 . 072Figure 3 . Heat map representing the effect of the binary combinations for rosette and leaf area ( A ) in reciprocal heterozygous crosses and ( B ) homozygous lines . C/W represents the percentage of the rosette size of the combinations compared to the WT . Percentages of the observed sizes of the cotyledons ( L0 ) until leaf 6 ( L6 ) and the complete rosette are shown compared to the expected if non-interacting value ( EXPni ) . The color code represents the range of differences with dark pink being the lowest and deep green being the highest value . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 07210 . 7554/eLife . 02252 . 073Figure 3—figure supplement 1 . Statistical output of the phenotypic data for the heterozygous combination EXP10OE-BRI1OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( b ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 07310 . 7554/eLife . 02252 . 074Figure 3—figure supplement 2 . Statistical output of the phenotypic data for the heterozygous combination SAUR19OE-BRI1OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 07410 . 7554/eLife . 02252 . 075Figure 3—figure supplement 3 . Statistical output of the phenotypic data for the heterozygous combination SAUR19OE-ami-ppd . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 07510 . 7554/eLife . 02252 . 076Figure 3—figure supplement 4 . Relative gene expression levels in the homozygous binary combinations and their controls . Each graph represents the relative expression of a gene of interest in Col-0 , the homozygous control and the homozygous combinations . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 07610 . 7554/eLife . 02252 . 077Figure 3—figure supplement 5 . Statistical output of the phenotypic data for the homozygous combination ami-ppd-SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 07710 . 7554/eLife . 02252 . 078Figure 3—figure supplement 6 . Statistical output of the phenotypic data for the homozygous combination samba-eod1-2 . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 07810 . 7554/eLife . 02252 . 079Figure 3—figure supplement 7 . Statistical output of the phenotypic data for the homozygous combination da1-1-SAUR19OE . Top left panel ( A ) : p1/w , p2/w , c/w: percentage of the area to the WT ( w ) of parent 1 ( p1 ) , parent 2 ( p2 ) and the combination ( C ) respectively . c/e: percentage of the area of the combination ( C ) to the expected if non-interacting value ( EXPni ) ( e ) . Top right panel ( B ) , corresponding FDRs for the percentages presented in the top left panel . The cotyledons ( L0 ) , first six leaves ( L1–L6 ) and the rosette ( R ) are represented . Bottom left panel ( C ) : graphs representing leaf areas for the WT ( green ) , the combination ( red ) and both single lines ( dark and light blue ) in mm2 . Bottom right panel: graph showing the leaf area of the combination ( red ) and the WT ( green ) compared to the calculated EXPni ( dotted red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 079 In order to explain the cause for the observed synergistic phenotype at a cellular level , we quantified cell numbers and cell size in the homozygous combination ami-ppd-SAUR19OE . In the ami-ppd line , in which PPD1 and PPD2 expression is downregulated , the increased leaf size results from a prolonged division of meristemoids ( White , 2006 ) , whereas overexpression of SAUR19 leads to cell enlargement ( Spartz et al . , 2012 ) . Samples of leaf 3 were harvested at 21 DAS , cleared and cell drawings of the abaxial epidermis were analyzed . As shown in Figure 4 , the larger leaves of SAUR19OE contain less but larger cells , whereas in leaves of ami-ppd more cells are produced . In the latter , an observed reduction in average cell area results from the presence of a larger amount of smaller cells surrounding the stomata which do not reach the mature wild-type size ( Figure 4B ) . In the homozygous ami-ppd-SAUR19OE line , we observed an increased cell number compared to the WT , but to a lower extend than in the ami-ppd line , and an increased cell area similar to that of SAUR19OE . Thus the effect of SAUR19OE allows for an increased cell expansion of the many small cells resulting from PPD downregulation . 10 . 7554/eLife . 02252 . 080Figure 4 . Cellular basis of the difference in leaf size observed for the homozygous line amippd-SAUR19OE and the corresponding controls . ( A ) The graphs represent the percentage difference of leaf area , cell number and cell area between a transgenic and the WT . ( n = 3; *p<0 . 05 ) . ( B ) Representative drawing of cells in the different lines . Cells are colored in function of their area . Red: cells smaller than 1 . 25 E−4 mm2 , light green: cell area ranging from 1 . 25 E−4 mm2 to 1 . 6 E−3 mm2 , medium green: cell area ranging from 1 . 6 E−3 mm2 to 3 . 2 E−3 mm2 , dark green: cells larger than 6 . 4 E−3 mm2 , stomata are marked in grey . DOI: http://dx . doi . org/10 . 7554/eLife . 02252 . 080
In order to identify potential interactions existing within the genetic network regulating leaf growth , we pairwised combined 13 gene perturbations each leading to an enhanced leaf size and looked for positive interactions resulting in an increased leaf area larger than the additive combination of the single perturbations . From this screen , we found that 61% of the paired perturbations showed epistasis: 38% of the studied gene combinations further enhanced leaf organ size synergistically and 23% negatively influenced leaf size . Studies using limited numbers of mutations , random or affecting a specific trait , also showed that epistasis is common , although lower levels of interactions were found ( Clark and Wang , 1997; Magwire et al . , 2010 ) . In D . melanogaster , for example , 35 of 128 ( 27% ) of random paired mutations showed epistasis ( Clark and Wang , 1997 ) . Larger-scale studies , in systems allowing automated quantitative assays , identified between 13 and 35% of epistatic effects ( Byrne et al . , 2007; St Onge et al . , 2007 ) . The large number of interactions we identified could be explained by the fact that we studied a set of perturbations , including loss and gain of function , leading to one particular phenotype , namely an increase of leaf area . In model systems permitting genome-wide genetic interactions assays , all genes are either knocked down or knocked out and these perturbations can therefore affect the studied trait , for example fitness , by increasing it or decreasing it . In D . melanogaster , the study of ten mutations leading to an increased life span showed that paired combinations have high levels of connections , with 21 significant epistatic interactions in males and/or females ( 47% ) observed ( Magwire et al . , 2010 ) . Interestingly , three genes , SAMBA , BRI1 and SAUR19 , were found to lead to a synergistic effect in the majority of combinations they were part of . Large-scale genetic interaction studies in yeast and nematodes have shown that most genes in a network have only a few interactions , while a limited number of genes show multiple interactions and are therefore considered as network hubs mediating across-process connections ( Lehner et al . , 2006; Baryshnikova et al . , 2010; Costanzo et al . , 2010 ) . Despite the relative small scale of the study presented here , our observations suggest that SAMBA , BRI1 and SAUR19 play a central role in the leaf growth regulatory networks . Two of these highly connected genes in synergistic combinations , BRI1 and SAUR19 , have a known role in hormone signaling . Interestingly , yeast studies have shown that highly connected genes in a genetic network tend to be pleiotropic and multi-functional ( Costanzo et al . , 2010 ) , similar to plant hormones which regulate multiple processes . BRI1 is a receptor of the brassinosteroid ( BR ) hormone which plays a crucial role in several biological processes , including leaf growth , as severe dwarfism is observed in bri1 mutants and other mutants of the BR biosynthesis and signaling pathways ( Clouse et al . , 1996; Vert et al . , 2008 ) . BRI1 is highly expressed in all organs during early seedling development ( Friedrichsen et al . , 2000 ) similarly to highly connected genes in yeast which show high mRNA levels ( Costanzo et al . , 2010 ) . Additionally , introduction of BRI1OE into P10-CKX3OE , which has a smaller rosette size than WT plants , results in positive epistatic effects on shoot growth ( Vercruyssen et al . , 2011 ) , highlighting the importance of this gene in leaf growth regulation . SAUR19 belongs to the family of SAUR genes known to be rapidly and strongly induced by auxin ( Hagen and Guilfoyle , 2002 ) , which plays a major role in the initiation of leaf primordia , the formation of vascular patterns and leaf shape , but also in the regulation of leaf cell expansion ( Chen et al . , 2001; Wilmoth et al . , 2005; Scarpella et al . , 2010 ) . SAUR19 is a positive regulator of cell expansion , most likely through the modulation of auxin transport ( Spartz et al . , 2012 ) . Our findings therefore suggest that alterations of BR or auxin signaling in the binary combinations could potentiate the effect of several growth-promoting genes . Interactions between BR and other plant hormones have been shown for several physiological and developmental processes ( Choudhary et al . , 2012; Li and He , 2013; Zhu et al . , 2013 ) . BR and auxin interactions exist at multiple levels , including hormone synthesis , transport , signal transduction , and gene transcription . For example , microarray studies have revealed similar effects of BR and auxin on a large number of genes , including a member of the SAUR family , SAUR15 ( Goda et al . , 2004; Nemhauser et al . , 2004; Walcher and Nemhauser , 2012 ) . Interestingly , exogenous application of both hormones leads to a synergistic induction of many common targets ( Nemhauser et al . , 2004; Vert et al . , 2008 ) . In addition , auxin can increase the biosynthesis of BRs ( Chung et al . , 2011; Yoshimitsu et al . , 2011 ) and the BR-regulated BIN2 kinase contributes to a synergistic increase in auxin-induced gene expression ( Vert et al . , 2008 ) . The overexpression of both BRI1 and SAUR19 , involved in BR perception and auxin transport , respectively , could therefore amplify the effect of both hormones , hereby leading to the observed synergism in leaf growth . Studies in yeast have shown that most genetic interactions occur between genes involved in the same biological process , except for highly connected genes ( Tong et al . , 2004; Costanzo et al . , 2010 ) . In agreement with these studies , we found that by combining AN3OE with GRF5OE , shown to interact in a yeast two-hybrid assay ( Horiguchi et al . , 2005 ) , the leaf size is increased more than expected . A similar effect is seen when BRI1OE and ami-ppd , both producing enlarged and curled leaves ( Wang et al . , 2001; White , 2006 ) , are combined . Moreover , PPD genes regulate the division of dispersed meristemoid cells in the leaf epidermis , which will give rise to the stomatal lineage ( White , 2006 ) and BRs have been shown to control stomatal development ( Gudesblat et al . , 2012; Kim et al . , 2012; Khan et al . , 2013 ) . In addition , in BRI1 overexpressing seedlings , PPD2 has been reported to be downregulated ( Gonzalez et al . , 2010 ) . In literature , the combination of da1-eod has been reported to show a positive epistatic effect on leaf growth . Both proteins are suggested to work in ubiquitin-mediated proteolysis that could modulate the activity of a shared , yet unknown target ( Li et al . , 2008 ) . However , not only combining growth-regulating genes that are interconnected can lead to larger phenotypes than expected , also combining cell proliferation with cell expansion leads to positive effects on leaf size as found in the combinations ami-ppd-SAUR19OE , GRF5OE-SAUR19OE and samba-EXP10OE . In addition , the combination of lines positively affecting distinct growth processes seems to allow compensating negative effects sometimes observed when constitutively expressing or strongly downregulating growth regulators , such as observed in ami-ppd-SAUR19OE ( Figure 4 ) . In plants overexpressing GRF5 and jaw-D , each promoting cell proliferation , a reduction in cell area has also been reported ( Gonzalez et al . , 2010 ) . Interestingly , when these genes are combined with SAUR19OE , a synergistic effect on growth can be observed , similar to ami-ppd-SAUR19OE . This suggests that the double transgenic line can acquire the benefits from both genes and therefore enhance leaf size more than expected . Such compensation could be lacking in the negative combinations we observed , therefore leading to the formation of smaller plants than expected . For example , by combining GRF5 and jaw-D , both producing more but smaller cells , the negative effect on leaf size could be caused by overstimulation of cell division that affects the overall growth as observed when E2Fa and DPa are overexpressed simultaneously ( De Veylder et al . , 2002 ) . These findings highlight the challenge of studying genetic interactions in multicellular organisms , compared to single cell systems such as yeast . Genetic interactions observed at the organ level can reflect connections between genes working in the same pathway , but also the interconnection of several processes such as cell division and cell expansion which occur in different cell types and tissues , at different rates and developmental stages . Although yeast is heavily used as a model to identify genetic interactions , it will be essential to also use multicellular organisms as a model for genetic interactions to capture the complex relationship between developmental processes . In this study we searched for binary combinations of growth-regulating genes exhibiting an increase in leaf growth larger than the addition of the two single transgenic parents . In plants and animals , the phenomenon of heterosis or hybrid vigor corresponds to the increased performance of a hybrid offspring compared to its parents ( Schnable and Springer , 2013 ) . Heterosis has been proposed to arise from various mechanisms such as intra-allelic dominance and intra-allelic over-dominance , but emerging evidence also exists for the contribution of inter-gene interactions , or epistasis ( Kaeppler , 2012; Chen , 2013; Schnable and Springer , 2013 ) . Our findings suggest that differences in expression of growth-promoting genes in natural variants could lead to synergistic effects in hybrids . For example , one could imagine that in one variant , PPD is lowly expressed , whereas in another variant SAUR19 is highly expressed . The combination of both genes in a cross of natural variants could lead to a synergistic increase in leaf size as observed in our study . Heterosis could therefore originate , in part , from the assembly of the effects of various pairwised combinations of growth-regulating genes . Another theory to explain heterosis describes the fact that hybrid vigor allows for the compensation of small negative alleles ( Kaeppler , 2012; Chen , 2013; Schnable and Springer , 2013 ) . In our study , we also found that negative effects of some perturbations can be compensated in pairwised combinations , allowing the appearance of a synergistic effect on growth , such as in the cross ami-ppd-SAUR19OE . So far , genetic engineering of crops mainly has been commercially successful for input traits , such as insect tolerance and herbicide resistance ( http://www . isaaa . org ) . Engineering quantitative , yield-related traits , such as drought tolerance and enhanced biomass production , turned out to be much more difficult . The current study illustrates that gene combinations have great promise to successfully engineer quantitative traits . Furthermore , the observation that genes stimulating cell proliferation combine remarkably well with genes enhancing cell expansion , argues for a need for further in-depth analysis of how single genes promote organ growth . A better understanding of the mode of action of growth- and/or yield-enhancing genes will allow for rationalizing which gene stacks have the highest probability to give successful results . Future prospects of combining multiple genes or even entire circuits of networks using synthetic biology approaches offer great perspectives to further enhance crop yield and to deliver sufficient food for the growing world demand .
Seeds of A . thaliana ( L ) Heyhn . ecotype Columbia-0 ( Col-0 ) and all mutants ( Table 1 ) were grown on soil and kept in the same growth room for 25 days , when flower stalks started to emerge . For all single insertion locus transgenic lines , binary crosses were made in one direction; for a selection of these lines , reciprocal crosses were made and homozygous lines were produced . All plants were grown on plates containing half-strength MS medium ( Murashige and Skoog , 1962 ) supplemented with 1% sucrose with a density of one plant per 4 cm2 . The seeds were stratified for 2 days at 4°C and placed in growth rooms kept at 21°C and 16-hr day/8-hr night cycles . Plants were grown in three experiments , consisting of 16 replicates per experiment . To ensure environmental conditions are similar between the experiments , they were performed consecutively in the same growth chamber on the same shelf . To prevent positional effects on plant growth , all plates were randomized every 2 days . We set out to grow all genotypes simultaneously in three repeats , but due to germination issues with some seed batches , a total of 5 experiments have been performed to obtain three repeats for each genotype , with the exception of the cross ANTOE-da1-1 for which we could obtain results in one repeat . At 21 DAS , individual leaves ( cotyledons and rosette leaves ) were dissected at the base of the petiole and their area was measured with ImageJ v1 . 45 ( NIH; http://rsb . info . nih . gov/ij/ ) . Total RNA was extracted from flash-frozen seedlings with TRIzol reagent ( Invitrogen , Belgium ) . To eliminate the residual genomic DNA present in the preparation , the RNA was treated by RQ1 RNAse-free DNase according to the manufacturer's instructions ( Promega , The Netherlands , http://www . promega . com ) and purified with the RNeasy Mini kit ( Qiagen , The Netherlands , http://www . qiagen . com ) . Complementary DNA was made with the iScript cDNA Synthesis kit from Biorad ( Biorad , Belgium , http://www . bio-rad . com ) according to the manufacturer's instructions . Q-RT-PCR was done on a LightCycler 480 ( Roche , Belgium , http://www . roche . com ) in 384-well plates with LightCycler 480 SYBR Green I Master ( Roche ) according to the manufacturer's instructions . Primers were designed with the Primer3 ( http://frodo . wi . mit . edu/ ) ( Supplementary file 2 ) . Data analysis was performed using the ΔΔCT method ( Pfaffl , 2001 ) , taking the primer efficiency into account . The data was normalized using six normalization genes ( UBQ10 , CDKA1 , CBP20 , AT1G13320 , AT2G32170 , and AT2G28390 ) according to the GeNorm algorithm ( Vandesompele et al . , 2002 ) . For the cellular analysis , samples of leaf 3 were cleared in 70% ethanol and mounted in lactic acid on a microscope slide . The total leaf blade area was measured for 10 representative leaves under a dark-field binocular microscope . Abaxial epidermal cells along the complete proximal–distal axis of the leaves were drawn with a microscope equipped with differential interference contrast optics ( DM LB with 403 and 633 objectives; Leica ) and a drawing tube . Photographs of leaves and scanned cell drawings were used to measure leaf and cell area , respectively , with ImageJ v1 . 45 ( NIH; http://rsb . info . nih . gov/ij/ ) , from which the cell numbers were calculated ( De Veylder et al . , 2001 ) . The leaf series data analysis yields the size of each individual leaf of the rosette . From this data , the rosette area was calculated by summating the area of all separate leaves . A mixed model analysis was performed on the log2 transformed rosette areas using Genotype as a fixed factor . Each experiment was repeated three times . This Experiment effect was included as a random factor in the model to account for correlation between measurements done within the same experiment . The Genotype*Experiment interaction was included in the model when it was found to be significant ( p<0 . 05 ) , based on a likelihood ratio test . For all described combinations the variation attributed to the Genotype was much larger than that attributed to the interaction between Genotype and Experiment . Severe outliers , caused by germination problems , were removed prior to the analysis . Least square means estimates for the rosette area were calculated . Significant differences between the rosette area ( RA ) of the cross and its parental lines , as well as with the reference plants , were determined using the described mixed model ( WALD-type III tests of fixed effects ) . To test for synergistic effects following null hypothesis was set up:log2 ( RA^cross ) =log2 ( RA^control1 ) +log2 ( RA^control2 ) −log2 ( RA^wildtype ) Through log-transformation of the data , we apply an additive model with a multiplicative scale ( Koornneef et al . , 1998; Phillips , 2008; Horn et al . , 2011 ) . As control lines the appropriate heterozygous parental lines were used . By rearranging terms we get:log2 ( RA^cross ) −log2 ( RA^control1 ) −log2 ( RA^control2 ) +log2 ( RA^wildtype ) =0 A FDR multiple testing correction was applied . Synergistic effects were assumed when the null hypothesis was rejected at a FDR level of 0 . 05 . The model was fit with the mixed procedure from SAS . To estimate repeatability ( broad sense heritabilities at the individual level ) , the mixed model was refit with genotype , experiment and genotype*experiment as random terms in the model ( Supplementary file 3 ) . The leaf series data was analyzed using repeated measurements analysis with either the hpmixed or mixed procedure from SAS . Data for leaves up to leaf 6 was included in the analysis . The four variance-covariance structures available in the procedure were tested and the best structure was determined based on the AIC values . For all combinations the unstructured structure was selected as the best . The mean model included the main effects Genotype and Leaf , and their interaction term . To account for dependencies of observations made within the same experiment , experiment was added as random factor in the model . Based on a likelihood ratio test the Genotype*Experiment interaction was incorporated in the model when p<0 . 05 . Several contrast hypotheses were set up . For all leaves , the area in the reference line was compared to that in the cross and both parental lines . Synergistic effects of the cross were determined for each leaf , as described previously . For the cross ANTOE-da1-1 , there was only one experiment that yielded results , therefore Experiment was not included as a factor in the model . All statistical analyses were performed with SAS 9 . 3 ( SAS Institute Inc . , 2011 , Cary , North Carolina ) . Residual diagnostics were carefully examined . | Different individuals of the same species are not usually identical . Humans , for example , have a range of different values for various traits , such as height and weight , and similar variations are seen in other animals and plants . Some of this variation can be explained by the individual animals or plants inheriting different versions of the same gene . Moreover , two or more genes can sometimes interact to produce a bigger ( or smaller ) effect on given trait than would be expected from combining the effects of each gene . This phenomenon is called epistasis . Leaf size in plants is a trait that is controlled by genes and by the environment . Genes can exert an effect by influencing how often the cells in the leaves divide and by influencing their expansion . In Arabidopsis thaliana , a small plant that has been widely studied by plant biologists , numerous genes influence leaf size . Now , in order to explore epistasis in plants , Vanhaeren , Gonzalez et al . have crossbred 13 Arabidopsis mutants that had leaves that were larger than those found in wild-type plants . The resulting offspring each inherited a different pair of mutant genes: one gene via the male pollen , and the other via the female egg cell . About 39% of the offspring plants had leaves that were even bigger than those of its parents . Moreover , in most cases the increase in leaf size was larger than expected . Vanhaeren , Gonzalez et al . found that three of the 13 genes were responsible for the biggest increases in leaf size . Also , some of the biggest increases were caused by plants inheriting two genes that both caused the cells to divide more . Other large increases in leaf size were caused by interactions between a gene affecting cell division and a gene affecting cell expansion . By uncovering combinations of plant genes that increase leaf size , it might be possible to develop agricultural crops with enhanced yields . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology"
] | 2014 | Combining growth-promoting genes leads to positive epistasis in Arabidopsis thaliana |
The delta-protocadherins ( δ-Pcdhs ) play key roles in neural development , and expression studies suggest they are expressed in combination within neurons . The extent of this combinatorial diversity , and how these combinations influence cell adhesion , is poorly understood . We show that individual mouse olfactory sensory neurons express 0–7 δ-Pcdhs . Despite this apparent combinatorial complexity , K562 cell aggregation assays revealed simple principles that mediate tuning of δ-Pcdh adhesion . Cells can vary the number of δ-Pcdhs expressed , the level of surface expression , and which δ-Pcdhs are expressed , as different members possess distinct apparent adhesive affinities . These principles contrast with those identified previously for the clustered protocadherins ( cPcdhs ) , where the particular combination of cPcdhs expressed does not appear to be a critical factor . Despite these differences , we show δ-Pcdhs can modify cPcdh adhesion . Our studies show how intra- and interfamily interactions can greatly amplify the impact of this small subfamily on neuronal function .
The delta-protocadherins ( δ-Pcdhs ) are a nine-member subfamily of the cadherin superfamily ( Hulpiau and van Roy , 2009; Nollet et al . , 2000 ) , and play diverse roles during neural development . Mutagenesis studies have shown that individual δ-Pcdhs are important for neural development , including hindbrain formation , axon guidance , and synaptogenesis ( Cooper et al . , 2015; Emond et al . , 2009; Hayashi et al . , 2014; Hoshina et al . , 2013; Leung et al . , 2013; Light and Jontes , 2017; Uemura et al . , 2007; Williams et al . , 2011 ) . In humans , mutations in PCDH19 are the causative basis of one form of epilepsy ( Dibbens et al . , 2008 ) , and other δ-Pcdhs are implicated in various neurological disorders ( Chang et al . , 2018; Consortium on Complex Epilepsies , 2014; Morrow et al . , 2008 ) . How does this relatively small gene family mediate these varied effects ? While significant effort has been devoted towards characterizing the role of individual δ-Pcdhs in neural development , almost nothing is known regarding how multiple family members function together . The δ-Pcdh subfamily has been further divided into the δ−1 ( Pcdh1 , Pcdh7 , Pcdh9 , and Pcdh11x ) and δ−2 ( Pcdh8 , Pcdh10 , Pcdh17 , Pcdh18 , and Pcdh19 ) subfamilies based on differences in the number of extracellular domains and also within the intracellular domain ( Redies et al . , 2005; Vanhalst et al . , 2005 ) . Double label RNA in situ hybridization studies indicate individual neurons express more than one δ-Pcdh ( Etzrodt et al . , 2009; Krishna-K et al . , 2011 ) . This suggests a model where different combinations of δ-Pcdhs may be expressed within different populations of neurons . Whether such combinations exist or how many δ-Pcdhs may be expressed per neuron is still not known . It seems reasonable , however , to postulate that combinatorial expression would greatly enhance the impact of δ-Pcdhs on cellular function . If such combinations exist , how they would influence or modify δ-Pcdh-mediated adhesion is also unknown . The importance of examining intrafamily δ-Pcdh interactions was recently underscored by a study examining the role of δ-Pcdh adhesion in PCDH19-GCE ( girls clustering epilepsy ) , a form of epilepsy limited to females . Pederick et al . demonstrated that mutations in PCDH19 , a δ−2 family member , affected cell sorting in both in vitro aggregation assays and in brains of mice . Furthermore , they also showed that humans with PCDH19-GCE exhibit abnormal cortical folding patterns ( Pederick et al . , 2018 ) . Importantly , they noted that PCDH19 is likely to be co-expressed with other δ-Pcdh family members , and tested how expressing PCDH10 and/or PCDH17 with PCDH19 affected sorting behavior in aggregation assays . In each case , the observed cell sorting behavior varied depending upon which δ-Pcdhs were co-expressed . This study demonstrated the importance of defining intrafamily interactions in order to understand how loss of Pcdh19 influences function . However , it did not define the extent of such combinations in vivo . It also did not establish any guiding principles for δ-Pcdh adhesion , or how different combinations influence adhesion . Here , we uncover principles used by the δ-Pcdhs to regulate combinatorial adhesion . We first used single color and double label RNA in situ hybridization to show that olfactory sensory neurons ( OSNs ) are likely to express different combinations of δ-Pcdhs . We next employed single cell RNA analysis to establish the scope of these combinations , and find individual OSNs express between zero and seven δ-Pcdhs . We then systematically address the impact of this combinatorial diversity on intrafamily interactions by utilizing cell aggregation assays . In striking contrast to what has been seen for the clustered protocadherins ( cPcdhs; Thu et al . , 2014 ) , we observed a range of potential adhesive behaviors . We were able to define fundamental principles that regulate these outcomes . In combination , these principles provide cells with a powerful means of fine tuning their adhesive interactions with other cells . Finally , we show that δ-Pcdhs can also modify the adhesive function of cPcdhs , which have been shown to be important for neuronal survival , dendritogenesis , synapse formation , and self-avoidance ( Lefebvre et al . , 2012; Molumby et al . , 2016; Wang et al . , 2002; Weiner et al . , 2005 ) . These results provide an initial glimpse into interfamily interactions among protocadherin subfamilies . Our studies therefore provide a framework for determining how combinations of δ-Pcdhs mediate adhesion , and also lay the foundation for understanding how different cadherin subfamilies integrate to regulate cell-cell adhesion .
We first performed single color RNA in situ hybridization to examine δ-Pcdh expression in the olfactory epithelium ( Figure 1—figure supplement 1A–G ) . All detectable δ-Pcdhs were expressed in a punctate pattern , indicating differential expression among OSNs . Interestingly , the expression pattern for any given δ-Pcdh was not uniform throughout the epithelium . For example , Pcdh1 is more highly expressed in the lateral epithelium , and more weakly medially ( Figure 1—figure supplement 1B , C ) . In both regions , the expression was clearly punctate , but greater numbers of OSNs in the lateral epithelium expressed Pcdh1 . In contrast , other δ-Pcdhs , such as Pcdh9 and Pcdh17 , show the opposite pattern , and are more strongly expressed medially with relatively low expression laterally ( Figure 1—figure supplement 1D–G ) . Differences between δ−1 and δ−2 family members could not be distinguished based upon these patterns . These patterns are essentially maintained as development proceeds , although subtle changes in expression did occur . One exception was Pcdh10 , whose expression we previously demonstrated to be dependent upon odorant-mediated activity ( Williams et al . , 2011 ) . The δ-Pcdhs are therefore expressed in regional patterns that overlap one another , suggesting combinatorial expression . We used double label RNA in situ hybridization to begin testing this hypothesis ( Figure 1A ) . We systematically assayed all expressed pairs to show that 5–35% of olfactory sensory neurons ( OSNs ) co-express any two δ-Pcdhs ( Figure 1—figure supplement 1H ) . Interestingly , the degree of co-expression varied within the family . For example , Pcdh1 and Pcdh7 were only co-expressed 8% of the time , while Pcdh8 and Pcdh9 were co-expressed 35% of the time . OSNs expressing the same odorant receptor project to common targets within the olfactory bulb ( Ressler et al . , 1994; Vassar et al . , 1994 ) . Mutant analysis of members of the δ-Pcdh and cPcdh subfamilies has previously shown these genes are important for OSN targeting ( Hasegawa et al . , 2008; Mountoufaris et al . , 2017; Williams et al . , 2011 ) . Interestingly , however , not all OSN populations were equally affected . Why some populations expressing a particular odorant receptor were more strongly affected in the mutant than those expressing a different receptor is unknown . We theorized that different OSN populations may express different combinations of δ-Pcdhs . Changes in these combinations would therefore affect cell adhesion mediated by the δ-Pcdhs . We therefore performed a second double label RNA in situ hybridization series to survey which δ-Pcdhs are co-expressed among OSNs expressing a given odorant receptor . For any one δ-Pcdh , we examined on average ~70 cells expressing a given odorant receptor to determine the degree of overlap ( Figure 1B , C ) . Confocal analysis showed all five OSN populations surveyed express varying proportions of different δ-Pcdhs ( Figure 1B , C ) . There were striking differences in expression of δ-Pcdhs among the different OSN populations , arguing for the presence of specific combinations of δ-Pcdhs within each population . Interestingly , we did not find a simple one-to-one correspondence between odorant receptor expression and δ-Pcdh expression . Instead , different OSN populations varied in the proportion of δ-Pcdhs they expressed . For example , Pcdh9 was expressed by more than half of all OSNs expressing Olfr558 . In contrast , ~12% of Olfr557 OSNs expressed Pcdh9 . The variation in δ-Pcdh expression within OSN populations indicates additional levels of regulation must exist . Nevertheless , different OSN populations clearly possess differences in the proportion of δ-Pcdhs expressed by those OSNs . Such differences could be important for defining how δ-Pcdhs mediate targeting . We next used the NanoString nCounter platform ( Geiss et al . , 2008 ) to more precisely define the extent of co-expression . We isolated 50 randomly selected OSNs , and performed single neuron RNA analysis for δ-Pcdhs and a subset of other genes . A heat map of the raw NanoString data showed strong heterogeneity among OSNs ( Figure 1D , Figure 1—source data 2 ) . To classify δ-Pcdhs as being ‘on’ or ‘off’ in a neuron , we used a constrained gamma-normal mixture model ( Ghazanfar et al . , 2016; Figure 1—figure supplement 1I ) . This revealed that individual OSNs expressed anywhere from zero to seven δ-Pcdhs ( Figure 1E ) , far exceeding prior estimates based on RNA in situ studies . We were unable to determine if there was any preference for co-expression among or between the δ−1 or δ−2 subfamilies . We performed several validation experiments ( see Validation of NanoString data , Figure 1F , and Figure 1—figure supplement 1J ) , including qRT-PCR on individual OSNs . The observed ‘on’ or ‘off’ expression pattern of this particular validation experiment was highly similar to our NanoString results ( Figure 1F ) . We chose NanoString because we hypothesized a targeted approach would be more sensitive than single cell RNA-seq , which is often limited by low capture efficiency of mRNA ( Islam et al . , 2011; Marinov et al . , 2014 ) . Subsequent comparison with single OSN RNA-seq data sets confirmed this hypothesis ( Figure 1—figure supplement 1K , L ) . To determine how δ-Pcdh combinations affect adhesion , we used K562 cell aggregation assays . K562 cells are commonly used to study adhesion mediated by cadherins because it is believed they do not express endogenous cadherins and are non-adherent ( Ozawa and Kemler , 1998; Schreiner and Weiner , 2010; Thu et al . , 2014 ) Our initial experiments showed extracellular and transmembrane domain ( ECTM ) constructs were easier to express than full-length constructs . Importantly , the ECTM domain was sufficient to drive homophilic adhesion ( Figure 2—figure supplement 1A ) . As our goal was to isolate the effects of adhesion on cell-cell interactions , we chose to use the ECTM domain for all subsequent experiments . As expected , the exogenously expressed protocadherins localized to sites of intracellular contact ( Figure 2—figure supplement 1B ) . We also confirmed that δ-Pcdh adhesion is highly sensitive to EDTA , consistent with being members of the calcium dependent cadherin superfamily ( Figure 2—figure supplement 1C ) . Although all expressed δ-Pcdhs induced cell aggregation ( Figure 2A ) , Pcdh10 formed very small aggregates relative to the others . We titrated the amount of DNA to try and normalize aggregate size ( Figure 2B ) . However , varying the amount of Pcdh10 DNA had little impact on aggregate size . We therefore excluded Pcdh10 from further experiments . We performed pair-wise assays by mixing cells expressing one δ-Pcdh ( fused to P2A-GFP ) with those expressing another ( fused to P2A-RFP ) . We found that cells expressing the same δ-Pcdh intermix ( Figure 2C , center diagonal ) while cells expressing different δ-Pcdhs segregate from one another . We interpret these results to indicate that δ-Pcdh adhesion is strictly homophilic . Identical results were found for the cPcdh subfamily using the same assay ( Thu et al . , 2014 ) . To determine how combinatorial expression of δ-Pcdhs affect adhesion specificity , we next performed mismatch coaggregation assays . In these experiments , cells expressing a single δ-Pcdh are mixed with a second population of cells expressing the same δ-Pcdh plus an additional , ‘mismatched’ δ-Pcdh . Prior studies on cPcdhs using this approach showed that a single mismatch causes one population to segregate from the other , even when several cPcdhs are expressed in common ( Thu et al . , 2014 ) . In contrast , this same assay suggested adhesive outcomes may be dependent on which δ-Pcdhs were co-expressed ( Pederick et al . , 2018 ) . To systematically define how mismatched δ-Pcdhs influence adhesive outcomes , we screened 42 possible mismatch pairs . We discovered a range of outcomes that could be grouped into three broad categories ( Figure 3A–D ) . In the first , the two populations intermixed ( Figure 3A , B ) . In the second , the populations interfaced ( Figure 3C ) , and in the last , the populations segregated from one another ( Figure 3D ) . We also noticed that interfacing and intermixing behaviors were not binary , but instead appeared to exist on a continuum . To better capture these differences , we developed a novel metric called the CoAggregation Index ( CoAg ) to quantify the degree of coaggregation ( see Materials and methods ) . Briefly , the index measures the proportion of red and green cells that share a common boundary within a given confocal image . In general , CoAg values below 0 . 1 indicate segregation , whereas values between 0 . 1–0 . 2 are typical of populations that interface . Above 0 . 2 , aggregates display increasingly higher degrees of intermixing . Thus , the CoAg index captures subtle differences in aggregation behavior not easily identified by eye . Ordering the CoAg values from our screen from high to low revealed a surprisingly linear range of behavior ( Figure 3E; mean CoAg values for a given experiment are indicated in the corner of each representative image ) . For comparison , the first column shows the CoAg value for Pcdh1 cells mixed with Pcdh7 cells ( e . g . complete segregation ) , as expected from cPcdh mismatch assays ( Thu et al . , 2014 ) . The red bar indicates complete mixing by matched populations . Reordering the CoAg values into a heat map strongly argued that different δ-Pcdh combinations produced different coaggregation behaviors ( Figure 3F ) . For example , we combined Pcdh1 cells with cells expressing Pcdh1+Pcdh7 or Pcdh1+Pcdh8 . In the first case , cells interfaced ( CoAg = 0 . 11; row 1 , column 2 ) , but in the second , they intermixed ( CoAg = 0 . 27; row 1 , column 3 ) . Although Pcdh1 was expressed by all populations , the presence of Pcdh7 vs . Pcdh8 led to differing behaviors . This suggested that , unlike the cPcdhs , the identity of the δ-Pcdh being tested is important for the outcome . This is further reinforced by the fact that strong asymmetry is observed across the diagonal in the heat map . For example , Pcdh19 cells segregate from Pcdh19+Pcdh7 cells ( CoAg = 0 . 02; Figure 3G ) . However , ‘across the diagonal , ’ Pcdh7 cells intermix with these same Pcdh19+Pcdh7 cells ( CoAg = 0 . 40 ) . Similarly , Pcdh19 cells intermix with Pcdh19+Pcdh9 cells ( CoAg = 0 . 23 ) but across the diagonal , Pcdh9 cells segregate ( CoAg = 0 . 07 ) . These results strongly suggest that coaggregation is dependent upon the identity of the mismatched δ-Pcdh . We obtained similar results using full-length constructs that could be expressed to generate an aggregation behavior ( data not shown ) . To compare how different δ-Pcdhs influence mismatch coaggregation , we generated a net mismatch score that revealed a potential hierarchy among δ-Pcdhs ( Figure 3H , see Materials and methods ) . We next considered if these variable behaviors were caused by differential surface expression of co-expressed δ-Pcdhs . Some prior studies control for overall expression ( e . g . from whole cell lysates ) , but not surface expression . To address this , we generated ECTM constructs fused to FLAG , GFP , or RFP , and used a cell-impermeant biotinylation reagent to label surface protein in live cells . Labeled proteins were then affinity purified and analyzed by western blotting for the various tags ( Figure 4—figure supplement 1A ) . Antibody signal intensities were calibrated to allow for cross-antibody comparisons . We re-tested all possible combinations of Pcdh1 , Pcdh7 , and Pcdh17 , as these three had the strongest net mismatch scores in our initial screen ( Figure 3H ) . For Pcdh1+Pcdh7 mismatch assays , we controlled for surface expression by carefully titrating DNA input ( Figure 4A ) , and examined aggregation behavior at 18 , 22 , 26 , and 44 hr post electroporation . As seen in our initial screen , Pcdh7 cells intermixed with Pcdh1+Pcdh7 cells across all time points , whereas Pcdh1 cells interfaced ( Figure 4B , C ) . We used 26 hr for all further tests , given no obvious differences in behavior beyond this point . We repeated the assay for Pcdh1+Pcdh17 , and found that Pcdh1 cells segregated ( CoAg = 0 . 07 , Figure 4D–F ) , while Pcdh17 cells intermixed ( CoAg = 0 . 42 ) . Interestingly , these results differ from our preliminary screen , where both Pcdh1 and Pcdh17 cells interfaced with Pcdh1+Pcdh17 cells . These results argue that controlling for surface level is important for interpreting coaggregation behavior , an aspect we explore below . Finally , we repeated our mismatch assay with Pcdh7 and Pcdh17 . We again found differences in behavior ( Figure 4—figure supplement 1B–D ) . However , we found that this pair was particularly sensitive to DNA input , as small changes could alter the result despite minor effects on surface expression ( Figure 4—figure supplement 1D ) . For one DNA input condition , Pcdh17 cells interfaced ( CoAg = 0 . 29 ) , while in the other they segregated ( CoAg = 0 . 08 ) . In contrast , Pcdh7 cells shifted towards intermixing . Nevertheless , these results confirm that differences in aggregation are dependent on δ-Pcdh identity . Finally , we titrated surface expression for cells co-expressing Pcdh1+Pcdh7+Pcdh17 ( Figure 4—figure supplement 1E ) . We tested all 3 vs 1 ( Figure 4—figure supplement 1F ) , 3 vs 2 ( Figure 4—figure supplement 1G ) and 2 vs 2 ( Figure 4—figure supplement 1H ) mismatch combinations . Differential adhesive behaviors were maintained as combinatorial depth increased , with the coaggregation outcome depending on which δ-Pcdhs were present ( Figure 4—figure supplement 1I ) . Our results argue that controlling for surface expression is important for understanding and interpreting differences in δ-Pcdh coaggregation behavior . In addition , our expression data ( Figure 1A , B and Figure 1—figure supplement 1A–G ) suggest that δ-Pcdh expression levels vary both within and between neurons . To further explore the role of expression , we established conditions where gradients of low , medium and high surface levels for Pcdh1 , Pcdh7 , and Pcdh17 could be reproducibly generated ( Figure 5A and Figure 5—figure supplement 1A ) . Medium levels were similar to those used in Figure 4 . Our mismatch assays involve mixing cells that express a single δ-Pcdh with those expressing two or more . We first asked what would happen if we altered surface expression in cells expressing a single δ-Pcdh . We found that Pcdh1 ( low , medium , and high ) cells all still interfaced with Pcdh1+Pcdh7 cells ( Figure 5B , C ) , while Pcdh7 ( low , medium , and high ) cells all still intermixed ( Figure 5—figure supplement 1B , C ) . We found identical results with a different pair of δ-Pcdhs ( Figure 5—figure supplement 1D–G ) . While the CoAg index varied slightly , the category of coaggregation behavior ( intermix , interface , or segregation ) did not . Thus , differences in mismatch coaggregation among δ-Pcdhs cannot be primarily explained based on variable expression in cells expressing one δ-Pcdh . We next asked if altering the relative proportion of δ-Pcdh expression within cells expressing two δ-Pcdhs would affect coaggregation . We created populations with high and low DNA input values for each δ-Pcdh ( e . g . Pcdh1High+Pcdh7Low and Pcdh1Low+Pcdh7High cells ) . We note that our goal was to simply alter the relative proportion of surface expression in these cells , and not to establish conditions where one δ-Pcdh was necessarily higher in expression than another . We found that varying the ratio of expression clearly altered coaggregation outcomes ( Figure 5D , E ) . Differences in coaggregation behavior are most easily seen by comparing results column by column . For example , in Figure 5D ( column 1 ) , Pcdh1 cells intermix with Pcdh1High+Pcdh7Low cells , but segregate from Pcdh1Low+Pcdh7High cells . The coaggregation behavior of Pcdh1 cells is therefore clearly affected by the ratio of Pcdh1:Pcdh7 in the co-expressing cells . In the complementary experiment ( column 2 ) , Pcdh7 cells intermixed with both Pcdh1High+Pcdh7Low and Pcdh1Low+Pcdh7Highcells . However , intermixing was clearly reduced in Pcdh1High+Pcdh7Lowcells . In column 3 , Pcdh1High+Pcdh7Low cells intermixed with Pcdh1High+Pcdh7Low cells , but less well with Pcdh1Low+Pcdh7High cells . The converse ( column 4 ) was observed for Pcdh1Low+Pcdh7High cells . Thus , relative surface levels of co-expressed δ-Pcdhs can influence aggregation behavior , even when there are no mismatches between populations . We tested eight additional pairs using this high/low DNA input approach , and found similar results ( Figure 5—figure supplement 1H ) . We confirmed a relative difference between high and low surface expression for a subset of pairs ( Figure 5—figure supplement 1I ) . We conclude that changing the relative ratio of expression in cells expressing two δ-Pcdhs has a much greater effect on coaggregation than varying expression in cells expressing one δ-Pcdh . Because differences in δ-Pcdh coaggregation behavior persisted despite controlling for surface expression , we next asked whether they possess differences in apparent adhesive affinity . Such differences have been argued to mediate segregation among classical cadherins , such as N- and E-cadherin ( Harrison et al . , 2010; Katsamba et al . , 2009 ) . We hypothesized that we could detect these potential differences by subjecting aggregates to higher shear forces . Cells expressing δ-Pcdhs with weaker apparent adhesive affinities should dissociate prior to those expressing δ-Pcdhs with stronger affinities . We generated cells expressing Pcdh1 , Pcdh7 or Pcdh17 at high surface levels ( Figure 5A , Figure 5—figure supplement 1A ) , and subjected them to gradual increases in rotational speed ( 15–220 RPM ) . Images were analyzed for aggregate size using a custom written code ( Aggregate Size Measurement ) . These populations began dissociating as speed increased . However , Pcdh7 cells maintained larger aggregates than Pcdh1 or Pcdh17 cells at all speeds ( Figure 6A , B ) . Furthermore , while Pcdh1 and Pcdh17 cells appeared to fully dissociate by ~200 RPM , Pcdh7 aggregates were still present even at 220 RPM . Because Pcdh1 , Pcdh7 , and Pcdh17 were at one end of our hierarchy ( Figure 3H ) , we compared Pcdh1 and Pcdh19 using the same approach . Similarly , we found that Pcdh1 cells maintained larger aggregates than Pcdh19 cells at all speeds ( Figure 6—figure supplement 1A–C ) . Varying expression levels also accentuated these differences . We generated cells expressing Pcdh7 or Pcdh17 at low , medium and high levels ( Figure 5A and Figure 5—figure supplement 1A ) . As expected , we found that higher surface levels generated larger aggregates that could better withstand increasing rotational speeds ( Figure 6—figure supplement 1D–G ) . We also found that Pcdh7 cells produced larger aggregates at all speeds compared to Pcdh17 cells . Even at 220 RPM , Pcdh7Low cells still maintained some aggregates . If Pcdh1 has weaker apparent adhesive affinity than Pcdh7 , this difference could explain why Pcdh1 cells interface with Pcdh1+Pcdh7 cells while Pcdh7 cells intermix in mismatch assays . Such differences should be accentuated by increasing shear force on aggregates . To test this , we repeated the Pcdh1+Pcdh7 mismatch assay . After allowing aggregates to form at 15 RPM , we increased the speed to 120 RPM . Despite the increased speed , Pcdh7 cells still intermixed with Pcdh1+Pcdh7 cells . However , Pcdh1 cells now segregated ( Figure 6C , D ) , consistent with weaker apparent adhesive affinity . To examine structural differences that could account for this varying behavior among δ-Pcdhs , we performed multiple sequence comparison by log expectation ( MUSCLE ) alignments . We found low sequence identity among δ-Pcdhs in extracellular domains ( EC ) 1–4 ( ~35%; Figure 6—figure supplement 1H ) . Prior work had shown that the adhesive interface of Pcdh19 was localized to EC1-4 ( Cooper et al . , 2016 ) . To test the importance of EC1-4 in adhesion mediated by other subfamily members , we deleted these domains ( Δ1–4 ) from Pcdh1 , Pcdh7 and Pcdh17 . Although the truncated proteins were still transported to the surface , they were unable to mediate adhesion ( Figure 6—figure supplement 1I , J ) . To determine how EC1-4 affect mismatch coaggregation , we mixed cells co-expressing Pcdh1+Pcdh7Δ1-4 with those expressing Pcdh1 or Pcdh7 alone . Pcdh7 cells could no longer intermix , and switched to a segregation behavior ( CoAg = 0 . 01; Figure 6—figure supplement 2A , B ) . Conversely , Pcdh1 cells switched from interfacing to intermixing ( CoAg = 0 . 25 ) . Next , we swapped the EC1-4 of Pcdh7 with that from Pcdh1 ( Pcdh7EC1-4:Pcdh1 ) . These cells now intermixed with Pcdh1 cells , but segregated from Pcdh7 cells ( Figure 6—figure supplement 2C , D , column 3 ) . Finally , Pcdh1 cells now intermixed with Pcdh7EC1-4:Pcdh1+Pcdh1 cells , while Pcdh7 cells segregated ( Figure 6—figure supplement 2C , D; column 4 ) . These results are consistent with EC1-4 mediating adhesive specificity . Our results argue that differences in apparent adhesive affinity and relative surface expression regulate coaggregation behavior . We therefore performed Monte Carlo simulations using a custom program ( cellAggregator , Ghazanfar , 2018 ) to see if we could model these factors in silico . We successfully captured the behavior of a subset of our experiments . The model functioned most optimally in predicting cells that will intermix . For example , the model correctly predicted that cells expressing identical δ-Pcdhs will intermix . Furthermore , the model also predicted the behavior of cells known to intermix in mismatch coaggregation assays . However , the model could not precisely recapitulate conditions where mismatched cells interfaced or segregated ( Figure 6E , far right column; for example mixing Pcdh1 cells with Pcdh1+Pcdh7 cells ) . Varying affinity differences , relative expression levels , or both still did not completely capture these behaviors . We anticipate other , as yet uncharacterized effects ( e . g . intracellular δ-Pcdh-δ-Pcdh interactions [Pederick et al . , 2018] ) must be incorporated into the model to better capture cell adhesive behavior . Our single cell RNA analysis showed individual OSNs express up to seven δ-Pcdhs . To test the impact of increasing the number of co-expressed δ-Pcdhs on mismatch aggregation , we generated populations of cells that co-expressed Pcdh7 with one to four additional δ-Pcdhs . To confirm changes in the relative expression of Pcdh7 vs the other co-expressed δ-Pcdhs , we measured surface expression levels ( Figure 7A ) and performed coaggregation assays with cells expressing only Pcdh7 . We found that each additional δ-Pcdh co-expressed with Pcdh7 led to a corresponding decrease in the CoAg index ( Figure 7B ) . Pcdh7 only cells shifted from intermixing towards interfacing as the relative proportion of Pcdh7 decreased . Quantification of surface expression showed that the percent of Pcdh7 with respect to total surface expression decreased from ~50% to 25% , almost perfectly mirroring the decline in CoAg index ( R2 = 0 . 94; Figure 7C ) . We repeated the experiment with Pcdh1 , and found a similar effect ( Figure 7—figure supplement 1A , B ) . In this case , increasing the number of co-expressed δ-Pcdhs shifted the behavior of Pcdh1 cells from interfacing to segregation . Finally , although we have focused on how δ-Pcdh subfamily members function in combination , individual neurons are likely to co-express multiple cadherin subfamily members . How δ-Pcdhs and these other subfamily members interact is not well understood . We first confirmed that cPcdh Pcdhb11 cells completely segregate from cells expressing δ-Pcdhs , demonstrating strict homophilic adhesion ( Figure 7—figure supplement 1C ) . We then generated populations co-expressing Pcdh7 and Pcdhb11 at three different relative expression levels for use in mismatch coaggregation assays ( Figure 7D ) . At the first two relative levels ( DNA input ratio of 3:4 and 1:2 ) , surface levels of Pcdh7 were ~45% of total ( Figure 7E ) . Under these conditions , Pcdh7 cells strongly intermixed while Pcdhb11 cells segregated ( Figure 7F , G ) . However , at a DNA ratio of 1:4 ( Pcdh7 ~20% of total ) , Pcdh7 cells still intermixed but Pcdhb11 cells could now interface . Thus , δ-Pcdhs influence the aggregation behavior of cells expressing this particular cPcdh . This raises the intriguing possibility that the two subfamilies may work in concert to specify adhesion .
The range of apparent adhesive affinities suggest that neurons can fine tune their overall adhesive profile by varying the repertoire of δ-Pcdhs expressed . One caveat is that we did not directly measure affinity using purified proteins . As our efforts are aimed at understanding how δ-Pcdhs mediate cell-cell interactions , we utilize the term apparent adhesive affinity to describe the functional impact of δ-Pcdhs on adhesion . Biophysical studies will be required to fully define such affinity differences . However , structural studies show cPcdhs possess varying adhesive affinities ( Goodman et al . , 2016; Rubinstein et al . , 2015 ) . Despite this , such differences do not appear to have a major impact in K562 assays ( Thu et al . , 2014 ) . While cell aggregation assays have been used for decades , the technical details have never been standardized . For example , cell type , speed of rotation , time of mixing , surface expression , and mode of quantitation all differ among past studies . We note that very few studies control for or report these factors , which in our hands are important for reproducible adhesive behavior . While such controls may not be necessary when cells essentially completely segregate from one another ( e . g . as for cPcdhs ) , such reproducibility was essential to our ability to identify and quantitate differences in adhesive outcomes among δ-Pcdh family members . Our aggregation assay results clearly contrast with a prior study of cPcdhs ( Thu et al . , 2014 ) . In this paper , two populations would only fully intermix if they expressed the same combinations of cPcdhs . If even one cPcdh differed between the two , the populations would completely segregate , regardless of the identity of the mismatched cPcdh . The observed results were always binary in nature , and produced either complete intermixing or complete segregation . In contrast , we were able to observe a range of coaggregation behaviors . This spectrum of adhesive outcomes illustrates how a comparatively small gene family can still have complex effects on cellular behavior . Biophysical analysis of complex formation may better illuminate the mechanism behind such differences . We note we did not identify any obvious differences between members of the δ−1 and δ−2 subfamilies in our assays . Members of both groups were expressed in overlapping patterns within the epithelium ( Figure 1—figure supplement 1 ) . In situ hybridization , NanoString , and qRT-PCR analyses also showed no obvious differences between subfamilies ( Figure 1 ) . In our mismatch aggregation assays , δ−1 and δ−2 members were distributed along the spectrum of our net mismatch score ( Figure 3 ) . For example , Pcdh1 , a δ−1 family member , had a roughly equivalent net mismatch score with Pcdh17 , a δ−2 family member . However , we note that δ−1 and δ−2 members are often co-expressed within neurons , leading to potential intracellular interactions that may not be captured in these assays . Further , how the varying number of extracellular domains between the two subfamilies influence adhesion is not known . Further structural studies will be needed to better define how these differences affect cell-cell interactions . We showed a simple solution to moderating high apparent adhesive affinity δ-Pcdhs is to vary relative expression level . These results are reminiscent of principles defined for classical cadherins . Steinberg’s differential adhesion hypothesis provides a commonly used framework for understanding how classical cadherins mediate cell sorting . In this model , cells sort from one another to reach an optimal thermodynamic equilibrium . This sorting can be driven by differences in adhesive affinity between cells , and/or by differences in expression level ( Foty and Steinberg , 2005; Friedlander et al . , 1989; Steinberg and Takeichi , 1994 ) . Thus , δ-Pcdhs appear to use some of the same principles as classical cadherins . However , Steinberg and colleagues typically focused on N- and/or E-cadherin , and did not , to our knowledge , examine the behavior of multiple classical cadherins in combination . The principles we define here therefore confirm similarities between the classical and δ-Pcdhs , and extend these canonical studies of cadherin function . We chose to use the ECTM domain for these experiments because expressing the full-length construct in K562 cells proved practically difficult . However , we demonstrated that the ECTM domain mediated homophilic adhesion to a degree similar to that of the full-length construct ( Figure 2—figure supplement 1 ) . As our goal was to study adhesive interactions among co-expressed family members , this allowed us to separate adhesion from intracellular signaling . In addition , the ECTM domain is typically used to study δ-Pcdh adhesion ( Chen et al . , 2007; Cooper et al . , 2016; Emond et al . , 2011 ) . Still , it is clear there are many aspects of δ-Pcdh function that are not addressed by this reductionist approach . Intracellular signaling events , heterologous extracellular interactions , and regulation of δ-Pcdh gene expression can all further tune the impact of δ-Pcdhs on cell-cell interactions . Indeed , our Monte Carlo simulation indicates we can capture many , but not all , behaviors associated with combinatorial expression . Most notably , not all interface or segregation behaviors could be adequately modeled ( Figure 6E ) . We expect that other , uncharacterized intracellular or extracellular interactions may explain these differences . In particular , Pederick et al . showed δ-Pcdhs can interact in cis ( Pederick et al . , 2018 ) . Such cis interactions have previously been proposed to be critical for cPcdh function ( Rubinstein et al . , 2017; Thu et al . , 2014 ) . If these cis interactions are also important for δ-Pcdh function , we anticipate that they may contribute towards adhesion of δ-Pcdhs in trans . Nevertheless , our studies lay the foundation for new models that can integrate these principles with those defined for other cadherin subfamilies , ultimately leading to a more complete determination of cadherin function within the nervous system . Our results represent a functional genomic approach towards understanding how combinations of cadherin expression identified via transcriptomic approaches impact cellular function . Our reductionist approach to understanding δ-Pcdh function has the fundamental advantage of allowing us to systematically test different combinations for their impact on adhesion . Such studies would be extremely difficult to execute in vivo , given the varied chromosomal locations of δ-Pcdhs and the technical complexity of manipulating multiple genes at once . Further , although K562 cells have been used extensively to study protocadherin function , they are not a neuronally derived line . An appropriate question would be to ask how our results apply towards understanding δ-Pcdh function in vivo . We believe there are two major applications of this study for understanding δ-Pcdh function . First , while δ-Pcdhs have been suspected to be expressed in combination in vivo based on double-label RNA in situ data , there has been no prior evidence demonstrating the extent of this expression . Our single cell NanoString and qRT-PCR data ( Figure 1D–F ) clearly demonstrate that multiple δ-Pcdhs are expressed per neuron , and show the variety and extent of such expression . Our round-robin RNA in situ hybridization studies ( Figure 1—figure supplement 1H ) are also consistent with this combinatorial expression . Further , our study of δ-Pcdh and odorant receptor overlap showed OSNs known to project to different targets clearly express different proportions of δ-Pcdhs ( Figure 1B ) . While the expression of δ-Pcdh vs . a given odorant receptor is not a simple , one-to-one correlation , there nevertheless were clear differences among OSNs expressing different odorant receptors . Thus , the combinatorial expression of δ-Pcdhs is not an entirely random event , as has been suggested for the cPcdhs ( Goodman et al . , 2016; Hirano et al . , 2012 ) . This is further supported by our single label RNA in situ studies , which clearly shows spatially restricted expression of δ-Pcdhs within the olfactory epithelium ( Figure 1—figure supplement 1B–G ) . Our results therefore demonstrate that δ-Pcdhs are combinatorially expressed in vivo , that 0–7 family members can be co-expressed within OSNs , and that this expression pattern is not stochastic . Second , our studies addressed the question of how these combinations could influence δ-Pcdh function . Our results argue that the particular combination expressed within a cell has a major impact on its adhesive profile . We therefore predict mutations in any one δ-Pcdh will not have uniform effects on all cells that express that particular δ-Pcdh , simply because different cells are likely to express different combinations . For example , we previously showed that mis- and over-expression of Pcdh10 in the olfactory system caused defects in glomerular target formation by OSNs expressing the Olfr9 odorant receptor , but not by those expressing Olfr17 ( Williams et al . , 2011 ) . A recently generated Pcdh19 mutant mouse in our lab also shows targeting defects of a subset of OSN populations ( data not shown ) . If Pcdh10 and Pcdh19 are expressed by multiple OSN populations ( Figure 1B ) , why are only a subset of OSNs affected in these mutants ? We speculate that this variation is due in part to the interactions between the mutated δ-Pcdh and the other , co-expressed δ-Pcdhs within a neuron . Furthermore , the two populations may express different levels of Pcdh19 , leading to different effects when Pcdh19 is mutated . A true understanding of how mutations in δ-Pcdhs mediate their effects would therefore be dependent on defining at a minimum what other δ-Pcdhs are co-expressed within affected cells . Loss of any one δ-Pcdh would alter the combination of δ-Pcdhs expressed and change the relative expression of co-expressed protocadherins . The changes that would occur as a result of these intrafamily interactions would therefore vary based on what δ-Pcdhs were co-expressed within the cell . This same K562 assay was used to examine a mouse mutant of Pcdh19 to understand why apparent cell sorting defects occurred in the cortex ( Pederick et al . , 2018 ) . Critically , this study postulated that co-expressed δ-Pcdhs might influence the observed sorting behavior . They found that K562 cell adhesion was indeed affected by different δ-Pcdh combinations . Although they did not correct for surface expression or draw any particular conclusions about principles that mediate their observed phenotypes , their results are consistent with ours in demonstrating the integral role of combinations in cell sorting . Our results therefore emphasize the importance of understanding what combinations exist within neurons in order to understand observed phenotypes . However , defining the particular combination of δ-Pcdhs expressed per neuron has been problematic . Single cell RNA-seq studies have been unable to adequately address what combinations are expressed within individual neurons . Our own analysis of three single OSN RNA-seq datasets ( Hanchate et al . , 2015; Saraiva et al . , 2015; Tan et al . , 2015 ) shows an average detection of ~1 δ-Pcdh per neuron , while our NanoString approach detects ~3 . 5 ( Figure 1—figure supplement 1K , L ) . Furthermore , our NanoString results were consistent with orthogonal validation assays using qRT-PCR and in situ hybridization . Thus , higher sensitivity approaches , similar to those used here , may be necessary to fully address what combinations are present within neurons . We would also like to highlight the importance of potential , interfamily interactions . We demonstrated co-expression of Pcdh7 with Pcdhb11 inhibits Pcdhb11 from intermixing with Pcdh7+Pcdhb11 cells ( Figure 7F , G ) . If , however , expression of Pcdh7 is reduced relative to Pcdhb11 , then these cells begin to display interfacing behavior . Thus , δ-Pcdhs can modify the behavior of other , co-expressed subfamily members . It seems reasonable that δ-Pcdhs , classical cadherins , cPcdhs , and other subfamily members are all likely to be co-expressed within individual neurons . How would interfamily interactions influence neuronal behavior in vivo ? Studies on cPcdhs have emphasized the sheer number of possible stochastic combinations that can be generated with this family . Our studies demonstrate that even greater adhesive complexity can be generated by superimposing the effects of δ-Pcdhs on cells expressing cPcdhs . Although we and others have begun establishing rules governing intrafamily interactions , it is likely that further complexity can be added via interactions between subfamilies . For example , δ-Pcdhs can bind and regulate classical cadherins ( Chen and Gumbiner , 2006; Chen et al . , 2009; Emond et al . , 2011 ) . Such interfamily interactions may well help to explain certain mutant phenotypes associated with the cPcdhs . In the retina , deletion of cPcdhs leads to neuronal death and to defects in dendritic self-avoidance . Interestingly , interactions between cPcdh subfamilies accentuates these effects ( Ing-Esteves et al . , 2018 ) , again underscoring the impact of combinatorial subfamily interactions . However , in the cortex , deletion of cPcdhs disrupts dendritic branching due to a failure to promote arborization ( Molumby et al . , 2016 ) . Thus , the same family has distinct effects in different regions of the nervous system . These differences were proposed to be due to context dependent effects . However , it is conceivable that interfamily interactions , such as those between the δ-Pcdhs and the cPcdhs , may also play a role in explaining these varying phenotypes . The fundamental principles defined here therefore enable new hypotheses to be generated regarding how mutations in protocadherins influence neuronal function .
All animal protocols were approved by the Cornell Institutional Animal Care and Use Committee . Non-Swiss Albino ( NSA ) mice of mixed sex were used for all single cell studies . For RNA in situ hybridization experiments , both NSA and C57Bl/6 mice were used . Mice were sacrificed at post-natal day 7 ( P7 ) for single cell and single label RNA in situ hybridization experiments , and embryonic day 17 . 5 ( E17 . 5 ) for double label experiments . Single and double label RNA in situ hybridization was performed essentially as described ( Williams et al . , 2011 ) . For single color studies at E17 . 5 and P7 , at least three independent heads were analyzed . For δ-Pcdh co-expression studies , three replicates were performed from three different heads for each gene . Imaging of double-label RNA in situ data was performed using a Zeiss ( Wetzlar , Germany ) LSM 510 confocal microscope , and multiple locations within each E17 . 5 olfactory epithelia were examined . Five optical slices ( each 3 μm thick ) from each location were used to assess co-expression . Positive co-expression was manually determined based on overlapping fluorescence signal observed in consecutive optical sections . Between 71 and 167 cells were analyzed per double label comparison . To quantify single label RNA in situ data , slides were scanned with a ScanScope ( Leica ) using a 20x objective . The OSN layer of each section was manually traced using HALO software ( Indica Labs , Corrales , New Mexico ) , and the percent positive area was determined using a built-in software module . For δ-Pcdh and odorant receptor co-expression studies , an average of 70 OSNs expressing a given odorant receptor were analyzed for co-expression with any one δ-Pcdh . Olfactory epithelia were dissected from P7 NSA mice and enzymatically dissociated for 1 hour using the Papain Dissociation Kit ( Worthington , Lakewood , NJ ) . The tissue was manually triturated , and the papain neutralized as per manufacturer’s instructions . Approximately 250 , 000 cells were then plated on coverslips coated with poly-ornithine , and the cells were allowed to recover at 37°C with 6% CO2 for 30 min in Modified Eagle’s Medium ( MEM ) . After recovery , the cells were gently washed three times with CO2 equilibrated MEM . The coverslip was then transferred to a 10 cm dish , where it was immobilized by applying small dabs of autoclaved Vaseline between the bottom of the coverslip and the 10 cm dish . The dish was flooded with 10 mL of equilibrated MEM , and individual OSNs isolated by manual aspiration under a 20X objective using a micromanipulator ( Eppendorf; Hauppauge , New York ) . Micropipettes for aspiration were prepared using a Sutter P-97 Flaming/Brown ( Novato , CA ) micropipette puller , and pre-filled with ~3 μL of MEM . After aspiration , the contents were transferred to a PCR tube by gently snapping the distal tip of the micropipette inside the tube and expelling the contents using a needle and syringe . Two different lysis buffers were utilized ( Cells-to-Ct or CellsDirect , Thermo-Fisher , Waltham , MA ) , with no apparent difference in lysis quality or NanoString results . Each tube was pre-loaded with 6 μL of CellsDirect lysis buffer ( containing lysis enhancer ) or Cells-to-Ct buffer ( containing DNAse I ) . As OSN isolation was performed at room temperature , neurons were collected from a given coverslip within 30 min . Cells processed in CellsDirect buffer were stored at −80°C until processing . Cells processed in Cells-to-Ct buffer were vortexed and then incubated at room temperature for five minutes . An additional 0 . 5 μL of stop solution was added and incubated for 2 min at room temperature before being stored at −80°C until further processing . Amplification reactions were done using the CellsDirect kit ( Thermo-Fisher ) essentially according to manufacturer’s instructions , with the following modifications . The 31 gene multiplex primer set was added to individual lysates ( 100 nM final ) in a final volume of 10 μL . Tubes were heated at 80°C for 10 min and chilled on ice for 3 min . 10 μL of 2x reaction buffer and 1 μL of SuperScript III/Platinum Taq ( Thermo-Fisher ) were added and tubes were reacted in a PCR machine at 50°C for one hour , followed by 85°C for 15 min to inactivate the reverse transcriptase . PCR amplification was then performed with an initial activation at 94°C for 2 min , followed by 18 cycles of 94°C for 30 s , 60°C for 30 s , and 72°C for 30 s . After amplification , 20 μL of 10 mM Tris 7 . 5 was added to each sample to bring up the total volume to 40 μL . Four μL of each sample was then screened by quantitative PCR to determine expression levels of Gapdh ( indicating successful capture and amplification ) and Ncam1 ( indicating an OSN ) . Taqman primers were designed to amplify regions internal to the 31 gene multiplex primer sequence , and samples were run on an ABI 7500 ( Thermo-Fisher ) . Only cells with Ct values ≤ 25 for both genes were used for the NanoString analysis ( Seattle , WA ) . See Figure 1—source data 1 . A custom codeset of 31 genes was designed that would detect a select subset of known axon guidance genes ( see Figure 1—source data 1 ) . Single cell cDNA was hybridized to the codeset in collaboration with NanoString . Genes were determined to be positively expressed using a constrained gamma-normal mixture model approach ( Ghazanfar et al . , 2016 ) . Briefly , ‘negative’ control genes ( e . g . Notch2 , Gfap and Cdh13 ) were used to estimate the distribution of the no or lowly expressed genes across all cells . Following this , for each cell a constrained gamma-normal mixture model was fit using the Expectation Maximization ( EM ) algorithm , constrained in the sense that the mean and variance of the no or lowly expressed component for that particular cell was the same as across all cells , allowing the highly expressed component to vary as required . This constrained gamma-normal mixture model allowed for ‘sharing’ of information across multiple cells , reducing the possibility of ill-fitting distributions to the cells’ expression patterns . Following model fitting , cells and genes were classed as ‘expressed’ if the corresponding posterior probability was 0 . 5 or above , and ‘not expressed’ otherwise . After this analysis , some cells were found to be Notch2 positive , and discarded from further study . Data from four codeset genes generated no useful information and were not utilized . OSNs were isolated and amplified in a manner identical to those used for NanoString analysis . Two uL of amplified cDNA from each single cell were used as template for each Taqman assay ( Gapdh , Ncam1 , Notch2 , and the δ-Pcdhs; Figure 1—source data 1 ) . All primer sets displayed efficiencies between 93–100% , except for Pcdh1 which had 83% efficiency ( improvement was not observed with multiple primer designs ) . Probes were designed to bind to regions distinct from those detected with the NanoString codeset . Genes were considered ‘on’ if we observed a Ct value less than or equal to 30 . EGFP-N1 ( Clontech ) vectors were modified to incorporate the TagRFP fluorophore and/or a P2A sequence . FLAG constructs were created in a pHAN vector modified to include a FLAG sequence at the 3’ terminus of the polylinker . ECTM domains of δ-Pcdhs and Pcdhb11 were then cloned into the appropriate vector . K562 cells were purchased from ATCC ( ATCC CCL-243 ) and tested mycoplasma negative . Low passage number cells ( 4–10 passages ) were maintained in RPMI +L glutamine with 10% calf bovine serum ( Gemini Bio , Sacramento , CA ) . Cells were grown to a density between 250–500 , 000 cells/mL prior to electroporation . For the electroporation , one million cells were removed , concentrated by centrifugation , and resuspended in 100 μL of Ingenio Electroporation Solution ( Mirus Bio , Madison , WI ) . Five to eight μg of cesium chloride or midi prepped ( Omega ) DNA for each δ-Pcdh to be expressed was added , and the cells electroporated using an Amaxa Nucleofector II ( Lonza; program T-016 , Cologne , Germany ) . Cells were allowed to recover for one hour at 37°C by immediate addition of 2 mLs of CO2 equilibrated media . After recovering , valproic acid ( VPA , 4 mM final; Sigma , St . Louis , MO ) was added to promote expression . Preliminary control experiments showed VPA did not affect cell adhesion , as cells electroporated with vector only remained non-adherent up to four days . For coaggregation experiments , equal volumes of cells from a given electroporation were mixed immediately following the recovery period and placed in individual wells of a 6-well ( 2 mLs/well ) or 24-well ( 0 . 5 mLs/well ) plate . Cells were gently and continuously agitated at 15 RPM overnight in a tissue culture incubator at 37°C with 6% CO2 . For initial aggregate size titration , 15–20 images were taken of each replicate using an inverted fluorescent microscope ( Nikon , Tokyo , Japan ) with a 10x objective . For speed and aggregate size experiments , ~6 field of views were captured at each speed for each replicate using a confocal microscope ( Zeiss LSM 510 ) with a 5x objective . For all other aggregation experiments , ~10–15 confocal images were captured of each replicate using a 10x objective . To generate the Coaggregation Index , confocal images were analyzed using custom code ( ‘CoAg’ ) written in Mathematica ( Wolfram Research , Champaign , IL ) . Briefly , each confocal image of an aggregate is parsed into squares just slightly larger than the area of a single cell . After removing all black squares from the image ( those containing no cells ) , the remaining squares are analyzed to calculate the percent of squares that contain more than one color . As a result , cells that completely segregate from one another will have a very low CoAg index because few squares will contain more than one color . In contrast , cells that interface will have higher CoAg indices as green and red cells abut one another , while those that intermix will have the highest index . K562 cells were electroporated and following a one hour recovery period , allowed to form aggregates at 15 RPM overnight . At 24–26 hr , images were captured of each replicate . To determine size of aggregates , images were analyzed using the particle size plugin in ImageJ . Aggregates smaller than three cells were removed from the analysis to prevent dividing cells and single cells not participating in aggregation from skewing the results . Aggregate pixel size was compared to the pixel area of one cell to approximate the number of cells per aggregate . K562 cells were electroporated and following the 1 hr recovery period , allowed to form aggregates at 15 RPM overnight . At 24–26 hr , images were captured to establish a 15 RPM baseline . Plates were then returned to the incubator and the speed increased for 1 hr to 120 RPM . Images were then acquired , and this process repeated at 160 , 200 and 220 RPM . Each image was then analyzed using a custom written code ( ‘Aggregate Size Measurement’ ) in Mathematica ( Wolfram Research , Champaign , IL ) to measure the pixel size of each aggregate , and aggregate pixel size was then converted to microns . For mismatch coaggregation assays , paired t-tests were performed between each paired population to determine statistical significance in Prism ( Graph Pad , La Jolla , CA ) . For aggregate speed and size analyses , analysis of variance ( ANOVA ) were performed in R . Surface biotinylation of live K562 cells was performed using the Pierce Cell Surface Isolation Kit ( Thermo-Fisher ) essentially as recommended . Volume of cell resuspension was reduced to 1 mL , and an additional 150 uL of lysis buffer was added to ensure complete mixing during incubation . Western blots were performed by loading 8 uL ( roughly 15% of the total elution from each biotinylation experiment ) onto 10% SDS polyacrylamide gels . All primary antibodies used were monoclonal in origin , and carefully titrated to establish working dilutions of equivalent detection so that samples across antibodies could be compared . To achieve this , we calibrated working monoclonal concentrations with purified RFP and GFP proteins . We also electroporated cells with the same δ-Pcdh fused to different tags to optimize antibody dilution to account for variation in signal intensity . The antibodies used were mouse anti-GFP ( 1:4 , 000 , Thermo-Fisher MA5-15256 ) , mouse anti-RFP ( 1:2 , 000 , Thermo-Fisher MA5-15257 ) and mouse anti-FLAG ( 1:6 , 000 , Thermo-Fisher MA1-91878 ) . We used the transferrin receptor ( TfR ) as a loading control for surface protein ( 1:1 , 000 , Thermo-Fisher 13–6800 ) . All antibodies were diluted in 20% glycerol upon receipt to promote cryostability . Estimation of band intensity was carried out using ImageJ . To investigate the aggregation behavior of cell populations expressing δ-Pcdhs of varying apparent adhesive affinities and expression , we performed Monte Carlo based simulations to describe cell binding interactions as a dynamic cell-cell network across discrete time steps using custom code ( cellAggregator ) . Two cell populations , green ( n = 25 ) and red ( n = 25 ) , were assigned properties of two hypothetical genes named A and B , corresponding to the coaggregation assay experiments conducted . For example , green cells could be designated as expressing high levels of A and low levels of B , and red cells as expressing low levels of B and high levels of A . The genes A and B were each also assigned binding affinities , for example , A possesses two times greater apparent adhesive affinity than B . The initial cell-cell network consists of the green and red cells as nodes in the network , and edges represent cell-cell binding interactions occurring . For each simulation , 100 time steps were performed . At each discrete time point , the cells are mixed and allowed to bind to other cells according to a ‘speed dating’ set up , where the majority of cell pairs ( arbitrarily set at 75% ) result in a cell-cell interaction . Allowing the majority ( as opposed to all cell pairs ) to bind avoids oscillatory network behavior . The probability that two cells would ‘speed date’ increased as the Euclidean distance between the force-directed network projection onto two dimensions decreased , that is nodes more closely connected were more likely to ‘speed date’ . Once ‘speed-dating’ begins , the cell pair would bind via the genes expressed by each cell , with unbound genes selected at random with a probability corresponding to the expression level . The duration of interaction ( number of time steps ) depended on the identity of genes . A-B interactions persisted for only a single time step , while B-B interactions persisted for three time steps , and A-A interactions persisted for three multiplied by the affinity ratio time steps . This differential length of time for cell-cell interactions is based on the idea that non-homophilic protocadherin interactions are unstable and do not persist ( A-B ) , and that some protocadherins may have different levels of apparent adhesive affinity , leading to more persistent or stable binding time ( e . g . A-A lasts more time steps than B-B if A is assigned greater affinity than B ) . The green or red color of the cells did not affect the binding of cell pairs . Instantaneous network coaggregation was measured by calculating the average proportion of different-color to same-color binding partners across all cells in the network for any one time step . Cells with no network partners were not included in this calculation . The in silico coaggregation behavior for the entire simulation was then determined as an average of all instantaneous network coaggregations in the simulation . This value did not include initial time steps ( arbitrarily set at 25% of the 100 total time steps ) to allow for the network to stabilize following the initial state of all cells being unconnected . This resulted in a single overall in silico coaggregation index value determined for the simulation scenario . A total of 100 time steps were simulated for each scenario , and each scenario was repeated five times . To model varying affinity between genes , the affinity values were allowed to range between 1 ( same affinity ) and 10 . The source code for performing the Monte Carlo simulation is available at https://github . com/shazanfar/cellAggregator ( copy archived at https://github . com/elifesciences-publications/cellAggregator ) and an interactive R Shiny application available at http://shiny . maths . usyd . edu . au/cellAggregator/ . Pcdh18 data was discarded due to an error in the codeset . However , Pcdh18 was not detected by RNA in situ hybridization experiments in the epithelium nor in subsequent single OSN qPCR experiments . Negative controls ( e . g . water or media only ) showed no signal following amplification , indicating a lack of contamination . To validate the NanoString data , we first performed a ‘pool-split’ experiment to determine technical reproducibility . RNA from 12 single cells were pooled and then split into multiple aliquots . Each aliquot was separately amplified and processed to assess technical reproducibility . Samples showed good correlation ( R2 = 0 . 62; data not shown ) . Second , we asked if averaging the expression patterns from single neurons approximated the pattern seen using bulk epithelial RNA . We found strong correlation ( R2 = 0 . 65 ) despite the fact we only analyzed 50 cells , and bulk RNA contains neurons , glia , and other cell-types ( data not shown ) . Finally , multiple discriminant analysis ( MDA ) showed that pool-split samples clustered with single cells while the water and bulk samples formed separate , discrete clusters ( data not shown ) . To address the concern that dissociation of whole epithelia would affect δ-Pcdh expression , we generated a proxy for in vivo expression by performing single color RNA in situ hybridization studies ( Figure 1—figure supplement 1A–G; no signal was detected for Pcdh11x or Pcdh18 ) . Interestingly , the pattern of expression was clearly variable among neurons , and unevenly distributed within the epithelium ( Figure 1—figure supplement 1B–G ) . We used this RNA in situ data to estimate the proportion of OSNs that express each δ-Pcdh ( Figure 1—figure supplement 1J; see Materials and methods ) . We found that our single neuron data and these in vivo estimates followed similar trends ( R2 = 0 . 58 ) , suggesting dissociation did not have an appreciable impact on our NanoString data . | Multicellular life depends on cells being able to stick together . The human body , for example , consists of trillions of cells grouped into tissues and organs . The brain alone contains some 87 billion neurons organized into complex networks . To stay together , cells use proteins on their surface called cell adhesion molecules ( CAMs ) . There are four major families of CAMs , each with multiple members , and the CAMs on one cell recognize and interact with the CAMs on another . But how does this process work ? One possibility is that different combinations of CAMs allow different cells to stick together . Bisogni et al . tested this idea by studying a family of CAMs called the delta-protocadherins . This family has nine members , each with its own gene . Before cells can use a gene to produce a protein , they must first use the gene’s DNA as a template to build an RNA molecule . By counting the number of different types of RNA molecules inside individual cells , Bisogni et al . showed that sensory neurons in the mouse each produce up to seven different delta-protocadherins . Further experiments revealed that cells fine-tune their interactions by varying the number , type and combination of delta-protocadherins on their surface . In addition , the delta-protocadherins also alter interactions between members of a related gene family , the clustered protocadherins . This further increases their ability to regulate how cells interact . In contrast to previous studies that focused on single molecules , Bisogni et al . have shown how combinations of molecules work together to influence cell adhesion . Deciphering this combinatorial code is key to understanding how interactions between cells go awry in disease . Mutations in the genes for CAMs often impair brain development . The reported findings may provide insights into how such mutations disrupt the CAM combinatorial code and alter cell to cell interactions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2018 | Tuning of delta-protocadherin adhesion through combinatorial diversity |
Initiation of transcription is a primary means for controlling gene expression . In bacteria , the RNA polymerase ( RNAP ) holoenzyme binds and unwinds promoter DNA , forming the transcription bubble of the open promoter complex ( RPo ) . We have determined crystal structures , refined to 4 . 14 Å-resolution , of RPo containing Thermus aquaticus RNAP holoenzyme and promoter DNA that includes the full transcription bubble . The structures , combined with biochemical analyses , reveal key features supporting the formation and maintenance of the double-strand/single-strand DNA junction at the upstream edge of the −10 element where bubble formation initiates . The results also reveal RNAP interactions with duplex DNA just upstream of the −10 element and potential protein/DNA interactions that direct the DNA template strand into the RNAP active site . Addition of an RNA primer to yield a 4 base-pair post-translocated RNA:DNA hybrid mimics an initially transcribing complex at the point where steric clash initiates abortive initiation and σA dissociation .
Transcription initiation is a major control point of gene expression . The initiation process is best understood in the bacterial system ( Saecker et al . , 2011 ) where the conserved ∼400 kD catalytic core of the RNA polymerase ( RNAP or E , subunit composition α2ββ′ω ) combines with the promoter-specificity factor σA to form the holoenzyme ( EσA ) , which locates promoter DNA and unwinds 12–14 base pairs ( bps ) of the DNA duplex to yield the transcription-competent open promoter complex ( RPo ) . In the presence of nucleotide substrates , RNA synthesis begins with the formation of an initial transcription complex ( RPITC ) . Before transitioning to a stable elongation complex , steric clash between the elongating RNA transcript and elements of σ set up abortive initiation , where the RNAP repeatedly generates and releases short transcripts without dissociating from the promoter ( McClure et al . , 1978; Murakami et al . , 2002a; Goldman et al . , 2009 ) . Eventually , the transcript reaches a length of around 17 nt , where σ dissociation and the transition to the stable elongation complex begins ( Nickels et al . , 2005 ) . The architecture of EσA recognition of the key −35 and −10 promoter elements was delineated by the structure of Thermus aquaticus ( Taq ) EσA bound to an upstream fork ( us-fork ) promoter fragment , but the low resolution ( 6 . 5 Å ) prevented the visualization of molecular details ( Murakami et al . , 2002b ) . Although high resolution crystal structures defined key , sequence-specific interactions of σ with the −35 element ( Campbell et al . , 2002 ) , the melted −10 element ( Feklistov and Darst , 2011 ) , as well as with downstream promoter DNA in the context of holoenzyme ( Zhang et al . , 2012 ) , these structures did not contain the full transcription bubble with the upstream double-strand/single-strand ( ds/ss ) DNA junction at the upstream edge of the −10 element where transcription bubble formation initiates . Structures of Escherichia coli ( Eco ) transcription initiation complexes containing a complete transcription bubble delineated the overall architecture of the full bubble , but the low resolution of the analyses ( between 5 . 5 and 6 Å resolution ) prevented a detailed description of protein/DNA interactions ( Zuo and Steitz , 2015 ) . Here , we determine crystal structures of Taq EσA bound to an us-fork promoter fragment , as well as a complete RPo ( Figure 1 , Figure 1—figure supplement 1 ) , refined using diffraction data extending to 4 . 00 and 4 . 14 Å-resolution , respectively ( Table 1 , Figure 1—figure supplement 2 ) , allowing visualization of key features that stabilize the upstream edge of the transcription bubble . The results also reveal functionally relevant holoenzyme interactions with duplex DNA just upstream of the −10 element and potential protein/DNA interactions that direct the DNA template strand ( t-strand ) into the RNAP active site . Addition of an RNA primer to yield a 4-bp post-translocated RNA:DNA hybrid mimics RPITC at the point where steric clash initiates abortive initiation and σA dissociation ( Murakami et al . , 2002a; Kulbachinskiy and Mustaev , 2006 ) . 10 . 7554/eLife . 08504 . 003Figure 1 . Structure of RPo . ( A ) Oligonucleotides used for 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; t-strand DNA ( bottom strand ) , light grey; RNA transcript , red . ( B ) Overall structure of RPo . The nucleic acids are shown as CPK spheres and color-coded as above . The Taq EΔ1 . 1σA is shown as a molecular surface ( αI , αII , ω , grey; β , light cyan; β′ , light pink; Δ1 . 1σA , light orange ) , transparent to reveal the RNAP active site Mg2+ ( yellow sphere ) and the nucleic acids held inside the RNAP active site channel . ( C ) Electron density and model for RPo nucleic acids . Blue mesh , 2Fo − Fc maps for nucleic acids ( contoured at 0 . 7σ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 00310 . 7554/eLife . 08504 . 004Figure 1—figure supplement 1 . ( Left ) Synthetic oligonucleotides used for us-fork ( −12 bp ) crystallization . The numbers above the sequence 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 . ( Right ) Electron density and model for nucleic acids . Blue mesh , 2Fo − Fc maps for nucleic acids ( contoured at 0 . 7σ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 00410 . 7554/eLife . 08504 . 005Figure 1—figure supplement 2 . Data and model quality for us-fork ( −12 bp ) and RPo complexes . Plots relating data quality with model quality at 4 . 0 Å-resolution ) 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 Taq EΔ1 . 1σA/us-fork ( −12 bp ) at 4 . 0 Å-resolution . ( Right ) Data for Taq EΔ1 . 1σA RPo ( with 4-nt RNA primer ) at 4 . 0 Å-resolution . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 00510 . 7554/eLife . 08504 . 006Figure 1—figure supplement 3 . Sequence alignment of regions 2–4 of selected bacterial RNAP primary ( Group I ) σ subunits . Sequence alignment of regions 2–4 ( Lonetto et al . , 1992 ) of bacterial RNAP primary ( Group I ) σ subunits . The sequences shown were selected from diverse phyla/groups taken from a much larger alignment of 1002 sequences . The σA sequences shown are from the following organisms chosen to represent the preceding phylum/group: Deinococcus-Thermus , Thermus aquaticus; γ-Proteobacteria , Escherichia coli; Actinobacteria , Mycobacterium tuberculosis; 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; Thermodesulfobacteria , Thermodesulfatator atlanticus . The sequences are shaded according to conservation within the sub-alignment; red shading indicates 100% identity , blue shading indicates >50% identity . The histogram at the top represents the sequence conservation within the entire 1002 sequence alignment ( red bar , 100% identity; orange , 83–99%; green , 68–83%; blue , 50–67% ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 00610 . 7554/eLife . 08504 . 007Table 1 . Table of crystallographic statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 007Taq EΔ1 . 1σA +Us-fork ( −12 bp ) Us-fork ( −11 bp ) Bubble/RNA ( RPo ) BubbleData collection Space groupP43212P43212P43212P43212 Combined datasets34104 Cell dimensions a ( Å ) 289 . 87288 . 23289 . 26290 . 76 b ( Å ) 289 . 87288 . 23289 . 26290 . 76 c ( Å ) 537 . 36535 . 25536 . 60540 . 84 Wavelength ( Å ) 1 . 0751 . 0751 . 0751 . 075 Resolution ( Å ) 50 . 03–4 . 01 ( 4 . 143–4 . 01 ) †49 . 43–4 . 60 ( 4 . 76–4 . 60 ) †34 . 96–4 . 14 ( 4 . 29–4 . 14 ) †40 . 00–4 . 74 ( 4 . 91–4 . 74 ) † Total reflections2 , 192 , 774 ( 167 , 274 ) 1 , 268 , 008 ( 123 , 590 ) 5 , 022 , 989 ( 367 , 167 ) 1 , 849 , 900 ( 143 , 237 ) Unique reflections185 , 025 ( 18 , 323 ) 125 , 012 ( 11 , 043 ) 172 , 210 ( 16 , 966 ) 116 , 874 ( 8115 ) Multiplicity11 . 5 ( 9 . 1 ) 10 . 1 ( 10 . 1 ) 29 . 2 ( 21 . 6 ) 15 . 8 ( 12 . 7 ) Completeness ( % ) 99 . 9 ( 98 . 6 ) 99 . 0 ( 100 . 00 ) 100 ( 99 . 8 ) 99 . 6 ( 97 . 0 ) <I>/σI6 . 68 ( 0 . 43 ) 5 . 57 ( 0 . 60 ) 9 . 4 ( 0 . 8 ) 8 . 11 ( 0 . 81 ) Wilson B-factor ( Å2 ) 133 . 90154 . 68101 . 16196 . 78 Rpim‡0 . 173 ( 2 . 136 ) 0 . 238 ( 1 . 816 ) 0 . 207 ( 1 . 264 ) 0 . 177 ( 2 . 047 ) CC1/2§0 . 988 ( 0 . 219 ) 0 . 975 ( 0 . 323 ) 0 . 983 ( 0 . 157 ) 0 . 974 ( 0 . 205 ) CC*§0 . 997 ( 0 . 601 ) 0 . 994 ( 0 . 698 ) 0 . 996 ( 0 . 521 ) 0 . 993 ( 0 . 584 ) Anisotropic scaling B-factors¶ a* , b* ( Å2 ) 18 . 1922 . 1515 . 4420 . 96 c* ( Å2 ) −36 . 37−44 . 3−30 . 88−41 . 92Refinement Rwork/Rfree0 . 2531/0 . 2961 ( 0 . 3712/0 . 4188 ) 0 . 2446/0 . 2800 ( 0 . 3464/0 . 3726 ) 0 . 270/0 . 308 ( 0 . 358/0 . 371 ) – CCwork/CCfree§0 . 918/0 . 900 ( 0 . 373/0 . 300 ) 0 . 923/0 . 904 ( 0 . 438/0 . 293 ) 0 . 897/0 . 890 ( 0 . 343/0 . 280 ) – No . atoms56 , 47856 , 50158 , 279– Macromolecule56 , 47256 , 49558 , 273– Ligand/ion666– Water000– Protein residues687168716875– B-Factors Protein139 . 60175 . 65137 . 7– Ligand/ion169 . 70175 . 69134 . 4– R . m . s deviations Bond lengths ( Å ) 0 . 0040 . 0050 . 003– Bond angles ( ° ) 0 . 911 . 120 . 80– Clashscore11 . 9122 . 8912 . 88– Ramachandran favored ( % ) 948892– Ramachandran outliers ( % ) 0 . 410 . 830 . 23–†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/ ) .
We combined Taq EΔ1 . 1σA ( Δ1 . 1σA: Taq σA lacking the N-terminal region 1 . 1 , which is dispensable for in vitro transcription . Region 1 . 1 is not expected to alter protein/DNA interactions in RPo ) with us-fork promoter DNA , which contains a ds −35 element and a mostly ss −10 element ( Figure 1—figure supplement 1 ) . The resulting complex ( 423 kD ) was crystallized and diffraction data were collected and analyzed ( Table 1 ) . The structure was determined by molecular replacement , which identified two complexes per asymmetric unit , and refined using data extending to 4 Å-resolution ( Table 1 , Figure 1—figure supplement 2 ) . The solvent content of the crystals was 82% and examination of the crystal packing revealed space for the expected position of additional promoter DNA . We therefore formed a complete RPo by combining Taq EΔ1 . 1σA with a duplex promoter DNA scaffold ( −36 to +12 with respect to the transcription start site at +1 ) but with a non-complementary transcription bubble generated by altering the sequence of the t-strand DNA from −11 to +2 . RPo crystallized in the same habit and diffraction data were analyzed to 4 . 7 Å-resolution ( Table 1 ) . In the resulting electron density maps , most of the ss t-strand DNA was poorly ordered and unable to be modeled . To stabilize the t-strand DNA , we added an RNA primer complementary to the ss t-strand DNA from +1 to −3 , yielding a 4 bp RNA:DNA hybrid ( Figure 1A ) . We crystallized the resulting complex ( 437 kD , which we call RPo hereafter ) , collected and analyzed diffraction data , and refined the structure using reflections to a minimum Bragg spacing of 4 . 14 Å ( Table 1 , Figure 1—figure supplement 2 ) . In RPo , good electron density for all of the nucleic acids included in the scaffold was observed ( Figure 1C ) . The protein/DNA contacts seen in the us-fork complex are essentially identical to the relevant subset of contacts in RPo . The extensive protein/DNA interface in RPo buries 6300 Å2 of molecular surface ( Figure 1B ) . Overall close contacts with the nucleic acids occur from −36 to −30 and −17 to +9 , consistent with hydroxyl-radical footprinting of RPo on promoters ( Schickor et al . , 1990; Ross and Gourse , 2009 ) . Protein/DNA interactions are absent in the −35/−10 spacer DNA from −29 to −18 . Despite the relatively low resolution of our analysis ( Table 1 ) , important protein side chain/nucleic acid interactions were resolved in electron density maps . Protein side chain/nucleic acid interactions specifically discussed in this paper are supported by unbiased simulated annealing omit maps shown for each case ( see below ) . The protein side chain/nucleic acid interactions specifically discussed in this paper occur via conserved ( often universally ) residues of the RNAP β′ or σA subunits . The level of conservation of relevant β′ residues , determined from an alignment of 834 bacterial RNAP β′ subunit sequences ( Lane and Darst , 2010 ) is tabulated in Table 2 . An alignment of 1002 diverse σA sequences was constructed ( Supplementary file 1; a sub-alignment of selected diverse sequences is shown in Figure 1—figure supplement 3 ) and the level of conservation of relevant σA residues is tabulated in Table 3 . 10 . 7554/eLife . 08504 . 008Table 2 . Conservation of RNAP β′ subunit residuesDOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 008Residue% Identity*Blosum62 score* , †Distribution of residues from alignment*β′Y3499 . 50 . 976Y 803; H 1; Q 1; F 2β′R3599 . 40 . 980R 829; K 5*Determined from an alignment of 834 bacterial RNAP β′ subunit sequences ( Lane and Darst , 2010 ) . †Blosum62 score calculated by PFAAT ( Johnson et al . , 2003 ) . 10 . 7554/eLife . 08504 . 009Table 3 . Conservation of σA residuesDOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 009Residue% Identity*Blosum62 score* , †Distribution of residues from alignment*σA Y21799 . 40 . 988996 Y; 5 H; 1 FσA R2201000 . 998–σA W2561000 . 998–σA W2571000 . 998–σA Q2601000 . 998–σA R2641000 . 998–σA R2741000 . 998–σA V2771000 . 998–σA H2781000 . 998–σA E2811000 . 998–σA R28899 . 70 . 993999 R; 3 KσA R29199 . 70 . 988997 R; 2 K; 1 H; 1 S; 1 L*Determined from an alignment of 1002 bacterial RNAP primary σ subunit sequences ( Supplementary file 1 ) . †Blosum62 score calculated by PFAAT ( Johnson et al . , 2003 ) . Starting from the upstream end of the promoter DNA , the −35 element interacts exclusively with σ4A in a manner consistent with the high-resolution ( 2 . 4 Å ) structure of the isolated σ4A/−35 element complex ( Campbell et al . , 2002 ) . The duplex DNA just upstream of the −10 element ( −17 to −13 ) interacts with β′ , σ3A , and σ2A ( Figure 1B ) . Previously , conserved residues of the β′-zipper ( β′Y34 and , to a lesser extent , β′R35; Table 2 ) that contribute to RPo stability by interacting with duplex spacer DNA were identified ( Yuzenkova et al . , 2011 ) . In the RPo structure , both β′Y34 and β′R35 are positioned to form polar interactions with the −17 nt-strand DNA ( −17 ( nt ) ) phosphate ( Figure 2A , C ) . 10 . 7554/eLife . 08504 . 010Figure 2 . Protein interactions with duplex DNA upstream of the transcription bubble and recognition of the extended −10 element . ( A ) ( Left ) Overall view of RPo structure ( similar to Figure 1B ) . The boxed area is magnified on the right . ( Right ) Magnified view showing protein interactions ( β′ and σA ) with duplex DNA from −18 to −12 . Proteins are shown as backbone worms ( β′ , light pink; σA , light orange ) with interacting side chains shown in stick format ( β′ , pink; σA , orange ) . Likely polar interactions are denoted with grey dashed lines . ( B ) Same as ( A ) ( right ) but rotated 180° about the x-axis . ( C ) Similar view as ( A ) ( right ) . Superimposed is the simulated annealing omit map ( grey mesh , 2Fo − Fc , contoured at 1σ ) , calculated from a model where the following protein segments were removed ( β′ 33–36; σA 259–292 ) and shown only within 2 Å of omitted atoms . ( D ) Similar view as ( B ) . Superimposed is the simulated annealing omit map ( grey mesh , 2Fo − Fc , contoured at 1σ ) , calculated from a model where the following protein segments were removed ( β′ 33–36; σA 259–292 ) and shown only within 2 Å of omitted atoms . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 010 We observe many interactions of σ3A and σ2A with the duplex DNA just upstream of the transcription bubble ( −17 to −12 ) , predominantly with the nt-strand facing the holoenzyme ( Figures 1B , 2 ) . Conserved H278 and R274 of σA ( corresponding to Eco σ70 H455 and R451; Figure 1—figure supplement 3; Table 3 ) are positioned to interact with the −17 ( nt ) and −16 ( nt ) phosphates , respectively ( Figure 2 ) . Substitution of either of these residues causes defects in promoter binding ( Barne et al . , 1997; Fenton et al . , 2000; Singh et al . , 2011 ) . Sequence-specific recognition of the duplex DNA upstream of the −10 element can occur through the extended −10 element ( T−15G−14 ) , which stabilizes RPo and can substitute for the −35 element ( Keilty and Rosenberg , 1987 ) . Conserved E281 of σ3A ( σ70 E458; Figure 1—figure supplement 3; Table 3 ) is positioned to recognize the −14 GC bp through a polar interaction with C−14 ( t ) , as predicted from allele-specific suppression genetics ( Barne et al . , 1997 ) ( Figure 2A , C ) . G−14 ( nt ) is also specifically recognized by conserved R264 ( σ70 R441; Figure 1—figure supplement 3; Table 3 ) of σ2A ( Figure 2B , D ) . Substitutions in the corresponding amino acid position of an alternative σ cause defects in promoter recognition ( Daniels et al . , 1990 ) . Methylation protection and interference indicates Eco Eσ70 makes close contacts with G−14 ( nt ) on an extended −10 promoter ( Minchin and Busby , 1993 ) . Conserved V277 ( σ70 V454; Figure 1—figure supplement 3; Table 3 ) may contact the T−15 ( nt ) methyl group , possibly explaining the preference for T at this position ( Figure 2B ) . The primary role of σ2 in −10 element recognition was first uncovered when substitutions of invariant Q260 ( σ70 Q437; Figure 1—figure supplement 3; Table 3 ) were shown to affect sequence-specific recognition of the −12 bp ( Kenney et al . , 1989; Waldburger et al . , 1990 ) . Modeling suggested that Q260 may H-bond with the major-groove edge of A−12 ( t ) ( Feklistov and Darst , 2011 ) . However , in our structures , the amide group of the Q260 side chain points away from the major-groove edge of A−12 ( t ) and cannot form H-bonds ( Figure 2B , D ) . We suggest that Q260 may form base-specific H-bonds with the −12 bp in an intermediate during the pathway to RPo formation ( Saecker et al . , 2011 ) , whereas our structures represent the final , transcription ready RPo , explaining the genetic data . Flipping of the A−11 ( nt ) base from the duplex DNA into its recognition pocket in σ2A is thought to be the key event in the initiation of promoter melting ( Chen and Helmann , 1997; Lim et al . , 2001; Heyduk et al . , 2006; Feklistov and Darst , 2011 ) . Strand opening propagates downstream to +1 , but in the upstream direction , the base-paired T−12 ( nt ) interacts with an invariant W-dyad of σ2A ( W256/W257 , σ70 W433/W434; Figure 1—figure supplement 3; Table 3 ) to maintain the ds/ss ( −12/−11 ) junction at the upstream edge of the transcription bubble ( Figure 3A , C , D , Figure 3—figure supplement 1 ) . The stabilization of the upstream ds/ss junction involves a previously unseen rearrangement of the W256 side chain . In all previous high resolution structures of σA/σ70 in many different contexts but never with an upstream ds/ss junction ( Malhotra et al . , 1996; Campbell et al . , 2002; Vassylyev et al . , 2002; Feklistov and Darst , 2011; Zhang et al . , 2012 ) , the W256 side chain makes an ‘edge-on’ interaction with W257 ( Figure 3B ) . In the presence of the upstream ds/ss junction , the W256 side chain rotates away from W257 , filling the space vacated by the flipped-out A−11 ( nt ) and forming a π-stack with the face of T−12 ( t ) otherwise exposed by the absence of A−11 ( nt ) ( Figure 3C , D , Figure 3—figure supplement 1 ) . The W-dyad forms a ‘chair’-like structure , with W256 serving as the back of the chair , and W257 as the seat , buttressing T−12 ( nt ) from the major groove side ( Figure 3A , C , D ) . The methyl group of the T−12 ( nt ) base approaches the face of the W257 side chain at a nearly orthogonal angle , possibly forming a favorable methyl π interaction ( Umezawa and Nishio , 1998; Brandl et al . , 2001 ) ( Figure 3C ) . 10 . 7554/eLife . 08504 . 011Figure 3 . Structures maintaining the upstream ds/ss junction of the transcription bubble and directing the t-strand DNA to the RNAP active site . ( A ) ( Left ) Overall view of RPo structure ( similar to Figure 1B ) . The boxed area is magnified on the right . ( Right ) Magnified view showing the upstream ds/ss junction of the transcription bubble in RPo ( the RNAP β subunit , which obscures the view , has been removed ) . RNAP is shown as a molecular surface , except side chains of key σA residues ( R217 , R220 , W256 , R288 , R291 ) are shown ( orange ) . The orthogonal directions of the ss nt- and t-strand DNA following the upstream ds/ss junction are denoted by black arrows . The dashed , curved line denotes the potential path of the t-strand −11 base from its position in the duplex DNA ( base-paired to A−11 ( nt ) ) to its position in the structure . ( B ) Structure of Taq σ2A bound to the ss , nt-strand −10 element ( PDB ID 3UGO ) ( Feklistov and Darst , 2011 ) showing the disposition of the universally conserved σA W-dyad ( Taq σA W256/W257 ) . Shown is the ss DNA from −14 to −7 ( −10 element colored yellow ) , the σ2 . 3A-helix ( light orange ) and the W-dyad ( orange side chains with transparent CPK atoms ) . W256 makes an edge-on interaction with the face of W257 , as observed in all other σ70/σA structures in many different contexts ( Malhotra et al . , 1996; Campbell et al . , 2002; Vassylyev et al . , 2002; Murakami et al . , 2002a , 2002b; Feklistov and Darst , 2011; Zhang et al . , 2012 ) . ( C ) Disposition of the W-dyad in RPo ( containing upstream ds/ss junction , shown schematically above ) . Only the nt-strand DNA from −14 to −7 , the σ2 . 3A-helix , and the W-dyad are shown ( as in B ) . ( D ) Same view as ( C ) . Superimposed is the simulated annealing omit map ( grey mesh , 2Fo − Fc , contoured at 1σ ) , calculated from a model where the following segments of σA were completely removed ( 216–221 , 255–258 , and 287–292 ) and shown only within 2 Å of omitted atoms . ( E ) Similar view as ( A ) ( right ) . Superimposed is the simulated annealing omit map ( grey mesh , 2Fo − Fc , contoured at 1σ ) , calculated from a model where the following segments of σA were removed ( 216–221 , 255–258 , and 287–292 ) and shown only within 2 Å of omitted atoms . Clear Fourier density for σA Y217 and R288 is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 01110 . 7554/eLife . 08504 . 012Figure 3—figure supplement 1 . Stereo view of RPo model and resulting electron density map ( grey mesh , 2Fo − Fc , contoured at 0 . 7σ ) . The view is similar to Figure 3A . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 012 Examination of the structure near the upstream ds/ss junction revealed the solvent-exposed aromatic face of a conserved σ2A Tyr side chain , Y217 ( σ70 Y394; Figure 3A , E; Figure 1—figure supplement 3; Table 3 ) , that does not appear to play an important role in the σ structure per se , but lies along the path the −11 ( t ) base could follow from its position in duplex DNA ( base-paired to A−11 ( nt ) ) to its position in the structure when orphaned by the flipped out A−11 ( nt ) ( dashed line , Figure 3A ) . The −11 ( t ) nucleotide is almost always a T , being complementary to A−11 ( nt ) , the most highly conserved position of the −10 element ( Shultzaberger et al . , 2007 ) . In the us-fork , the −11 ( t ) nucleotide is absent ( Figure 1—figure supplement 1 ) , whereas in RPo , the −11 ( t ) nucleotide is an ( atypical ) A , being part of the engineered non-complementary transcription bubble ( Figure 1A ) . In RPo , the A−11 ( t ) base is not stacked on Y217 but instead is about 12 Å away , flipped up alongside the σ3A−3 . 0 α-helix , sitting between R288 and R291 ( Figure 3A; Figure 1—figure supplement 3; Table 3 ) . We reasoned that we may not observe the orphaned −11 ( t ) base stacked on Y217 for two reasons that are not mutually exclusive . First , Y217 may play an important role in stabilizing the melted state of the −11 bp during an intermediate of the normal promoter melting pathway ( Saecker et al . , 2011 ) . Second , structural modeling suggested that the A−11 ( t ) purine base present in the synthetic promoter construct ( Figure 1A ) may be too bulky to stack on Y217 , which sits at the bottom of a narrow trough in the σ2A structure ( Figure 3A ) . To investigate the role of Y217 further , we crystallized Taq EΔ1 . 1σA with an us-fork template containing a complementary A:T bp at the −11 position ( us-fork ( −11 bp ) ; Figure 4A ) . To avoid model bias , we determined the structure by molecular replacement using the Taq EΔ1 . 1σA/us-fork ( −12 bp ) structure ( lacking the −11 ( t ) base; Figure 1—figure supplement 1 ) . The structure was modeled and refined ( 4 . 6 Å-resolution , Table 1 , Figure 4—figure supplement 1 ) , and the unbiased density maps revealed clear difference density for the T−11 ( t ) base stacked on Y217 ( Figure 4B ) . 10 . 7554/eLife . 08504 . 013Figure 4 . The σA Y217 may stack on the T−11 ( t ) base orphaned by the flipped out A−11 ( nt ) base . ( A ) Synthetic oligonucleotides used for us-fork ( −11 bp ) crystallization . The numbers above the sequence 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 element ( Keilty and Rosenberg , 1987 ) purple . The nt-strand DNA ( top strand ) is colored dark grey; the t-strand DNA ( bottom strand ) , light grey; the RNA transcript , red . ( B ) The T−11 ( t ) base orphaned by the flipped out A−11 ( nt ) stacks on σA Y217 in the us-fork ( −11 bp ) structure . The 4 . 6 Å-resolution electron density map ( contoured at 0 . 7σ ) is shown ( grey mesh ) . Also superimposed is the simulated annealing omit map ( green mesh , Fo − Fc , contoured at 3σ ) , calculated from a model where σA Y217 was mutated to Ala and the T−11 ( t ) nucleotide was deleted . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 01310 . 7554/eLife . 08504 . 014Figure 4—figure supplement 1 . Data and model quality for us-fork ( −11 bp ) complex . Plots relating data quality with model quality at 4 . 6 Å-resolution ) 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 . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 014 A functional role for W256 in promoter melting was first proposed by Helmann and Chamberlin ( 1988 ) . Ala substitution of the corresponding Trp in Bacillus subtilis σA gave rise to severe promoter melting defects in vitro and corresponding cold phenotypes in vivo ( Juang and Helmann , 1994; Panaghie et al . , 2000 ) . The functional role of Y217 has not , to our knowledge , been previously examined . We investigated the effects of individual Ala substitutions in Eco σ70 W433 and Y394 ( Taq W256 and Y217 ) on the kinetics of RPo formation ( Roe et al . , 1984; Buc and McClure , 1985 ) using a recently reported fluorescence assay ( Ko and Heyduk , 2014 ) . The assay relies on a Cy3 fluorophore attached to the promoter nt-strand at position +2; fluorescence yield in this context is sensitive to the local environment and increases more than twofold upon RPo formation . Unlike previously used non-equilibrium methods ( EMSA , filter binding ) , this assay allows detection of promoter melting at equilibrium and does not depend on the use of competitors , such as heparin . For these assays , we used one of the most thoroughly characterized promoters , λ PR ( Saecker et al . , 2002 , 2011 ) . Control assays showed that under saturating conditions , both σ70 substitutions ( W433A and Y394A ) associated with core RNAP and supported abortive transcription as well as wild-type σ70 ( data not shown ) , confirming their structural integrity . The multistep process of promoter opening can be described by a simplified kinetic scheme ( Figure 5A ) ( McClure , 1980 ) where an initial promoter complex ( RPi ) existing in rapid equilibrium with free promoter and RNAP ( binding step described by a dissociation constant Kd ) is converted in a rate-limiting step to RPo ( isomerization described by the rate constant k2 ) . Fluorescence traces of RPo formation under pseudo first-order conditions ( Roe et al . , 1984 ) recorded at increasing RNAP concentrations were fit to single-exponentials and yielded observed rate constants ( kobs ) for RPo formation ( Figure 5B ) . Nonlinear fits to the resulting hyperbolic curves ( Figure 5C ) allowed the determination of Kd and k2 ( Saecker et al . , 2002 ) ( Figure 5D ) . 10 . 7554/eLife . 08504 . 015Figure 5 . Functional role of Eco σ70 W433 and Y394 in RPo formation . ( A ) Simplified , two-step kinetic scheme for RPo formation ( Roe et al . , 1984; Buc and McClure , 1985 ) ( R , RNAP; P , promoter; RPi , intermediate complex ) . ( B ) Representative time trace of fluorescence increase ( from Cy3 labelled promoter DNA ) during RPo formation . The solid red line illustrates the non-linear regression fit to a single-exponential model ( see ‘Materials and methods’ ) , which described >90% of the fluorescence amplitude rise . ( C ) The RNAP-concentration dependence of the observed rate ( kobs ) of RPo formation detected by Cy3 fluorescence ( Ko and Heyduk , 2014 ) for Eco holoenzymes with σ70 ( wt ) as well as σ70 carrying substitutions W433A or Y394A . Error bars denote standard errors of the mean for ≥three independent measurements . ( D ) Summary of effects of σ70 W433A and Y394A substitutions on thermodynamic and kinetic parameters of RPo formation . The data was normalized to the % observed with wild-type Eσ70 . ( E ) Equilibrium binding of ss nt-strand oligos of λ PR promoter −10 element detected in the fluorescent RNAP beacon assay ( Feklistov and Darst , 2011; Mekler et al . , 2011 ) to Eco holoenzymes with σ70 , as well as σ70 carrying substitutions W433A or Y394A . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 01510 . 7554/eLife . 08504 . 016Figure 5—figure supplement 1 . RPo dissociation data . ( Left ) Representative time trace of fluorescence decay after rapid mixing of pre-formed Eco RPo ( with wild-type σ70 ) into 1 . 1 M NaCl ( Gries et al . , 2010 ) . The solid line illustrates the non-linear regression fit to a single-exponential model . ( Right ) Representative dissociation curves for holoenzymes containing wild-type σ70 and W433A and Y394A substitutions . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 016 Neither σ70 W433A nor Y394A had a significant effect on Kd for RPi formation , but the substitutions decreased the rate of isomerization by about twofold to threefold ( at 37°C , Figure 5D ) . At suboptimal temperature ( 25°C ) the effect of the W433A substitution was more pronounced , resulting in an ∼sevenfold reduction in isomerization rate . Neither σ70 W433A nor Y394A significantly altered the affinity of holoenzyme binding to ss oligos comprising the nt-strand of the −10 element ( Tomsic , 2001 ) ( Figure 5E ) . W256 appears to make the primary contribution to maintaining the ds/ss junction at the upstream edge of the transcription bubble ( Figure 3A ) , suggesting that this residue may play an important role in preventing transcription bubble collapse and dissociation of RPo . To probe the roles of both σ70 W433 and Y394 in maintaining RPo stability , we rapidly destabilized preformed RPo with 1 . 1 M NaCl ( Gries et al . , 2010 ) and followed the loss of RPo by monitoring the decay of fluorescence intensity with time ( Figure 5—figure supplement 1 ) . The dissociation curves are complex , reflecting the detection of a short lived intermediate ( expected under these conditions ) ( Gries et al . , 2010 ) by this assay . Although a full analysis is beyond the scope of this study , the overall apparent rate of RPo decay ( koffapp ) was determined from single-exponential fits of the decay curves . The σ70 W433A and the Y394A variants both gave a ∼fourfold higher rate of RPo dissociation under high salt conditions than did wild-type σ70 ( Figure 5D , Figure 5—figure supplement 1 ) . Downstream from the point of melting , the two DNA strands are directed on orthogonal paths ( black arrows , Figure 3A ) . The nt-strand ( −11 to −4 ) drapes across the surface of σ2A , directed by phosphate backbone interactions and notable base-specific recognition of A−11 ( nt ) and T−7 ( nt ) of the −10 element , and G−6 ( nt ) of the discriminator ( Feklistov and Darst , 2011; Zhang et al . , 2012 ) . Further downstream , interactions of the nt-strand from −3 to +2 occur exclusively with the RNAP β subunit , including base-specific recognition of G+2 ( nt ) ( Zhang et al . , 2012 ) . At the point of melting , a ∼90° turn of the t-strand backbone ( between −12 and −11 ) may be effected by electrostatic interactions between conserved basic residues of σ2A ( R220; Figure 1—figure supplement 3; Table 3 ) and σ3A ( R288 , R291 ) and four t-strand backbone phosphates in a row ( −13 , −12 , −11 , −10 ) encompassing the turn ( Figure 3A ) . Strong simulated annealing omit 2Fo − Fc density is associated wth σ3A R288 , confirming its role in interacting with the −13 ( t ) phosphate ( Figure 3E ) . The σ2A R220 and σ3A R291 give weaker difference density so their role in interacting with the −12 ( t ) and −11 ( t ) phosphate groups is tentative . The turn directs the t-strand away from the nt-strand and towards the RNAP active site ( Figure 3A ) . The ss t-strand DNA from −9 to −5 is guided towards the RNAP active site through a tunnel formed between the RNAP β1-lobe ( called the protrusion in eukaryotic RNAP II; Cramer et al . , 2001 ) and the σ3 . 2-loop ( also referred to as the σ-finger ) , an extended linker that loops into and out of the RNAP active-site channel ( Murakami et al . , 2002a; Zhang et al . , 2012 ) , connecting the σ3 and σ4 domains ( Figure 6 ) . 10 . 7554/eLife . 08504 . 017Figure 6 . Structural role of the σ3 . 2-loop . ( Left ) Overall view of RPo structure , colored as in Figure 1 except σA is orange . The RNAP β and β′ subunits are transparent to reveal the RNAP active site Mg2+ ( yellow sphere ) and the nucleic acids held inside the RNAP active site channel . The ss nt-strand DNA is omitted for clarity . The boxed area is magnified on the right . ( Right ) Magnified view showing a cross-section of the RNAP active site channel . For clarity , the RNAP β , β′ , and σ2A domains are shown mostly as outlined shapes , with β transparent . The ss t-strand DNA ( −11 to −4 ) is directed towards the RNAP active site through a tunnel between the σ3 . 2-loop and the β1-lobe . The 4-nt RNA transcript ( −3 to +1 ) contacts the distal tip of the σ3 . 2-loop . Further elongation of the RNA would require displacement of the σ3 . 2-loop . DOI: http://dx . doi . org/10 . 7554/eLife . 08504 . 017 Previous structural analyses predicted that the σ3 . 2-loop would physically occupy the path of the elongating RNA and must be displaced for full RNA extension to occur ( Vassylyev et al . , 2002; Murakami et al . , 2002a ) . Indeed , the upstream edge of the post-translocated 4-nt transcript fits snugly between the RNAP active site and the distal tip of the σ3 . 2-loop , which contacts the upstream RNA:DNA bp at −3 , and the t-strand bases at −4 and −5 ( Figure 6 ) . Extension of the RNA transcript and translocation to form a 5 bp post-translocated RNA:DNA hybrid cannot occur without displacement of the σ3 . 2-loop ( Basu et al . , 2014 ) , marking the point in transcription initiation ( translocation of the 4–5 bp RNA:DNA hybrid from pre- to post-translocated ) where steric clash between the elongating RNA transcript and the σ3 . 2-loop begins effecting abortive initiation and σ release ( Murakami et al . , 2002a; Nickels et al . , 2005; Kulbachinskiy and Mustaev , 2006 ) .
Our results clarify the role of the universally conserved W-dyad of housekeeping ( also called primary or group 1 ) σ's ( Gruber and Bryant , 1997 ) in the promoter opening pathway , particularly for Taq σA W256 ( Eco σ70 W433 ) , which rotates into the DNA duplex and serves as a steric mimic of the flipped-out A−11 ( nt ) base by a stacking mechanism ( Figure 3A , C , D ) . The bacterial RNAP σ subunit can be added to the list of proteins using a wedge residue ( usually an aromatic side chain ) to invade the DNA duplex to stabilize the extrahelical conformation of a flipped-out base ( Lau et al . , 1998; Davies et al . , 2000; Yang et al . , 2009; Yi et al . , 2012 ) . We also identified another conserved σA aromatic residue ( Taq σA Y217 ) that plays an important role in the promoter opening pathway , possibly by stacking with T−11 ( t ) orphaned when the conserved A−11 ( nt ) base flips out ( Figure 4B ) . The kinetic studies reveal that both aromatic residues ( W256 and Y217 ) act in a context dependent manner—they are not important for the initial promoter binding step ( Figure 5D ) nor for binding the ss −10 element DNA ( Figure 5E ) : instead W256 and Y217 act to increase the rate of the isomerization ( promoter opening step ) itself ( Figure 5E , D ) , possibly by making contacts unique to the transition state that lower the energy barrier between RPi and RPo in the two-step kinetic scheme ( Figure 5A ) . Since the initial promoter binding step ( formation of RPi , Figure 5A ) is not affected by the σ70 W433A substitution ( Figure 5D ) , we surmise that RPi does not feature the stacking interaction formed by W433A on the T−12 ( nt ) base ( exposed by the flipping-out of A−11 ( nt ) ) . Since the −11 bp is thought to be the first bp disrupted in the promoter opening pathway ( Chen and Helmann , 1997; Lim et al . , 2001; Heyduk et al . , 2006; Feklistov and Darst , 2011 ) , this implies that RPi is a closed complex ( RPc ) comprising duplex promoter DNA . The effects of σ70 W433A that we observed are consistent with previous observations using nonequilibrium methods ( Fenton et al . , 2000; Tomsic , 2001; Fenton and Gralla , 2003; Schroeder et al . , 2009 ) . These observations support the critical role of σA W256 and Y217 ( σ70 W433 and Y394 ) in formation and stability of RPo . In addition to the housekeeping σ ( σA in Taq or σ70 in Eco ) that controls transcription of the majority of cellular genes ( with consensus −35 and −10 elements of TTGACA and TATAAT , respectively; Shultzaberger et al . , 2007 ) , bacteria rely on alternative σ′s to direct RNAP to highly specialized promoters ( with alternative −35 and −10 elements ) controlling operons in response to environmental and physiological cues ( Gruber and Gross , 2003; Feklistov et al . , 2014 ) . Although the W-dyad is universally conserved in housekeeping σ's ( Gruber and Bryant , 1997 ) , it is not a conserved feature of alternative σ's ( Lonetto et al . , 1992; Helmann , 2002; Campbell et al . , 2003 ) ; bulky hydrophobic residues are favored at the corresponding positions of alternative σ's ( but rarely W ) . The W-dyad is likely to be the optimal configuration for supporting the upstream ds/ss junction of the transcription bubble , giving the housekeeping σ′s a powerful DNA-melting capacity , allowing them to function on thousands of highly divergent , nonoptimal promoter sequences . Alternative residues supporting the upstream ds/ss junction of the transcription bubble may weaken the ability of RNAP with alternative σ's to form RPo , fine-tuning their specificity ( Feklistov et al . , 2014 ) . The residue corresponding to Taq σA Y217 ( σ70 Y394 ) appears to be conserved as either Y or F among σ70-family alternative σ's suggesting that this residue plays a key role common to all σ′s . Zuo and Steitz ( 2015 ) soaked crystals of Eco transcription initiation complexes ( containing a full transcription bubble ) with NTP substrates to generate short transcripts ( with 5′-triphosphate ) in crystallo . A pre-translocated 4-nt transcript did not reach the σ3 . 2-loop , whereas a pre-translocated 5-nt transcript appeared to just reach and interact with the σ3 . 2-loop . Attempts to generate longer transcripts resulted in severe degradation of the crystals , suggesting significant conformational changes of the RNAP that were incompatible with the crystal packing either due to transcript/σ3 . 2-loop interactions , ‘scrunching’ of the t-strand DNA ( Kapanidis et al . , 2006; Revyakin et al . , 2006; Roberts , 2006 ) , or both . The upstream edge of our post-translocated 4-nt transcript is equivalent to the pre-translocated 5-nt transcript observed by Zuo and Steitz ( 2015 ) : in both cases the upstream edge of the RNA just contacts the σ3 . 2-loop and the conformation of the σ3 . 2-loop is very similar indicating that , at least in this case , the presence or absence of the 5′-triphosphate does not alter the gross interaction of the elongation transcript with the σ3 . 2-loop . In vitro , RNAP initiates efficiently with dinucleotide primers lacking a 5′-triphosphate without obvious defects in σ release or promoter escape . Basu et al . ( 2014 ) were able to generate a 6-nt pre-translocated transcript ( containing a 5′-triphosphate ) in crystals of Tth transcription initiation complexes with a downstream-fork promoter template that lacks duplex DNA upstream of the −10 element and is therefore unable to ‘scrunch’ the t-strand DNA . In this case , the 5′-nt of the transcript displaces the σ3 . 2-loop , which is not modeled and presumably disordered . Other conformational changes of the RNAP or changes in σ/RNAP interactions were not observed . In vitro , the rate-limiting step of bacterial RNAP transcription is often the isomerization step to open the promoter and form RPo ( McClure , 1980 , 1985; Amouyal and Buc , 1987 ) . The kinetics of the many steps of the transcription cycle in vivo have not been characterized , but many transcription units are clearly controlled at the initiation step ( Paul et al . , 2004 ) . In bacteria , recognition of the promoter −10 element and DNA opening are directly coupled ( Feklistov and Darst , 2011; Liu et al . , 2011 ) , with the Trp stacking interaction ( Figure 3A , C ) playing a key role . In contrast to tight coupling between promoter recognition and transcription bubble formation at most bacterial promoters , in eukaryotes promoter recognition , RNAP II recruitment , and promoter opening appear to be uncoupled . The preinitiation complex ( PIC ) is the molecular assembly through which eukaryotic RNAP II locates and utilizes a promoter , which may be pre-recognized by basal transcription factors . RPo formation requires ATP hydrolysis by the Ssl2 ( XPB ) subunit of TFIIH , which translocates downstream DNA into RNAP II against fixed upstream contacts to force DNA melting ( Kim et al . , 2000; Grünberg and Hahn , 2013 ) . This contrasts with the spontaneous unwinding driven by RNAP/promoter DNA interactions alone during bacterial RPo formation ( Liu et al . , 2011 ) . Although there are clear similarities between σ and the eukaryotic basal transcription factor IIB in the contacts made to the 5′ RNA , hybrid junction , and ss-tDNA , there is no structural similarity between σ and TFIIB ( Kostrewa et al . , 2010; Liu et al . , 2010; Sainsbury et al . , 2013 ) . These contacts may play similar roles in aiding promoter escape by helping eject σ or TFIIB from the RNAP active site cleft , but it is currently unclear whether any eukaryotic basal transcription factor stabilizes an upstream fork-junction by interactions similar to the σ-mediated Trp stacking ( Figure 3A , C ) . Further , although effects on RPo formation may help regulate some eukaryotic promoters ( Kouzine et al . , 2013 ) , other steps , including removal of nucleosomes and promoter-proximal pausing ( Boeger et al . , 2003; Adelman and Lis , 2012 ) appear to be rate-limiting at many eukaryotic promoters . Even when promoters are nucleosome-free , assembly of the PIC , rather than promoter opening , may be rate-limiting . Further mechanistic and structural studies of RNAPII on promoters with diverse architectures , including both TATA-containing and TATA-less promoters , are needed for a better understanding of the steps in RNAPII initiation . The structures of RPo determined here reveal how the RNAP holoenzyme recognizes the extended −10 element , stabilizes the transcription bubble , directs the t-strand DNA into the RNAP active site , and how the RNA:DNA hybrid initiates σA release . Supported by the real-time kinetic data , the structures elucidate the roles of individual aromatic amino acid residues in nucleation of the transcription bubble and maintenance of RPo stability , in part through previously unobserved stacking mechanisms . The results also provide a basis for more incisive investigations of RPo formation and transcriptional regulation ( Bae et al . , 2015 ) .
Taq core RNAP and Δ1 . 1σA were prepared as described previously ( Murakami et al . , 2003 ) . Promoter DNA strands ( Oligos Etc . ) were annealed in 10 mM Tris–HCl , pH 8 . 0 , 1 mM EDTA , 0 . 2 M NaCl and aliquots were stored at −20°C . For crystallization , aliquots of purified Taq core RNAP and Δ1 . 1σA were thawed on ice and buffer-exchanged into crystallization buffer ( 20 mM Tris–HCl , pH 8 . 0 , 0 . 2 M NaCl ) . Taq Δ1 . 1σA-holoenzyme was formed by adding 1 . 2-fold molar excess of Δ1 . 1σA to the core RNAP and the mixture was incubated for 15 min at room temperature . A 1 . 5-fold molar excess of promoter DNA was then added to the holoenzyme along with MgCl2 ( 10 mM final ) and incubated for 15 min at room temperature . When present , a fivefold molar excess of RNA primer ( GE Dharmacon , Lafayette , CO , United States ) was also added . The final RNAP concentration was adjusted to 25 μM . Crystals were grown by vapor diffusion at 22°C by mixing 1 μl of sample with 1 μl of reservoir solution ( 20 mM MgCl2 , 20 mM Tris–HCl , pH 8 . 0 , 1 . 6 M ammonium sulfate ) in a 48-well hanging drop tray ( Hampton Research , Aliso Viejo , CA , United States ) . Thin rod-shaped crystals ( typically , 30 × 30 × 300 μm ) appeared after about 5 days . The crystals were transferred into reservoir solution supplemented with 25% ( vol/vol ) glycerol in two steps for cryo-protection , then flash frozen by plunging into liquid nitrogen . X-ray diffraction data were collected at Brookhaven National Laboratory National Synchrotron Light Source ( NSLS ) beamline X29 and at Argonne National Laboratory Advanced Photon Source ( APS ) NE-CAT beamlines 24-ID-C and 24-ID-E . 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 1C , Figure 1—figure supplements 1 , 2C , D , 3D , E , Figure 3—figure supplement 1 , Figure 4B ) . Initial electron density maps were calculated by molecular replacement using Phaser ( McCoy et al . , 2007 ) from a starting model of Taq Δ1 . 1σA-holoenzyme determined at 2 . 8 Å-resolution ( unpublished ) . Two RNAP/DNA complexes were clearly identified in the asymmetric units . The models were first improved using rigid body refinement of each RNAP molecule and subsequently of 20 individual mobile domains using PHENIX ( Adams et al . , 2010 ) . At this point , the electron density maps showed strong connected difference density for the nucleic acids , allowing unambiguous placement using COOT ( Emsley and Cowtan , 2004 ) . Detailed nucleic acid modeling was facilitated using available models of complexes with promoter fragments: σ4A/−35 element DNA complex at 2 . 4 Å ( 1KU7 [Campbell et al . , 2002] ) , RNAP-holoenzyme/us-fork DNA at 6 . 5 Å-resolution ( 1L9Z [Murakami et al . , 2002b] ) , σ2A/nt-strand −10 element DNA at 2 . 1 Å ( 3UGO [Feklistov and Darst , 2011] ) , RNAP-holoenzyme/downstream-fork DNA at 2 . 9 Å ( 4G7H [Zhang et al . , 2012] ) , RNA/DNA hybrid at 2 . 5 Å ( 2O5I [Vassylyev et al . , 2007] ) . The resulting models were improved using deformable elastic network ( DEN ) refinement ( Schröder et al . , 2010 ) with noncrystallographic symmetry ( NCS ) restraints using CNS 1 . 3 ( Brunger et al . , 1998 ) performed on the Structural Biology Grid portal ( O'Donovan et al . , 2012 ) , followed by iterative cycles of manual building with COOT ( Emsley and Cowtan , 2004 ) and refinement with PHENIX ( Adams et al . , 2010 ) . In the RPo structure , the ss t-strand DNA from −11 to −4 was only modeled in one complex of the asymmetric unit . In the other complex , strong , connected Fourier difference density for this segment of DNA was observed but the density was relatively featureless and we were unable to model this segment of the DNA . In the us-fork ( −11 bp ) complex , the t-strand T−11 was modeled in only one complex of the asymmetric unit . In the other complex , density for this base was absent . We follow the criteria of Karplus and Diederichs ( 2012 ) , who showed that the Rmerge statistic commonly used to evaluate data quality is ‘seriously flawed’ and should not be used ( Diederichs and Karplus , 1997 ) , and that the commonly used criteria of <I>/σI > 2 also results in the loss of much useful crystallographic data ( Karplus and Diederichs , 2012 ) . Karplus and Diederichs ( 2012 ) showed , using objective and unbiased analyses , that inclusion of weak X-ray diffraction data ( Rmerge values >> 1 . 0 and <I>/σI << 1 ) resulted in improved structural models . An improved statistic , CC* ( essentially a Pearson correlation coefficient ) , was introduced that provides a single statistically valid guide for deciding whether diffraction data are useful . Since most of the analyses described herein were performed from the RPo structure , we justify the inclusion of diffraction data to 4 . 14 Å-resolution for this case . Data in the highest resolution shell ( 4 . 29–4 . 14 Å ) are very weak when examined by standard criteria ( high Rpim values and <I>/σI = 0 . 8 , Table 1 ) , but have good multiplicity ( 21 . 6 ) and completeness ( 99 . 8% ) , and yield a CC1/2 of 0 . 157 , which is significantly different from zero for the large sample size ( 16 , 966 unique reflections ) at exceedingly low p values ( Rahman , 1968 ) . That the highest resolution shells contain useful data and not noise is reflected in the observation that the Rfree and Rwork for the model refinement do not diverge ( Figure 1—figure supplement 2 , Figure 4—figure supplement 1 ) . Inclusion of higher resolution data resulted in unacceptably low completeness in the highest shells due to the data anisotropy . In the final 2Fo − Fc electron density maps , numerous protein side chains were resolved , including many that appeared to form important protein/nucleic acid interactions . To confirm these protein side chain positions , we produced unbiased difference Fourier maps using a simulated annealing omit procedure . Protein segments flanking the side chains in question were removed completely from the structural model , and the modified models were subjected to simulated annealing refinement using PHENIX ( Adams et al . , 2010 ) . We used the following annealing temperatures ( K ) , 1000; 2500; 5000; 10 , 000 . All temperatures gave the same result ( recovery of electron density for the omitted side chains ) , but the 5000 and 10 , 000 K refinements gave rise to obvious local structural distortions ( expected for such high annealing temperatures with our low-resolution data ) so the unbiased 2Fo − Fc maps were calculated from the 2500 K annealing refinements ( Figures 2C , D , 3D , E ) . | 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’ . However , it is not clear how RNA polymerase holoenzyme interacts with DNA in the open promoter complex . Bae , Feklistov et al . have now used X-ray crystallography to reveal the three-dimensional structure of the open promoter complex with an entire transcription bubble from a bacterium called Thermus aquaticus . The experiments show that there are several important interactions between RNA polymerase holoenzyme and promoter DNA . In particular , the sigma factor inserts into a region of the DNA at the start of the transcription bubble . This rearranges the DNA in a manner that allows the DNA to be exposed and contact the main part of the RNA polymerase . If the holoenyzyme fails to contact the DNA in this way , the holoenzyme does not bind properly to the promoter and transcription does not start . These findings build on previous work to provide a detailed structural framework for understanding how the RNA polymerase holoenzyme and DNA interact to form the open promoter complex . Another study by Bae et al . —which involved some of the same researchers as this study—reveals how another protein called CarD also binds to DNA at the start of the transcription bubble to stabilize the open promoter complex . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | Structure of a bacterial RNA polymerase holoenzyme open promoter complex |
In animals , microtubules and centrosomes direct the migration of gamete pronuclei for fertilization . By contrast , flowering plants have lost essential components of the centrosome , raising the question of how flowering plants control gamete nuclei migration during fertilization . Here , we use Arabidopsis thaliana to document a novel mechanism that regulates F-actin dynamics in the female gametes and is essential for fertilization . Live imaging shows that F-actin structures assist the male nucleus during its migration towards the female nucleus . We identify a female gamete-specific Rho-GTPase that regulates F-actin dynamics and further show that actin–myosin interactions are also involved in male gamete nucleus migration . Genetic analyses and imaging indicate that microtubules are dispensable for migration and fusion of male and female gamete nuclei . The innovation of a novel actin-based mechanism of fertilization during plant evolution might account for the complete loss of the centrosome in flowering plants .
Flowering plants have evolved a double fertilization process that requires the fusion of two sperm cells with the two female gametes , the egg cell and the central cell . This is achieved in a series of complex processes ( Kawashima and Berger , 2011 ) . Sperm cells are immotile and delivered to the female gametes by the pollen tube . The pollen tube is guided towards the female gametes ( pollen tube guidance ) and interacts with a synergid cell ( two specialized cells associated with the female gametes ) to release sperm cells ( pollen tube reception ) into the area proximal to the female gametes . After the fusion between the female and male gametes ( plasmogamy ) , the male gamete nuclei migrate towards the female gamete nuclei and fuse ( karyogamy ) , thus completing fertilization . Several factors and molecular mechanisms controlling these early events such as pollen tube guidance and reception have been recently characterized ( Higashiyama et al . , 2001; Palanivelu et al . , 2003; Kasahara et al . , 2005; Escobar-Restrepo et al . , 2007; Okuda et al . , 2009; Kessler et al . , 2010; Tsukamoto et al . , 2010; Takeuchi and Higashiyama , 2012; Leydon et al . , 2013; Liang et al . , 2013; Ngo et al . , 2014 ) , and mechanisms governing plasmogamy have been identified ( Mori et al . , 2006; Von Besser et al . , 2006; Mori et al . , 2014; Sprunck et al . , 2012 ) . However , how gamete nuclei migrate and fuse in flowering plants still remain largely elusive . In most animals , microtubules organized by the centrosome control female pronucleus migration towards the male pronucleus ( Schatten , 1994; Reinsch and Gonczy , 1998 ) . By contrast , the essential components of the centrosome have been lost before the emergence of flowering plants ( Carvalho-Santos et al . , 2011 ) , and the sperm cell nucleus moves towards the female gamete nucleus without the participation of centrosomes in flowering plants . During plant evolution , the basal plants developed actin-based organelle movements ( Madison and Nebenfuhr , 2013 ) , which coincided with the centrosome loss in somatic cells in the basal land plants ( Vaughn and Harper , 1998 ) . However , the genomes of bryophytes , ferns , and most likely certain gymnosperms encode components of the centrosome and these organisms produce a centrosome specifically in spermatogenous cells ( Vaughn and Harper , 1998 ) . Indeed , microtubules are essential for gamete nuclear migration in the fern Marsilea vestita ( Kuligowski et al . , 1985 ) , and these observations imply that gamete nuclear migration without the centrosome in flowering plants evolved separately from the actin-based organelle movement mechanism in somatic cells , leading to the question of how flowering plants control gamete nuclear migration without a centrosome . Immunofluorescence approaches revealed that corona structures of actin filaments around the sperm cells appear at the time of sperm cell release from the pollen tube prior to plasmogamy in many flowering plants ( Huang and Russell , 1994; Huang and Sheridan , 1998; Huang et al . , 1999; Fu et al . , 2000; Ye et al . , 2002 ) . Changes in F-actin organization in the egg cell during fertilization are also evident ( Huang et al . , 1999; Fu et al . , 2000 ) , and indeed , an involvement of F-actin in gamete nuclear migration has been suggested during in vitro fertilization in rice ( Ohnishi et al . , 2014 ) . Here , we report that in contrast to animals , microtubules are dispensable for fertilization and F-actin is the primary factor controlling sperm cell nucleus migration in Arabidopsis . Upon sperm nucleus entry in the central cell , F-actin generates an actin aster-like structure around the sperm cell nucleus while it migrates towards the female nucleus . The F-actin dynamic movement towards the central cell nucleus is established prior to fertilization and is controlled by a central cell-specific Rho-GTPase and myosin . Additional observations of F-actin function during egg cell fertilization suggest similar conclusions . Hence , flowering plants innovated a novel actin-based mechanism of fertilization that made the centrosome dispensable .
To investigate the requirement and function of F-actin during fertilization , we expressed the fluorescent reporter Lifeact-Venus ( Riedl et al . , 2008; Era et al . , 2009 ) under the control of the egg cell-specific EC1 promoter ( Sprunck et al . , 2012 ) to visualize the actin cytoskeleton in the egg cell ( Figure 1A , B ) . Lifeact-Venus marked cables were disassembled after treatment with the actin polymerizing inhibitor Latrunculin A ( LatA; Figure 1C ) . Pharmacological analysis by applying inhibitor drugs is useful to dissect out the cytoskeleton function at the cellular level . However , treatment with actin polymerization inhibitors disrupts functions in all cells when applied to tissues such as ovules and thus prevents the analysis on specific cytoskeleton functions in a specific cell-type . To overcome this problem , the semi-dominant negative ACTIN transgene ( DN-ACTIN; Kato et al . , 2010 ) was introduced to disrupt F-actin specifically in female gametes . DN-ACTIN contains one amino acid substitution in the hydrophobic loop of Arabidopsis ACT8 , which causes instability and fragmentation of actin filaments , leading to incomplete yet strong disruption of actin cytoskeleton ( Kato et al . , 2010 ) . Consistent with the effect of DN-ACTIN reported previously , the filamentous structures shown in the wild-type ( WT ) egg cell became much shorter and generated aggregates in the egg cell expressing DN-ACTIN ( Figure 1D ) . In WT plants , fertilization leads to karyogamy followed by decondensation of the chromatin from the male nucleus ( Figure 1E; Ingouff et al . , 2007 ) . Egg cell fertilization initiates embryo development while the fusion of the other sperm cell with the central cell leads to endosperm development ( Figure 1A , F ) . By contrast , fertilization of the egg cell expressing DN-ACTIN failed as the sperm nucleus did not fuse with the egg cell nucleus and the sperm chromatin remained condensed ( Figure 1G; Line 1 , 35% defects in DN-ACTIN [n = 104] compared to 0% defects in WT [n = 98] ) . Karyogamy was prevented only in the egg cell expressing DN-ACTIN but not in the central cell , resulting in a seed containing endosperm without an embryo [Figure 1H; Line 1 , 27% defects in DN-ACTIN ( n = 110 ) compared to 0% defects in WT ( n = 389 ) ] . Taken together , these results suggest that actin cytoskeletons are required for egg cell fertilization . Consistently , other independent transgenic lines showed similar seed developmental arrest [Line 2 , 20% defects ( n = 125 ) ; Line 3 , 22% defects ( n = 114 ) ] . Not all ovules of DN-ACTIN expressing lines showed the fertilization defect , likely because a certain fraction of actin filaments was still functional . 10 . 7554/eLife . 04501 . 003Figure 1 . F-actin is required for egg cell fertilization . ( A ) Cartoon of Arabidopsis mature ovule . cc , central cell; cz , chalaza; ec , egg cell; mp , micropyle; sy , synergid . ( B–D ) Egg cell actin cables ( B ) become disassembled in LatA treatment ( C ) and in DN-ACTIN ( D ) . ( E and F ) Successful fertilization marked by decondensation of the sperm cell chromatin ( ssc , red ) into the egg cell nucleus ( dashed oval ) ( E ) , resulting in a normal embryo in WT ( F ) . ( G and H ) Egg cell expressing DN-ACTIN shows arrests in sperm cell nuclear migration ( G ) and embryo development ( H ) . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 003 We further investigated the requirement of F-actin during fertilization in the central cell that is about five times larger than the egg cell , thus enabling more detailed F-actin dynamics visualization . We expressed Lifeact-Venus under the control of the central cell-specific FWA promoter ( Kinoshita et al . , 2004 ) . F-actin in the WT mature central cell showed that F-actin is organized in three major distinct structures: ( i ) a ring-shaped actin network at the micropylar end of the central cell surrounding the synergid cells and the egg cell; ( ii ) long actin cables at the periphery of the central cell , extending from the chalazal region to the boundary with the egg cell; and ( iii ) a track formed of a bundle of actin cables , running from the micropylar actin ring to the central cell nucleus ( Figure 2A; Figure 2—figure supplement 1 ) . As expected , these structures disassembled in response to LatA treatment ( Figure 2—figure supplement 2 ) . During fertilization , the track of actin cables became associated with the migrating sperm cell nucleus in the central cell from the site of fusion between the sperm cell and the central cell to the central cell nucleus ( Figure 2B , C ) . Time-lapse image analysis further confirmed F-actin association to the sperm cell nucleus during its migration in the central cell ( Figure 3; Video 1 ) . Upon completion of karyogamy , the micropylar actin ring and bundles of F-actin associated with the sperm cell nucleus disassembled ( Figure 2D; Video 2 ) , suggesting that F-actin assembles transient dynamic structures dedicated for fertilization . The constitutive expression of DN-ACTIN in the central cell prevented fertilization [Figure 2—figure supplement 2; Line 1 , 30% defects in DN-ACTIN ( n = 120 ) compared to 0% defects in WT ( n = 153 ) ] and endosperm development [Figure 2—figure supplement 2; Line 1 , 25% defects in DN-ACTIN ( n = 412 ) compared to 0% defects in WT ( n = 233 ) ] . Other independent transgenic lines expressing DN-ACTIN in the central cell also showed similar rates of endosperm developmental defect ( Line 2 , 22% defects [n = 300]; Line 3 , 21% defects [n = 434] ) . Occasionally , a few nuclear divisions were observed in the endosperm ( 1–2% ) , resembling a phenotype described in ovules fertilized by cdka;1 sperm cells that are unable to undergo karyogamy ( Aw et al . , 2010 ) , thus supporting further that sperm cell fusion triggers central cell division . Taken together , these results suggest that F-actin also plays an important role in fertilization of the central cell . 10 . 7554/eLife . 04501 . 004Figure 2 . F-actin in the central cell associates with the sperm cell nucleus during migration . ( A–D ) Z-Stacked ( top ) and single plane ( bottom ) images of F-actin in the WT central cell during fertilization . The central cell F-actin before fertilization ( A ) , at the onset ( B ) , and completion ( C ) of nuclear migration , and karyogamy ( D ) were shown . White arrows and arrowheads point to sperm chromatins in the egg and central cells , respectively . White-dashed ovals display the position of the central cell nucleus . Purple arrowheads point to the F-actin ring structure present at the micropylar end of the central cell . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 00410 . 7554/eLife . 04501 . 005Figure 2—figure supplement 1 . Unique F-actin structures in the mature central cell before fertilization . F-actin structures of a ring-shaped network at the micropylar end of the central cell ( purple ) , long cables at the periphery of the central cell and extending from the chalazal region to the boundary with the egg cell ( cyan ) , and tracks running from the micropylar actin ring to the central cell nucleus ( red ) are marked on the duplicated image on the right . White-dashed oval displays the position of the central cell nucleus . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 00510 . 7554/eLife . 04501 . 006Figure 2—figure supplement 2 . Effects of LatA and DN-ACTIN on F-actin and fertilization in the central cell . Unlike LatA treatment , DN-ACTIN generates short fragmented F-actin cables and aggregates . Sperm cell nuclear migration becomes disrupted in the central cell ( arrow head ) , whereas sperm chromatin already becomes decondensed in the egg nucleus ( arrow ) , the sign of successful fertilization . As a result , no endosperm development is initiated , although the central cell becomes enlarged . ccn , central cell nucleus; ec , egg cell . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 00610 . 7554/eLife . 04501 . 007Figure 3 . Actin cables generate an aster-like structure around the sperm cell nucleus during migration in the central cell . ( A–H ) 1-min interval time-lapse image montage during sperm cell nuclear migration . After gamete fusion , the sperm cell nuclei ( red ovals ) start moving towards opposite directions ( A–C ) . In the central cell , actin cables ( green ) become associated with the sperm cell nucleus ( arrowhead ) after its entry ( D and E ) and assemble an aster-like structure around the sperm cell nucleus ( F–H ) . Inlet in panel G shows actin cables ( white ) around the sperm cell nucleus in the central cell without sperm chromatin fluorescence . See also Video 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 00710 . 7554/eLife . 04501 . 008Video 1 . Time-lapse image of F-actin dynamics and sperm cell nuclear migration in the WT central cell . Sperm cell chromatins ( red ovals ) are marked by pH3 . 10::H3 . 10-mRFP1 and F-actin in the central cell ( green ) is marked by pFWA::Lifeact-Venus . Autofluorescence marks the central cell outline ( red line ) . Images were taken at 30 s intervals . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 00810 . 7554/eLife . 04501 . 009Video 2 . Time-lapse image of F-actin micropylar ring-like structure in the central cell during fertilization . Sperm cell chromatins ( red ovals ) are marked by pH3 . 10::H3 . 10-mRFP1 and F-actin in the central cell ( green ) is marked by pFWA::Lifeact-Venus . Arrowheads point the micropylar F-actin ring in the central cell . Autofluorescence marks the central cell outline ( red line ) . Images were taken at 1-min intervals . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 009 Arabidopsis female gametes undergo a maturation phase once they become specified ( Drews and Koltunow , 2011 ) , and the FWA promoter is active soon after the central cell is determined . Therefore , we could not rule out the possibility that the constitutive expression of DN-ACTIN in the central cell by the FWA promoter might impair the maturation of the central cell , which in turn would be responsible for fertilization failure . To overcome this problem , we designed a dexamethasone ( DEX ) -inducible expression system ( Samalova et al . , 2005 ) to activate DN-ACTIN after central cell maturation ( Figure 4A , B ) . As the activator construct , the LhGR , constituted from the steroid-binding domain of a rat glucocorticoid receptor ( GR , residues 508–795 ) , the mutated DNA-binding lac repressor ( lacIHis17 , residues 1–330 ) , and yeast GAL4 transcription activation domain II ( Gal4-II , residues 768–881 ) , was flanked by the FWA promoter . The reporter construct consists of the nuclear-localized mCherry fluorescent protein ( NLS-mCherry ) and DN-ACTIN genes , connected by a self-cleaving 2A peptide derived from porcine teschovirus-1 ( P2A ) , and this transgene was flanked by the four tandem repeats of the lac operator with the cauliflower mosaic virus 35S minimal promoter ( TATA; Figure 4A ) . During translation , the P2A sequence causes ribosome skipping , generating two individual proteins from a single transgene ( El Amrani et al . , 2004; Kim et al . , 2011 ) . As a result , two proteins , NLS-mCherry and DN-ACTIN , are produced in the central cell by DEX application ( Figure 4B ) , and the NLS-mCherry can be used as a marker for spatio-temporal DEX induction . Consistent with the result from the constitutive expression of DN-ACTIN by the FWA promoter , DN-ACTIN induced by DEX in the mature central cell disrupted F-actin organization and prevented sperm cell nucleus migration towards the female nucleus and karyogamy ( Figure 4C , E ) . Endosperm development was also prevented in DEX-treated plants ( Figure 4D , F ) . These results confirmed that F-actin is essential for the migration of the sperm cell nucleus to the central cell nucleus . 10 . 7554/eLife . 04501 . 010Figure 4 . Spatio-temporal control of F-actin disruption demonstrates importance of F-actin for fertilization . ( A ) Schematic representations of the transcription activator ( top ) and reporter ( bottom ) constructs of DEX inducible expression strategy . The transcription activator LhGR is placed under the control of the FWA promoter . The reporter construct consists of NLS-mCherry and DN-ACTIN genes , connected by a self-cleaving 2A peptide derived from porcine teschovirus-1 ( P2A ) , flanked by the four lac operators with the CaMV 35S minimal promoter ( TATA ) . ( B ) Schematic representations of DN-ACTIN induction in the mature central cell by DEX . When DEX is applied , LhGR ( blue square ) localizes to the central cell nucleus . As a result , the reporter transgene becomes activated specifically in the central cell . During translation , the P2A peptide is recognized for the ribosomal skip , resulting in two separate proteins: NLS-mCherry and DN-ACTIN . NLS-mCherry proteins are transferred to the nucleus ( red nucleus ) , and DN-ACTIN ( purple ovals ) functions in the same cell for F-actin disruption . cc , central cell; eg , egg cell . ( C–F ) DN-ACTIN induction in the mature central cell disrupts nuclear migration and endosperm development . Mock treatment showing sperm chromatin decondensation ( cyan ) in both the egg and central cells ( C ) , leading to normal embryo and endosperm development ( D ) . Central cell DN-ACTIN induction marks specifically the central cell nucleus ( dashed oval ) and causes F-actin disruption ( E ) , preventing sperm cell nuclear migration ( arrowhead ) while karyogamy takes place in the egg cell ( arrow ) , resulting in embryo but no endosperm development ( F ) . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 010 We observed that , upon entry into the central cell , the sperm cell nucleus becomes surrounded by an increasing number of F-actin cables ( Figure 3A–D ) , followed by the formation of a dynamic aster-like structure around the sperm cell nucleus ( Figure 3E–H ) . This aster-like structure of F-actin was only observed around the sperm cell nucleus in the central cell after plasmogamy ( 100% , n = 48 ) and was never observed in unfertilized central cells ( 0% , n = 72 ) . Consistently , the time-lapse imaging showed the presence of aster-like F-actin structure only during sperm cell nucleus migration [100% ( n = 5 ) ; 0% in unfertilized ovule ( n = 14 ) ] . The F-actin aster-like structure migrated together with the sperm cell nucleus towards the central cell nucleus ( Video 1 ) , suggesting the involvement of specific actin regulators in this process . In plants , F-actin assembly is regulated by formins and ARP2/3 complex under the control of plant-specific Rho-GTPases ( ROP; Craddock et al . , 2012; Henty-Ridilla et al . , 2013 ) . Transcriptional profiling ( Borges et al . , 2008; Le et al . , 2010; Wuest et al . , 2010; Belmonte et al . , 2013 ) suggested that ROP8 is specifically active in the central cell and the endosperm ( Figure 5A ) . Indeed , we found that the ROP8 promoter is specifically active in the central cell ( Figure 5B ) and the fusion protein Venus-ROP8 associates to the plasma membrane of the central cell ( Figure 5C ) . Consistently , the ectopic expression of GFP-ROP8 also showed its association to the plasma membrane in root epidermal cells ( Figure 5—figure supplement 1 ) , indicating that ROP8 sub-localization is controlled by general machinery common between the gametophytic central cell and somatic root epidermal cell . To understand the function of ROP8 in the central cell during fertilization , we investigated F-actin organization in central cells expressing constitutively active and dominant-negative forms ( Yang , 2002 ) of ROP8 ( CA-ROP8 and DN-ROP8 , respectively ) under the control of the ROP8 promoter . Although F-actin structures in both CA-ROP8 and DN-ROP8 appeared similar to WT ( Figure 5—figure supplement 2 ) , only DN-ROP8 lines showed that the sperm cell chromatin in the central cell remained condensed and unfused with the central cell nucleus , when the sperm cell chromatin had already fused with the egg cell nucleus and appeared decondensed ( Figure 5D; Line 1 , 42% defects [n = 98] , Line 2 , 36% defects [n = 102] ) . This phenotype was similar to the fertilization defect observed in central cells expressing DN-ACTIN ( Figure 4 ) . 10 . 7554/eLife . 04501 . 011Figure 5 . ROP8 is specific to the central cell and involved in fertilization . ( A ) Arabidopsis ROP encoding genes expression heat map constructed from public data . CZ-END , chalazal endosperm; CZ-SC , chalazal seed coat; EP , embryo proper; G-SC , general seed coat; MP-END , micropylar endosperm; PE-END , peripheral endosperm . ( B and C ) H2B-Venus ( B ) and Venus-ROP8 ( C ) under the ROP8 promoter show expression specifically in the central cell and its association to the plasma membrane ( cyan ) , respectively . Autofluorescence marks the central cell outline ( red line ) . ( D ) DN-ROP8 shows central cell fertilization defect . Sperm chromatin in the egg cell ( arrow ) is already decondensed , but sperm chromatin remains condensed in the central cell ( arrowhead ) . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 01110 . 7554/eLife . 04501 . 012Figure 5—figure supplement 1 . Ectopic expression of ROP8 in the root also shows its association to the plasma membrane . sGFP:ROP8 expression under the control of the CaMV35S promoter in root epidermal cells . BF , bright field . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 01210 . 7554/eLife . 04501 . 013Figure 5—figure supplement 2 . The expression of neither constitutively-active ROP8 ( CA-ROP8 ) nor dominant-negative ROP8 ( DN-ROP8 ) under the control of the ROP8 promoter affect the F-actin structure ( green ) in the central cell . Autofluorescence marks the central cell outline ( red line ) . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 013 We measured the dynamics of F-actin around the central cell nucleus , where the sperm cell nucleus migration occurs , and observed a predominant F-actin movement from the cell periphery towards the nucleus ( Video 3 ) , hereafter referred to as inward movement ( Figure 6A , B ) . By contrast , F-actin movement was extremely reduced in central cells expressing DN-ROP8 ( Figure 6C–E; Videos 3; Video 4; Video 5; 45% and 30% defects in DN-ROP8 Line 1 [n = 11] and Line 2 [n = 10] , respectively , compared to 0% defect in WT [n = 7] and CA-ROP8 Line 1 [n = 5] and Line 2 [n = 6] ) . Taken together , these results show that ROP8 controls novel constant inward dynamics of F-actin movement in the central cell and that this ROP8-dependent inward F-actin movement is required for the proper transport of the sperm cell nucleus to the central cell nucleus . Although F-actin flows towards the central cell nucleus , we did not detect accumulation of F-actin around the nucleus ( Video 3 ) , suggesting that F-actin disassembly takes place around the central cell nucleus . 10 . 7554/eLife . 04501 . 014Video 3 . Time-lapse image of F-actin dynamics in the mature WT central cell . F-actin in the central cell ( white ) were marked by pFWA::Lifeact-Venus . Images were taken at 30-s intervals . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 01410 . 7554/eLife . 04501 . 015Figure 6 . Controls of F-actin dynamics in the central cell . ( A ) A scheme of the mature Arabidopsis ovule . F-actin ( light green cables ) velocities of the entire central cell and around the nucleus ( dashed box ) were analyzed . ccn , central cell nucleus; cz , chalaza; ec/sy , egg/synergid cells complex; mp , micropyle . ( B ) F-actin velocity around the central cell nucleus . Inward and outward represent F-actin movement direction towards and away from the nucleus , respectively . a . u . , arbitral unit . ( C ) The representative probability distributions of velocities integrated over the entire central cell area and time . Solid lines represent the fitted gamma distribution . ( D ) The mean velocities of central cell F-actin movement . Error bars represent the standard deviation of five to seven biological replicates . n . s . , not significant; *** , p < 0 . 001; ** , p < 0 . 01 , student t-test . ( E ) Stacks of central cell F-actin time-lapse . F-actin movements are marked by rainbow colors while white color results from overlapping multiple colors , indicating less movement . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 01510 . 7554/eLife . 04501 . 016Video 4 . Time-lapse image of F-actin dynamics in the mature CA-ROP8 central cell . F-actin in the central cell ( white ) was marked by pFWA::Lifeact-Venus . Images were taken at 30-s intervals . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 01610 . 7554/eLife . 04501 . 017Video 5 . Time-lapse image of F-actin dynamics in the mature DN-ROP8 central cell . F-actin in the central cell ( white ) is marked by pFWA::Lifeact-Venus . Images were taken at 30-s intervals . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 017 In plants , the motor protein myosin generates force to transport organelles along actin filaments within a cell ( Madison and Nebenfuhr , 2013 ) , including the nuclear movement in somatic cells ( Tamura et al . , 2013 ) . To investigate an involvement of myosin , we applied the myosin inhibitors 2 , 3-butanedione monoxime ( BDM ) and N-ethylmaleimide ( NEM; Hoffmann and Nebenfuhr , 2004 ) to dissected ovules containing the mature central cell . In the presence of either myosin inhibitor , F-actin movement significantly reduced as observed in central cells expressingDN-ROP8 ( Figure 6C–E; Videos 6; Video 7; 100% defect with BDM [n = 11] and NEM [n = 7] , compared to 0% defect in mock treated ovules [n = 7 and 5 , respectively] ) . These results indicate that myosin plays a role in the dynamics of F-actin in the central cell and contributes to a net inward F-actin movement . 10 . 7554/eLife . 04501 . 018Video 6 . Time-lapse images of F-actin dynamics in the mature WT central cell in the presence of the myosin inhibitor BDM ( 50 mM ) . F-actin in the central cell ( white ) is marked by pFWA::Lifeact-Venus . Images were taken at 30-s intervals . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 01810 . 7554/eLife . 04501 . 019Video 7 . Time-lapse images of F-actin dynamics in the mature WT central cell in the presence of the myosin inhibitor NEM ( 0 . 35 mM ) . F-actin in the central cell ( white ) is marked by pFWA::Lifeact-Venus . Images were taken at 30-s intervals . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 019 Although F-actin appears to be the main factor controlling male gamete nuclear migration , it is still possible that microtubules are also involved in male gamete nuclear migration in flowering plants even in the absence of a centrosome . To investigate the requirement of microtubules in Arabidopsis fertilization , we used a null mutant allele of PORCINO ( POR ) that encodes a subunit of TUBULIN FOLDING COFACTOR C ( Steinborn et al . , 2002 ) . Self-pollinated por/+ heterozygous plants produce about 25% defective embryos that do not assemble microtubule and abort , indicating that viable por gametes fuse to produce por/por homozygous seeds that are not viable ( Steinborn et al . , 2002 ) . We expressed the TagRFP-TUBULIN ALPHA-5 in the central cell to visualize microtubules in por/+ plants that produce 50% por and 50% WT gametes . In contrast to WT , about half of the female gametes from por/+ mutants did not show microtubule structures ( Figure 7A , B , F ) . This proportion was consistent with the requirement of POR function for microtubule assembly . WT and por mutant female gametes fertilized with either WT or por sperm cells showed successful karyogamy leading to the decondensation of chromatin from the sperm cell nucleus ( Figure 7C , D; 97% in WT [n = 102] and 96% in por/+ [n = 96] ) . Taken together , these results show that fertilization takes place despite defective microtubule assembly and as a result , por/por embryo and endosperm development initiates . Mutations in genes encoding the other subunits ( A , D and E ) of TUBULIN FOLDING COFACTOR cause a seed abortion phenotype similar to that described in por ( Steinborn et al . , 2002 ) , further demonstrating that microtubules are not absolutely required for fertilization in Arabidopsis . Additionally , we investigated whether F-actin and microtubules interact to coordinate nuclear migration . However , microtubule organization was not affected in central cells expressing DN-ACTIN . Conversely in por central cells , the actin cytoskeleton retained its organization similar to WT ( Figure 7E , F ) . These observations support the idea that the actin cytoskeleton is the primary factor controlling sperm cell nuclear migration during fertilization in Arabidopsis in contrast to most animal species in which microtubules play a major role in pronuclear migration . 10 . 7554/eLife . 04501 . 020Figure 7 . Effect of microtubule disruption on cytoskeleton in the central cell . ( A and B ) Microtubule structures ( green ) in the mature WT ( A ) and the tubulin folding cofactors c ( por ) /+ heterozygous mutant ( B ) central cells marked by TagRFP-TUA5 under the control of the AGL80 promoter . ( C and D ) In por/+ plants , fertilization is successful in all ovules even though they do not display organized microtubules ( D ) like WT ( C ) . Arrows and arrowheads point to the sperm cell chromatin ( red ) , already decondensed in the egg cell and central cell nucleus , respectively . ( E ) F-actin in por/+ heterozygous mutant visualized by the Lifeact-Venus under the control of the FWA promoter . ( F ) Percentages of ovules that show cytoskeleton structures in the mature central cell of different genetic backgrounds . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 04501 . 020
Nuclear movement in both gamete and somatic cells has been extensively studied in animals . Pronuclear migration in the fertilized egg is microtubule dependent , but both actin filament and microtubules are indispensable for nuclear positioning in many somatic cells ( Reinsch and Gonczy , 1998; Gundersen and Worman , 2013 ) . In plants , nuclear positioning during cell division and photo-relocation as well as tip growth of root hair and pollen tube has been investigated ( Takagi et al . , 2011; Higa et al . , 2014 ) , and actin filaments play a critical role in these nuclear positioning events in plant somatic cells . By contrast , very little was known about gamete nuclear migration in plants . During evolution , the centrosome loss in spermatogenous cells of land plants did not coincide with either the centrosome loss in somatic cells or the initiation of the actin-based organelle movement system in plants , indicating that gamete nuclear migration in flowering plants might depend on a distinct mechanism . Here , we identified that the dynamic F-actin movement towards the central cell nucleus from the cell periphery constantly occurs prior to the sperm cell nucleus entry , and this F-actin movement is essential for sperm cell nucleus migration in Arabidopsis . Upon entry into the central cell , the sperm nucleus becomes surrounded by F-actin , generating an aster-like structure that migrates together with the sperm cell nucleus to join the central cell nucleus for karyogamy . We also found that this dynamic F-actin movement in the central cell is controlled by gamete-specific Rho-GTPase ROP8 and is myosin dependent . On the contrary , microtubules are dispensable for sperm cell nuclear migration in both female gametes . Taken together , our results show that the sperm cell nuclear migration in flowering plants is controlled by a novel actin-based mechanism that generates a constant F-actin inward movement towards the central cell nucleus . The dynamic movement of F-actin towards the central nucleus , F-actin aster-like structure around the sperm cell nucleus during migration , and precise spatio-temporal disruption of F-actin after central cell maturation by the DEX system suggest direct involvement of F-actin for sperm cell nuclear migration . However , it still remains to be determined whether F-actin directly interacts with the sperm cell nucleus for migration . Further investigations should be made to rule out the possibility that the sperm cell nuclear migration defect by DN-ACTIN expression is not due to general perturbation of cellular structure . We also describe that F-actin is organized as unique structures in the central cell . In addition to the F-actin aster-like structure and the F-actin track that associate with the sperm cell nucleus migration , the central cell features a ring-shaped actin network at the micropylar end that surrounds the egg apparatus ( the synergid cells and the egg cell ) . The F-actin ring might maintain the position of the egg apparatus for fertilization or it might act as a clamp to maintain mechanical constraints that are required for efficient sperm cell delivery . Upon completion of karyogamy , all three organized F-actin structures become disassembled , suggesting that these unique actin structures are specialized to achieve fertilization . Rho-GTPases are well-studied signaling factors controlling cytoskeleton dynamics ( Yang , 2002 ) . In the WT central cell , the constant inward movement of F-actin from the cell periphery to the nucleus occurs prior to fertilization , and this constant inward movement becomes significantly reduced in DN-ROP8 , but not in CA-ROP8 transgenic lines . These results suggest that ROP8 might be constantly active in the entire central cell plasma membrane , generating a constitutive dynamic inward movement of F-actin . This could explain why DN-ROP8 , but not CA-ROP8 transgenic lines affected F-actin dynamics significantly ( Figure 6 ) . In general , the activation/inactivation of ROPs is controlled spatially at the plasma membrane , generating subdomains within a cell for downstream events such as secondary cell wall formation ( Oda and Fukuda , 2012 ) and pavement cell indentation ( Xu et al . , 2010 ) . Although the machinery by which ROP8 is associated to the plasma membrane is general and the activity of ROP8 in the central cell should be investigated further , this putative constant active mode of ROP8 in the central cell might shed light on a new facet of Rho-GTPase function in plant cells . F-actin function is also important for sperm cell nuclear migration in the Arabidopsis egg cell described here and in the rice egg cell shown by in vitro fertilization assay ( Ohnishi et al . , 2014 ) . It is plausible that a distinct ROP plays a similar role during egg cell fertilization , but it is also possible that other mechanisms control F-actin-dependent migration of the male nucleus to the egg cell nucleus . According to observations in somatic cells ( Tamura et al . , 2013 ) , myosin was expected to tether the nucleus and enable its migration as a cargo . However , the impairment of F-actin movement in the central cell halts sperm cell nuclear migration even in the presence of the WT-like structure of actin filaments shown in DN-ROP8 . Hence , myosin might not transport the sperm cell nucleus towards the female nucleus as a cargo . Our observations rather suggest that myosin affects F-actin dynamics directly . In vitro , myosin also creates bipolar filaments and cross-links two actin filaments , generating a non equilibrium dynamic F-actin movement ( Mizuno et al . , 2007 ) . This type of dynamics is similar to the inward F-actin movement observed in the central cell , suggesting that myosin is directly involved in the dynamic inward movement of F-actin in this cell type . Yet , we have not identified which specific myosin is involved and detailed myosin function in sperm cell nuclear migration remains to be determined . Along with the development of actin-based cellular dynamics in plants , the centrosome loss in somatic cells was established in bryophytes during land plant evolution ( Vaughn and Harper , 1998 ) . The diversification and expansion of actin nucleator formin genes are already evident in bryophytes ( Grunt et al . , 2008 ) ; it is still not yet clear how ROP signaling pathway genes in plants initiated and diversified during land plant evolution . However , several higher land plant features such as pollen tube tip growth ( Guan et al . , 2013 ) , secondary cell wall formation ( Oda and Fukuda , 2012 ) , pavement cell indentation ( Xu et al . , 2010 ) , and actin-based sperm cell nuclear migration described here are controlled by specific ROP signaling pathways . These observations suggest that extended roles of actin cytoskeleton at least by ROP signaling pathways through land plant evolution might have paved the way to the complete loss of microtubule–centrosome-dependent processes in flowering plants together with a dramatic impact on plant cell biology including reproduction mechanisms , cell divisions and cell structures .
All Arabidopsis plant lines used in this study are Columbia-0 ( Col ) ecotype , except the porcino ( por ) mutant in Landsberg erecta ( Ler ) . Seeds were first germinated under short day conditions at 16°C ( 8 hr light and 16 hr dark ) and kept for 1–2 weeks . Plants were then shifted to long day conditions at 20°C ( 16 hr light and 8 hr dark ) . The pH3 . 10::H3 . 10-mRFP1 ( Ingouff et al . , 2007 ) and pEC1::H2B-mRFP1 ( Ingouff et al . , 2009 ) lines have been described previously . The p35S::sGFP-ROP8 and pAGL80::tagRFP-TUA5 transgenic lines are gifts from Dr Zhenbiao Yang at UC Riverside , USA and Dr Shuh-ichi Nishikawa at Niigata University , Japan , respectively . All fragments were amplified by PCR using the KOD-plus ver . 2 PCR kit ( TOYOBO , Japan ) . Primer sequences for PCR are listed in Supplementary file 1A and all constructs ( Supplementary file 1B ) were generated by the Multisite Gateway Technology ( Invitrogen , CA , USA ) . The Lifeact-Venus , the omega translational enhancer Ω ( Gallie , 2002 ) , Ω-Venus , Ω-H2B , the dominant-negative form of ACTIN 8 ( DN-ACTIN ) , the constitutively-active and dominant-negative forms of ROP8 ( CA-ROP8 and DN-ROP8;Yang , 2002 ) , NLS-mCherry , and the ribosomal skip peptide ( El Amrani et al . , 2004; Kim et al . , 2011 ) P2A-DN-ACTIN fragments were generated by Overlap Extension PCR ( Horton , 1993 ) . The multisite gateway binary vector pAlligatorG43 was generated by replacing the EN35S expression cassette of pAlligator2 ( Bensmihen et al . , 2004 ) with the multisite gateway cassette . The GFP selection marker gene of pAlligatorG43 was replaced with mCherry to generate pAlligatorR43 . Other multiple gateway vectors , pK7m24 GW , 3 , pH7m24 GW , 3 , pB7m24 GW , 3 , have been described previously ( Karimi et al . , 2007 ) . All constructs were transformed into Col except por mutant ( Ler ) using the floral dip method ( Clough and Bent , 1998 ) . Fluorescence was acquired with laser-scanning confocal microscopy ( 63×/1 . 3 glycerol immersion objective lens; Leica SP5 , Germany ) for Venus ( excitation 514 nm and emission BP 520–585 nm ) and for mRFP1 and mCherry ( excitation 594 nm and emission BP 600–650 nm ) , and a microscope ( 40×/1 . 30 and 60×/1 . 40 oil immersion objective lens; binning of 2; Nikon Ti-E , Japan ) equipped with a disk-scan confocal system ( Yokogawa CSU-X1 , Japan ) and CCD and CMOS camera ( Photometrics CoolSNAP-HQ2 , AZ , USA; Hamamatsu Photonics ORCA-Flash4 . 0 , Japan ) . Time-lapse images were acquired every 0 . 5–1 min using multiple z-planes ( 7–10 planes , 1–1 . 5 µm intervals , objective piezo drive system ) with the spinning disk confocal microscopy . Digital image processing was performed with LAS AF Lite ( Leica , Germany ) , Metamorph ( Molecular Devices , CA , USA ) , ImageJ ( http://rsbweb . nih . gov/ij/ ) , and/or Adobe Photoshop ( Adobe Systems , CA , USA ) . The experimental strategy was modified from previously described ( Samalova et al . , 2005 ) . Mature emasculated pistils ( 36 hr after emasculation ) , harboring pFWA::Lifeact-Venus , pFWA::LhGR , and pOPx4::NLS-mCherry-P2A:DN-ACTIN , were submerged into the DEX working solution ( DEX [D1756; Sigma-Aldrich , MO , USA] , 40 µM; Silwet L-77 [VIS-02; LEHLE Seeds , TX , USA] , 0 . 0001% in ddH2O ) . Pollens , harbouring pH3 . 10::H3 . 10:Clover , were pollinated 12 hr after the DEX induction . 8–10 hr after pollination , ovules were dissected and fertilization phenotypes were observed using confocal microscopy . The method was modified from previously described ( Hamamura et al . , 2011 ) . In brief , pollen growth medium ( 14% sucrose , 0 . 001% boric acid , 1 . 27 mM Ca ( NO3 ) 2 , 0 . 4 mM MgSO4 , 1 . 5% low gelling temperature agarose [A9414; Simga-Aldrich , MO , USA] , pH 7 . 0 with KOH ) was poured into a well of a glass bottom dish ( D141410; Matsunami Glass IND . , LTD . , Japan ) to make a thin layer . Stigmas from emasculated pistils were cut with a 27-gauge needle ( Terumo , Japan ) at the junction between the style and ovary and placed horizontally on the pollen growth medium , followed by hand-pollination with pollen from the pH3 . 10::H3 . 10-mRFP1 . After 3-hr incubation at 22°C in the dark , fresh ovules were dissected from the pFWA::Lifeact-Venus and were placed in front of growing pollen tubes on the medium and incubated further 2 hr . The spinning disk confocal microscopy was used to generate time-lapse images of double fertilization . Particle image velocimetry ( PIV ) of the F-actin dynamics was performed on MATLAB using the PIVlab package ( version 1 . 32 ) developed by William Thielicke and Eize J Stamhuis . PIV analysis was performed on the central cells as follows . We first defined the region of interest ( ROI ) and overlaid a mask representing the shape of the central cell using the first image of the Video to restrict analysis only to the cell . Sizes of interrogation areas were optimized in order to calculate frame to frame displacement of F-actin . Higher spatial resolution was achieved by employing multi-pass window deformation technique , with a final size of 16 × 16 pixels ( 1 . 7 × 1 . 7 μm ) and 50% overlap . Once the velocity vector field was obtained , we then analyzed the vectors in two different ways . To analyze the F-actin movement towards or away from the nucleus , we restricted our analysis to around the nucleus . The component of velocity vectors around the nucleus in the direction of the center of nucleus was calculated and categorized into inward ( towards the nucleus ) and outward ( away from the nucleus ) velocity . The probability distribution of inward and outward velocities integrated over space and time was plotted as shown in Figure 6B . To compare the average F-actin dynamics in the central cell of different genetic backgrounds and in the presence of myosin inhibitor BDM or NEM , the mean velocity of F-actin was measured . The probability distributions of velocities integrated over the entire central cell area and time were plotted as Figure 6C . The probability distribution of the velocity f ( v ) is then fitted by the gamma distribution ( solid line in Figure 6C ) . f ( x;k , θ ) =xk−1e−xθΓ ( k ) θk;x≥0 and k , θ>0 Where k is a shape parameter , θ is a scale parameter , and Γ ( k ) is the gamma function evaluated at k . Γ ( k ) =∫0∞xk−1e−xdx The mean velocity of the probability distribution is represented by kθ . The summary of the mean velocity of different genetic backgrounds and in the presence of myosin inhibitor BDM or NEM was shown in Figure 6D . The error bars shown in Figure 6D are SEM . The p-values are calculated using student's t-test . Ovules 2 days after emasculation were dissected and submerged into the assay medium ( 5% sucrose , half strength Murashige and Skoog salt , pH 5 . 7 ) in a glass bottom dish ( D141410; Matsunami Glass IND . LTD . , Japan ) . LatA ( L5163; Sigma-Aldrich , MO , USA; 10 mM in DMSO ) , BDM ( B0753; Sigma-Aldrich , MO , USA; 500 mM in the assay buffer ) , and NEM ( E1271; Sigma-Aldrich , MO , USA; 500 mM in ethanol ) were supplemented to generate fresh 100 µM LatA , 50 mM BDM , and 0 . 35 mM NEM assay media , respectively . The spinning disk confocal microscopy was used to generate time-lapse images of F-actin dynamics . Affymetrix Arabidopsis ATH1 22k GeneChip raw data of the female gametophyte cells ( Wuest et al . , 2010 ) , young seed compartments ( Belmonte et al . , 2013 ) , adult tissues ( Le et al . , 2010 ) , and male gametophyte cells ( Borges et al . , 2008 ) were obtained from ArrayExpress in EBI ( https://www . ebi . ac . uk/arrayexpress/ ) and Gene Expression Omnibus in NCBI ( http://www . ncbi . nlm . nih . gov/geo/ ) websites . The raw data were processed and normalized using GCRMA in R . Hierarchical clustering analysis on ROP genes expression was carried out using CLC Main Workbench 7 . 02 ( CLC bio , Denmark ) . | Sexual reproduction involves combining the genetic material from two parents to create an offspring . The genetic material in the male sperm cell and the female egg cell is contained in the nucleus of each cell . Once these two cells fuse at fertilization , their nuclei must then navigate towards each other and fuse . When an animal egg cell is fertilized , cable-like protein filaments called microtubules guide the two nuclei into contact . These microtubules are organized by a cellular structure called a centrosome . However , flowering plants do not have centrosomes; as such , it was unclear how genetic material from the sperm and egg cells is brought together after fertilization in flowering plants . To investigate this , Kawashima et al . turned to a flowering plant commonly used in research , called Arabidopsis thaliana , and found that microtubules are not needed to guide the nuclei of the sperm and the egg cell after fertilization . Instead , another cable-forming protein—called F-actin—fulfills a similar role in Arabidopsis cells . F-actin filaments often connect together to form a network; and when Kawashima et al . disrupted the F-actin in Arabidopsis egg cells , the nucleus of the sperm cell failed to fuse with that of the female . Pollen from Arabidopsis plants actually contains two sperm cells . One sperm cell fertilizes the egg cell; the other fertilizes the so-called ‘central cell’ , which develops into a tissue that nourishes the plant embryo . Kawashima et al . found that the fertilization of both of these cells requires an intact F-actin network . By looking more closely at F-actin networks in the larger central cell , Kawashima et al . discovered that the sperm nucleus becomes surrounded by a star-shaped structure of F-actin cables and that this F-actin structure migrates together with the sperm nucleus . The F-actin network constantly moves inward , from the edges of the cell towards the nucleus , prior to fertilization . This movement is essential for guiding the sperm nucleus towards the central cell nucleus . Kawashima et al . also found that this continual movement of the F-actin network depends on a small signaling protein found in the central cell , called ROP8 . It also involves a motor protein that normally transports “cargo” , such as proteins and other molecules , inside cells by walking along the F-actin networks . However , rather than transporting the sperm nucleus as cargo , Kawashima et al . believe that the motor protein instead helps to maintain the inward movement of the F-actin network . One of the next challenges will be to investigate the molecular mechanism that underlies this motor protein's involvement in this dynamic F-actin network . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology"
] | 2014 | Dynamic F-actin movement is essential for fertilization in Arabidopsis thaliana |
Natural selection of variants within the Arabidopsis thaliana circadian clock can be attributed to adaptation to varying environments . To define a basis for such variation , we examined clock speed in a reporter-modified Bay-0 x Shakdara recombinant inbred line and localized heritable variation . Extensive variation led us to identify EARLY FLOWERING3 ( ELF3 ) as a major quantitative trait locus ( QTL ) . The causal nucleotide polymorphism caused a short-period phenotype under light and severely dampened rhythm generation in darkness , and entrainment alterations resulted . We found that ELF3-Sha protein failed to properly localize to the nucleus , and its ability to accumulate in darkness was compromised . Evidence was provided that the ELF3-Sha allele originated in Central Asia . Collectively , we showed that ELF3 protein plays a vital role in defining its light-repressor action in the circadian clock and that its functional abilities are largely dependent on its cellular localization .
The rotation of the earth leads to environmental changes in ambient light and temperature that define a 24-hr cycle . During these diurnal cycles , light is correlated with warmth , while darkness is correlated with coolness . Thus , to ensure fitness in response to such regular environmental oscillations , many organisms have evolved an internal timing mechanism to predict these cyclic environmental cues , and this is termed the circadian clock . This internal clock synchronizes major developmental and physiological processes . For example , a considerable proportion of metabolism is under clock control ( Graf et al . , 2010; Haydon et al . , 2011; Sanchez et al . , 2011; Sanchez-Villarreal et al . , 2013 ) , and seasonal developmental programs require a circadian oscillator ( Thines and Harmon , 2011 ) . The coordination of molecular and physiological processes with the external environment thus provides an adaptive advantage under diverse climatic conditions ( Dodd et al . , 2005 ) . Circadian rhythms are produced by molecular oscillators , which are comprised of interlocking feedback loops having several components ( Bujdoso and Davis , 2013 ) . In Arabidopsis , the central loop consists of two closely related single-myb transcription factors , CIRCADIAN CLOCK ASSOCIATED 1 ( CCA1 ) and LATE ELONGATED HYPOCOTYL ( LHY ) , as well as the pseudo response regulator ( PRR ) TIMING OF CAB EXPRESSION 1 ( TOC1 ) . In the morning , CCA1 and LHY repress TOC1 by directly binding to its promoter , resulting in the evening accumulation of TOC1 , which , in turn , represses CCA1 and LHY expression ( Gendron et al . , 2012; Huang et al . , 2012 ) . The core oscillator is further tuned by an evening- and morning-phased loop . The evening loop is proposed to include GIGANTEA ( GI ) and TOC1 , where GI activates the expression of TOC1 , and TOC1 represses GI ( Locke et al . , 2006; Zeilinger et al . , 2006; Huang et al . , 2012 ) . Two members of the pseudo response regulator gene family , PRR7 and PRR9 , repress CCA1 and LHY during the day , and this establishes the morning loop ( Locke et al . , 2006; Zeilinger et al . , 2006 ) . Recently , an evening complex comprising EARLY FLOWERING 3 ( ELF3 ) , EARLY FLOWERING 4 ( ELF4 ) , and LUX ARRYTHMO ( LUX ) has been found to repress the morning loop by specifically binding to the PRR9 promoter to mediate transcription repression ( Nozue et al . , 2007; Kolmos et al . , 2009; Dixon et al . , 2011; Helfer et al . , 2011; Nusinow et al . , 2011; Chow et al . , 2012; Herrero and Davis , 2012; Herrero et al . , 2012 ) . Ambient light and temperature are two important environmental factors , termed zeitgebers , which reset the clock in a process referred to as entrainment ( McClung and Davis , 2010; McClung , 2011 ) . PRR7 and PRR9 have been reported to play a role in temperature compensation , which is the resistance of period change under differing mean ambient temperatures ( Salomé and McClung , 2005; Salome et al . , 2010 ) . However , diurnal temperature regulation of the circadian clock is still poorly understood ( McClung and Davis , 2010 ) . In contrast , light has been established as a key factor for the entrainment of the clock where continuous irradiation shortens free-running period length in a fluence-rate-dependent manner as a process termed parametric entrainment ( Aschoff , 1979; Covington et al . , 2001; Johnson et al . , 2003 ) . Most genetic components of the clock were originally discovered from a light-entrained oscillator . Several studies have shown that many observable rhythms in plants dampen in constant darkness ( DD ) , where dampening is defined as a reduction in circadian amplitude . For example , many genes have been associated with circadian rhythms of mRNA abundance that are expressed robustly under constant light ( LL ) , but dampen in DD ( Watillon et al . , 1993; Zhong et al . , 1997 ) . Furthermore , the cyclic protein accumulation of several clock-regulated components , such as GI and ELF3 ( Liu et al . , 2001; David et al . , 2006 ) , was found to be reduced when shifted to darkness . Thus , the effects of various light-input signals on the Arabidopsis core oscillator remain to be elucidated . Molecular and genetic studies support the involvement of ELF3 in light gating to the circadian clock , and ELF3 is also a required component of the core oscillator ( Covington et al . , 2001; Thines and Harmon , 2010; Kolmos et al . , 2011; Herrero et al . , 2012 ) . The elf3 mutant was originally isolated in a screen for lines displaying photoperiod-independent early flowering ( Zagotta et al . , 1992 , 1996 ) . Further characterization of elf3 revealed other severe phenotypes , including defects in clock-regulated leaf movement rhythms , rhythmic hypocotyl elongation , and arrhythmic expression of gene expression under the free running condition of LL and in DD ( Hicks et al . , 1996; Reed et al . , 2000; Thines and Harmon , 2010; Kolmos et al . , 2011; Herrero et al . , 2012 ) . Natural variation at ELF3 has been detected that could be associated to an alteration in shade-avoidance responses , which included increased elongation of the hypocotyl , delay of cotyledon opening in seedlings , increased elongation of stems and petioles , and reduced developmental timing in adults , and to changes in circadian function ( Tajima et al . , 2007; Jimenez-Gomez et al . , 2010; Coluccio et al . , 2011; Undurraga et al . , 2012 ) . ELF3 was cloned and reported to encode a nuclear protein of unknown function , and it was proposed to work as a transcription factor ( Hicks et al . , 2001; Liu et al . , 2001 ) . Based on the accumulation of ELF3 protein upon shifting plants to LL , and its decrease when plants were shifted to DD , it was concluded that its abundance is dependent on ambient light ( Liu et al . , 2001 ) . To extend this hypothesis based on the reported light-dependent phenotypes of elf3 , McWatters et al . ( 2000 ) proposed that ELF3 fulfills the so-called zeitnehmer concept , in that this factor acts to bridge light perception to the clock . This idea was further strengthened when ELF3 was found to physically interact with the phytochromeB photoreceptor ( Liu et al . , 2001 ) . The previously noted role of ELF3 in light input to the circadian clock ( Kolmos et al . , 2011 ) is not altered by new findings that ELF3 is core to the oscillator . Many elements in the clock were isolated in screens for rhythm mutants using bioluminescence readout of the molecular oscillator . Worldwide , the species distribution of Arabidopsis ranges from the equator to extreme northern latitudes near the Arctic Circle ( Koornneef et al . , 2004 ) . Therefore , as an alternative to induced mutants , another source of genetic variation can be found among naturally occurring populations of Arabidopsis . More specifically , environmental variations across and within local populations of Arabidopsis act as a discriminatory force on the gene pool from which only a few genetic variants of adaptive phenotypes will be selected and pass through reproduction . Arabidopsis populations based on Recombinant Inbred Lines ( RILs ) have been derived from parental accessions . Such populations are advantageous in mapping novel gene interactions . For example , quantitative trait locus ( QTL ) mapping was used to explain circadian parameters associated with natural variation in the circadian rhythmicity of leaf movement ( Swarup et al . , 1999; Michael et al . , 2003; Edwards et al . , 2005; Anwer and Davis , 2013 ) . More recently , the rhythm of bioluminescence from modified firefly LUCIFERASE ( LUC ) genes coupled to the clock-controlled promoter was successfully used to map QTLs for circadian parameters ( Darrah et al . , 2006; Boikoglou et al . , 2011 ) . Thus , the luciferase-based system can be employed to accurately measure variation in circadian-rhythm parameters within RIL populations to detect phenotypically expressed genetic variation in circadian clock genes ( Anwer and Davis , 2013 ) . In this study , we characterized a natural allele of ELF3 to determine its effect on rhythmicity in the Arabidopsis circadian clock . Classical QTL mapping in a modified Bay-0 x Shakdara mapping population , followed by positional cloning , revealed an allele of ELF3 ( ELF3-Sha ) that accelerates circadian oscillations in a light-dependent manner and a dampened oscillator in darkness . We determined that the periodicity phenotype of ELF3-Sha results from a single encoded amino-acid change A362V , which is an allelic variant that is largely confined to a latitudinal stripe in Central Asia . Furthermore , we identified that the circadian abnormalities in ELF3-Sha are associated to cellular localization defects of ELF3 protein that results in an inability to properly function under light and in darkness . Thus , by characterizing ELF3-Sha , we clarified the molecular mode-of-action of ELF3 in the circadian clock .
A total of 71 RILs derived from the genetic cross of Bayreuth-0 and Shakdara ( BxS ) were modified to harbor the CCR2 promoter , also termed GRP7 ( Loudet et al . , 2002 ) , fused to luciferase ( CCR2::LUC ) . These lines were synchronized to respective photic and thermal entrainment , and free-running periodicity was assessed under identical conditions of constant light and constant temperature . After applying both entrainment protocols , extensive variation in periodicity was observed ( Figure 1A; Supplementary file 1 , Supplementary file 2 ) ; however , the two parental ecotypes displayed similar periodicity , irrespective of preceding entrainment . Compared to photic entrainment , a significant shift for shortened free-running periodicity of CCR2 after thermal entrainment was observed for the parental lines , as well as RILs ( Figure 1A–C; Table 1 ) . The periodicity differences after respective photic vs thermal entrainment ( Period Light Dark ( LD ) –Period TMP [TMP] ) extended from minus 0 . 56 hr to plus 1 . 66 hr ( Figure 1B , C; Table 1 ) , with parental lines displaying periodicity differences of plus 1 . 84 and plus 2 . 15 hr for Sha and Bay-0 , respectively ( Boikoglou et al . , 2011 ) . The averaged periodicity differences of BxS lines were moderate ( 0 . 433 , p<0 . 001 ) ( Table 1 ) . Thus , despite the transgressive variation observed in the RILs for CCR2 periodicity and periodicity differences ( LD-TMP ) , the averaged periodicity differences seen in the RILs were similar in magnitude as the periodicity differences seen in the parental lines . 10 . 7554/eLife . 02206 . 003Figure 1 . Illustrative and statistical features of CCR2 periodicity post-photic vs post-thermal entrainment in BxS populations . ( A ) Normal frequency distribution of CCR2 periodicity in BxS individuals . Blue-colored bars represent periodicity after photic entrainment , and magenta-colored bars represent periodicity after thermal entrainment . Bay-0 and Sha denote the periodicity of CCR2 in the parental genotypes . Note the skew of temperature-entrained plants to shorter periodicity , when compared to photic-entrained plants . ( B ) Periodicity differences of CCR2 in BxS RIL lines . The x-axis denotes 54 RILs of which periodicity was assayed in both in LD and TMP entrainment ( see Supplementary file 1 and Supplementary file 2 ) . The y-axis represents the periodicity differences of thermal minus photic-entrained lines ( calculated as PeriodLD–PeriodTMP ) . ( C ) A scatter plot for TMP vs LD periodicities from the BxS RILs described in the tables Supplementary file 1 and Supplementary file 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 00310 . 7554/eLife . 02206 . 004Table 1 . Circadian periodicity analysis of the BxS RILs after photic or thermal entrainmentDOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 004ZeitgeberMean*S . E95% confidence intervalLower boundUpper boundBxSLD26 . 7020 . 03526 . 63326 . 771TMP26 . 2690 . 03426 . 20226 . 335LD-TMPBxSPairwise comparisonsMean difference ( hr ) 0 . 433S . E0 . 047Significance<0 . 001CorrelationsCorrelation coefficient0 . 86Significance0 . 001LD stands for photic-zeitgeber and TMP for thermal-zeitgeber . *denotes the modified population marginal mean for the 95% Confidence Interval . S . E . denotes standard error . LD-TMP denotes the pairwise comparisons such that TMP period is subtracted from LD period . CCR2 free-running periodicity followed a normal distribution regardless of the entrainment zeitgeber , where lower and higher extreme periodicities were measured in thermal and photic entrainment , respectively ( Figure 1A ) . The normally distributed phenotypes allowed us to test the fixed effects of entrainment ( E ) by the random effects of genotypes ( G ) , and we found statistically significant G by E interactions ( Table 2 ) . The factor with the most significant effect in period variation was environment , with genotypes having a lesser , albeit highly significant effect ( Table 2 ) . Moreover , when the genotypic effect was compared between RILs and transformants , a far greater variation in periodicity could be observed in RILs ( Table 2 ) . Transformation-derived variation ( position effects ) was thus not a major component of detected variation . These findings suggest that both parents had been differentially selected in a number of loci for both zeitgeber inputs . 10 . 7554/eLife . 02206 . 005Table 2 . Statistical analysis of CCR2 circadian periodicity after the two zeitgeber protocols for BxS populationDOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 005Overall modelUnivariate BxSFactorFPGenotype7 . 753<0 . 001Environment36 . 413<0 . 001Environment*Genotype2 . 183<0 . 001Univariate LDCV LDUnivariate TMPCV TMPFPFPBxSRIL10 . 387<0 . 00118 . 0437 . 382<0 . 00116 . 498TRANS ( RIL ) 1 . 513<0 . 0011 . 915<0 . 001F = mean sum of squares\error sum of squares . P = significance value of the F-ratio . Genotype denotes RIL , ENVIRONMENT denotes the different entrainments , TRANS denotes independent transformants of each genotype . *denotes the testing of an interaction between two factors , whereas B ( A ) denotes the testing main factor A in which a factor B is nested . CV is the coefficient of genetic variation , LD stands for photic and TMP stands for thermal entrainment . NS denotes nonsignificance . The significant effect of genotypes in periodicity prompted us to calculate trait heritabilities . These were found to be similar for photic ( 0 . 76 ) and thermal entrainment ( 0 . 73 ) , respectively ( Table 3 ) . Our efforts then focused on mapping the genetic components of differential periodicity of the two inputs . This resulted in the identification of three large-effect QTLs for photic entrainment on chromosomes 1 , 2 , and 4 . Two co-localized QTLs were found on chromosomes 2 and 4 ( Table 3 ) . The QTL on Chromosome 2 ( Chr2 ) displayed the largest phenotypic effect and we thus pursued its identification . 10 . 7554/eLife . 02206 . 006Table 3 . Localization of the main QTLs for the BxS population after photic vs thermal zeitgeber protocolsDOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 006ZeitgeberH2ChromosomePositionLOD score ( * ) % expl varianceFp value2a ( cM ) ( h ) LD0 . 76I63 . 73 . 2114 . 97 . 7040 . 007−0 . 614II34 . 55 . 7227 . 320 . 784<0 . 0011 . 003IV69 . 93 . 8917 . 114 . 003<0 . 0010 . 833TMP0 . 73II34 . 54 . 5625 . 525 . 033<0 . 0010 . 988IV69 . 93 . 2616 . 26 . 5010 . 0140 . 508H2 denotes broad sense heritability . ( cM ) denotes centiMorgan . ( * ) LOD- score threshold was determined at 2 . 4 . % expl variance is the percent of explained variance . F = mean sum of squares\error sum of squares . P denotes the significance value of the F-ratio . 2a denotes the additive effect of Bay allele when the effect of the Sha allele on period is subtracted . ( h ) denotes the effect in hours . − denotes that the Sha allele displayed longer period than that of Bay allele . To validate the identity of the Chr2 QTL , we generated heterogeneous-inbred families ( HIFs ) , and near isogenic line ( NIL ) . Three HIFs were made ( 57 , 92 and 343; see ‘Materials and methods’ for construction details ) , each harboring either the Bay allele ( HIF-Bay-0 ) or Sha allele ( HIF-Sha ) at the Chr2 locus . An analysis of the free-running period of these lines revealed that HIF-Sha always displayed a shorter period as compared to HIF-Bay-0 . This period-shortening effect was independent of the preceding LD or TMP entrainment condition ( Figure 2A ) . The complexities of the HIF genomic structure created the possibility for equally complex epistatic interactions that cannot be simply defined . Therefore , to circumvent the possibility of such interactions , a NIL with a small introgression of Sha at this Chr2 locus ( NIL-S ) , in an otherwise homogeneous Bay genetic background , was generated . This line was measured for periodicity . Consistent with the HIFs , NIL-S displayed a 2-hr shorter period , as compared to Bay-0 , which was only observed under LL . In DD , no significant period difference was detected ( Figure 2B ) . However , in DD , the free-running profile of CCR2::LUC in NIL-S was different from the Bay parental line , so that , in NIL-S , CCR2::LUC rhythmicity lost robustness , whereas CCR2::LUC in Bay-0 displayed rhythms for 7 days ( Figure 3A ) . Interestingly , under both LL and in DD , the Sha parental line displayed a CCR2::LUC profile similar to that of Bay-0 , and different from NIL-S ( Figures 2B and 3A ) . 10 . 7554/eLife . 02206 . 007Figure 2 . Map-based cloning identifies ELF3 as a candidate for chr2 QTL . Period estimates of CCR2::LUC expression in ( A ) three independent HIFs ( 57 , 92 , and 343 ) harboring either the Bay-0 or the Sha allele at QTL confidence interval ( B ) parental accessions and NIL-S with introgression of Sha at QTL confidence interval in otherwise homogeneous Bay-0 background , under LL and in DD . ( C ) Schematic diagram showing the fine mapping strategy of chr2 locus . Black and gray bars represent Bay-0 and Sha genotypes , respectively . The names below the bars represent the molecular markers used for genotyping , and the numbers above correspond to their physical position on the genome . The crosses represent the position of the recombination event . Two recombinants , 89-S and 539-B , were found to have a recombination event surrounding a 40-Kb region , where nine annotated genes are located , as indicated below the solid bar . ( D ) Free-running profile of CCR2::LUC expression in recombinants 89-S , 539-B and Bay-0 under continuous red and blue light ( LL ) . ( E ) Period estimates of rhythm shown in ( D ) . All error bars indicate SEM , where n ≥ 24 . Mean values that are significantly different from Bay-0 wild type are indicated by * , ** , or *** for p-values ( ANOVA ) <0 . 05 , 0 . 01 , or 0 . 001 , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 00710 . 7554/eLife . 02206 . 008Figure 3 . Transgenic complementation confirms ELF3 as QTG underlying Chr2 QTL . Free-running profile of CCR2::LUC expression in Bay-0 , Sha , and NIL in darkness ( A ) , Left y-axis shows the luminescence measures of Bay and NIL , and the right y-axis shows the luminescence measures of Sha . NIL with introgression of Sha at ELF3 could not sustain the robust rhythms of CCR2::LUC after 4 days in darkness . Error bars represent SEM , n ≥ 24 . ( B ) Free-running period estimates of CCR2::LUC and LHY::LUC expression in T2 transgenic lines harboring either ELF3-Bay-YFP or ELF3-Sha-YFP in Ws-2 background . The data are the average of three independent , single-insert lines displaying similar rhythm profile . Error bars represent SEM , n = 96 . ( C ) Sequence comparison of ELF3-Bay and ELF3-Sha . Schematic representation of ELF3 ( AT2G25930 ) . Vertical bars show the position of the nucleic acid transition . The letters above the bars represent the nucleic acid in Bay , and letters below represent the nucleic acid in Sha . The numbers above the letters represent the position of the nonsynonymous change . The letters in parenthesis show the amino-acid change . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 008 We next employed a fine-mapping strategy to isolate the Chr2 QTL and delineated the region to a 40-Kb interval with nine annotated genes ( Figure 2C ) . ELF3 ( AT2G25930 ) was the only gene in this region with a previously reported function in the Arabidopsis circadian clock . Therefore , we compared the sequence of ELF3-Bay and ELF3-Sha and identified two nonsynonymous changes in ELF3-Sha: an encoded alanine-to-valine transition at position 362 ( A362V ) , and in the C-terminal encoded glutamine stretch , Sha had 8 more than Bay ( Figure 3C; and as reported [Tajima et al . , 2007; Coluccio et al . , 2011] ) . Thus , ELF3 appeared as a strong candidate for the Chr2 QTL . We then tested whether ELF3-Sha is involved in clock-controlled physiological processes under normal growth conditions . This was particularly relevant as it has been reported that the ELF3 allele in Sha affects shade-avoidance responses ( Tajima et al . , 2007; Jimenez-Gomez et al . , 2010; Coluccio et al . , 2011 ) . To determine the sensitivity of ELF3-Sha to day length , as a basis of comparing to an elf3 null allele , we examined the flowering time of the HIFs and NIL-S under long- and short-day growth conditions , comparing the results to the Bay-0 wild type . While we observed no large differences in flowering time under long days , HIF 89-S showed modestly advanced flowering time under short days , compared to HIF 539-B ( Figure 4A ) . Next , we assessed the effect of ELF3-Sha on hypocotyl length by measuring the hypocotyl length of HIFs , NIL-S , Bay-0 , Sha , and the null allele elf3-1 under short days , constant red ( RR ) , and constant blue ( BB ) light . Similar to the near lack of flowering-time phenotype under long days , we did not observe a substantial difference in the hypocotyl length of HIFs , NIL-S , and Bay-0 under RR or BB . Under short days , however , HIF 89-S displayed a marginally elongated hypocotyl length , compared to HIF 539-B , but this response was four times less than that seen in the null elf3 ( Figure 4B ) . Overall , a NIL that harbored ELF3-Sha in a Bay background did not display substantial physiological effects under normal growth conditions , indicating that ELF3-Sha maintains developmental activity lost in null elf3 alleles . 10 . 7554/eLife . 02206 . 009Figure 4 . Flowering time and hypocotyl length measurements for ELF3-Sha . ( A ) Flowering time of HIF 89-S , HIF 539-B , NIL , and Bay-0 under long day ( 16L:8D ) and short day ( 8L:16D ) . The flowering time was counted as the number of days at the appearance of 1 cm bolt . ( B ) Hypocotyl length of HIF 89-S , HIF 539-B , NIL , elf3-1 , Sha , and Bay-0 under short day ( 8L:16D ) , under RR ( 15 μmol m−2s−1 ) , and under BB ( 15 μmol m−2s−1 ) . Error bars represent the standard deviation . Significance as described in Figure 2 compared to Bay-0 , except HIF 89-S , which was compared to HIF 539-B . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 009 Although fine-scale mapping provided strong evidence favoring ELF3 as the gene responsible for the short-period phenotype we observed , any one of the nine genes in the 40-kb region were candidates in the context of this study . Therefore , to generate genetic materials to test if ELF3-Sha was indeed the causal QTL , we cloned ELF3 , along with its own promoter from Bay-0 and Sha , fused these with YFP . These constructs were then transformed to the null allele elf3-4 in the Ws-2 background , harboring CCR2::LUC and LHY::LUC reporter genes , respectively . The free-running period analysis of T2 transgenic lines harboring ELF3-Sha-YFP displayed a consistent short period under LL , compared to the lines harboring ELF3-Bay-YFP , for both LHY::LUC and CCR2::LUC ( Figure 3B ) . Based on these complementation experiments , we confirmed ELF3 as quantitative trait gene ( QTG ) underlying the Chr2 periodicity QTL . To define the causal polymorphism in ELF3-Sha , we investigated the role of an encoded A362V residue change compared to the difference in the number of encoded glutamines in ELF3-Bay and ELF3-Sha . In this experiment , we separately cloned the promoter and coding regions of ELF3-Bay ( SpBc ) and ELF3-Sha ( SpSc ) and then induced encoded A362V and V362A changes in ELF3-Bay ( SpBa2v ) and ELF3-Sha ( SpSv2a ) coding regions , respectively . These constructs were then recombined in all possible promoter::coding combinations ( illustrated in Figure 5A ) . The free-running period of LHY::LUC in T2 transgenic lines harboring different coding regions , under the control of the ELF3 promoter of the Sha accession , was analyzed under LL . We observed that lines SpSc and SpBa2v that contained encoded valine displayed a shorter period as compared to the SpBc and SpSv2a lines harboring encoded alanine , irrespective of the number of glutamines present ( Figure 5B ) . We further confirmed these results in the transgenic lines expressing ELF3-Bay and ELF3-A362V under the Bay-0 ELF3 promoter . The free-running profile of LHY::LUC was analyzed under LL and in DD . Both ELF3-Bay and ELF3-A362V displayed a robust rhythm of LHY::LUC expression under LL , albeit with lower amplitude observed in ELF3-A362V ( Figure 5C ) . However , in DD , the expression profile of LHY::LUC in ELF3-Bay and ELF3-A362V was distinctly different . ELF3-A362V failed to maintain robust rhythms after 4 days in darkness and displayed an acute dampening of LHY::LUC expression , whereas ELF3-Bay displayed robust rhythm of LHY::LUC , even after 6 days in darkness ( Figure 5D ) . The period analysis of LHY::LUC expression in ELF3-Bay and ELF3-A362V revealed no statistically significant period difference in DD , whereas under LL , ELF3-A362V displayed a shorter period compared to ELF3-Bay ( Figure 5E ) . These results are consistent with those described above for both HIFs and NIL-S and support that the encoded A362V polymorphism is the functionally encoded variant in Sha sufficient to explain the Chr2 periodicity QTL in the BxS population ( Figure 2A , B ) . Thus , we could explain the molecular basis of a clock periodicity QTL to an absolute level of a single nucleotide , colloquially defined as the ‘quantitative trait nucleotide’ ( QTN ) . 10 . 7554/eLife . 02206 . 010Figure 5 . Molecular basis of ELF3-Sha phenotypes . ( A ) Schematic diagram explaining the different promoter-coding combinations used in ( B ) . The Sha promoter of ELF3 fused with different coding regions is shown ( for details see 'Materials and methods' ) . The encoded amino-acid residue at position 362 , along with the number of encoded glutamines , is shown . ( B ) Period estimates of the LHY::LUC expression in the lines explained in ( A ) . Note that the lines SpSc and SpBa2v with Valine in the coding part displayed period acceleration , irrespective of the number of glutamines . Error bars represent SEM , n = 48 . Significance as explained in Figure 2 , compared to SpBc . Free-running profile of LHY::LUC expression in T2 transgenic lines harboring either ELF3-Bay or ELF3-A362V in Ws-2 genetic background under LL ( C ) and in DD ( D ) . A single nucleotide exchange was induced in ELF3-Bay to change the encoded alanine residue at position 362 to valine ( ELF3-A362V ) . The data are the average of three independent single-insert lines with similar rhythm profile . ( E ) Period estimates of the lines shown in ( C ) and ( D ) . Error bars represent SEM , n = 96 . Mean values that are significantly different from Bay-0 wild type are indicated by * , ** , or *** for p-values ( ANOVA ) <0 . 05 , 0 . 01 , or 0 . 001 , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 010 To further dissect the effect of light and darkness on ELF3-Sha , we measured LHY::LUC expression in ELF3-Bay and ELF3-A362V for 15 days with a defined regime of extended light and darkness . We observed that both ELF3-Bay and ELF3-A362V displayed robust LHY::LUC expression so long as they remained under light ( Figure 6A ) . However , when the plants were transferred to darkness , LHY::LUC expression dampened more rapidly in ELF3-A362V . For ELF3-Bay , an acute reduction of LHY::LUC expression was seen such that proper rhythm was maintained throughout the dark period , while no detectable rhythmic expression of LHY::LUC was observed in ELF3-A362V ( Figure 6A ) . To test if the loss of rhythm of LHY::LUC in ELF3-A362V was light-dependent or resulted from a ‘permanent’ defect in the clock caused by continuous darkness , we subsequently returned these plants to light after an interval of 4 days in darkness . Both ELF3-Bay and ELF3-A362V recovered robust rhythms of LHY::LUC after the 11-day treatment , including the 4-day period in darkness ( Figure 6A ) . These results suggested that light is necessary to sustain a robust oscillator in the context of ELF3-A362V . 10 . 7554/eLife . 02206 . 011Figure 6 . Alterations in the ELF3-Sha oscillator and clock resetting . ( A ) Free-running profile of LHY::LUC expression in ELF3-Bay and ELF3-A362V in a 15-day continuous experiment under consecutive light and in dark conditions . The plants were entrained for 7 days under 12 hr:12 hr light dark cycles , followed by transfer to LL and measurement of LHY::LUC expression for 6 days . On day 7 , plants were transferred to darkness , and the measurement of LHY::LUC was continued in DD for four more days . On day 11 , the plants were transferred to light conditions again , and the expression profile of LHY::LUC was measured for an additional 4 days . Open bars in the graph represent time in LL , and closed bar represents time in DD . Error bars represent SEM and are shown on every third reading . ( B ) Period and R . A . E . analysis of profiles shown in ( A ) n = 48 . ( C ) Phase shifts in dark-adapted seedlings after resumption to LL . ELF3-Bay and ELF3-A362V plants entrained for 7 days under light/dark cycles ( LD ) were transferred in DD for 1 day and then replicate samples were released into LL at 4-hr intervals , monitoring the phase of LHY expression in LL to determine the state of the oscillator in the preceding DD interval . Phase difference plot ( Phase ELF3-Bay–ELF3-A362V ) for 3 days in DD is shown . Third peak under LL was used for phase analysis . n = 36 . Experiment was repeated three times with similar results . ( D ) Frequency demultiplication assay . After 7 days of entrainment under 12L:12D ( T = 24 ) cycles , the LHY:LUC profile was monitored under 12L:12D ( T = 24 ) for 1 day and then 6L:6D ( T = 12 ) for 4 days . The shaded boxes indicate the duration of the LD cycles . For clarification , the LHY::LUC profiles from day-3 to day-5 is magnified and shown below . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 011 Rhythmic robustness depends on the precision of the circadian clock , and relative amplitude error ( R . A . E . ) is a measure of such precision ( Kolmos et al . , 2009 ) . Therefore , to test the precision of LHY::LUC expression rhythms in ELF3-Bay and ELF3-A362V , we estimated the period and R . A . E . of these lines under LL and in DD . Under LL , ELF3-A362V displayed a ∼1 . 5-hr short period compared to ELF3-Bay , with both lines showing high precision ( R . A . E . <0 . 20; Figures 6B and 7A , ZT40-120 ) . However , in darkness , while both lines had similar period estimates , ELF3-A362V did not maintain precision and displayed high R . A . E . ( R . A . E . >0 . 75 ) , compared to ELF3-Bay ( R . A . E . <0 . 35; Figures 6B and 7B , ZT160-240 ) . Interestingly , the loss of precision of LHY::LUC rhythm was recovered in ELF3-A362V when the plants were returned to the light at ZT250 ( Figures 6B and 7C , ZT260-340 ) . This restoration of precision led us to hypothesize that in darkness the clock remains partially functional in ELF3-Sha in the context of this low precision . To confirm this , we performed an experiment where the phase of ELF3-Bay and ELF3-A362V lines was estimated after resumption of plants to LL that were previously subjected to different intervals of darkness . We reasoned that if the ELF3-A362V plants are truly arrhythmic in DD , then the circadian phase after resumption of light should be determined solely by the exposure to light and not by the duration of the dark period . For ELF3-Bay , with a more functional clock in the dark , the phase will be determined in part by the duration of dark exposure . We found that LHY::LUC peaked at a similar time in both ELF3-Bay and ELF3-A363V , which in part was determined by the duration of the dark period . This revealed the presence of a partially functional oscillator during the dark period in both these lines . Further , no significant phase difference between these lines was detected at any time point tested ( Figure 8A ) . Interestingly , consistent with the above results ( Figure 6A ) , once the plants were transferred back to the light , robust rhythms of LHY::LUC were restored ( Figure 8B ) . Notably , in a phase–difference graph ( Phase ELF3-Bay minus Phase ELF3-A362V ) , a consistent pattern of phase oscillations was detected , supporting the existence of a functional oscillator in both ELF3-Bay and ELF3-A362V that display differences in their respective resetting behavior ( Figure 6C ) . This confirms that ELF3-Sha contributes to light-input to the clock , but displays differences in its ‘zeitnehmer’ entrainment capacity . Taken together , we concluded that ELF3-Sha requires light input to maintain precision of the circadian clock; however , darkness does not fully abolish the low-amplitude oscillation in ELF3-Sha plants and that entrainment processes appear altered . 10 . 7554/eLife . 02206 . 012Figure 7 . Alanine sustains robust oscillator in darkness . ( A–C ) Scatter plot for R . A . E . against period showing estimates of individual ELF3-Bay and ELF3-A362V lines shown in ( Figure 5A ) . Only lines with R . A . E . < 1 . 0 were plotted . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 01210 . 7554/eLife . 02206 . 013Figure 8 . Circadian oscillator does not abolish in ELF3-Sha in darkness . ( A ) Phase shifts in dark-adapted seedlings after resumption to LL . ELF3-Bay and ELF3-A362V plants entrained for 7 days under light/dark cycles ( LD ) were transferred in DD for 1 day and then replicate samples were released into LL at 4-hr intervals , monitoring the phase of LHY expression in LL to determine the state of the oscillator in the preceding DD interval . Phase was calculated relative to dawn ( ZT00 ) , n = 36 . ( B ) The LHY::LUC profile in ELF3-Bay and ELF3-A362V after 128 hr in darkness . Note the robust rhythms of LHY::LUC in both lines . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 013 In Arabidopsis , light input to the circadian clock follows Aschoff's rule who noted that the activity phase shortens in nocturnal organisms exposed to constant light conditions and lengthens in diurnal organisms ( Aschoff , 1979 ) . These trends were termed alpha compression and alpha expansion , respectively . This process is also known as parametric entrainment ( Aschoff , 1979; Johnson et al . , 2003 ) . ELF3 , as a repressor of light signaling to the clock , is expected to be involved in such parametric entrainment ( Covington et al . , 2001 ) . In the past , the availability of null mutants precluded testing such a hypothesis . However , the fact that ELF3-Sha is a functional allele could be advantageous in determining if ELF3 is involved in parametric entrainment . To test this , we measured the free-running period of Bay and NIL-S under an array of fluences , including RR , BB , and ‘white’ ( RB ) light . Under diverse intensities and qualities of light , we found that NIL-S displayed a shorter period than Bay . Moreover , both Bay and NIL-S followed Aschoff's rule , displaying shortening of periodicity with increase in fluence rate ( Figure 9A–C ) . From these results , one of two conclusions may be drawn: ( 1 ) that ELF3 is not directly involved in parametric entrainment or ( 2 ) that the ELF3-Sha allele is fully functional in discriminating between different light intensities . 10 . 7554/eLife . 02206 . 014Figure 9 . ELF3-Sha is short-period under a range of light intensities . Free-running period of Bay and NIL-S under different intensities of ( A ) red ( B ) blue and ( C ) red+blue lights ( ∼3:1 ) . Plants were entrained for 7 days under 12 hr LD cycles ( white light ) before transferring to the respective light conditions . Neutral density filters were used to control the light intensities . Light intensities were measured in µmol m−2s−1 and were transformed to log10 scale shown in the graphs . DD represents darkness . Error bars represent SEM , n = 36 . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 014 An important characteristic of clock entrainment is termed frequency demultiplication , which describes the resetting properties of the oscillator . Wild-type plants with a functional clock entrained under 24-hr environmental cycles ( T24 ) maintain the same rhythms when transferred to shorter cycles ( T12 ) . In this way , plants resist the resetting of the oscillator by sporadic environmental changes during a diurnal cycle . However , if the short cycles ( T12 ) persist , a robust clock should reset to the new environment . Conversely , when transferred from T24 to T12 , mutants with a dysfunctional clock promptly follow the shorter cycles ( Nozue et al . , 2007; Thines and Harmon , 2010; Kolmos et al . , 2011 ) . To test the demultiplication ability of ELF3-Sha , we monitored LHY::LUC expression for 7 days in ELF3-Bay , ELF3-A362V , Ws-2 , and elf3-4 lines under T12 cycles ( 6 hr light and 6 hr dark ) after initially entraining under T24 cycles ( 12 hr light and 12 hr dark ) . As previously reported for null elf3 ( Kolmos et al . , 2011 ) , we detected a driven 24 hr rhythm , and , thereafter , LHY::LUC completely followed the T12 cycles in this mutant , revealing , in turn , a dysfunctional clock ( Figure 6D ) . As expected , both Ws-2 and ELF3-Bay initially resisted the T12 cycles for 3 days , displaying driven T24 rhythms . However , after 3 days , both Ws-2 and ELF3-Bay showed a robust adaptation to the T12 cycles , which were observed as strong LHY::LUC peaks matching the peaks expected for T12 cycles ( Figure 6D ) . Interestingly , ELF3-A362V did not show this effect in T12 cycles , and strong LHY::LUC peaks completely matched the T24 cycles for all 5 days ( Figure 6D ) . Since both Ws-2 and ELF3-Bay responded to persistent T12 cycles , while ELF3-Sha did not , we concluded from these results that ELF3-Sha failed to perceive persistent new environmental signals , revealing that the circadian oscillator in ELF3-Sha is indeed defective in proper entrainment resetting . Under continuous light and in darkness , it has been reported that elf3 loss-of-function alleles display arrhythmia ( Hicks et al . , 1996; Thines and Harmon , 2010 ) . Accordingly , expression profiling revealed that elf3 had reduced expression of the core oscillator genes CCA1 and LHY , but high levels of the evening genes TOC1 and GI ( Fowler et al . , 1999; Kikis et al . , 2005; Dixon et al . , 2011 ) . As these null elf3 studies were conducted in the context of arrhythmia , placing ELF3 in the clock network has been difficult . However , as ELF3-Sha displayed rhythmicity ( Figure 5C ) , we monitored the luciferase reporter expression profile of the central clock genes CCA1 , LHY , TOC1 , GI , PRR7 , and PRR9 in NIL-S and compared it with Bay-0 wild type and the null mutants elf3-1 and elf3-4 under both LL and DD . Consistent with previous reports ( Kikis et al . , 2005; Thines and Harmon , 2010; Dixon et al . , 2011; Kolmos et al . , 2011; Herrero et al . , 2012 ) , both null mutants , elf3-1 and elf3-4 , displayed arrhythmia with reduced levels of CCA1::LUC and LHY::LUC and high levels of TOC1::LUC , GI::LUC , PRR7::LUC , and PRR9::LUC , compared to Bay-0 ( Figure 10 ) . Consistent with the short-period phenotype of CCR2::LUC in NIL-S ( Figure 2B ) , all clock genes also displayed a short-period phenotype ( Figure 10A–F ) . The comparison of expression profiles of these clock genes in NIL-S , elf3-1 , elf3-4 and Bay-0 revealed an intermediate effect of ELF3-Sha . Specifically , in NIL-S , the expression of CCA1::LUC and LHY::LUC was higher than that of elf3-1 and elf3-4 , but lower than that of Bay-0 . Similarly , the expression of TOC1::LUC , GI::LUC , PRR7::LUC , and PRR9::LUC in NIL-S was lower than that of elf3-1 and elf3-4 , but higher than that of Bay-0 ( Figure 10A–F ) . Similar to LL , profiles of luciferase expression for clock genes with reduced levels in DD were observed in NIL-S , elf3-1 , elf3-4 and Bay-0 , except that of TOC1::LUC in NIL-S , which was high and comparable to the null mutants ( Figure 10G–L ) . A continuous increase in the expression of GI::LUC in NIL-S was also observed . No significant period difference was observed in NIL-S and Bay-0 in any clock gene under DD . Consistent results were also obtained when we confirmed the luciferase expression data by monitoring the transcript profile of these genes under LL ( Figure 11 ) . Unlike elf3-1 and elf3-4 loss-of-function mutants , our cumulative results indicate that ELF3-Sha is a hypomorphic allele capable of sustaining the oscillation network . 10 . 7554/eLife . 02206 . 015Figure 10 . Compromised clock network in ELF3-Sha . Luciferase expression profile of different clock genes in NIL , elf3-1 , elf3-4 and Bay-0 under LL ( left panel , A–F ) and in DD ( right panel , G–L ) . ( A and G ) CCA1::LUC , ( B and H ) LHY::LUC , ( C and I ) TOC1::LUC , ( D and J ) GI::LUC , ( E and K ) PRR7::LUC , and ( F and L ) PRR9::LUC . Error bars represent SEM and are shown on every third reading . Note that the NIL-S displayed an intermediate expression of all clock genes relative to the null mutants elf3-1 and elf3-4 compared to Bay-0 . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 01510 . 7554/eLife . 02206 . 016Figure 11 . Transcript accumulation pattern of different clock genes under LL . Transcript accumulation of different clock genes in Bay-0 , Sha , NIL-S and elf3-4 under LL . ( A ) CCA1::LUC , ( B ) LHY::LUC , ( C ) PRR7::LUC , and ( D ) PRR9::LUC , ( E ) GI::LUC , and ( F ) TOC1::LUC . Error bars represent the standard deviation of three technical repeats . Expression levels are normalized for PROTEIN 19 PHOSPHATASE 2a subunit A3 ( PP2A ) . Growth conditions , quantitative RT-PCR , and primer sequences were previously described ( Kolmos et al . , 2009; Kolmos et al . , 2011 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 016 A higher expression of morning genes , PRR7 and PRR9 , and evening genes , TOC1 and GI , suggests a repressive role of ELF3 . Thus , the short-period phenotype of ELF3-Sha could be caused by a reduced level of ELF3 transcript or protein accumulation . To test this , we first measured the transcript accumulation of ELF3 and then assessed the amount of protein generated . We found that the mean level of ELF3-Sha transcript was slightly higher relative to ELF3-Bay . Correspondingly , higher ELF3-Sha protein accumulation was detected in comparison to ELF3-Bay , as measured by the ELF3-YFP signals ( Figure 12A–D ) . As increases in ELF3 drives a long period ( Covington et al . , 2001; Herrero et al . , 2012 ) , elevated protein levels in ELF3-Sha thus did not explain the short-period phenotype . Importantly , a more detailed comparison of ELF3-Bay-YFP and ELF3-Sha-YFP revealed differences in cellular localization . Specifically , the formation of distinct nuclear bodies , a characteristic of ELF3 ( Herrero et al . , 2012 ) , was markedly attenuated in ELF3-Sha , whereas these nuclear bodies were clearly observable in ELF3-Bay ( Figure 12A , B ) . Additionally , the preferential nuclear localization of ELF3 was compromised in ELF3-Sha , which displayed a markedly elevated amount of cytoplasmic content , compared to ELF3-Bay ( Figure 12A–D ) . Quantification of YFP revealed that the nuclear–cytoplasmic ratio of ELF3-Sha was four times less than that of ELF3-Bay ( Figure 12E ) . As such , we proposed that these localization defects of ELF3-Sha underlie its oscillator defects . 10 . 7554/eLife . 02206 . 018Figure 12 . Sub-cellular localization defects of ELF3-Sha . ( A ) and ( B ) show maximum intensity projection of ELF3-YFP localization in root cells of ELF3-Bay-YFP ( A ) and ELF3-Sha-YFP ( B ) . Arrows indicate the nuclei that are magnified four times and shown in small boxes at the bottom of ( A ) and ( B ) . Note that ELF3 forms distinct nuclear foci in ELF3-Bay , whereas in ELF3-Sha , YFP signal for ELF3 is diffused in the nucleus . Scale bar is 20 µm . ( C ) and ( D ) display the YFP intensity distribution of ( A ) and ( B ) in visual-thermal units , respectively . Note that the ELF3 cytoplasmic contents were higher in ELF3-Sha as compared to ELF3-Bay . ( E ) shows the nucleus-to-cytoplasmic fluorescence ratio of ELF3-Bay-YFP and ELF3-Sha-YFP , as calculated by ImageJ . Error bars represent SEM , n = 3 . Significance as described in Figure 2 . The representative data of three independent experiments and three independent lines are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 018 Having shown that ELF3-Sha is defective in proper cellular localization , we next sought to identify if the localization aberrations cause faulty regulation of ELF3-Sha . This is particularly relevant as it has been shown that ELF3 accumulates under light , but quickly dissipates in darkness by proteasomal machinery involving the action of COP1 and GI . Furthermore , both these proteins physically interact with ELF3 to form similar nuclear bodies that are presumed to be their point of interaction ( Yu et al . , 2008 ) . As such , it could be anticipated that the lack of formation of such nuclear bodies in ELF3-Sha might result in its dysregulation . To examine such a possibility , we monitored the patterns of ELF3 transcript and protein accumulation over a diurnal cycle under LD compared to the same patterns under LL and in DD . Under LD , we found only a minor increase in ELF3-Sha transcript and protein levels compared to ELF3-Bay ( Figure 13A , B ) . However , under LL , the ELF3-Sha transcript was significantly higher at all time-points compared to ELF3-Bay , which also resulted in elevated levels of ELF3-Sha protein ( Figure 13C , D ) . Furthermore , compared to LD where ELF3 accumulation decreased during the evening hours , we found elevated protein levels of both ELF3-Bay and ELF3-Sha during subjective night ( ZT16 and ZT20 ) . These results were consistent with previous reports that ELF3 accumulation increases in light ( Liu et al . , 2001; Yu et al . , 2008 ) . In DD , similar to LL , ELF3-Sha transcript levels were higher compared to ELF3-Bay . However , we did not detect any significant difference in ELF3-Bay and ELF3-Sha protein levels during subjective day , whereas during subjective night ( ZT16 and ZT 20 ) , we found that ELF3-Sha accumulation was considerably lower than ELF3-Bay ( Figure 13E , F ) . Taken together , a higher accumulation of ELF3-Sha under LL and overall lower levels in DD , despite higher transcript abundance , led us to conclude that ELF3-Sha protein dissipates in darkness comparatively more rapidly than ELF3-Bay . 10 . 7554/eLife . 02206 . 017Figure 13 . ELF3 cyclic accumulation is altered in ELF3-Sha . Accumulation pattern of ELF3 transcript and encoded protein in ELF3-Bay-YFP and ELF3-Sha-YFP lines under LD ( A and B ) , LL ( C and D ) and in DD ( E and F ) . For LD , plants were grown under 12L:12D cycles for 6 days , and starting the next day at ZT0 , plants were harvested every 4 hr for RNA extraction and then , separately , scanned under the microscope for cellular studies . For LL and DD , after initial entrainment , plants were transferred under white light or in darkness for 1 day , followed by harvesting the samples for RNA extraction or taking pictures for their respective zeitgeber time . This microscopic experiment was performed three times with similar results . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 017 Finally , we sought to identify the evolutionary history of the ELF3-Sha locus . To accomplish this , we retrieved the ELF3 coding sequence ( CDS ) of 251 accessions from the ‘1001 Genome-Project’ ( ‘Materials and methods’ ) and performed several population-genetics analyses . The sequence comparison revealed the presence of 59 polymorphic sites segregating at the ELF3 locus , including encoded A362V . Out of these 59 polymorphic sites , 16 were synonymous and 43 were nonsynonymous . The A362V was detected in 15 accessions , and in a phylogenetic analysis , all of these grouped together in a single clade , suggesting a common origin ( Figure 14A ) . Therefore , we looked at the geographical distribution of these accessions and found that they all were distributed in Central Asia between 37°N and 54°N , with two exceptions at Nemrut and Rubenzhoe collected from Turkey and the Ukraine , respectively ( Figure 14B; Table 4 ) . Interestingly , similar to Sha that naturally grows at high altitudes ( e . g . , Pamir Mountains , Tajikistan ) , most of these accessions were also high-altitude accessions collected in the mountains ( Table 4 ) . These results suggest that ELF3-Sha originated from Central Asia and that this allele might have been maintained during species migration preferentially by altitude-associated individuals . 10 . 7554/eLife . 02206 . 019Figure 14 . Distribution of the ELF3 locus . ( A ) Neighbor-joining tree showing the phylogenetic relationship of ELF3 coding sequence of 251 accessions . All accessions harboring ELF3-Sha allele were grouped in a single clade indicated by red-filled circles . Scale represents the distance calculated by interior–branch test . Sites with gaps/missing data were not included in the analysis . ( B ) Geographical distribution of ELF3-Sha allele . Accessions harboring ELF3-Sha are shown with pink marks , whereas blue marks represent ELF3-Bay allele . Locations of Bay and Sha are shown with yellow and green marks , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 01910 . 7554/eLife . 02206 . 020Table 4 . Geographical location of the 15 accessions with ELF3-Sha alleleDOI: http://dx . doi . org/10 . 7554/eLife . 02206 . 020AbbreviationNameStock IDCountryLocationLatitudeLongitudeAltitudeICE130Kolyvan/Kly-4CS76384RussiaKolyvan51 . 3282 . 55505ICE138Lebjashje/Leb-3CS76426RussiaAltaijskij Kraj51 . 6580 . 82301ICE70Borskoje/Borsk-2CS76421RussiaSamarskaja Oblast53 . 0451 . 7575ICE71Shiguljovsk/Shigu-1CS76375RussiaSamarskaja Oblast53 . 3349 . 48181ICE72Shiguljovsk/Shigu-2CS76374RussiaSamarskaja Oblast53 . 3349 . 48181Kar1KarakolCS76522KyrgyzstanSusamyr village/West Karakol river42 . 374 . 362503KondaraKondaraCS76532TadjikistanKhurmatov38 . 568 . 5815Kz9KazakhstanCS76537KazakhstanKaragandy49 . 6773 . 33535NemrutNemrut-1CS76398TurkeyNemrut Dag38 . 6442 . 232249Neo6Neo6CS76560TajikistanJawshangoz village37 . 3572 . 463467RubenzhoeRubenzhoe-1CS76594UkraineRubezhnoe49 . 0138 . 3656ShaShakdaraCS76382TadjikistanPamiro-Alay38 . 9069 . 053439SorboSorboCS22653Tadjikistan38 . 8269 . 481751Westkar4West KarakolCS76629KyrgyzstanKarakol valley42 . 3742187Zal1Zal1CS76634KyrgyzstanTchong-Kemin valley/Djachyl-Kul lake42 . 876 . 352230
ELF3 has been established as a required component for the generation of circadian rhythms and the perception of light and temperature inputs ( Covington et al . , 2001; Thines and Harmon , 2010; Kolmos et al . , 2011 ) . However , most of the elf3 alleles used in these studies were identified in mutagenesis screens , and many of them displayed arrhythmicity under free-running conditions of LL and DD . Here , we report the cloning and characterization of ELF3-Sha as a natural allele of ELF3 , which displayed a light-dependent short-period phenotype . Unlike elf3 loss-of-function mutants , this allelic state displayed robust circadian rhythms under continuous light . However , in prolonged darkness , the expression profile of many clock genes was dramatically damped ( Figure 10 ) . The association between the clock phenotypes of ELF3-Sha and their cellular basis led us to conclude that proper cellular distribution of ELF3 is required to perform normal repressive function in the circadian clock ( Figures 12 and 13 ) which in turn , depends on its cyclic accumulation . Importantly , we could track the origin of ELF3-Sha to Central Asia , providing more insight about the molecular evolution of this allele . The identification and validation of quantitative trait loci ( QTLs ) by genetic mapping is a well-established procedure . However , to understand the molecular basis of a locus by characterizing QTL to a quantitative trait nucleotide ( QTN ) level is still an arduous task ( Koornneef et al . , 2004 ) . Here , we reported the characterization of a natural allele of ELF3 ( ELF3-Sha ) to a QTN level of understanding . In the quantitative analysis of circadian periodicity of a modified BxS population , we identified a QTL and further validated it in HIFs and a NIL ( Table 1; Figure 2A , B ) . Our periodicity results were consistent with previous studies where the ELF3 in Sha was found to modulate the shade-avoidance response , including the action of shade on the oscillator ( Jimenez-Gomez et al . , 2010; Coluccio et al . , 2011 ) . In those studies , HIFs were the main genetic resource used . The authors of those papers proposed ELF3 to be a candidate locus for the detected shade-avoidance QTL , and transgenic efforts were used to complement that notion ( Jimenez-Gomez et al . , 2010 ) . Here , we generated the appropriate NIL and showed that the introgression of ELF3-Sha to Bay-0 resulted in severe circadian alterations , including a curtailed period length ( Figure 2B ) . The light-dependent acceleration of circadian oscillations of ELF3-Sha could be associated with reduced ELF3 functionality ( McWatters et al . , 2000; Covington et al . , 2001; Kolmos et al . , 2011 ) . The intermediate expression profile of several clock genes in ELF3-Sha , when compared to wild-type and null mutants , confirmed that it is a hypomorphic allele ( Figures 10 and 11 ) . This characteristic of ELF3-Sha is distinctly different from elf3 null mutants , which displayed complete arrhythmia under LL ( Covington et al . , 2001; Hicks et al . , 2001; Thines and Harmon , 2010; Kolmos et al . , 2011 ) . However , the functionality of the oscillator in these mutants in darkness remained disputable . Based on the observation that the circadian rhythms persist in the elf3-1 mutant , it was proposed that ELF3 solely acts as a light zeitnehmer , or timekeeper , and that it is not a required component of the oscillator . In contrast , a recent study reported that the circadian oscillator was abolished in constant darkness ( DD ) in temperature-entrained elf3-1 seedlings , emphasizing that ELF3 is an integral part of the oscillator as well as acting as a zeitnehmer ( Thines and Harmon , 2010 ) . As these studies were mainly based on null mutants , a conclusive statement about ELF3 function in sustaining the oscillator in DD is hard to draw . Consequently , in this paper , we thus analyzed the profile of several clock genes in the functionally stable ELF3-Sha mutant in DD , comparing it with those in elf3-1 and elf3-4 null mutants . Consistent with a previous report ( Thines and Harmon , 2010 ) , we could not detect any sign of rhythmicity in null elf3 mutants . Interestingly , in the hypomorphic ELF3-Sha , a rapid dampening in the expression of several clock markers in DD was observed ( Figures 3A , 5D , 6A , and right panels of Figure 10G–L ) . Thus , our results conclusively demonstrate that ELF3 is required for robust oscillation in darkness and further support the idea that ELF3 is an integral part of the oscillator ( Thines and Harmon , 2010 ) . Although ELF3-Sha lost robustness and precision in prolonged darkness , it is important to note that it quickly recovered when transferred back to the light ( Figures 6A , B , 7A–C ) . Consequently , we further propose that the function of ELF3 is collectively defined by both the light–dark boundary and by circadian clock regulation . elf3 loss-of-function mutants under free-running conditions were previously shown to display major defects in the expression of the core oscillator genes PRR7 , PRR9 , CCA1 , LHY , TOC1 , and GI ( Kikis et al . , 2005; Dixon et al . , 2011; Kolmos et al . , 2011 ) . To monitor the expression of these clock genes in ELF3-Sha , we used the robust luciferase reporter-expression system and measured these expression profiles in Bay-0 , NIL-S , and the nulls elf3-1 and elf3-4 . Our results were consistent with previous reports showing the arrhythmic low levels of CCA1 and LHY and high levels of TOC1 and GI in elf3 loss-of-function mutants . In contrast , under LL , ELF3-Sha was rhythmic for all genes studied , and it displayed an intermediate level of expression compared to wild type and loss-of-function elf3 ( Figure 10A–F , left panels ) . Specifically , NIL-S displayed higher expression of TOC1 and GI compared to Bay-0 , but lower expression than elf3-1 and elf3-4 . In contrast , expression of CCA1 and LHY was lower in NIL-S , compared to Bay-0 , but higher than elf3-4 . Consistent with the previous finding that CCA1 and LHY regulate PRR7 and PRR9 ( Farre et al . , 2005; Nakamichi et al . , 2010 ) , a higher transcript abundance of PRR7 and PRR9 was observed in NIL-S ( Figures 10E , F and 11C , D ) . These data , along with recent findings showing the evening complex ( ELF3 , ELF4 , and LUX ) directly binds to the PRR9 promoter to repress transcription , support the repressive action of ELF3 in the core oscillator ( Helfer et al . , 2011; Nusinow et al . , 2011; Chow et al . , 2012; Herrero et al . , 2012 ) . Higher expression levels of the evening genes TOC1 and GI , as well as dramatically dampened oscillations of ELF3-Sha in darkness , cannot be simply explained by only considering ELF3 at PRR9 . Considering the double loss-of-function mutant cca1-11 lhy-21 is rhythmic , an additional repressive role of ELF3 by targeted degradation of GI is conceptually plausible ( Ding et al . , 2007; Yu et al . , 2008 ) . Taken together , our data support a hypothesis which holds that the EC-containing ELF3 has more than one entry point in the circadian clock ( Kolmos et al . , 2009 , 2011; Herrero and Davis , 2012 ) . ELF3 is both a cytosolic and nuclear localized protein . It appears to be multifunctional in that it has several binding partners . These include phyB , COP1 , ELF4 , and GI ( Yu et al . , 2008; Nusinow et al . , 2011; Herrero et al . , 2012 ) . Different domains of ELF3 specifically interact with these different proteins . Both phyB and COP1 interact with the N-terminal domain ( Liu et al . , 2001; Yu et al . , 2008; Kolmos et al . , 2011 ) , whereas ELF4 and GI interact with the middle domain ( Yu et al . , 2008; Herrero et al . , 2012 ) . Further , all these proteins co-localize in the nucleus where they form distinct nuclear bodies ( Mas et al . , 2000; Yu et al . , 2008; Chen et al . , 2010; Herrero et al . , 2012 ) . Such nuclear foci could be suspected as interaction points of binding proteins ( Yu et al . , 2008; Herrero and Davis , 2012; Herrero et al . , 2012; Kim et al . , 2013 ) . Thus , formation of fewer ELF3-nuclear foci from encoded ELF3-Sha might be the result of a defect in the binding of one of its interacting proteins . Since the A362V variant is located in the middle domain of ELF3 , such a hypothesis could be particularly attributed to either GI and/or ELF4 . The rhythmic accumulation of both ELF3 and GI depends upon the activity of COP1 that mediates ubiquitination and targeted degradation of these proteins in dark conditions ( Yu et al . , 2008 ) . It is noteworthy here that ELF3 is essential for this process . In the absence of ELF3 , COP1 cannot interact with GI and thus cannot initiate its decay . However , GI is not required for COP1-mediated ELF3 degradation ( Yu et al . , 2008 ) . When these data are taken together , the formation of a COP1-ELF3-GI complex could be considered a plausible active mechanism controlling the rhythmic accumulation of these proteins . Under these conditions , A362V mutation in ELF3-Sha would result in attenuated binding affinity with GI , disturbing the balance of the COP1-ELF3-GI complex and , in turn , resulting in the rapid decay of ELF3 . Both the lower accumulation of ELF3-Sha protein , despite the generation of higher transcript ( Figure 13E , F ) , and aberrant oscillator behavior of ELF3-Sha in prolonged darkness support this hypothesis . The short-period phenotype of ELF3-Sha under LL could also be explained by an early decay of ELF3-Sha during light/dark entrainment . As ELF3 directly represses PRR9 ( Dixon et al . , 2011; Herrero et al . , 2012 ) , the early depletion of ELF3-Sha results in an early expression of PRR9 , an event that sets the pace of the oscillator during the preceding entrainment , which is maintained through LL . Based on these results and our observation that ELF3-Sha is defective in proper clock resetting ( Figure 6D ) , we propose that ELF3 plays a pivotal role in defining entrainment properties of the circadian clock , which , in turn , depend on cyclic accumulation of ELF3 protein , but not on its absolute levels or , indirectly , its transcript abundance . This idea could be further supported by the observation that ELF3 protein rhythmically oscillates and drives a long period in plants that constitutively overexpress ELF3 ( Covington et al . , 2001; Dixon et al . , 2011; Herrero et al . , 2012 ) . Discounting the important role that ELF3 plays in clock entrainment , it can be provisionally excluded that ELF3 is involved in parametric entrainment because we could not detect any differential effect of fluence rate on ELF3-Sha periodicity compared to ELF3-Bay ( Figure 9A–C ) . Fluence rate curves based on ELF3-ox also support this notion ( Covington et al . , 2001 ) . Studies on ELF3-Sha as a natural allele disrupted in normal circadian behavior provide a perspective on its repressive action on clock periodicity . Notably , in the context of the Ws-2 and Bay-0 genomes , we could show that ELF3-Sha is a hypomorphic allele defective in proper localization of encoded ELF3 protein . This defect resulted in two distinct phenotypes: one displaying a light-dependent short period and the other exhibiting loss of rhythm robustness in darkness . These phenotypes were clearly observable in both the Bay-0 and Ws-2 genetic background . It is notable that Sha parental line did not itself display such obvious circadian defects ( Figures 2B and 3A ) . Related to this , the transcript profile of clock genes in Sha followed the same pattern as observed in Bay-0 , and such pattern was distinct from that seen in NIL-S ( Figure 11 ) . Moreover , ELF3-Sha did not affect hypocotyl length in NIL-S and Sha in the same way . Under all light qualities tested , the hypocotyl length of NIL-S was similar to that of Bay-0 , but was significantly longer when compared to Sha ( Figure 4B ) , leading to the possibility that other segregating QTLs within Sha wild type genetically interact with ELF3-Sha . Such similar background-dependent effects of natural alleles have been reported in Arabidopsis for seed longevity , axillary bud formation , and flowering time ( Sugliani et al . , 2009; Huang et al . , 2012; Undurraga et al . , 2012; Méndez-Vigo et al . , 2013 ) . Our efforts to track the history of the ELF3-Sha revealed that this allele is present in a group of genetically related accessions that are predominately distributed in a geographical area of Central Asia ( Figure 14A , B ) . This confined location of ELF3-Sha leads to the possibility that this allele provides local adaptive advantage to these accessions under their respective environmental conditions , and thus , has been positively selected during species migration . Such a hypothesis can only be confirmed in further studies using realistic environmental conditions in which these accessions are derived .
The Recombinant Inbred Lines ( RILs ) used were derived from the Bayreuth-0 ( Bay-0 ) by Shakdara ( Sha ) RIL collection ( termed here BxS ) ( Loudet et al . , 2002 ) . Multiple , independent T1 transgenic CCR2::LUC reporter lines were obtained from 71 lines after floral dipping ( Boikoglou et al . , 2011; Davis et al . , 2009 ) . T2 progeny were used for circadian rhythm experiments ( Supplementary files 1 and 2 ) . To confirm the chromosome 2 periodicity QTL in BxS , three heterogeneous inbred families ( HIFs ) were generated . For this , RILs 57 , 92 , and 343 were used as recipients of a pollen donation from Bay-0 CCR2::LUC . These F1 lines were respectively backcrossed twice to the given RIL . In these three BC2 populations from the separate RIL crosses , the plants were self-crossed , and in the derived BC2F2 populations , lines that were homozygous for Bay or for the Sha alleles , at four marker positions , were identified . Multiple F2 versions of each of these derived lines were isolated . F3 seeds that harbored CCR2::LUC were collected from these plants for periodicity tests , as described below . The selected F3 HIFs containing Sha at QTL interval were further backcrossed three times to Bay-0 to generate NIL-S . The introgression of Sha at QTL interval and the homogeneous Bay-0 background were confirmed with genome-wide SSLP markers . For fine mapping , the progeny of HIF343 , heterozygous at QTL interval , was screened with different SSLP , CAPS , and dCAPS markers . Out of 1100 plants screened , 14 plants with a recombination event between markers elf100L and elf100R were selected and self-fertilized to obtain homozygous recombinant lines . The progeny of these homozygous recombinant lines was then used for circadian-periodicity assay . The detail of all the markers used for genotyping is given in Supplementary file 3 . The mutant lines used in this study were as follows: elf3-1 and elf3-4 ( Zagotta et al . , 1992 , 1996; Hicks et al . , 2001 ) . Both mutant lines were backcrossed four times to the Bay-0 wild type to homogenize the accession background to Bay-0 . Homozygous plants were subsequently identified in the BC4F2 population using specific markers ( Supplementary file 3 ) . Different clock-marker lines , CCR2::LUC , CCA1::LUC , LHY::LUC , TOC1::LUC , and GI::LUC , used in the study were generated by initial transformation of the respective marker into Bay-0 , followed by crossing the T2 transformants to the target genotype: Bay-0 , NIL-S , elf3-1 , and elf3-4 . The homozygous lines obtained by the self-fertilization of BC1F2 were used for the circadian assays . To generate ELF3 Bay transgenic lines , the ELF3 gene , along with ELF3 native promoter , was amplified from Bay-0 genomic DNA and cloned into pPZP211 vector . In ELF3 A362V , a nucleotide change encoding Alanine to Valine ( A362V ) was induced using the QuikChange method ( Stratagene , California , USA ) . The multi-gateway technology ( Invitrogen , California , USA ) was used to generate the YFP-tagged lines with different promoter-coding combinations . Initially , the promoter and coding regions of ELF3 were separately amplified from Bay-0 and Sha genomic DNA and recombined into pDONR4-P1R or pDONR201 donor vectors , respectively . The respective nucleotide change encoding either the A362V or V362A amino acid was then induced using the QuikChange method ( Stratagene ) . The YFP tag was separately cloned into pDONRp2r-p3 donor vector . These three donor vectors were then recombined in different combinations into pPZP211 destination vector . The final vector was then transformed into Agrobacterium tumerfaciens ( strain ABI ) . For all transformations , the improved floral-dip method was used ( Davis et al . , 2009 ) . Based on Mendelian segregation , T2 transgenic lines having single insertion were selected on kanamycin . All the transgenic lines harboring ELF3-Bay , ELF3-A362V , ELF3-Bay-YFP , ELF3-Sha-YFP , SpSc , SpBc , SpBa2v and SpSv2a are in elf3-4 Ws-2 genetic background . All oligonucleotides used for cloning and QuikChange are listed in Supplementary file 3 . For luciferase assays , seeds were surface-sterilized and plated on MS medium containing 3% sucrose . Following ∼3 days stratification at 4°C , seedlings were entrained for 7 days , either under 12L:12D cycles ( ∼100 μE light ) with constant temperature of 22°C ( LD ) or under 12 hr at 16°C: 12 hr of 22°C temperature cycles with constant light ( TMP ) ( ∼100 μE light ) . The bioluminescence measurement and data analysis are as described ( Hanano et al . , 2006 , 2008 ) . For hypocotyl assays , seedlings were grown on MS medium ( 2 . 2g/L pH 5 . 7 ) without sucrose , as described ( Davis et al . , 2001 ) . Hypocotyl length was determined for seedlings grown under SD ( 8L:16D ) , BB , or RR for 7 days ( light intensity SD: 120 ìmol m-2s-1; light intensity RR and BB: 15 ìmol m-2s-1 ) . Seedlings were scanned , and hypocotyl elongation was measured using the ImageJ 64 program , V1 . 43b ( Wayne Rasband , National Institutes of Health , USA , http://rsb . info . nih . gov/ij ) . For flowering time measurement , plants were grown on soil containing a 3:1 mixture of substrate and vermiculite in a temperature-controlled greenhouse environment with 16L:8D long-day and 8L:16D short-day cycle . The flowering time was scored at the time of bolting ( 1 cm above rosette leaves ) as the total number of days to bolt ( Domagalska et al . , 2010 ) . In total , 60 and 65 BxS lines were assayed for CCR2 rhythmic periodicity after light and temperature entrainment , respectively . Period mean was subsequently used for QTL mapping , which was performed with MapQTL 5 . 0 ( Kyazma BV , Wageningen , The Netherlands ) . Mapping settings used were as in Boikoglou et al . ( 2011 ) . The statistical analyses , including broad sense heritability were calculated as reported ( Keurentjes et al . , 2007; Boikoglou et al . , 2011 ) . For all microscopic work , the Zeiss LSM700 confocal microscope from Carl Zeiss was used , as in Herrero et al . ( 2012 ) . Briefly , the plants were grown on MS medium containing 1 . 5% sucrose . Following ∼3 days stratification at 4°C , seedlings were entrained for 6 days under 12L:12D cycles ( ∼100 μE light ) with constant temperature of 22°C . The following day , the plants were either put under constant light ( LL: ∼100 μE light ) or in darkness for another day . On day 7 , starting at ZT0 , the plants were scanned , and the photographs were taken every 4 hr for 1 day . For the comparison of ELF3 Bay-YFP and ELF3 Sha-YFP lines , one slide from each line was prepared , and both slides were put together in the microscope . The plants from each slide were then scanned , one after another , within 30 min and with the same microscope settings . The microscope settings for the Figure 13 data set were as follows: Image size: x = 512 , y = 512 , z = 20; Channels: 3 , 8-bit , Zoom = 1 . 0; Objective: Plan-Aprochromat 40x/1 . 30 Oil; Pixel dwell: 2 . 55 µs; Master gain: ch1 = 972 , ch2 = 847 , ch3 = 162; Digital gain: ch1 = 1 . 20 , ch2 = 1 . 0 , ch3 = 1 . 50; Digital offset = ch1 = −18 . 0 , ch2 = 2 . 0 , ch3 = −24 . 42; Pinhole = 156 µm; and laser: 488 nm with 10 . 0% strength . For the Figure 12 data set , microscope settings were as follows: Image size: x = 512 , y = 512 , z = 20; Channels: 3 , 8-bit , Zoom = 0 . 5; Objective: Plan-Aprochromat 63x/1 . 40 Oil; Pixel dwell: 2 . 55 µs; Master gain: ch1 = 1096 , ch2 = 928 , ch3 = 393; Digital gain: ch1 = 1 . 20 , ch2 = 0 . 61 , ch3 = 1 . 40; Digital offset = 0 . 0 , Pinhole = 156 µm; and laser: 488 nm with 10 . 0% strength . The ELF3 CDS sequences of accessions were downloaded from the 1001 Genome-Project ( http://www . 1001genomes . org/ ) . The geographical coordinates for the accessions were obtained from the SALK database ( http://signal . salk . edu/atg1001/index . php ) . These coordinates were used to map the geographical position of the accessions using Google maps ( maps . google . com ) ( Table 4 ) . For sequence alignment and phylogenetic analysis , MEGA 4 . 0 software was used ( Tamura et al . , 2007 ) . The phylogenetic relationships between the sequences were determined using the neighbor-joining ( NJ ) method and applying the interior–branch test ( Saitou and Nei , 1987 ) . | Life on Earth tends to follow a daily rhythm: some animals are awake during the day and asleep at night , whilst others are more active at night , or during the twilight around dawn and dusk . For many living things , these cycles of activity are driven by an internal body clock that helps the organism to adapt to the daily cycle of light and dark—and similar internal clocks also exist in plants . These internal clocks define daily—or circadian—cycles whereby multiple genes are switched ‘on’ or ‘off’ at different time points in every 24-hr period . And , because light and ambient temperatures also vary with time of the day , many organisms use these external signals as cues to reset their own internal clocks . Moreover , the hours of daylight and temperature vary around the world , and also with the seasons , so plants and animals must be able to change how these external signals influence their internal clocks so that they stay in tune with the day/night cycle . However , it is not clear how they do this . To explore this question , Anwer et al . grew plants that were from a cross between two types of the model plant Arabidopsis thaliana from different environments: one from Germany , and the other from Tajikistan in Central Asia . These offspring were also genetically engineered so that an enzyme that could give off light was produced under the control of the internal clock . Anwer et al . found that the plants continued to glow and fade with an almost daily rhythm even after external cues , such as changes in temperature or light , had been removed . Different offspring plants consistently glowed and faded with different rhythms such that some had , for example , a 21-hr day and others a 28-hr day . These differences were caused by many genes that differed from the original German and Tajikistan parent plants , and Anwer et al . ‘mapped’ one of these genetic differences to a single gene . Offspring that inherited a version of a gene called ELF3 from the Tajikistan parent had internal clocks that ran faster when the plant was under the light . These plants also gradually stopped glowing as brightly as the German parent when they were kept in the dark , suggesting that their internal clocks were ‘ticking more softly’ . It was already known that the ELF3 gene affected the circadian clock in plants , and Anwer et al . thus concluded that the plants with Tajikistan version of this gene , called ELF3-Sha , were also less able to reset their internal clocks to synchronize in response to external cues . Anwer et al . also showed that the normal ELF3 protein is more likely to be found in the nucleus of a plant cell than the ELF3-Sha version , which might suggest that this protein is involved in switching genes off . Further research is now needed to uncover exactly how the ELF3 protein does this to keep the plant's internal clock ‘ticking’ correctly . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"cell",
"biology"
] | 2014 | Natural variation reveals that intracellular distribution of ELF3 protein is associated with function in the circadian clock |
Engineering light-sensitivity into proteins has wide ranging applications in molecular studies and neuroscience . Commonly used tethered photoswitchable ligands , however , require solvent-accessible protein labeling , face structural constrains , and are bulky . Here , we designed a set of optocontrollable NMDA receptors by directly incorporating single photoswitchable amino acids ( PSAAs ) providing genetic encodability , reversibility , and site tolerance . We identified several positions within the multi-domain receptor endowing robust photomodulation . PSAA photoisomerization at the GluN1 clamshell hinge is sufficient to control glycine sensitivity and activation efficacy . Strikingly , in the pore domain , flipping of a M3 residue within a conserved transmembrane cavity impacts both gating and permeation properties . Our study demonstrates the first detection of molecular rearrangements in real-time due to the reversible light-switching of single amino acid side-chains , adding a dynamic dimension to protein site-directed mutagenesis . This novel approach to interrogate neuronal protein function has general applicability in the fast expanding field of optopharmacology .
The vast majority of excitatory neurotransmission in the mammalian brain is mediated by ionotropic glutamate receptors ( iGluRs ) , which are classified into AMPA , kainate , and NMDA receptors ( Traynelis et al . , 2010 ) . NMDA receptors ( NMDARs ) stand out from other iGluRs by their unique capability to induce long-term synaptic plasticity that underlies higher cognitive functions such as learning and memory . They are also targets of therapeutic interest since their dysfunction is associated with numerous neurological and psychiatric disorders such as schizophrenia , mental retardation , and epilepsy ( Paoletti et al . , 2013 ) . At the structural level , NMDARs are massive ( >550 kD ) tetrameric complexes , typically composed of two GluN1 and two GluN2 subunits . Representative of their receptor family , NMDARs show a layered organization with an amino-terminal domain ( NTD ) and an agonist-binding domain ( ABD ) extruding into the extracellular space , an intracellular carboxy-terminal domain ( CTD ) , and a transmembrane domain ( TMD ) that contains the selective ion channel pore ( Karakas and Furukawa , 2014; Lee et al . , 2014 ) . A main feature of native NMDARs is their broad molecular heterogeneity that translates into a wide variety of receptor subtypes with distinct biophysical , pharmacological , and signaling properties . This is further diversified by their differential location between brain regions , developmental stages , and even subcellular localizations , supporting the idea that each receptor subpopulation is tailored to match the strict requirements of specific neuronal functions ( Paoletti et al . , 2013 ) . Up to now , putative contributions made by the individual NMDAR subunits to specific neuronal functions are largely unknown or controversial . The GluN2 subunits , of which there are four subtypes ( GluN2A-D ) , are the key determinants of the receptor’s functional diversity ( Paoletti , 2011; Wyllie et al . , 2013; Glasgow et al . , 2015 ) . Among those , GluN2A and GluN2B are the major subunits in the adult brain , endowing receptors with distinct charge transfer capacities during activity-dependent synaptic transmission . For instance , the influence of the GluN2A/2B ratio on the polarity of synaptic plasticity – whether long-term potentiation ( LTP ) or long-term depression ( LTD ) is induced – is still debated ( Yashiro and Philpot , 2008; Paoletti et al . , 2013 ) . Likewise , the role of NMDAR at synaptic and ( GluN2B-enriched ) extrasynaptic sites in cell survival remains largely contentious ( Parsons and Raymond , 2014 ) . Thus , the modality of GluN2-subtype-driven fine-tuning of information processing in distinct brain regions , within individual neurons , and at individual excitatory synapses remains to be resolved . By providing precise ways to manipulate endogenous signaling proteins , optopharmacological approaches open new avenues for deciphering the molecular basis of physiological function . Optopharmacology combines the power of optics , endowing high spatiotemporal resolution , with that of genetics and pharmacology , to achieve unique photocontrol on the molecular receptor level ( Kramer et al . , 2013 ) . One attractive strategy to engineer light-responsiveness relies on the introduction of azobenzene moieties . Upon exposures to light of different wavelengths , azobenzene groups undergo a reversible photoswitching between two different configurations – a compact cis- and a stretched trans-configuration ( Beharry and Woolley , 2011; Broichhagen and Trauner , 2014 ) . The use of azobenzene offers high quantum yield , minimal photo-bleaching , excellent photo-stability , and has the major benefit of functional reversibility . In the last decade , chemical tethering of photoswitchable azobenzene-coupled ligands has been shown to be an elegant and powerful method to engineer light-responsiveness in neuronal receptors . This strategy has allowed to photo-agonize or -antagonize different classes of glutamate receptors ( Volgraf et al . , 2006; Reiner et al . , 2015; Berlin et al . , 2016 ) and diverse members of other neurotransmitter receptor families ( Tochitsky et al . , 2012; Lemoine et al . , 2013; Browne et al . , 2014; Lin et al . , 2015 ) . This approach , however , requires proper protein conjugation with the azobenzene-coupled ligand and is restricted to pharmacologically-characterized sites that are accessible to the external solvent , excluding intracellular and transmembrane protein domains . The incorporation of photoreactive groups into the receptor itself , using unnatural amino acids ( UAAs ) , provides an alternative to achieve optical control over receptor activity . The UAA methodology relies on the re-assignment of a stop codon ( usually the Amber stop codon ) by a suppressor tRNA aminoacylated with the desired UAA ( Wang et al . , 2001; Chin , 2014; Leisle et al . , 2015; Liu and Schultz , 2010 ) . Orthogonal tRNA/synthetase pairs enable efficient and direct incorporation of genetically-encoded UAAs into receptors expressed in simple cell lines , neuronal cultures , brain slices , and even whole organisms ( Wang et al . , 2007; Kang et al . , 2013; Klippenstein et al . , 2014; Zhu et al . , 2014; Ernst et al . , 2016 ) . Important in the context of our study , light-triggered inactivation of AMPA and NMDA receptors was recently demonstrated following incorporation of the photocrosslinking UAAs Azido-phenylalanine ( AzF ) and Benzoyl-phenylalanine ( BzF ) at specific interface sites ( Klippenstein et al . , 2014; Zhu et al . , 2014; Tian and Ye , 2016 ) . Another type of UAAs , decorated with a photocage , was shown to provide control over neuronal excitability following incorporation into K+-channels ( Kang et al . , 2013 ) . Photocrosslinking or photocaged UAAs offer excellent site tolerance and molecular specificity ( even at solvent-inaccessible sites ) , however , they require prolonged exposures to light ( time scale of several seconds or more ) , have a possibly fragile photostability ( in case of AzF ) , and crucially , lack reversibility . In this study , we combined the advantages of UAA incorporation and azobenzene photochemistry by implementing azobenzene-based photoswitchable UAAs ( PSAAs ) ( Bose et al . , 2006; Hoppmann et al . , 2014 ) in order to achieve fast and reversible photocontrol over a set of NMDAR subunits . We demonstrate that genetically encodable PSAAs provide an efficient and flexible approach to install precise and reversible light-sensitivity on a key neuronal receptor . By revealing novel structural determinants involved in controlling NMDAR function , we further provide evidence that PSAAs are powerful tools to probe receptor biophysics . In particular , our approach allows targeting transmembrane pore sites whose conformational changes during receptor activity remain poorly understood . The development of photoswitchable NMDARs with genetic encodability , as presented here , should be valuable to manipulate NMDAR signaling and unmask GluN2-specific roles in neuronal function . We expect these principles to be generally applicable to other brain proteins , enabling optical investigation of a range of receptors and ion channels both in recombinant and native systems . Differences and advantages of the PSAA approach compared to classical mutagenesis are discussed as well as the technical requirements for its in vivo application .
Recently , we and others have successfully exploited the genetic code expansion methodology to design light-sensitive iGluRs through incorporation of photocrosslinking UAAs ( Klippenstein et al . , 2014; Zhu et al . , 2014; Tian and Ye , 2016 ) . Here , following similar principles , we expanded the approach to generate NMDARs containing azobenzene-based photoswitchable UAAs , which have the major advantage of reversibility ( Figure 1 ) . To test the feasibility of genetically-encoded PSAAs in NMDARs , we co-expressed specific NMDAR subunits containing an introduced Amber stop codon with a genetically engineered orthogonal tRNA/aminoacyl synthetase pair derived from the methanogenic archea Methanosarcina mazei ( Mm ) in HEK cells ( Figure 1a ) . This tRNA/synthetase pair , evolved from the pyrrolysine tRNA/synthetase pair ( tRNAPyl–MmPSCAA-RS ) , allows incorporation of azobenzene-based PSAAs into E . coli and mammalian cells in a site-specific manner ( Hoppmann et al . , 2014 ) . The simple photoisomerization from the planar ( trans ) to the bent ( cis ) configuration of the azobenzene moiety results in a pronounced change in shape , with a reduction in the end-to-end distance between the two rings of 3 . 5 Å ( Beharry and Woolley , 2011 ) ( Figure 1b and c ) . The Alkene moiety of PSAA provides an additional functionality for click reactions and formation of photobridges with nearby cysteines ( Hoppmann et al . , 2014 , 2011 ) . Here , we reasoned that the single side-chain flip of the azobenzene moiety may be sufficient to drive structural changes and impact receptor functionality when introduced at specific subunit locations ( Figure 1c ) . Thus , we introduced Amber mutations , one at a time , at various sites within the GluN1 or GluN2 subunit and assessed receptor functionality using electrophysiological patch-clamp recordings ( Table 1 ) . Three phenotypes were observed in our mutational analysis: ( i ) absence of measurable NMDAR-mediated currents; ( ii ) presence of currents , but lack of photosensitivity; ( iii ) presence of photosensitive currents . We interpreted the first scenario as indicative of improper receptor expression and focused on the latter , demonstrating successful PSAA incorporation into mature , cell surface expressed NMDARs . Promisingly , within each domain of the receptor ( NTD , ABD , and TMD ) , at least one position was identified that endowed significant light-sensitivity ( Table 1 ) , highlighting the efficacy of the approach . 10 . 7554/eLife . 25808 . 003Figure 1 . General principle for genetic encoding photoswitchable UAAs into membrane receptors . ( a ) Schematic representation of the UAA methodology applied to NMDARs . A gene encoding a membrane protein of interest ( here shown for the NMDAR GluN1 subunit; blue ) and containing an introduced Amber stop codon ( TAG , red ) at a desired position is co-transfected into HEK cells with vectors encoding a WT GluN2 subunit and an orthogonal suppressor tRNA ( yellow ) / aminoacyl synthetase ( RS; grey ) pair . The cells are incubated in the presence of the photoswitchable UAA ( PSAA , green asterisks ) in the culture medium . Within the cell , the orthogonal RS specifically aminoacylates the suppressor tRNA with the PSAA . At the ribosome level , the PSAA is incorporated in response to the Amber codon on the mRNA by the complementary anticodon CUA on the suppressor tRNA ( red ) . After release from the ribosome , the full-length GluN1 polypeptide chain , site-specifically carrying the PSAA , assembles into a functioning receptor expressed at the cell surface . ( b ) Structure of the PSAA . The photoisomerizable azobenzene unit is highlighted in green . Additionally , a functional alkene group , which allows a covalent click reaction with nearby cysteines , is attached to the azobenzene unit . ( c ) Diagram of a NMDAR GluN1/GluN2 heterodimer carrying a PSAA at the ABD dimer interface . Toggling between the trans- and cis-configuration is induced by light exposure at 460 or 365 nm , respectively . This simple change in side chain geometry , when placed at a receptor key moving site , allows inducing remarkable structural changes that can impact receptor activity . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 00310 . 7554/eLife . 25808 . 004Table 1 . Screening for expression and photosensitivity of NMDAR Amber mutants . Overview of all Amber positions screened for PSAA insertion within the NTD , ABD , and TMD . PSAA was introduced to either the GluN1 , GluN2A , or GluN2B subunit . A total of 27 different combinations were tested . Among those , 13 gave co-agonist evoked currents ( green boxes ) , the rest did not result in functional receptors ( orange boxes ) . Eight subunit combinations assembled into functional receptors , whose activity could be reversibly modulated by light ( ↓ UV-triggered photoinactivation; ↑ UV-triggered photopotentiation ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 004GluN1GluN2AGluN2BExpressionPhotomodulationWTwtwtwtwtNTDY109wt↓Y109wt↓wtY281wtY282ABDP521wtP532wt↓P532wtY535wt↓wtP527wtE530wtP528wtE531E698wtTMDF554wt↑W563wt↑W578wtW608wtW611wtwtS632W636wtwtY645wtY646Y647wt↑Y647wtF654wt↑F654wtF817wt The activation of NMDARs requires binding of two different agonists , glutamate and glycine ( or D-serine ) , a unique feature among ligand-gated ion channels . These co-agonists bind at appropriate clamshell-like domains that close thereupon to transduce the signal to the ion channel pore . Within a tetrameric GluN1/GluN2 receptor , the ABDs assemble as two heterodimers with the glycine-binding GluN1 ABD and the glutamate-binding GluN2 ABD , pairing through back-to-back interactions within each dimer ( Furukawa et al . , 2005; Karakas and Furukawa , 2014; Lee et al . , 2014 ) ( Figure 2a ) . To endow NMDARs with optical sensitivity , we initially targeted the GluN1/GluN2A ABD heterodimer interface for PSAA incorporation since this region critically controls receptor functionality ( Furukawa et al . , 2005; Gielen et al . , 2008; Borschel et al . , 2011 ) . Replacing the highly conserved proline residue GluN1-P532 by the PSAA resulted in functional receptors , as represented by large and robust co-agonist activated currents ( Figure 2—figure supplement 1 ) as well as a pronounced modulation of receptor activity induced by application of light ( Figure 2 ) . Specifically , following receptor activation with saturating co-agonist concentrations , illumination with UV light ( 365 nm ) to switch the PSAA from the trans- to the cis-isomer , produced a remarkable current inhibition of 48 . 9 ± 1% ( n = 23; Figure 2b and g ) . Crucially , this effect could be fully reversed by blue light ( 460 nm ) that regenerates the trans-version of the PSAA . The degree of photoinactivation was similar when UV illumination occurred prior to agonist application ( 48 . 4 ± 1 . 6% [n = 7]; Figure 2c and d ) or during application of competitive antagonists ( Figure 2—figure supplement 2 ) , indicating that the effect is not dependent on the functional state of the receptor . Application of blue light in the resting state was ineffective to alter the current amplitude ( Figure 2c and d ) , in agreement with the PSAA azobenzene moiety adopting essentially 100% of the trans-isomer in the dark ( Beharry and Woolley , 2011 ) . Importantly , both wavelengths of light had virtually no effect on wild-type ( WT ) GluN1/GluN2A receptors ( Figure 2e–g ) . To confirm specific incorporation of PSAA , we performed identical transfections of mutant NMDAR subunits and the tRNA/synthetase pair , but omitted to supplement the culture medium with the PSAA . In such conditions , the majority of transfected cells yielded no or tiny NMDAR-mediated currents ( few tens of pA at most ) that were completely insensitive to UV or blue light , indicating that the Amber codon suppression by endogenous amino acids is negligible ( Figure 2—figure supplement 3a and b ) . In PSAA-containing receptors , the degree of UV-induced current inhibition neither depended on the initial level of the peak current amplitude nor on the extent of receptor desensitization ( Figure 2—figure supplement 3c and d ) . Overall , these results reveal a high yield of PSAA incorporation at the GluN1-P532 site and strongly support the presence of a single homogenous ( PSAA-containing ) receptor population at the cell surface . Further evidence for specific PSAA introduction was obtained by exposing P532PSAA mutant NMDARs to the positive allosteric modulator ( PAM ) GNE-6901 . This compound binds the GluN1-GluN2 ABD upper lobe dimer interface ( Hackos et al . , 2016 ) , adjacent to the PSAA incorporation site . Whereas WT receptors were strongly potentiated upon PAM application , GluN1-P532PSAA mutant receptors were entirely unaffected by PAM , but retained light sensitivity in the presence of PAM ( Figure 2—figure supplement 4 ) , indicating the disruption of the PAM-binding site by PSAA . We found that PSAA-substitution at the ABD site GluN1-Y535 also resulted in a significant UV-induced photoinactivation . The extent of photomodulation was , however , less pronounced compared to GluN1-P532 . On that account and due to the lower expression levels of the GluN1-Y535PSAA mutant receptors ( Figure 2—figure supplement 1 ) , we chose the more robust GluN1-P532 site to examine the photomodulation properties in further detail . 10 . 7554/eLife . 25808 . 005Figure 2 . Reversible photomodulation of NMDARs with PSAA at the GluN1-P532 site . ( a ) Structure of a heteromeric GluN1/GluN2B receptor ( left panel ) . The two GluN1 subunits are shown in blue , the two GluN2 subunits in orange . One receptor dimer is highlighted by cartoon representation . The typical layered organization contains the N-terminal domains ( NTDs ) , the agonist-binding domains ( ABDs ) , and the transmembrane domain ( TMD ) . The C-terminal domain ( CTD ) is absent in the structure . The Amber mutation site GluN1-P532 at the GluN1/GluN2 ABD heterodimer interface is shown as green spheres . The top view of one ABD dimer highlights the location of the PSAA insertion ( right panel ) . ( b ) Representative current trace of GluN1-P532PSAA/GluN2A receptors showing reversible photomodulation following 5 s of UV and blue light during the application of saturating co-agonists . In this example , the UV-driven current reduction was 46% . ( c ) UV ( 2 s ) in the resting state on the same cell as in b gave a current inhibition of 51% . ( d ) The photoinactivation degree was similar when UV was applied in the ‘mixed’ state ( 50% ) . ( e ) As in b , for WT GluN1/GluN2A receptors . Nearly no modulation of the current amplitude was observed ( 2% inhibition with UV ) . ( f ) As in c , for WT receptors ( 13% inhibition with UV ) . ( g ) Summary of UV inhibition degrees for the PSAA-mutant and WT receptors following UV exposures in different states . For the PSAA-mutant , mean photoinactivation values are ( in % ) : Inhactive = 48 . 9 ± 1 ( n = 23 ) , Inhrest = 48 . 4 ± 2 ( n = 7 ) , Inhmixed = 49 . 7 ± 2 ( n = 14 ) . For WT receptors: Inhactive = 1 . 3 ± 0 . 6 ( n = 18 ) , Inhrest = 4 ± 5 ( n = 3 ) , Inhmixed = 7 . 9 ± 2 ( n = 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 00510 . 7554/eLife . 25808 . 006Figure 2—figure supplement 1 . : Summary of expression levels for wild-type receptors and all GluN1 Amber mutant receptors that exhibited photomodulation following introduction of PSAA . The maximal peak current amplitudes , recorded in the dark state , were as follows ( in pA ) : 2450 ± 280 ( n = 18 , WT ) , 760 ± 110 ( n = 23 , P532 ) , 230 ± 90 ( n = 9 , Y535 ) , 430 ± 90 ( n = 5 , F554 ) , 50 ± 10 ( n = 3 , W563 ) , 120 ± 40 ( n = 14 , Y647 ) , 210 ± 60 ( n = 12 , F654 ) , 1030 ± 310 ( n = 6 , Y109+GluN2A ) , IPeak320 ± 80 ( n = 10 , Y109+GluN2B ) . The current amplitudes represent the expression levels 24–72 hr following PEI transfection . The graph shows the peak currents of those cells that underwent statistical analysis following any kind of light protocols . Blank recordings were not taken into account in this illustration . All GluN1 mutants were co-expressed with GluN2A-WT subunits unless otherwise noted . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 00610 . 7554/eLife . 25808 . 007Figure 2—figure supplement 2 . The presence of the competitive antagonists APV and DCKA does not impact the UV-mediated current inhibition for GluN1-P532PSAA/GluN2A mutant receptors . ( a ) Representative trace showing the degree of inhibition of PSAA-mutant receptors in the absence ( left part ) and presence ( right part ) of APV ( 20 μM ) and DCKA ( 5 μM ) in the extracellular solution . The cell was exposed to the competitive antagonists for 6 s following their wash out for 10 s . The orange line indicates the lack of change in the extent of UV-driven photoinactivation . ( b ) The current inhibition degrees ( in % ) in a regular washing solution were 43 . 4 ± 2 ( n = 5 ) , and following antagonist application 45 . 3 ± 2 . 5 ( n = 5; p=0 . 27 ) . The degrees of inhibition were equal for the application of APV/DCKA on either the cis- or trans-isoform of the PSAA . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 00710 . 7554/eLife . 25808 . 008Figure 2—figure supplement 3 . High efficiency of PSAA-incorporation at the GluN1-P532Amber site . ( a ) Example traces from mutant GluN1/GluN2A receptors incubated either with ( black trace ) or without ( green ) PSAA in the culture medium . The unspecific readthrough of the Amber stop codon was minimal in the absence of PSAA . UV and blue light were equivalently applied during both experiments , but reversible photoinactivation was only seen for receptors incubated with the PSAA ( here , 44% photoinhibition ) . ( b ) Pooled peak current amplitudes ( in pA ) for WT ( 24–48 hr post transfection ) and Amber mutants ( 48 hr post transfection ) incubated in the presence or absence of PSAA . On average , WT receptors gave currents amplitudes of 2200 ± 640 pA ( n = 11 ) . For PSAA-harboring receptors , seven cells gave co-agonist induced currents of 460 ± 110 pA , and one cell resulted in a blank recording . Amber mutant receptors lacking the PSAA had an average current amplitude of 50 ± 15 pA ( n = 6 ) , nine further cells did not express at all . Crucially , UV or blue light never modulated mutant receptor activity with unspecifically incorporated endogenous amino acids . In contrast , receptors site-specifically carrying the PSAA showed a clear UV-triggered photoinactivation ( mean 45 ± 1 . 5%; n = 7 ) . ( c ) The degree of photomodulation shows minimal dependence on peak current amplitude ( in pA , measured immediately before the application of UV ) . A linear fit gave a correlation coefficient of R2 = 0 . 37 ( n = 23 ) . ( d ) Similarly , the degree of photomodulation is minimally affected by the extent of current desensitization ( in %; compared to the absolute peak current ) . The correlation coefficient is R2 = 0 . 13 ( n = 23 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 00810 . 7554/eLife . 25808 . 009Figure 2—figure supplement 4 . PSAA incorporation at the GluN1-P532 Amber site disrupts PAM action . ( a ) Example trace demonstrating the strong potentiating effect of the positive allosteric modulator ( PAM ) GNE-6901 ( 30 µM ) on WT GluN1/GluN2A receptors ( here , 2 . 3-fold potentiation ) that were alternatively exposed to UV or blue light . ( b ) As in a , for GluN1-P532PSAA/GluN2A receptors . In contrast to WT receptors , PAM does not potentiate mutant receptors with the PSAA switched into the trans-state . Moreover , the degree of photoinhibition is comparable in the absence and presence of PAM ( 37% vs . 32% , respectively ) . ( c ) As in b , for mutant receptors with the PSAA switched into the cis-state . Again , PAM has no potentiating effect and did not prevent UV photoinhibition ( 43% photoinhibition prior PAM application , 39% during the perfusion with PAM ) . ( d ) Mean fold-potentiation by PAM . Values are 2 . 3 ± 0 . 3 ( n = 6 ) for WT receptors and 0 . 9 ± 0 . 04 ( n = 7 ) for PSAA mutant receptors ( ***p<0 . 001 ) . For mutant receptors , cells were pooled regardless of the photoisomeric state of PSAA before applying PAM ( trans as in panel b , cis as in panel c ) . ( e ) Mean current photoinhibition degrees ( in % ) for the PSAA mutant . In the absence of PAM , the current was inhibited by 43 . 2 ± 2% ( n = 7 ) by UV . In the presence of PAM , the extent of UV photoinactivation was similar with 41 . 5 ± 2% ( n = 7 , p=0 . 31 ) . Again , for mutant receptors , cells were pooled regardless of the photoisomeric state of PSAA before applying PAM ( trans as in panel b , cis as in panel c ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 009 We first quantified the light responsiveness of the receptor by measuring the current inhibition and re-activation kinetics induced by the wavelength-dependent photoisomerization of the azobenzene moiety . Forward and backward transitions between the fully active and the UV-inactivated state were both satisfactorily fitted by monoexponetial functions with decay time constants in the tens to hundreds of ms range ( τuv = 500 ± 30 ms [n = 23]; τBLUE = 45 ± 1 ms [n = 23]; Figure 3a ) . Additionally , we examined the re-activation kinetics when stepping back to green light ( 520 nm ) that revealed a much slower switch rate ( τGREEN = 1350 ± 140 ms [n = 5] , Figure 3a ) , associated with partial current recovery ( 76 ± 7% [n = 5] , compared to full recovery with blue light ) . This partial recovery , accompanied by its slowness , is indicative of a mixture of PSAA cis- and trans- forms at 520 nm and is consistent with the green light providing less energy and being less efficient than blue light to reverse PSAA from the cis- to the trans-isomer ( Figure 3—figure supplement 1 ) ( Beharry and Woolley , 2011; Hoppmann et al . , 2014 , 2011 ) . We further studied the dependence of the UV-induced photoinactivation on the duration and power of the UV irradiation ( Figure 3—figure supplement 2 ) . Both the extent and time course of photoinactivation were measured . Since isomerization events of free azobenzene groups are known to occur on a picosecond timescale ( Beharry and Woolley , 2011 ) , substantially faster than most biological processes , we tested if short light pulses ( <5 s , the duration used so far ) were sufficient to trigger equal photoresponses . By varying the light duration between 100 ms and 5 s , we identified a UV interval length of 1 s to be the minimum for maximal current inactivation ( Figure 3—figure supplement 2a–d ) . Similar experiments including variations of UV light intensities ( 10–100% ) revealed a strong impact on the photoinactivation kinetics – with lower intensities resulting in slower kinetics – but much less on the degree of inhibition ( Figure 3—figure supplement 2e–h ) . Importantly , however , there was no difference in the photomodulation properties ( time-course , degree of current inhibition ) between a UV intensity of 75% and full LED power indicating that the conditions used in most experiments ( 1–5 s of UV at full power ) enabled full photoregulation . Strikingly , although light-induced changes in receptor activity were all relatively fast on a biological time scale , they were surprisingly slow in regard to the ( ultrafast ) chemistry of azobenzene photoisomerization ( see above ) . This difference could stem from protein conformation changes needed to allow isomerization to local interactions with nearby residues of the PSAA embedded into the protein . 10 . 7554/eLife . 25808 . 010Figure 3 . Photomodulation properties of GluN1-P532PSAA/GluN2A receptors . ( a ) Kinetics of photoinhibition and its recovery . Photoinactivation following UV ( 5 s ) was slower ( τ = 520 ms ) compared to the re-activation induced by blue light ( τ = 35 ms ) , as obtained by monoexponential fits . Much slower re-activation kinetics ( τ = 1300 ms ) and an intermediate state of recovery are observed upon green light exposure ( left panel ) . The average exponential time constants for the different wavelengths of light ( 5 s each ) are ( in ms ) : τ365nm = 500 ± 30 ( n = 23 ) , τ460nm = 45 ± 1 ( n = 23 ) , τ520nm = 1350 ± 140 ( n = 5 ) ( right panel ) . ( b ) Representative current trace demonstrating the persistence of photoinhibition following a brief ( 3 s ) UV illumination . ( c ) The cis-isoform , induced by UV ( 1 s ) , inhibited 40% of the current and was stable over minutes . Further co-agonist activation , interrupted by periods in the dark , did not affect the degree of the photoinhibition . A short pulse of blue light ( 1 s ) allowed full recovery to the initial current amplitude . ( d ) Stability of the photoresponse during prolonged UV exposures ( 30 s ) , as shown in the example trace ( left ) . The mean current inhibition degrees ( in % ) following 5 and 30 s of UV are: Inh5s = 42 ± 3 , Inh30s = 41 ± 2 ( n = 3; right panel ) . ( e ) Example trace demonstrating the stability of the photoresponse over 20 PSAA cis-trans toggling events . The degree of receptor modulation remained stable throughout the recording ( left panel ) . On average , the first UV pulse gave a current reduction of 40 ± 1% . The averaged extent of UV inhibition following 20 toggling events was 39 ± 2% ( n = 6; right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 01010 . 7554/eLife . 25808 . 011Figure 3—figure supplement 1 . : The UV/Vis spectra of the different PSAA photoisomers . ( a ) UV/Vis spectrum recorded in the extracellular ( Ringer ) solution after 5 min of illumination at 365 , 460 , or 525 nm . No difference in the spectra after irradiating at 460 and 525 nm was observed ( here , the 460 and 525 nm spectra are superimposed ) . Typically , two absorbance peaks were observed representing π-π* and n-π* transitions . Upon trans-to-cis isomerization the π-π* absorption decreased , while the n-π* absorption increased . ( b ) As in a , but recorded in isopropanol after 5 min of illumination at 365 , 460 , or 525 nm . Here , irradiating at 525 nm gave a higher cis/trans ratio ( 69% ) compared to irradiation at 460 nm ( 74% ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 01110 . 7554/eLife . 25808 . 012Figure 3—figure supplement 2 . Impact of UV duration and intensity on the photomodulation properties of GluN1-P532PSAA/GluN2A receptors . ( a ) Example current trace of the dependence of receptor photoinactivation on the UV duration . Different lengths of UV pulses between 100 ms and 5 s at full power were given . ( b ) The current inhibition degrees ( in % ) following illumination with different UV durations are: Inh5s = 47 ± 1 ( n = 5 ) , Inh1s = 47 ± 2 ( n = 5 ) , Inh500ms = 37 ± 2 ( n = 5 ) , Inh250ms = 28 ± 3 ( n = 3 ) , Inh100ms= 17 ± 2 ( n = 4 ) . At least 1 s of UV light is required to induce the maximal inhibitory effect . ( c ) Photoinhibition kinetics . The mean exponential time constants for the different UV durations used are: τ5s = 370 ± 30 ( n = 5 ) , τ1s = 340 ± 20 ( n = 5 ) , τ500ms = 250 ± 10 ( n = 5 ) , τ250ms = 150 ± 20 ( n = 3 ) , τ100ms = 140 ± 10 ( n = 4 ) . ( d ) The photoinhibition ( in % ) is correlated to the inactivation time constants ( in ms ) obtained using different lengths of UV pulses ( R2 = 0 . 92 ) . ( e ) Example current trace demonstrating the dependence of the photoinactivation on the UV light intensities . ( f ) The intensity of the UV-LED did not impact the degree of current inhibition . The mean inhibition extents ( in % ) are: Inh100% = 40 ± 3 ( n = 6 ) , Inh75% = 40 ± 1 ( n = 4 ) , Inh50% = 38 ± 3 ( n = 5 ) , Inh25% = 40 ± 3 ( n = 6 ) , Inh10% = 37 ± 3 ( n = 6 ) . ( g ) In contrast , the UV-LED intensity strongly impacted the inactivation kinetics . Mean values of exponential time constants are ( in ms ) : τ100% = 400 ± 40 ( n = 6 ) , τ75% = 490 ± 90 ( n = 4 ) , τ50% = 660 ± 40 ( n = 5 ) , τ25% = 1300 ± 160 ( n = 6 ) , τ10% = 2100 ± 240 ( n = 6 ) . ( h ) The current inhibition ( in % ) shows no correlation to the photoinactivation time constants for the different intensities of UV used ( R2 = 0 . 52 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 01210 . 7554/eLife . 25808 . 013Figure 3—figure supplement 3 . Absence of photoresponses in GluN1/GluN2A WT receptors . ( a ) Example current trace showing that a short pulse ( 1 s ) of UV or blue light does not affect the activity of WT receptors . Overall , the current amplitude remained stable over minutes . The orange line indicates the minor peak current run down during the long-lasting recording . ( b ) Equally , long exposures of UV ( 30 s ) did not modulate WT receptor activity . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 013 We next assessed the robustness and stability of the photoresponses . Once initiated by a short pulse of UV , the photoinactivation remained stable for extended periods of time ( minutes ) , well beyond the presence of UV light and irrespective of the presence or absence of agonists ( Figure 3b and c ) . Thus , the slow spontaneous thermal cis-to-trans isomerization of azobenzene as seen in solution ( Beharry and Woolley , 2011; Hoppmann et al . , 2011 ) is preserved in the context of the protein-embedded PSAA . The photoinactivated state was also highly stable during elongated exposures to UV ( 30 s; Figure 3d ) . Following such a prolonged photoinactivation , immediate resetting to full receptor activity could still be achieved by a brief pulse of blue light , demonstrating minimal photobleaching of the azobenzene moiety and lack of phototoxicity . Again , similar light protocols performed on WT receptors showed no photosensitivity ( Figure 3—figure supplement 3 ) . Finally , repetitive cycles of illumination showed that the photoresponses could sustain multiple rounds of cis-to-trans photoisomerization with no detectable fatigability ( Figure 3e ) . In summary , all observed photoresponses were fully reversible , thermally bi-stable , and displayed reproducible kinetics in a millisecond-to-second time range . The GluN1/GluN2 ABD dimer interface is a central structural determinant of NMDAR activity , which impinges on several key receptor properties , including agonist sensitivity , deactivation kinetics , as well as channel open probability ( Furukawa et al . , 2005; Gielen et al . , 2008; Borschel et al . , 2011 ) . We therefore determined the influence of GluN1-P532PSAA photoswitching on NMDAR gating parameters . To examine the channel maximal open probability ( Po ) , we measured the kinetics of current inhibition by the open channel blocker MK-801 , a method classically used to index receptor channel Po ( Zhu et al . , 2014 ) . As indicated by the modestly slower MK-801 blocking rate , the basal receptor Po was slightly reduced by introducing the PSAA at the GluN1-P532 site per se ( dark state; ~1 . 3 fold slower kinetics; p=0 . 27 ) . Critically , under UV exposure , MK-801 inhibition kinetics occurred at ~2 fold slower rates , indicating a pronounced decrease in Po in the photoinactivated state of PSAA-mutant channels ( Figure 4a and Figure 4—figure supplement 1 ) . In contrast , no change in Po was observed for WT receptors upon exposure to UV or blue light ( Figure 4—figure supplement 1 ) . Thus , the PSAA introduction and modulation by light at position GluN1-P532 confers direct control of receptor channel Po . 10 . 7554/eLife . 25808 . 014Figure 4 . Reversible photomodulation of channel activity and glycine sensitivity of GluN1-P532PSAA/GluN2A receptors . ( a ) Normalized example current traces demonstrating the distinct MK-801 inhibition kinetics in the dark and UV-state ( τDARK = 470 ms , τUV = 3200 ms; left panel ) . Pooled values ( right panel ) under various light conditions for WT and PSAA-mutant receptors ( *p=0 . 016 ) . ( b ) Glycine dose-response traces under UV or blue illumination ( left panel ) . Glycine DRCs ( right panel ) for WT ( dashed line ) and PSAA-mutant ( full lines ) in the dark ( black line ) or under UV or blue light . EC50 values are ( in μM ) : EC50 , DARK = 0 . 5 ( n = 5 for each point ) for WT and EC50 , DARK = 1 . 5 ( n = 3–12 ) , EC50 , BLUE = 1 . 5 ( n = 3–8 ) , EC50 , UV = 10 . 5 ( n = 3–7 ) for PSAA-mutant receptors . ( c ) As in b , sensitivity to glutamate . EC50 values are: EC50 , DARK = 4 . 4 ( n = 8–9 ) for WT and EC50 , DARK = 14 . 6 ( n = 4–11 ) , EC50 , BLUE = 13 . 3 ( n = 3–8 ) , EC50 , UV = 17 . 1 ( n = 3–8 ) for PSAA-mutants . ( d ) Representative current trace demonstrating the photomodulation properties under various glycine concentrations ( 1 , 10 , 100 or 1000 μM ) . The mean photoinhibition degrees are ( in % , right panel ) : Inh1 = 94 . 3 ± 2 ( n = 3 ) , Inh10 = 76 . 1 ± 2 ( n = 8 ) , Inh100 = 49 . 9 ± 2 . 5 ( n = 8 ) , Inh1000 = 33 . 5 ± 2 ( n = 3; ***p<0 . 001; **p<0 . 01 ) . ( e ) Glycine deactivation kinetics . Example trace showing glycine dissociation under UV or blue light ( τdeact , UV = 75 . 4 ms , τdeact , BLUE = 131 . 3 ms; left panel ) . The mean weighted deactivation time constants are ( in ms; right panel ) : τdeact , DARK = 184 ± 16 ( n = 5 ) and τdeact , UV = 166 ± 30 ( n = 4; p=0 . 69 ) for WT , and τdeact , DARK = 111 ± 9 ( n = 6 ) and τdeact , UV = 81 ± 4% ( n = 7; **p<0 . 01 ) for PSAA-mutants . ( f ) As in e , for glutamate ( τdeact , UV = 77 . 7 ms , τdeact , BLUE = 79 . 2 ms ) . The deactivation time constants are ( in ms ) : τdeact , DARK = 157 ± 25 ( n = 15 ) and τdeact , UV = 172 ± 30 ( n = 8; p=0 . 92 ) for WT and τdeact , DARK = 113 ± 36 ( n = 5 ) and τdeact , UV = 115 ± 21% ( n = 6; p=0 . 97 ) for PSAA-mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 01410 . 7554/eLife . 25808 . 015Figure 4—figure supplement 1 . MK-801 inhibition kinetics for GluN1-P532PSAA/GluN2A mutant receptors . ( a ) Representative current trace showing the inhibition of WT GluN1/GluN2A receptors by 1 µM MK-801 in the dark . In this example , the receptor activity was inhibited with an exponential time constant of τ = 600 ms ( green line , mono-exponential fit ) . ( b ) As in a , under UV light ( τ = 700 ms ) . ( c ) Inhibition by MK-801 of GluN1-P532PSAA/GluN2A receptors in the dark ( τ = 470 ms ) . ( d ) As in c , under UV light ( τ = 3200 ms ) . ( e ) Summary of the single-exponential time constants of MK-801 inhibition under different light conditions ( normalized values are shown in Main Figure 4a ) . For WT , the inhibition half-times are ( absolute values , in ms ) : τDARK = 690 ± 140 ( n = 5 ) , τUV=620±50 ( n = 7 ) , τBLUE=610±90 ( n = 5 ) . For the P532PSAA mutant , the values are: τDARK=940±160 ( n = 9 ) , τUV=1820±290 ( n = 12; *p=0 . 016 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 015 To explore the inter-relationship between agonist occupancy and photomodulation , we next generated full agonist dose-response profiles in the dark or under UV or blue light . Introducing the UAA per se induced a similar reduction of both glutamate and glycine sensitivity , compared to WT receptors ( ~3 fold increase in EC50; Figure 4b and c ) . Upon illumination , however , a striking difference between the two co-agonists arose . The sensitivity to glutamate was found to be equal in all light conditions tested ( dark , UV , and blue ) . In contrast , the sensitivity to glycine was profoundly affected by the change in wavelength . Specifically , inducing the cis-form by UV strongly decreased the sensitivity to glycine ( EC50 increased by ~20 fold compared to WT receptors ) , an effect that could be subsequently reversed with blue illumination . In accordance with these light effects on glycine potency , the extent of UV-induced current inhibition was highly dependent on glycine concentrations , displaying an inverse correlation ( the lower the concentration the larger the photoinhibition; Figure 4d ) . Remarkably , at 1 µM glycine , the reduction in current amplitude was extensive , approaching nearly complete receptor silencing ( 94 . 3 ± 1% inhibition [n = 3] ) . Finally , to gain further insights into the effects of azobenzene photoswitching on the agonist-receptor interaction , we measured glutamate and glycine deactivation kinetics following short ( 800 ms ) jumps into saturating agonist concentrations under different light exposures . Glycine deactivation kinetics was significantly fastened when switching from dark to UV , in good agreement with the decreased glycine sensitivity observed at equilibrium ( see Figure 4b ) . This effect was fully reversed with blue light ( Figure 4e ) . For glutamate , the deactivation kinetics were indistinguishable between dark , UV , and blue states ( Figure 4f ) , as expected from the insensitivity of the glutamate dose-response curves to light ( see Figure 4c ) . Overall , these results show that photoswitching the azobenzene side chain at GluN1-P532 to the cis-state inhibits receptor activity through a dual effect , a drop in gating efficacy ( decrease in channel Po ) and a reduction in co-agonist affinity ( increase in glycine EC50 ) . Also , these findings demonstrate the power of PSAAs to optically manipulate key receptor gating properties in a precise and reversible manner . We next targeted the NTD region for PSAA introduction . This region , which lies the most distal to the channel pore , forms a major regulatory region that critically influences NMDAR gating and pharmacology in a subunit-specific manner ( Paoletti , 2011; Gielen et al . , 2009; Yuan et al . , 2009 ) . In particular , we previously showed that introduction of the irreversible photocrosslinker AzF at the GluN1-Y109 site at the NTD upper lobe dimer interface allows to irreversibly photoinhibit GluN1/GluN2B , but not GluN1/GluN2A receptors ( Zhu et al . , 2014 ) . Building on this , we turned back to our GluN1-Y109 Amber cDNA mutant and co-expressed it with either the GluN2A or GluN2B subunit and in the presence of the PSAA . Hereby , UV exposure resulted in inhibition of activity of both types of receptors , an effect that was reversed by blue light illumination ( Figure 5 ) . Importantly , however , the photoinhibition profiles were markedly distinct between the two mutants , with a modest reduction in current amplitude for mutant GluN2A receptors ( 21 ± 3% [n = 6] ) , and a profound inhibition for mutant GluN2B receptors ( 79 ± 2% [n = 10]; Figure 5 ) . Therefore , rearrangements at the GluN1-GluN2B NTD dimer interface can directly modulate the level of receptor activity , in line with recent structural studies ( Tajima et al . , 2016 ) . Also , subtle conformational perturbations at this distal interface , triggered by minimal shape modification of a single azobenzene side chain , are sufficient to alter the downstream gating machinery . Finally , these results point to important structural and functional differences in the NTD region between GluN2A and GluN2B receptors ( Zhu et al . , 2014; Tian and Ye , 2016; Romero-Hernandez et al . , 2016 ) , the two main NMDAR subtypes in the adult brain . 10 . 7554/eLife . 25808 . 016Figure 5 . Reversible and subunit-specific photocontrol of GluN1-Y109PSAA/GluN2A receptors . ( a ) Structure of a heteromeric GluN1/GluN2B receptor ( left panel ) with PSAA inserted at the Amber mutation site GluN1-Y109 ( green spheres ) at the NTD heterodimer interface . The top view of one ABD dimer highlights the location of the PSAA insertion ( right panel ) . ( b ) Representative current trace of GluN1-Y109PSAA/GluN2B receptors . In this example , the reduction of activity upon UV was 78% . ( c ) As in b , but for GluN1Y109PSAA/GluN2A receptors . Here , the UV-mediated photomodulation was only 28% . ( d ) Summary of peak current inhibition degrees upon exposure to UV for the GluN1-Y109PSAA mutant co-expressed with either GluN2A or GluN2B . On average , the photoinactivation ( in % ) is: InhGluN2B = 79 ± 2 ( n = 10 ) , InhGluN2A = 21 ± 3 ( n = 6 ) , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 016 With the major advantage of site tolerance using UAAs , we next placed our photoswitchable probe to various prominent sites within the TMD ( Table 1 ) . Primarily , we chose aromatic positions to maintain steric properties within the ion channel region and concentrated on the upper part of the TMD , where the channel gate resides ( Chang and Kuo , 2008; Karakas and Furukawa , 2014; Lee et al . , 2014 ) . We identified a cluster of bulky hydrophobic residues sitting within a TMD cavity at the back of the M3 bundle crossing , to be tolerant to PSAA introduction and to be highly responsive to light exposure ( Figure 6a ) . This cluster includes four GluN1 residues – F554 ( in the pre-M1 loop ) , W563 ( in the M1 helix ) , and Y647 and F654 ( in the M3 helix ) – most of which are highly conserved within the iGluR family ( Alsaloum et al . , 2016 ) ( Figure 6—figure supplement 1 ) . Exposure to UV light with the PSAA placed at one of these sites invariably produced a potentiation of receptor activity , with a site-dependent degree of current amplitude increase . This effect could be fully and rapidly reversed by shining blue light ( Figure 6b–e ) . The photopotentiation observed for the TMD mutants contrasts the photoinhibition previously seen at NTD and ABD sites , thus demonstrating that PSAA-mediated receptor tuning can be bi-directional . The light-induced potentiation effects were particularly massive at GluN1-Y647 and GluN1-F654 sites ( fold potentiation , 4 . 3 ± 0 . 4 [n = 14] and 3 . 9 ± 0 . 2 [n = 12] , respectively ) , both fundamental parts of the M3 SYTANLAAF motif sequence ( Figure 6—figure supplement 1 ) . For these mutants , ‘leakage’ through unspecific Amber codon suppression was absent or minimal ( Figure 6—figure supplement 2 ) and the extent of photopotentiation was similar for all states tested ( with UV applied in the resting , active , or ‘mixed’ state; Figure 6f ) . This demonstrates that the cis-trans azobenzene interconversion occurs with similar efficacy irrespective of the receptor’s functional state . Quantifying the on-relaxation kinetics of current potentiation , we observed marked site-dependence within the hydrophobic amino acid cluster ( Figure 6g ) , indicating that the photopotentiation effect is strongly influenced by the local steric environment . Finally , in accordance with the previously described photoresponse properties at extracellular sites , the TMD photoresponses invariably displayed high ( bi- ) stability , full reversibility , and excellent repeatability ( Figure 6h and i ) . 10 . 7554/eLife . 25808 . 017Figure 6 . Reversible NMDAR photopotentiation with PSAA at various pore sites . ( a ) Structure of the GluN1/GluN2 TMD represented by a side ( left panel ) and top view ( right panel ) . The four sites of the hydrophobic cluster , where PSAA was incorporated , are illustrated as spheres in different colors . ( b ) Example recording for GluN1-647PSAA/GluN2A mutants . UV illumination in the presence of the co-agonists resulted in a 3 . 4-fold potentiation of receptor activity . ( c ) Same cell as in b , with light applied in the resting state . Here , the UV-driven potentiation was 3 . 9-fold . ( d ) Same cell as in b , with light applied in the ‘mixed’ states . In this example , the current increased by 5-fold . ( e ) As is b , for the GluN1-654PSAA/GluN2A mutant . Here , UV light induced a 4 . 6-fold potentiation . ( f ) The fold-potentiation of activity differed between the mutants , but not between their functional states . On average , the photopotentiation in the active state is: PotF554 = 1 . 4 ± 0 . 1 ( n = 5 ) , PotW563 = 2 . 3 ± 0 . 1 ( n = 3 ) , PotY647 = 4 . 3 ± 0 . 4 ( n = 14 ) , PotF654 = 3 . 9 ± 0 . 2 ( n = 12 ) . UV in the resting state resulted in PotY647 = 3 . 4 ± 0 . 2 ( n = 5 ) and PotF654 = 3 . 6 ± 0 . 5 ( n = 6 ) ; and in the ‘mixed’ state in PotY647 = 3 . 9 ± 0 . 3 ( n = 8 ) and PotF654 = 3 . 7 ± 0 . 5 ( n = 6 ) . The dashed line indicates a potentiation of 1 ( i . e . no potentiation ) . ( g ) Kinetics of photopotentiation . The mean photopotentiation exponential time constants are ( in ms ) : τF554= 1640±170 ( n = 4 ) , τW563 = 240 ± 20 ( n = 3 ) , τY647 = 1020 ± 80 ( n = 14 ) , τF654 = 290 ± 30 ( n = 8 ) . ( h ) Example trace showing that the UV-induced potentiation , once initiated , is highly stable even after UV light has been turned off ( here , for 70 s ) . ( i ) Example trace demonstrating the stability of the photoresponse over numerous trans-cis illumination cycling events . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 01710 . 7554/eLife . 25808 . 018Figure 6—figure supplement 1 . Multiple sequence alignment in the TMD region across multiple iGluR family subunits ( NMDA , AMPA , and kainate receptor subunits ) . A section of the entire sequence around the TMD is shown . The TMD is shown in blue , the helices pre-M1 and M1 to M3 are highlighted as blue boxes . The four highly conserved GluN1 positions of the TMD hydrophobic cluster that were tested in this study are highlighted in red . All subunits are from rat . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 01810 . 7554/eLife . 25808 . 019Figure 6—figure supplement 2 . No or minor unspecific Amber codon suppression at TMD sites . ( a ) Example traces from GluN1-Y647Amber/GluN2A receptors incubated either with ( black trace ) or without ( green ) PSAA in the cell culture medium . No unspecific suppression of the Amber codon by endogenous amino acids was detected for this mutant . UV and blue light were equivalently applied in the –PSAA and +PSAA conditions . In the presence of the PSAA , a typical reversible photopotentiation was observed ( in this example , 3 . 8-fold current potentiation ) . In the absence of the PSAA , the blank recording did not show any increase of current upon UV light . ( b ) As in a , for the GluN1-F654Amber mutant . A minor unspecific background in the minus PSAA-condition ( –PSAA ) was detected . Crucially , this ‘leakage’ current did not show any light-related effects , in contrast to receptors harboring the PSAA at the F654 site ( here , 3 . 7-fold current potentiation ) . ( c ) Pooled peak current amplitudes for both mutants ( 48 hr post transfection ) incubated in the presence or absence of PSAA . Supplementing the medium with the PSAA resulted in average current amplitudes of 84 ± 35 pA ( n = 10 ) for GluN1-Y647Amber/GluN2A receptors . Only a few cells gave no agonist-mediated currents ( n = 3 blanks ) . In the absence of the PSAA , no currents were detected in all cells tested ( n = 12 blanks ) . For the GluN1-F654 Amber mutant , the current amplitudes were 196 ± 142 pA ( n = 8 ) when PSAA was supplemented . The amount of blank recordings was minor ( n = 1 blank ) . Cells lacking the PSAA gave mean current amplitudes of 150 . 5 ± 48 pA ( n = 6 ) , six further cells did not express at all ( n = 6 blanks ) . UV or blue light never modulated mutant receptor activity with unspecifically incorporated endogenous amino acids . In contrast , receptors showed a 3 . 2 ± 0 . 3 fold potentiation with PSAA at the Y647 position ( n = 3 ) , and a 4 ± 0 . 2 fold potentiation when PSAA was introduced to the F654 site ( n = 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 019 To dissect the mechanism underlying the UV-triggered photopotentiation effect , we then performed single-channel recordings on outside-out patches pulled from cells expressing either GluN1-Y647PSAA or GluN1-F654PSAA receptors , the two mutants showing the strongest UV-mediated effects . As a control , we first verified that WT GluN1/GluN2A single-channels properties ( level of activity , conductance ) were insensitive to light changes ( dark , UV , blue; Figure 7—figure supplement 1 ) . In accordance to the observations on the macroscopic level , the GluN1-F654PSAA mutant showed strong light responsiveness characterized by a massive increase of channel activity during UV exposure , with no or little change in single-channel conductance ( Figure 7a ) . Quantification using all-points histograms showed that the unitary conductance was unaltered compared to WT channels ( 59 . 6 ± 3 pS [n = 4] for mutant vs 63 . 3 ± 2 . 4 pS [n = 5] for WT , p=0 . 38; Figure 7—figure supplement 1 ) and that the increase in single-channel activity ( N*Po ) could account almost entirely for the current enhancement seen at the whole-cell level ( Figure 7c ) . On the other hand , the behavior of the GluN1-Y647PSAA mutant was strikingly different . A predominant existence of the trans-isomer in the dark or blue state resulted in strongly disrupted single-channel properties , creating low-conductive and excessively brief channel openings . Although all-points amplitude histograms failed to identify well-defined single peaks , detailed inspection of the recordings consistently revealed a few rare openings suggestive of a diminished unitary conductance ( although the channel may not have reached full amplitude because of filtering; Figure 7b ) . Flipping the azobenzene side chain into the bent cis-form upon UV caused a release from this constrained state as manifested by a sudden and dramatic increase in channel activity ( Figure 7c ) . This large increase in activity was accompanied by the appearance of typical WT-like large conductance openings ( Figure 7b ) . Whole-cell noise analysis further supported light-triggered modifications of the single-channel conductance of GluN1-Y647PSAA receptors , but not of GluN1-F654PSAA channels ( Figure 7—figure supplement 2 ) . 10 . 7554/eLife . 25808 . 020Figure 7 . Optical modulation of receptor channel gating and permeation . ( a ) Example single-channel recording from an outside-out patch expressing GluN1-F654PSAA/GluN2A mutant receptors in the presence of 100 µM glutamate and glycine and exposed to different light conditions ( dark , UV , blue ) . All-points amplitude histograms were produced from 11 s of total recording time for each light condition and fitted with multiple Gaussian components ( black lines ) , allowing calculation of single-channel elementary conductance ( γel ) and of receptor activity ( N*Po ) . In this example , the values for γel are ( in pS ) : γel , DARK = 64 . 3 , γel , UV = 66 , γel , BLUE = 65 . Here , applying UV light resulted in a 2-fold increase in N*Po compared to the dark state . Data filtered at 2 kHz for illustration . ( b ) As in a , for the GluN1-Y647PSAA/GluN2A mutant . Glutamate concentration was 0 . 05 µM , glycine 100 µM . Values of γel in UV is 69 . 4 pS . Here , applying UV light resulted in a 5 . 8-fold increase in N*Po compared to the dark state . Top panel , inset: example single-channel openings from WT and GluN1-Y647PSAA mutant in the dark . ( c ) Pooled change in N*Po between UV and dark state for WT and mutant receptors . Values are ( from left to right ) : 0 . 9 ± 0 . 08 ( n = 5 ) , 3 . 6 ± 0 . 63 ( n = 4 ) and 5 . 3 ± 0 . 56 ( n = 5 ) . ( d ) Example magnesium dose-response traces from GluN1-Y647PSAA/GluN2A receptors under UV or blue light ( left panel ) . Mg2+-DRCs under the three different light conditions ( right panel ) . IC50 values are ( in μM ) : IC50 , DARK = 103 ( n = 4–6 for each point ) ; IC50 , BLUE = 102 ( n = 3–7 ) ; IC50 , UV = 41 . 5 ( n = 4–10 ) . The dashed line corresponds to the magnesium sensitivity of WT receptors in the dark . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 02010 . 7554/eLife . 25808 . 021Figure 7—figure supplement 1 . Wild-type GluN1/N2A receptors do not show any light-dependent modulation of single-channel properties . ( a ) Example single-channel recording of an outside-out patch expressing WT GluN1/N2A receptors in the presence of 100 µM glycine and 0 . 05 µM glutamate and exposed to different light conditions ( dark , UV , blue ) . All-points amplitude histograms were produced from 11 s of total recording time for each light condition . Histograms were fitted with multiple Gaussian components ( black lines ) , the amplitude of which enables the calculation of single-channel elementary conductance ( γel ) and receptor activity ( N*Po ) in the dark , UV , and blue state . In this example , the values for the conductance are ( in pS ) : γel , DARK = 54 . 8 , γel , UV = 59 . 1 , γel , BLUE = 58 . 1 . There was no major change in regard of the channel activity when switching wavelength illumination . In this patch , a slow and continuous decrease in channel activity was observed throughout the recording ( ‘run-down' ) . The data were filtered at 2 kHz for illustration . ( b ) Pooled elementary conductance levels . For WT , the averaged values are: γel , DARK = 63 . 3 ± 2 . 4 ( n = 5 ) , γel , UV = 64 . 1 ± 1 . 7 ( n = 5 ) , γel , BLUE = 61 . 5 ± 2 . 2 ( n = 3 ) . For the GluN1-F654PSAA mutant: γel , DARK = 59 . 6 ± 3 ( n = 4 ) , γel , UV = 60 . 7 ± 2 ( n = 4 ) , γel , BLUE = 60 . 4 ± 3 . 4 ( n = 3 ) . For the GluN1-Y647PSAA mutant , only conductance levels under UV illumination could be obtained: γel , UV = 66 . 4 ± 1 . 8 ( n = 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 02110 . 7554/eLife . 25808 . 022Figure 7—figure supplement 2 . Estimation of the single-channel conductance by whole-cell noise analysis . ( a ) Example of a macroscopic current as typically obtained from GluN1-Y647PSAA/GluN2A receptors . This trace was sampled at 20 kHz and low-pass filtered at 2 kHz ( for reasons of illustration ) . The cell was kept in the dark and exposed to UV for 8 s each ( in the presence of saturating co-agonists ) . In this example , the calculated single-channel conductances , as obtained from the whole-cell noise ( low-pass filter at 10 kHz ) , are ( in pS ) : γnoise , DARK = 24 . 2 , γnoise , UV = 54 . The ratio between γnoise , UV and γnoise , DARK is Rγ=2 . 23 for this particular cell . ( b ) Pooled conductance levels of GluN1-Y647PSAA/GluN2A receptors that were obtained from whole-cell noise recordings sampled at 20 kHz and low-pass filtered at 10 kHz . The averaged values are: γnoise , DARK = 26 . 9 ± 5 . 7 , γnoise , UV = 56 . 7 ± 5 . 7 ( n = 4; ***p<0 . 001 ) . ( c ) The averaged ratio between γnoise , UV and γnoise , DARK for GluN1-Y647PSAA/GluN2A receptors is Rγ=2 . 1 ± 0 . 2 ( n = 4 ) , and Rγ=1 . 1 ± 0 . 3 ( n = 6 ) for GluN1-F654PSAA/GluN2A receptors ( ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 02210 . 7554/eLife . 25808 . 023Figure 7—figure supplement 3 . External magnesium block is not affected by light for WT and GluN1-F654PSAA receptors . ( a ) Example magnesium dose-response traces under illumination of either UV or blue light for WT GluN1/GluN2A receptors . ( b ) Magnesium dose-response curves ( Mg2+-DRCs ) for WT receptors , obtained from traces as in a . IC50 values are ( in µM ) : IC50 DARK = 34 . 6 ( n = 4–7 ) ; IC50 BLUE = 32 . 6 ( n = 3–7 ) ; IC50 UV = 35 . 8 ( n = 4–7 ) . ( c ) As in a , for the GluN1-F654PSAA mutant . ( d ) Mg2+-DRCs for GluN1-Y654PSAA/GluN2A receptors , obtained from traces as in c . IC50 values are ( in µM ) : IC50 DARK = 30 ( n = 3–7 ) ; IC50 BLUE = 34 . 3 ( n = 3–6 ) ; IC50 UV = 33 . 3 ( n = 5–8 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 02310 . 7554/eLife . 25808 . 024Figure 7—figure supplement 4 . MK-801 inhibition kinetics of TMD mutant receptors . ( a ) Representative trace showing the inhibition of GluN1-Y647PSAA/GluN2A receptors by 1 µM MK-801 in the dark state . In this example , the receptor activity was inhibited with an exponential time constant of 4400 ms ( green line , mono-exponential fit ) . ( b ) As in a , under UV light ( τ = 3300 ms ) . ( c ) Inhibition by MK-801 of GluN1-F654PSAA mutants in the dark ( τ = 3970 ms ) . ( d ) As in c , under UV light ( τ = 1000 ms ) . ( e ) Summary of the single-exponential time constants of MK-801 inhibition at different light conditions . The mean values are ( absolute values , in ms ) : τDARK = 5070 ± 580 ( n = 5 ) and τUV = 4150 ± 910 ( n = 7; p=0 . 41 ) for the GluN1-Y647PSAA mutant;τDARK = 3070 ± 560 ( n = 5 ) and τUV = 730 ± 70 ( n = 9; ***p<0 . 001 ) for the GluN1-F654PSAA mutant . ( f ) The mean extent of block produced by MK-801 in the dark and UV state . Values are ( in % ) : for GluN1-Y647PSAA mutant , BlockDARK = 69 ± 5 ( n = 5 ) , BlockUV = 86 ± 3 ( n = 7; *p=0 . 02 ) ; for the GluN1-F654PSAA mutant , BlockDARK = 90 ± 3 ( n = 5 ) , BlockUV = 97 ± 1 ( n = 9; p=0 . 07 ) . WT GluN1/N2A receptors were nearly entirely blocked by MK-801 with: BlockDARK = 97 ± 1 ( n = 5 ) , BlockUV = 98 ± 1 ( n = 7; p=0 . 53 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 024 We gained further evidence that photoswitching GluN1-Y647PSAA directly impacts the channel permeation properties by investigating the photosensitivity to two well-characterized NMDAR pore blockers , Mg2+ and MK-801 . While GluN1-Y647PSAA receptors displayed similar Mg2+-sensitivity as WT receptors under UV light ( IC50 , UV = 41 . 5 µM [n = 4–10] vs IC50 , DARK = 34 . 6 [n = 4–7] for WT ) , their Mg2+-sensitivity was significantly decreased under dark or blue light , as reflected by the 2 . 5-fold increase in IC50 ( IC50 , DARK = 103 µM [n = 4–6]; IC50 , BLUE = 102 µM [n = 3–7]; Figure 7d and Figure 7—figure supplement 3a and b ) . These effects were fully reversible , allowing rapid alternations between high and low Mg2+-sensitivities in the same receptor population by simple PSAA toggling . On the other hand , Mg2+ inhibited WT and GluN1-F654PSAA receptors with similar potency and in a light-independent manner ( Figure 7—figure supplement 3 ) . Because Mg2+-inhibition is largely independent of channel activity ( Qian et al . , 2002 ) , the light-induced changes in Mg2+-sensitivity of GluN1-Y647PSAA mutants are likely attributable to changes in how Mg2+-ions interact with the pore . Similarly , inhibition of GluN1-Y647PSAA receptors by the large organic cation MK-801 revealed a pronounced photosensitivity . While under UV current the extent of inhibition by MK-801 ( 1 µM ) was >85% , switching the azobenzene side chain to trans resulted in a marked reduction of MK-801 channel block ( 69% maximal inhibition; Figure 7—figure supplement 4 ) . For GluN1-F654PSAA receptors , the degree of MK-801 block was also affected by light , although to a lesser extent , and reached values closer to WT levels ( ≥90%; Figure 7—figure supplement 4c–f ) . These results , combined with the single-channel and noise analysis data , indicate that the ‘ease’ with which ions flow through or block the GluN1-Y647PSAA channel pore can be manipulated by light . They also identify GluN1-Y647 as a novel residue critically involved in determining the energetics and size of the channel open state ( see Discussion ) .
In this study , using a novel approach of chemical optogenetics , we have generated a family of NMDARs that can be accurately and reversibly controlled by light . We installed light sensitivity by genetically encoding azobenzene-containing UAAs that act as site-specific nanoscale photoswitches . We demonstrate that PSAA incorporation occurs with high fidelity and endows robust photoregulation combining full reversibility , high temporal precision , and molecular specificity . We identified several positions in different receptor regions that allowed to either photoinhibit or photopotentiate receptor activity through various mechanisms . Remarkably , these effects were driven by a minimal photoswitching event involving toggling between the elongated trans- and the bent cis-isoform of single azobenzene side-chains . The changes of receptor activity , detected on both the macroscopic and the single-channel level , were invariably reversible , sharply time-locked to the remote illumination events , and remarkably stable over long periods ( minutes ) . By offering a novel approach to gain site-specific and reversible photocontrol on target proteins , our work significantly expands the repertoire of useful tools to engineer and manipulate biomolecules . Using this PSAA methodology , we were able to demonstrate its utility on studying NMDAR mechanisms , unveiling the contribution of specific side chains to distinct receptor properties such as agonist sensitivity , channel open probability , and permeation . Reviewing our entire set of mutations , we found PSAA to be successfully incorporated in 50% of the cases ( 13/27 positions; Table 1 ) , highlighting the feasibility and robustness of the methodology . PSAA azobenzene moieties are bulkier compared to natural substitutions as utilized in classical mutagenesis , nonetheless , our high rate of incorporation , as demonstrated by electrophysiologically detectable receptor functionality , argues for their overall good tolerance and comparable insertion to classical amino acid substitutions . Overall , the measured peak currents from the functional Amber mutant receptors tested were sufficiently large and stable , to enable long-lasting optical experiments ( see Figure 2—figure supplement 1 ) . Comparing the group of functional Amber mutants tested between each other , we have observed stop codon sites that were better tolerated , resulting in higher expression levels and comparable Po properties to wild-type channels ( e . g . GluN1-P532 ) . Other positions ( e . g . those within the TMD hydrophobic cluster ) showed a higher invasiveness by PSAA , producing ion channels with lower expression rates and disrupted channel open probabilities compared to wild-type receptors ( all of these changes were photoisomerization state-dependent however ) . Generally , the genetically-encoded single-arm photoswitch , as provided by PSAA , facilitates the functional screening process , since it does not require residue proximity ( as for cysteine disulfide bridges ) nor any posttranslational protein labelling process ( as for photoswitchable tethered ligands or PTLs ) . Some of the Amber stop codon sites tested did not tolerate PSAA incorporation at all , as detected by the lack of agonist-induced currents ( see Table 1 ) . For instance , many GluN2 mutants did not lead to functional receptors , as detected by the lack of agonist-induced currents . In particular , sites within GluN2A and GluN2B homologous to the permissive GluN1-P532 or -Y535 positions at the ABD dimer interface were not tolerated . A hydrophobic pocket , located at the GluN2A ABD hinge , is required to stabilize this interface through intrusion of the GluN1-Y535 residue ( Furukawa et al . , 2005; Yi et al . , 2016 ) . This hydrophobic pocket is absent on the GluN1 subunit though . We thus speculate that introduction of a bulky hydrophobic PSAA in the GluN2 ABD hinge prevents proper receptor assembly by disrupting ABD dimerization through steric clashes . We note that PSAA introduction at two other extracellular positions within GluN2 ( GluN2A-Y281 and GluN2B-Y282 ) , both not situated at any inter-subunit contact sites or exposed interfaces , were tolerated and folded into functional receptors . Recapitulatory , the use of PSAAs , as shown in our study , extends the properties of classical mutagenesis with a remote , real-time and dynamic control of side-chain geometry , thus adding a critical novel dimension to protein mutagenesis . PSAA introduction enabled rapid and reproducible light-dependent toggling of the receptor between two functional states . Importantly , addition of this new light-controllable allosteric switch occurred without compromising the overall receptor function , in particular its gating machinery . Therefore , PSAA provides an artificial control mechanism that is orthogonal to the natural one provided by evolution ( ligand binding ) , an attractive feature for biological investigations . Introduction of the PSAA to the P532 Amber site in the GluN1 ABD hinge resulted in a pronounced , although incomplete ( ~50% inhibition at saturating agonist concentrations ) , receptor inhibition upon UV exposure . This effect was fully reversible with blue light , allowing temporary precise alternations between a fully and a partially activated state of the receptor . The ability to reversibly control the receptor channel activity by simple photoswitching of an azobenzene side chain located far from the ion channel pore provides a striking example of how highly localized atomic scale conformational changes can propagate through long distance to impact protein functionality . Our nanoscale photoswitch installed at the ABD dimer interface also echoes recent pharmacological studies showing that side chain mobility in this region is instrumental in mediating action of positive and negative allosteric modulators and their respective stabilization of the active or inhibited receptor state ( Hackos et al . , 2016; Yi et al . , 2016 ) . A straightforward explanation of the partial nature of UV inhibition of GluN1-P532PSAA receptors is the presence of mixed subunits harboring either a PSAA or a natural amino acid , the later contributing to light-insensitiveness . This possibility could be excluded , however , based on our control experiments performed in the absence of the UAA showing negligible unspecific background ( at the GluN1-P532 site and all other positions tested; Figure 2—figure supplement 3 and Figure 6—figure supplement 2 ) . Since the absorption spectra of the azobenzene trans- and cis-states overlap substantially ( Beharry and Woolley , 2011; Hoppmann et al . , 2014 , 2011 ) , UV irradiation produces a photostationary state with <100% of the cis-isoform . This limited cis-isomer induction is also unlikely to account for the incomplete nature of UV inhibition of maximally-activated GluN1-P532PSAA receptors . Indeed , nearly 95% of photoinhibition was achieved when the glycine concentration was reduced to 1 μM ( Figure 4d ) . This indicates that under our conditions where light duration and intensity are not limiting ( Figure 3—figure supplement 2 ) , the vast majority of the azobenzene side chains had switched to the cis-state following UV exposure , in good agreement with previous estimations with PSAA introduced into small model peptides ( Hoppmann et al . , 2011 ) . Hence , the incomplete UV inhibition of GluN1-P532 receptors unlikely stems from an inefficient PSAA photochemistry . Rather , it likely finds its origin in the biological mechanism underlying the UV photoinhibition . By studying two key gating parameters ( channel activity , agonist sensitivity ) , we show that flipping the GluN1-P532PSAA azobenzene group to the cis-configuration triggers a drop in channel Po accompanied by a reduction in glycine , but not glutamate , sensitivity . Thus , at saturating glycine concentrations , only the decrease in channel Po impacts the receptor activity , resulting in 50% current inhibition . In contrast , at subsaturating glycine concentration , both effects are in play , resulting in greater current inhibition . The strong effect of the GluN1-P532 PSAA trans-cis isomerization on receptor gating further supports the critical importance of the ABD dimer interface in controlling NMDAR activity ( Furukawa et al . , 2005; Gielen et al . , 2008; Borschel et al . , 2011 ) . It also reveals that PSAA at the GluN1-P532 site acts as a photomodulator of glycine binding . The strategic location of GluN1-P532 at the GluN1 ABD interlobe hinge is ideally suited to influence glycine binding by controlling GluN1 ABD clamshell conformational dynamics . We propose that the PSAA in its trans-state ( dark and blue condition ) stabilizes the GluN1 ABD in a closed-cleft ( i . e . active ) conformation . Induction of the cis-isoform destabilizes this active conformation , favoring clamshell opening , which in turn accelerates glycine dissociation ( Figure 8a and b ) . This proposed allosteric mechanism bears striking resemblance with the mode of action of TCN compounds , a family of negative allosteric modulators of glycine binding that bind the GluN1/GluN2A ABD dimer interface and make direct atomic interactions with several GluN1 ABD hinge residues including GluN1-P532 ( Hansen et al . , 2012; Hackos et al . , 2016; Yi et al . , 2016 ) . At glutamatergic synapses , NMDAR activity is strongly dependent on the level of co-agonists ( glycine , D-serine ) in the receptor’s vicinity . Until now , the exact physiological co-agonist concentrations at synaptic sites as well as their dynamic changes during neuronal activity are unknown ( Mothet et al . , 2015 ) . We envision using the glycine dependence of GluN1-P532PSAA receptor photomodulation as an original means to assess glycine levels in the synaptic cleft . Designing a PSAA-containing GluN2 subunit in which glutamate sensitivity could be manipulated would provide another promising optochemical tool to control synaptic strength in a subunit-dependent manner by simple toggling between cis- and trans-configurations . 10 . 7554/eLife . 25808 . 025Figure 8 . Proposed mechanism for photomodulation of the gating and permeation properties of NMDARs carrying PSAAs . ( a ) Toggling of the PSAA azobenzene moiety between the trans- and cis-isomers in dependence of blue or UV light . For simplicity , the two isomers are replaced by symbols as indicated . ( b ) Schematic representation of the mechanism underlying the photomodulation at the ABD level . For clarity , only one dimer is represented and the NTD region is omitted . PSAA in its trans-configuration ( dark and blue condition ) , when placed at the GluN1 ABD clamshell hinge ( P532 position ) , intrudes into a hydrophobic pocket located at the GluN2A ABD hinge region ( Furukawa et al . , 2005 ) . This interaction stabilizes the GluN1 ABD in a closed-cleft conformation ( i . e . active state ) as well as ABD GluN1/GluN2 heterodimer interface , resulting in active ( i . e . open ) channels ( Furukawa et al . , 2005; Gielen et al . , 2008 ) . Flipping the GluN1-azobenzene moiety to the cis-configuration by UV illumination destabilizes this active conformation , favoring GluN1 clamshell opening , which in turn accelerates glycine dissociation . This effect is accompanied by a weakening of the ABD dimer interface resulting in a 50% drop of channel Po ( UV-inhibited state ) . ( c ) Schematic representation of the mechanism underlying photomodulation in the pore domain . The TMD pore region is viewed from the top . PSAA is placed at the GluN1-Y647 site , which is part of a highly conserved hydrophobic cluster that sits at the back of the channel lumen and connects the M3 helix bundle ( channel gate , inner ring ) to the peripheral M1/M4 helix outer ring . PSAA in its extended trans-configuration prohibits adequate channel opening by preventing full expansion of inner M3 ring towards the periphery . The channel is thus locked in a constrained ‘open’ state , highly unstable ( low Po ) and with low conductance . Upon UV , however , activity can be drastically enhanced , an effect that originates from the relief of the steric hindrance allowing the inner M3 ring to accomplish its full movement towards the outer ring ( UV-activated state ) . This motion generates an increase in ion channel Po and conductance . Thus , channel opening is energetically more favorable in the more compact cis-configuration than the rigid and elongated trans-configuration of the PSAA . DOI: http://dx . doi . org/10 . 7554/eLife . 25808 . 025 Exploiting the key advantage of UAA site tolerance , we identified a cluster of conserved aromatic residues in the TMD region that allowed to photocontrol the receptor’s ion channel behavior . These residues , situated in the upper part of the TMD , form a ‘hydrophobic cluster’ facing opposite to the channel lumen and strategically connecting the GluN1 channel gate region ( M3 ) to the peripheral GluN1 M1 helices . Placing the PSAA at sites within this patch resulted in receptors whose activity could be enhanced by light . The photopotentiation was particularly pronounced at GluN1-Y647 and -F654 , two M3 residues sitting directly at the back of the channel gate . Interestingly , despite their similar extent of photopotentiation , the two mutants showed a remarkable difference in their macroscopic current noise level in the photopotentiated state , which was noticeably larger for GluN1-Y647PSAA receptors ( compare current traces in Figure 6b and e ) . Our single-channel recordings reveal that this effect likely originates from the dual increase of channel Po and conductance for the GluN1-Y647PSAA variant , while only channel Po is affected for the GluN1-F654 mutant . Thus , the GluN1 hydrophobic cluster emerges as a novel structural determinant that controls the energetics of channel gating . Although the structural basis of the cis-trans effects on channel Po remains to be elucidated , we speculate that steric hindrance plays a major role . During channel gate opening , the hydrophobic cluster is likely compressed to accommodate the helical packing due to the expansion ( Kazi et al . , 2013; Tajima et al . , 2016 ) of the inner M3 bundle towards the M1/M4 outer ring . Accordingly , the rigid and elongated trans-configuration of the PSAA renders channel opening energetically less favorable than the smaller cis-configuration , thus accounting for the current increase when shining UV light . We speculate that in the case of GluN1-Y647PSAA , which locates deeper in the ion channel pore than GluN1-F654PSAA , steric hindrance imposed by the extended trans-isomer might even have more dramatic effect by preventing full expansion of the M3 bundle . On one hand , this strong structural constraint would greatly decrease the stability of the channel open state , accounting for the very brief openings . On the other hand , ion permeation would become less favorable , thus resulting in the decreased single-channel conductance and pore block effects . Probably , photoswitching to the cis-state relieves much of this constraint allowing the channel to fully dilate and regain its ‘large’ WT-like 60 pS conductance ( Figure 8a and c ) . These results demonstrate the first use of single photo-active amino acids to directly control transmembrane domain motions and ion channel properties . The M2 loop , which sits deep in the pore and forms the region of highest constriction , is well known to have critical role in determining single-channel conductance and Mg2+-block of NMDAR channels ( Wollmuth et al . , 1998; Siegler Retchless et al . , 2012 ) . Our data now reveal that the upper M3 channel gate region can also participate in controlling ion permeation . Finally , our data point to the transmembrane cavity between the inner M3 ring and the outer M1/M4 ring as a site with strong modulatory potential . A number of allosteric modulators targeting glutamate receptors have been proposed to bind the TMD region of glutamate receptors outside of the channel lumen ( where channel blockers act ) ( Traynelis et al . , 2010; Zhu and Paoletti , 2015 ) . The transmembrane cavity described here represents an ideal locus to accommodate such ligands and fine-tune receptor activity . With its site flexibility , reversibility , and genetic encodability , we anticipate that the PSAA nanoswitch approach described here will prove useful in native settings to optically manipulate neuronal proteins and advance our understanding of the molecular basis of brain function . With the high spatiotemporal precision conferred by light , we are able to control receptors on a time scale ( ms to s ) compatible with their biological activity , and in regard of NMDARs , with their rapid activity-dependent changes in subunit composition , which can occur on a time scale of minutes ( Sanz-Clemente et al . , 2013 ) . Also , the genetic encodability affords high molecular ( i . e . subunit ) specificity and cell-type targeting . Finally , by involving allosteric mechanisms , rather than direct orthosteric interactions as for photoswitchable tethered ligands ( PTLs; Reiner et al . , 2015 ) , the PSAA approach provides the ability to fine-tune receptor activity without interfering with its natural activation . Thus , highly accurate optical manipulation of NMDAR-mediated excitatory signals in defined neuronal circuits should not only enable to resolve the contributions of the two prominent subunits GluN2A and GluN2B to synaptic plasticity , but also of the less common GluN2C and GluN2D subunits . For instance , the photocontrol of GluN2A or GluN2B glutamate deactivation kinetics by PSAAs could assess how this gating parameter sculpts the time window for synaptic plasticity and integration . Similarly , we envision evaluating the contribution of GluN2D ultra slow deactivation kinetics ( Vicini et al . , 1998; Vance et al . , 2012 ) to neuronal excitability , still an unresolved issue . ‘On-demand’ facilitation or inhibition of specific NMDAR subpopulations should also provide critical insights into how NMDAR subunits influence opposite forms of synaptic plasticity ( LTP and LTD; see Yashiro and Philpot [2008] ) . By allowing acute and reversible silencing of specific receptor entities , we expect the PSAA approach to surpass the classical gene knock-out experiments both by its temporal accuracy and lack of compensatory mechanisms . PSAAs should also be well applicable in interrogating the reversing of NMDAR hypofunction in animal disease models . Reduced NMDAR signaling is associated with several neurological disorders , including schizophrenia and cognitive impairments ( Paoletti et al . , 2013; Yuan et al . , 2015; Burnashev and Szepetowski , 2015 ) . Accordingly , there is currently great interest in developing methodologies to boost GluN2A or GluN2B function specifically to determine whether enhancing one pool of receptors , or both , leads to beneficial outcomes . Finally , we foresee great potential for PSAAs , and UAAs in general , for studying intracellular receptor mechanisms and synapse biology . With their major advantage of site tolerance , PSAAs can be incorporated at specific sites in the subunit cytoplasmic tails , which may allow to gain optical control on the receptor association with scaffolding proteins or signaling partners . Real-time photo-manipulation of the synaptic anchoring of NMDARs or their coupling to key cellular messengers ( e . g . GluN2B-CamKII; Sanz-Clemente et al . , 2013 ) would certainly represent a major technical advance for clarifying the contentious roles ( Parsons and Raymond , 2014; Zhou et al . , 2015 ) of NMDAR subunit trafficking and signaling in synaptic plasticity and neuronal survival . The instantaneous and reversible on-off photoswitch provided by PSAAs is bound to afford superior spatiotemporal control of the receptor intracellular interactome compared to conventional approaches based on slowly-acting interfering ligands or genetic modifications . Implementing optochemical approaches to study brain proteins in native situations is challenging ( Kramer et al . , 2013 ) , but feasibility is within reach , be it for photosensitive ligands ( Szobota et al . , 2007; Caporale et al . , 2011; Pittolo et al . , 2014; Lin et al . , 2015; Laprell et al . , 2015; Berlin et al . , 2016; Levitz et al . , 2016 ) or UAAs ( Ernst et al . , 2016; Han et al . , 2017; Zheng et al . , 2017; Chen et al . , 2017 ) . There are two main requirements for in vivo incorporation of UAAs in mammals: ( i ) efficient delivery and expression of the genes encoding the orthogonal tRNA/synthetase pair and the mutated target protein of interest; and ( ii ) sufficient bioavailability of the UAA at the desired tissue or cell type . Successful attempts to address these challenges have recently emerged . In utero electroporation of plasmid DNAs coupled to direct injection of the UAA into the embryonic brain provides a first possible route ( Kang et al . , 2013 ) . Gene transfer using adeno-associated viral ( AAV ) vectors offers an alternative way for efficient delivery of the UAA genetic machinery in mammalian cells , tissues , and brains of living mice ( Ernst et al . , 2016; Zheng et al . , 2017 ) . The generation of transgenic mice with a heritable expanded genetic code - i . e . mice directly incorporating the necessary tRNA/synthetase genes for Amber stop codon suppression in their genome - is the most recent development in the UAA field , which will certainly greatly facilitate in vivo UAA applications in the future . Mouse strains allowing incorporation of the photocrosslinker AzF ( Chen et al . , 2017 ) or the post-translationally modified lysine Nε-acetyl-lysine ( AcK; Han et al . , 2017 ) have just become available . We expect this strategy to be extended to other tRNA/synthetase pairs in a custom-made manner , including the pair utilized in the current study enabling site-specific incorporation of PSAAs . UAA delivery in the whole animal can be achieved by intraperitoneal injection , or , directly into target tissues for spatio-specific induction of UAA-engineered protein expression ( Han et al . , 2017 ) . UAA supplementation in the mouse drinking water provides another easy-to-use and viable option , even when targeting brain proteins as recently shown with a lysine derivative ( Ernst et al . , 2016 ) . Depending on the UAA chemistry , intracranial UAA perfusion ( also performed by Ernst et al . , 2016 ) may be required to reach sufficient UAA levels in neural cells . The bioavailability of PSAAs is currently unknown , but in case of poor brain-blood-barrier crossing , direct brain injection using a multimodal optogenetic cannula for dual drug and light delivery ( Canales et al . , 2015 ) would advantageously combine PSAA supplementation with its light-controlled photoisomerization . The recent development of red-shifted azobenzene derivatives ( Samanta et al . , 2013; Kienzler et al . , 2013 ) , including a PSAA version controllable by visible light ( Hoppmann et al . , 2015 ) , should further enhance biocompatibility and offers promising perspectives for in vivo applications of azobenzene-based photoswitches .
The rat GluN1-1a ( named GluN1 herein ) , rat GluN2A , and mouse GluN2B subunits were expressed from pcDNA3-based vectors . Site-specific Amber codon ( TAG ) mutations for the insertion of the photoswitchable amino acid ( PSAA ) were introduced by means of site-directed Quikchange mutagenesis and confirmed by sequencing , as previously described ( Zhu et al . , 2014 ) . A list of all Amber mutations tested in this study is given in Table 1 . The incorporation of PSAA was driven by an orthogonal tRNA/synthetase pair , derived from a Methanosarcina mazei ( Mm ) pyrolysine tRNA/synthetase pair ( tRNAPyl–MmPSCAA-RS ) and specifically engineered for suppressing the introduced stop mutation ( Hoppmann et al . , 2014 ) . The designed synthetase recognized exclusively the trans-version of the PSAA , which is also the thermodynamically more stable isoform in the dark state . Wild-type eGFP or eGFP carrying an Amber mutation at the Y37 site both served as transfection controls . Wild-type and mutant NMDARs were expressed in HEK-293 cells ( obtained from ATCC Inc . ) . The cells were incubated in DMEM containing 10% FBS and 1% Penicillin/Streptomycin ( complete medium ) . The transfection of NMDAR subunits was performed using polyethylenimine ( PEI ) in a DNA/PEI ratio of 1:3 ( v/v ) , as previously described ( Klippenstein et al . , 2014 ) . For recordings of heteromeric NMDARs , the following vectors were co-transfected: ( i ) a WT subunit ( either GluN1 , GluN2A , or GluN2B ) ; ( ii ) an Amber mutant subunit ( with a TAG codon within GluN1 , GluN2A , or GluN2B ) ; ( iii ) the tRNA/synthetase pair; and ( iv ) eGFP or eGFP-Y37TAG as a control of transfection . Typically , the total amount of DNA was 1 . 5 μg per 0 . 8 cm cover slip and the mass ratio was 1:2:1:1 . PSAA was obtained from two different sources: ( i ) in-house chemical synthesis at UCSF ( laboratory of Lei Wang ) , and ( ii ) customized synthesis by Enamine Ltd . ( Kiev , Ukraine ) . No difference in photoregulation was observed between either compound source . Stock solutions of 30–50 mM PSAA were prepared by dissolving the PSAA through sonication in 0 . 1 N NaOH . Complete medium , supplemented with the final PSAA concentration of 0 . 3–0 . 5 mM ( pH 7 . 3 ) and 150 μM of the competitive antagonist D-APV ( to avoid cell toxicity caused by receptor overexpression ) , was added immediately after transfection . Single-channel currents of outside-out patches expressing TMD-Amber mutant or wild-type receptors were recorded 12–24 hr post transfection . Patch pipettes had a resistance of 10–12 MΩ and were filled with the same solution as for whole-cell recordings . The external perfusion solution was identical to that used in whole-cell recordings except that CaCl2 concentration was lowered to 0 . 5 mM in order to minimize the appearance of NMDAR channel sub-conductance states ( Premkumar et al . , 1997 ) . NMDARs were activated by applying saturating glycine ( 100 μM ) and variable glutamate concentrations ( 100 , 0 . 05 , or 0 . 02 μM ) , depending on the level of channel activity . Recordings were performed at −80 mV . The currents were acquired at a sampling rate of 20–50 kHz and low-pass filtered at 2 kHz . Computer-controlled light pulses during electrophysiological recordings were provided from sensitive high power LEDs ( Prizmatix ) . The three following LEDs were used: Mic-LED-365 ( UV , 200 mW ) , UHP-Mic-LED-460 ( blue , 2W ) and UHP-Mic-LED-520 ( green , 900 mW ) . The LED port was directly coupled via a microscope adaptor to the fluorescence port of an inverted DIAPHOT 300 Nikon microscope . The output beams of the three Mic-LEDs were joined by a dichroic beam combiner module and the resulting light was applied to the center of the recording dish through a 10X objective ( Leitz Wetzlar ) . The light intensity of all three LEDs was set to 100% , unless otherwise noted . Pulses of UV , blue , or green light had a minimal duration of 100 ms . The UV/Vis spectra of the free PSAA ( 100 μM ) were measured using a UVIKON 922 spectrophotometer ( Kontron Instruments ) in 1 x 1 × 4 cm quartz cuvettes . The spectra were measured at two different conditions: ( i ) with PSAA in the extracellular ( Ringer ) solution to be as close as possible to the recording conditions , and ( ii ) in isopropanol to mimic a hydrophobic environment ( since PSAA is buried in the protein ) ( Ye et al . , 2009 ) . To induce the different photoisomers , the dissolved PSAA was directly irradiated in the cuvette at 365 nm ( 30 mW ) , 460 nm ( 241 mW ) , and 525 nm ( 35 mW ) using a Cool LED pE-4000 illumination system . The UV spectra of the pure trans-isomer were recorded upon storage in the dark . Current traces obtained in electrophysiological recordings were transferred to Igor Pro ( version 6 . 3 ) or Clampfit ( version 10 . 5 ) for display and analysis . The kinetics of photoinactivation and photopotentiation were obtained by fitting currents with a single exponential function as follows: Y = A*exp ( -t/τ ) + C , with A as the initial current amplitude , Tau ( τ ) the decay time constant , and C the steady-state level . The inhibition by MK-801 was also fit with a monoexponential function . Although for WT receptors and a subset of mutant receptors ( e . g . the GluN1-P532PSAA mutant ) , double exponential functions yield better fits of the MK-801 inhibition ( as previously described Buck et al . [2000] ) , for the tested TMD mutant receptors with very slow rates of MK-801 channel block ( especially in the dark state ) , double component fits were unable to provide interpretable values . Accordingly , and for comparative purposes , only single exponential fits were used when studying MK-801 inhibition . Agonist DRCs were fit with the following Hill equation: Irel = 1/ ( 1 + ( EC50 / [ago] ) nH ) , with EC50 and the Hill coefficient ( nH ) as free parameters . For Mg2+-DRCs the following equation was used: Irel = 100* ( 1- ( a/ ( 1+[IC50]/[Mg2+ ) nH , ) ) , where a is the maximal current inhibition by Mg2+ . IC50 , a , and the Hill coefficient ( nH ) were set as free parameters . Glutamate and glycine deactivation responses were fitted with a double exponential function with a fast and a slow component as previously described Furukawa et al . ( 2005 ) : Y = Af*exp ( -t/τf ) + As*exp ( -t/τs ) , with Af and As as the amplitudes and τf and τs as the decay time constants of the fast and slow decay components , respectively . The weighted deactivation time constants were calculated with: τw = ( Af/ ( Af+As ) ) *τf + ( As / ( Af + As ) ) *τs . For the analysis of whole-cell noise , currents were sampled at 20 kHz and low-pass filtered at 10 kHz . Noise analysis was performed from segments of at least 6 s of steady currents in the dark and under UV illumination ( ≥6 s each , same cell ) . The apparent single-channel conductance was estimated from the whole-cell noise using the following equation ( Cull-Candy et al . , 1988; Smith et al . , 1999 ) : γnoise = var ( I ) /[ ( mI ) * ( 1-Po ) * ( ΔV ) ] , where mI is the mean current amplitude evoked by the agonists , Po the receptor channel open probability , ΔV the transmembrane voltage driving force ( 60 mV , with a holding potential of −60 mV and assuming a reversal potential of 0 mV ) , and var ( I ) the variance of the current around the mean . Po values in the dark , for mutant ( mut ) GluN1-F654PSAA and GluN1-Y647PSAA receptors , were estimated from MK-801 inhibition kinetics obtained in the current work: Po ( mut , dark ) = [Po ( wt ) ]/[ τon , MK801 ( mut ) /τon , MK801 ( wt ) ] , where Po ( wt ) is the maximal channel open probability of wild-type ( wt ) GluN1/GluN2A receptors ( 0 . 5; Paoletti , 2011 ) . Po values under UV illumination were calculated as follows: Po ( mut , UV ) = mI ( UV ) /mI ( dark ) . This ratio was determined for each individual cell . Relative changes in γnoise ( Rγ ) between UV and dark were then calculated , for each individual cell , using the formula: Rγ = γnoise ( UV ) /γnoise ( dark ) . Statistical significance was assessed with Student's t-test , using either pairwise comparisons for different values from a single cell , or unpaired tests for comparisons between different mutants or cells . Error bars represent the standard deviation of the mean value ( SD ) . For structural representations of the receptor , a homology model was built using the two full-length structures pdb 4PE5 ( Karakas and Furukawa , 2014 ) and pdb 4TLM ( Lee et al . , 2014 ) as templates . | Nerve cells communicate with each other by releasing chemicals , also known as neurotransmitters , from one cell to the next . Once released , these neurotransmitters bind to specific docking stations , called receptors , which are located on the surface of the neighboring cell . Due to changes in neurotransmitter release or the receptor number , the connections between neurons can either strengthen or weaken over time . This process , called synaptic plasticity , forms the basis of learning and memory . One of the key players in synaptic plasticity are NMDA receptors , and if these receptors are faulty , it can cause disorders such as schizophrenia or epilepsy . NMDAs are a large family of receptors that have many receptor subtypes , each with specific properties . Every subtype is composed of four varying subunits . It is still unclear how these different receptor subtypes contribute to synaptic plasticity and new methods are needed to resolve this puzzle . An emerging strategy to study brain receptors is to engineer them so that they can be controlled with light . One approach to provide light-sensitivity uses molecules that act as ‘light switches’ . These switches change their shape when exposed to specific colors of light and this way , turn a receptor on or off . However , commonly used light switches are often very large , meaning that they can only be introduced at specific sites in a receptor , and have limited ability to change the shape of a receptor . Klippenstein et al . have now generated a small light switch molecule with the size of a single amino acid side-chain that , in theory , could replace any of the usual amino acids in the NMDA receptor . Different locations for the light switch were tested to identify those that changed the activity of the receptor . When the receptors were stimulated with light , the light switch changed its shape , which in turn influenced the shape of the receptor . This meant that , depending on which amino acid in the receptor had been replaced with the light switch , light could be used to control the receptor activity in different ways . This new approach of using integrated light switches allows NMDA receptors to be controlled in a fast and reversible manner using something as simple as a beam of light . Further research will use the toolset of light-controllable receptors to study how the different NMDA receptor subtypes affect synaptic plasticity in the normal and diseased brain . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"tools",
"and",
"resources",
"neuroscience"
] | 2017 | Optocontrol of glutamate receptor activity by single side-chain photoisomerization |
Numerous brain regions have been shown to have neural correlates of gradually accumulating evidence for decision-making , but the causal roles of these regions in decisions driven by accumulation of evidence have yet to be determined . Here , in rats performing an auditory evidence accumulation task , we inactivated the frontal orienting fields ( FOF ) and posterior parietal cortex ( PPC ) , two rat cortical regions that have neural correlates of accumulating evidence and that have been proposed as central to decision-making . We used a detailed model of the decision process to analyze the effect of inactivations . Inactivation of the FOF induced substantial performance impairments that were quantitatively best described as an impairment in the output pathway of an evidence accumulator with a long integration time constant ( >240 ms ) . In contrast , we found a minimal role for PPC in decisions guided by accumulating auditory evidence , even while finding a strong role for PPC in internally-guided decisions .
Gradual accumulation of evidence for or against different choices has been implicated in many types of decision-making , including value-based decisions ( Basten et al . , 2010; Milosavljevic et al . , 2010; Cavanagh et al . , 2011; Hunt et al . , 2012; Solway and Botvinick , 2012 ) , social decisions ( Krajbich and Rangel , 2011 ) , economic decisions ( Gluth et al . , 2012 ) , gambling decisions ( Busemeyer and Townsend , 1993 ) , memory-based decisions ( Ratcliff , 1978 ) , numerical comparison decisions ( Sigman and Dehaene , 2005 ) , visual search decisions ( Purcell et al . , 2010; Heitz and Schall , 2012 ) , and perceptual ( Gold and Shadlen , 2007; Ratcliff et al . , 2007; Mante et al . , 2013 ) decisions . It is therefore considered a core decision-making process . Although neural correlates of evidence accumulation have been reported in several interconnected primate brain regions—such as PPC ( Shadlen and Newsome , 2001; Roitman and Shadlen , 2002; Hunt et al . , 2012 ) , prefrontal cortex ( Hunt et al . , 2012 ) including frontal eye fields ( FEF; Kim and Shadlen , 1999; Purcell et al . , 2010; Ding and Gold , 2012; Heitz and Schall , 2012; Mante et al . , 2013 ) , striatum ( Ding and Gold , 2010 ) , and superior colliculus ( Horwitz and Newsome , 1999; Ratcliff et al . , 2007 ) —the specific roles of these different brain regions in decisions driven by accumulation of evidence have not yet been distinguished . We recently developed a rat model of gradual accumulation of evidence for decision-making , using a task that allows detailed quantitative modeling of the accumulation and decision processes ( ‘Poisson Clicks’ task; Brunton et al . , 2013 ) . In separate work from our laboratory using the Poisson Clicks task , electrophysiological recordings in rat PPC and Frontal Orienting Fields ( FOF; Erlich et al . , 2011 ) revealed classic neural correlates of evidence accumulation ( Figure 1 of Hanks et al . , 2015 ) . Specifically , we found neurons in these rat regions that ramp up their activity during the stimulus , and the slope of that ramp is correlated with the strength of the momentary evidence just as one would expect from neurons whose firing rates represent the accumulation of evidence over time , and just as previously reported in monkey regions that have been suggested as analogous to the rat PPC and FOF ( primate PPC: Shadlen and Newsome , 1996 , 2001; Roitman and Shadlen , 2002; and monkey FEF: Ding and Gold , 2012; Mante et al . , 2013; for PPC analogy , see Whitlock et al . , 2008; Reep and Corwin , 2009; Wilber et al . , 2014; for FOF/FEF analogy see Erlich et al . , 2011 ) . 10 . 7554/eLife . 05457 . 003Figure 1 . Poisson clicks accumulation task trials and interleaved side LED trials . Each accumulation task trial begins with the onset of the center LED , which signals to the rat to enter the center port . The subject holds his nose in the center port for 2 s , until the center LED offset , which is the go cue . The majority of trials ( 90% ) are accumulation trials . On accumulation trials , clicks play from the right and left speakers ( right + left click rate = 40 clicks/s ) , terminating with the go cue . After the go cue reward is available at the side port associated with the greater number of clicks . The stimulus duration on each trial is set by the experimenter to be in the range 0 . 1–1 s . On Side LED trials , no sound is played during the fixation period and one of the side ports is illuminated once the rat withdraws from the center port to indicate that reward is available there . Accumulation and side LED trials are randomly interleaved , as are left and right trials . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 003 In addition to having neural correlates of accumulating evidence ( Hanks et al . , 2015 ) , several properties of the rat FOF suggest it as a candidate for a causal role in decisions driven by accumulation of evidence . Accumulation of evidence involves both maintaining a memory of evidence accrued so far and addition of new evidence to the memory , and is therefore linked to short-term memory processes . The rat FOF has delay activity that correlates with short-term memory , and plays a causal role in short-term memory for future orienting responses ( Erlich et al . , 2011 ) . Furthermore , the rat FOF is well-situated to play an important role in perceptual decision-making , since it receives inputs from multiple sensory cortices ( Condé et al . , 1995 ) , and it projects to the superior colliculus ( SC; Stuesse and Newman , 1990 ) , a subcortical region that , in both rodents and primates , is involved in controlling orienting motions ( Isa and Sasaki , 2002; Felsen and Mainen , 2008 ) and is thought to be involved in decisions reported through such orienting motions . Moreover , the rat FOF is reciprocally connected with the rat PPC ( Reep et al . , 1987 , 1994 ) , which is currently considered a critical , central node in rodent perceptual decision-making ( Carandini and Churchland , 2013 ) . The rodent PPC itself also has neural correlates of accumulating evidence ( Hanks et al . , 2015 ) , and it shares with the FOF some of the key properties that suggest a causal role in decisions driven by accumulation of evidence . The rodent PPC has delay activity that correlates with performance on short-term memory tasks ( Nakamura , 1999; Harvey et al . , 2012 ) , and , as shown through inactivations , plays a causal role in short-term memory for orienting acts ( Harvey et al . , 2012 ) and like FOF receives input from many sensory cortices as well as top-down input from prefrontal cortex , including FOF ( Wilber et al . , 2014 ) . For these reasons , the FOF and the PPC are the most prominent candidate regions in rodent association cortex for being important nodes in orienting decisions guided by accumulation of evidence . We focus on these two areas here . We implanted bilateral cannula in both FOF and PPC of rats trained to perform the Poisson Clicks task , and inactivated these regions with the GABA-A agonist muscimol while the rats performed the task . Consistent with expectations drawn from neural correlates in the rat FOF , inactivation of the FOF impaired performance in the task . We used quantitative modeling to characterize which aspect of the accumulation and decision process was impacted by inactivation of the FOF . The results of these analyses revealed a specific location for the FOF in the causal circuit underlying the Poisson Clicks behavior: the behavioral impairment caused by FOF infusions could be parsimoniously and quantitatively explained as an impairment in the premotor output pathway of an evidence accumulator with a long accumulation time constant ( 240 ms or more ) . It is possible that the decision itself ( i . e . , the categorization of the graded accumulator value into a discrete choice , which is a process subsequent to graded evidence accumulation ) could occur in the FOF . In contrast , we found that PPC inactivations had a relatively minor effect on the Poisson Clicks task . This was true even while the same PPC inactivations had strong effects on interleaved ‘free-choice’ trials , in which no sensory evidence was provided and rats were rewarded regardless of their choice of response . Our data thus suggest that the PPC plays a minimal causal role in decisions guided by accumulation of auditory evidence , while playing an important role in internally-guided decisions . Together , our findings from inactivations of the PPC and the FOF provide important constraints on the neural circuitry underlying decisions guided by accumulation of auditory evidence in the rat .
We trained male Long-Evans rats ( n = 14 rats ) on the Poisson Clicks accumulation task ( Figure 1 , Brunton et al . , 2013 ) . On each trial of this task , illumination of the center LED indicated that the rat should place its nose in the center port and remain there while click trains with Poisson-generated inter-click-intervals were played from the left and right speakers . The rats learned to report which side had played the greater total number of clicks by nose-poking into the corresponding side port ( Figure 2A ) . We refer to these trials as ‘accumulation trials’ . 10 . 7554/eLife . 05457 . 004Figure 2 . Behavioral evidence of accumulation . ( A ) Behavior as a function of total right minus total left clicks . For very easy trials ( large click differences ) performance is ≈90% correct . The circles ( with very small error bars ) are the mean ±95% binomial confidence intervals across accumulator trials from all rats 1 day before an infusion session ( n = 47 , 580 trials across 14 rats ) . The thick line is the psychometric curve generated by the accumulator model fit to these trials . ( B ) The time-constant of accumulation as fit by the model for each rat in the experiment . The median ( 810 ms ) is marked by a thin gray line . ( C ) Chronometric plot generated using the same data as in panel ( A ) . The rats' performance increases with longer duration stimuli , consistent with an accumulation strategy . The circles and error bars are the mean ±95% binomial confidence intervals across trials on the easiest ( blue ) , middle ( purple ) and hardest ( magenta ) thirds of trials defined by the absolute value of the ratios of left vs right click rates . The thick lines are the model generated chronometric curves . ( D ) Reverse correlation analyses showing that clicks throughout the stimulus were used in the rats' decision process , supporting the long accumulation time constants in ( B ) . The thick dark red and green lines are the means ± std . err . across trials for where the rats went right and left . Thin light red and green lines are the model generated reverse correlation . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 004Figure 2—figure supplement 1 . 9-parameter Accumulator Model ( reproduced from Brunton et al . , 2013 ) . At each timepoint , the accumulator memory a ( black trace ) represents an estimate of the ‘Right’ vs ‘Left’ evidence accrued so far . At stimulus end , the model decides ‘Right’ if a > Þ , the decision boundary , and ‘Left’ otherwise , where Þ is a free parameter . Light grey traces indicate alternate runs with different instantiase . a The decision variable . Right ↑ ( left ↓ ) pulses change the value of a by positive ( negative ) impulses of magnitude C . σi2 parameterizes noise in the initial value of a . σa2 a diffusion constant , parameterizing noise in a . σs2 parameterizes noise when adding the evidence from a Right or Left pulse: variance σs2 is added to the amplitude C of the evidence contributed by each click . λ parameterizes consistent drift in the memory a . In the ‘leaky’ or forgetful case ( λ < 0 , illustrated ) , drift is towards a = 0 , and later pulses impact the decision more than earlier pulses . In the ‘unstable’ or impulsive case ( λ > 0 ) , drift is away from a = 0 , and earlier pulses impact the decision more than later pulses . The memory's time constant τ = 1/λ . B the height of the ‘sticky’ decision bounds and parameterizes the amount of evidence necessary to commit to a decision . φ , τϕ parameterize sensory adaptation by defining the dynamics of C . Immediately after a click , the magnitude C is multiplied by φ . C then recovers towards an unadapted value of 1 with time constant τϕ . Facilitation is thus represented by ϕ > 1 , while depression is represented by ϕ < 1 ( inset ) . Þ the decision boundary . These properties are implemented by the following equations: if |a| ≥ B then da/dt = 0; else ( 1 ) da=σadW+ ( δt , tR·ηR·C−δt , tL·ηL·C ) dt+λadt , whereδt , tR , L are delta functions at the times of the pulses . η are i . i . d . gaussian variables drawn from N ( 1 , σs ) . dW is a white noise Wiener process . The initial condition a ( t = 0 ) is drawn from the gaussian N ( 0 , σi ) . Adaptation dynamics are given by: ( 2 ) dCdt=1−Cτϕ+ ( ϕ−1 ) C ( δt , tR+δt , tL ) . In addition , a lapse rate parameterizes the fraction of trials on which a random response is made . Ideal performance ( a = #right clicks − #left clicks ) would be achieved by λ=0 , B=∞ , σa2=σs2=σi2=0 , ϕ=1 , Þ = 0 . © 2013 AAAS . All Rights Reserved . 2013AAASFigure 2—figure supplement 1 and legend text reproduced from Brunton BW , Botvinick MM , Brody CD . 2013 . Rats and humans can optimally accumulate evidence for decision-making . Science 340 , 95–98 . doi:10 . 1126/science . 1233912 . Reprinted with permission . 10 . 7554/eLife . 05457 . 005Figure 2—figure supplement 2 . Behavioral evidence of accumulation in individual rats . ( A ) Behavior as a function of total right minus total left clicks . For very easy trials ( large click differences ) performance is ≈90% correct . The thick line is the average performance of the 14 rats in the study , the thin gray lines are the performance of individual rats . ( B ) The time-constant of accumulation as fit by the model for each rat in the experiment . The median ( 810 ms ) is marked by a thin gray line . ( C ) Chronometric plot showing that rats performance increases with longer duration stimuli , consistent with an accumulation strategy . The thin lines are the performance of individual rats ( n = 14 ) on the easiest ( blue ) , middle ( purple ) and hardest ( magenta ) thirds of trials defined by the absolute value of the ratios of left vs right click rates . The thick lines show the means across rats . ( D ) Reverse correlation analyses showing that clicks throughout the stimulus were used in the rats' decision process supporting the long accumulation time constants in ( B ) . Thin light red and green lines are the reverse click rate correlation of individual rats ( n = 14 ) . The thick dark red and green lines are the means across rats . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 005 In order to control for motor effects of inactivations , the accumulation trials were randomly interleaved , in most sessions , with trials that we refer to as ‘side LED’ trials . On side LED trials no sounds were played during fixation . Immediately after the end of fixation , one of the two side ports was illuminated , indicating availability of reward at the lit port ( Figure 1 ) . The right and left side LED trials , together , comprised ≈10% of the total trials . To demonstrate that subjects accumulated the sensory evidence provided by the auditory clicks , we fit an accumulator model using the individual click times and the rats' choices on each trial ( Figure 2—figure supplement 1; see also Brunton et al . , 2013 ) . Different parameter value regimes of this model can implement many different strategies , such as responding based on the first few clicks , or last few clicks , or to a burst of clicks , and many others . Consistent with previous results , maximum likelihood fits resulted in best-fit parameters associated with a gradual evidence accumulation strategy . Most importantly for this study , this strategy was characterized by a long accumulator time-constant , just under 1 s ( Figure 2B , Table 1 ) , which is the duration of the longest stimuli used here . As expected for a gradual accumulation strategy in which clicks from the entire stimulus are weighted equally , performance improved for longer stimuli with the same underlying click rates ( Figure 2C; Ratcliff and Rouder , 1998; Usher and McClelland , 2001; Brunton et al . , 2013 ) , and a psychophysical reverse correlation analysis ( Kiani et al . , 2008; Raposo et al . , 2012; Brunton et al . , 2013 ) indicates that rats used clicks from all times of the stimuli to make their decision ( Figure 2D ) . 10 . 7554/eLife . 05457 . 006Table 1 . Best-fit parametersDOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 006Ratnameλσa2σs2σinit2BϕτϕÞlapseB1151 . 4090 . 113102 . 1300 . 52314 . 8490 . 1750 . 0640 . 1570 . 094T0551 . 2260 . 00111 . 2480 . 04316 . 0140 . 2530 . 3510 . 1180 . 078T0570 . 8100 . 03174 . 4780 . 02715 . 0600 . 1560 . 0930 . 0200 . 075T0581 . 0870 . 00017 . 6120 . 00015 . 8750 . 0250 . 276−0 . 1220 . 051T0610 . 6200 . 00096 . 5450 . 50216 . 0380 . 3800 . 0410 . 2360 . 066T062−0 . 0980 . 00049 . 3610 . 61915 . 7610 . 1390 . 0470 . 5180 . 083A0652 . 0470 . 00037 . 6850 . 20715 . 7290 . 1470 . 092−0 . 4650 . 031A0660 . 3490 . 00015 . 5650 . 00012 . 7050 . 0720 . 4620 . 0410 . 170A077−2 . 7390 . 197128 . 58622 . 8019 . 2530 . 1840 . 0310 . 8860 . 001A078−2 . 0700 . 000104 . 6880 . 00018 . 0860 . 2830 . 0260 . 0620 . 063A060−1 . 5420 . 00054 . 7860 . 00015 . 4160 . 0100 . 1150 . 1800 . 245A0622 . 2580 . 296156 . 8600 . 48616 . 8390 . 5270 . 0760 . 4660 . 119A083−0 . 79047 . 44131 . 7881 . 38416 . 2820 . 0150 . 0590 . 0330 . 107A0841 . 3710 . 06470 . 2671 . 69015 . 0110 . 0160 . 0860 . 4670 . 110Meta-Rat1 . 2270 . 00157 . 6140 . 04316 . 0420 . 2210 . 1090 . 0650 . 102BiFOF−4 . 144*62 . 423237 . 6421 . 75422 . 0130 . 0820 . 0390 . 737*0 . 010BiPPC1 . 3310 . 53142 . 1750 . 00014 . 8600 . 5120 . 175−0 . 2490 . 321This table shows the values of the parameters which maximize the likelihood of the full 9-parameter accumulator model for each rat , as well as for the ‘meta-rat’ ( made from taking all of the control days that were 1 day before an infusion , n = 47 , 580 trials ) , the fit to the bilateral FOF data ( n = 1809 ) , and the fit to the bilateral PPC data ( n = 1569 ) . *indicate parameters that were significantly different from the control ‘Meta-Rat’ . We report the results of five different types of inactivations: unilateral FOF , bilateral FOF , unilateral PPC , bilateral PPC , and combined bilateral FOF + unilateral PPC inactivations , for a total of 26 , 521 trials from 161 infusions into the FOF and PPC of 14 rats ( Figure 3A and Figure 3—figure supplement 1 ) . We initially performed muscimol inactivations of the FOF and PPC in 6 rats performing the Poisson Clicks Task ( group 1 ) . In order to verify the results and perform follow-up and control experiments , we performed inactivations in two further groups ( group 2 , n = 4; group 3 , n = 4 ) . The specific order and outcome of the infusions in each rat is shown in Figure 3—figure supplement 2 . 10 . 7554/eLife . 05457 . 007Figure 3 . FOF Infusions . ( A ) Top-down view of rat cortex with the locations of the FOF and the PPC , into which cannulae were implanted . ( B ) Bilateral infusion of muscimol into the FOF results in a substantial impairment on accumulation trials but has no effect on side LED trials . In black are data from control sessions 1 day before an infusion ( n = 8 sessions , 4 rats ) . In blue are data from bilateral FOF infusions ( n = 8 sessions , 4 rats , 75 ng per side ) . The circles with error bars indicate the mean ± s . e . across sessions . Accumulation trials are binned by #R − #L clicks , spaced so there are equal number of trials in each bin . The lines are a 4-parameter sigmoid fit to the data . ( C ) Unilateral infusion of muscimol into the FOF results in a profound ipsilateral bias on accumulation trials but has no effect on side LED trials . In black are data from control sessions 1 day before an infusion ( n = 34 sessions , 12 rats ) . In red are data from right FOF infusions ( n = 17 sessions , 12 rats , 150 or 300 ng ) . In green are data from left FOF infusions ( n = 17 sessions , 12 rats , 150 or 300 ng ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 00710 . 7554/eLife . 05457 . 008Figure 3—figure supplement 1 . Cannula coordinates and histology . ( A–C ) The targets of cannula implants for group 1 , 2 , and 3 with the list of the rats in each group . ( D ) A birds-eye view of rat T061's brain after fixation and removal from the skull . The AP and ML locations of the cannula are clearly visible by eye . Each line of the brain blocker marks 1 mm . The blue cross marks the approximate location of Bregma . ( E ) Coronal section of a rat brain that had been infused through the implanted FOF cannula with two colors ( ‘red’ on the left and ‘blue’ on the right ) of Alexa Fluor-conjugated cholera toxin-B subunit , a fluorescent tracer . The rat was perfused 1 week after infusion of tracer . The tracer has labelled cells along the AP axis of the FOF . Shown here is a section from 2 . 5 mm anterior to Bregma ( Paxinos and Watson , 2004 ) . Note , that in the nomenclature of Paxinos and Watson ( 2004 ) the area that we describe as the FOF is considered to be part of M2 . In the bottom left corner , the top of another coronal section overlaps with the shown section . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 00810 . 7554/eLife . 05457 . 009Figure 3—figure supplement 2 . Timeline of bias for each rat . Each point of the figure is the bias for a single session ( %Right − %Left Correct ) . The number at the beginning of the x-axis indicates the days passed since surgical implantation with cannula . Control non-infusion days are shown as black dots . Right infusions are shown in red , Left infusions are shown in green . Bilateral infusions are shown in blue . For the simultaneous bilateral FOF and unilateral PPC infusions , the color indicates the side of the PPC infusion as in Figure 4C . Stars indicated PPC infusions , Diamonds indicate FOF infusions . Hollow markers indicate an infusion session where the subject did not perform enough trials to analyze . The bottom x-labels describe the details ( side , region and dose ) of each infusion . The top x-labels indicate the number of days passed from cannula surgery . If infusions generate an ipsilateral bias then red markers should be above zero and green markers below zero . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 00910 . 7554/eLife . 05457 . 010Figure 3—figure supplement 3 . FOF infusions cause profound impairment in the clicks task . The psychometric data and GLMM model fits for bilateral FOF infusions in each rat ( n = 4 ) . Open circles are binned data from accumulation trials and the small points are the predictions of the GLMM fits at sampled data points . ( A ) In black are the isoflurane control fits and in blue are the bilateral infusion fits . In every rat the slope of the bilateral infusions is shallower than in the isoflurane controls . In three of four rats there was also a shift , likely due to the challenge of performing perfectly balanced bilateral infusions . In three of four rats ( All but A066 ) the difference between performance between bilateral FOF and isoflurane is significantly different . ( B ) The psychometric data and GLMM model fits for unilateral FOF infusions in each rat ( n = 12 ) . Open circles are binned data from accumulation trials and the small points are the predictions of the GLMM fits at sampled data points . In red are the right infusion data and fits and in green are the left infusion data and fits . In every rat the right infusions result in more rightward responses on accumulation trials than the left infusions . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 010 We placed cannulae in the center of the location currently identified as FOF ( +2 . 0 mm AP , ±1 . 3 mm ML from Bregma , Figure 3—figure supplement 1A ) . These are the same coordinates used by Erlich et al . ( 2011 ) , and are also the coordinates at which neural correlates of accumulation of evidence were observed in the Poisson Clicks task ( Hanks et al . , 2015 ) . For the first bilateral FOF inactivation session we infused 300 ng of muscimol per side , for a total of 600 ng . After these infusions rats did not perform the task . We subsequently used a smaller dose of 75 ng per side ( Note: this is half the dose used in the bilateral PPC experiment described below ) . This resulted in a significant 10 . 3% decrement on performance on accumulation trials ( p = 0 . 018 , GLMM test; Figure 3B ) . The effect was individually significant in 3/4 rats ( Figure 3—figure supplement 3A ) . Side LED trials were unimpaired ( p > 0 . 5; Figure 3B ) , indicating that the impairment on accumulation trials was not simply a motor effect . Unilateral infusions of muscimol into the FOF resulted in a profound bias towards ipsilateral responses in the Poisson Clicks task ( Figure 3C ) . Averaged across all unilateral FOF infusion sessions , the ipsilateral bias ( defined as ipsilateral % correct − contralateral % correct ) , was 52 ± 7% ( mean ± s . e . ) for accumulation trials ( t-test across 12 rats t11 = 7 . 27 , p < 10−4; two rats in group 3 failed to perform sufficient numbers of trials during FOF inactivations to be included in this analysis ) . Unilateral FOF infusions reduced performance to chance on even the easiest contralateral accumulation trials ( Figure 3C; green data points for #R − #L ≫ 0 , and red data points for #R − #L ≪ 0 ) . The ipsilateral bias induced by FOF inactivation was highly reproducible: 87% ( 26/30 ) of individual infusion sessions resulted in a positive ipsilateral bias ( sign-test , p < 0 . 001 ) , and in every single rat there were more rightward responses after rightward infusions than leftward infusions ( Figure 3—figure supplement 3B ) . Importantly , as in the bilateral inactivations , there was no significant effect on side LED trials ( t-test t5 = 1 . 55 , p > 0 . 15; Figure 3C , side LED Trials ) , nor did unilateral FOF inactivations have an effect on the response time on side LED trials ( repeated-measures ANOVA , F ( 1 , 3 ) = 0 . 65 , p > 0 . 4 ) . This indicates that the effect of FOF inactivation was not simply an overall motor effect . Furthermore , the inactivations produced no observable effects outside of the behavioral task . Infused animals appeared normal in their home cages both immediately after the infusion and after the behavioral session . Our localized inactivations thus contrast with previous literature , in which large permanent unilateral lesions of the rat prefrontal cortex ( including but extending well beyond the FOF ) , produced persistent , ipsiversive circling in the lesioned animals ( Crowne and Pathria , 1982 ) . Which aspects of the evidence accumulation and decision process were impaired by the FOF inactivations ? To address this question , we took advantage of the accumulator model of Brunton et al . ( 2013 ) , which uses the knowledge the precise time of each click in each individual trial , as well as the rat's decision on each trial , to estimate 9 parameters that characterize the accumulation and decision processes ( Figure 2—figure supplement 1 ) . Each parameter quantifies a specific aspect of the decision process . For example , τ , the time constant of the accumulator ( also described by the parameter λ = 1/τ ) , characterizes the time period over which the subject accumulates evidence . A negative value of λ indicates a leaky accumulator , a positive value of λ indicates an unstable accumulator , and perfect accumulation would have τ = ∞ ( i . e . , λ = 0 ) . Another example parameter is the lapse rate , which quantifies the fraction of trials in which the subject behaves as if it had ignored the clicks that were played and had instead made its decision randomly . Deviations from the perfect values ( λ = 0 , lapse = 0 ) in either of these parameters can give rise to psychometric curves with shallow slopes , and both types of imperfections can produce curves qualitatively similar to the experimental curve obtained after bilateral FOF inactivation ( Figure 4A ) . But the similarity between the psychometric curves for the two imperfections is partly due to the fact that these psychometric curves ignore the specific timing of individual clicks . In contrast , because the two imperfections would have very different signatures in terms of how clicks at different times affect the rats' decisions , and click timing is fully taken into account in the behavioral model , the model can clearly distinguish the two imperfections . Using the control data , the best-fitting parameter values for the behavioral model had τ ≈ 0 . 8 s–indicating that subjects accumulated information over almost the entire stimulus duration but were on average slightly unstable–and a lapse rate of ≈0 . 1 ( black cross , Figure 4B ) . For data from bilateral inactivation sessions , the maximum likelihood parameter values ( center of blue likelihood peak , Figure 4B ) changed significantly . The inactivation data now had a dramatically different and much shorter accumulation time constant of the opposite sign to the control data , τ ≈ −0 . 24 s ( leaky accumulation over only a quarter of a second ) . In contrast , the best-fitting lapse rate remained essentially unchanged from control . As described above , attempting to fit the inactivation data by keeping the time constant unchanged and increasing the lapse rate could qualitatively match the psychometric curve ( magenta line , Figure 4A ) , but it produced an extremely poor fit relative to the full behavioral model ( magenta cross in low likelihood region , Figure 4B ) . Thus , after bilateral FOF inactivation , the subjects behaved as if their lapse rate was unchanged , and their accumulator had become much leakier ( Table 1 ) . 10 . 7554/eLife . 05457 . 011Figure 4 . Bilateral FOF inactivation is best fit as a reduction in the time-constant of accumulation . ( A ) When analyzed in terms of the psychometric function , changes to either lapse rate alone or accumulation time constant alone can match the bilateral FOF inactivation data . The black line shows the psychometric curve from control data , collected 1 day before bilateral FOF sessions ( n = 1526 trials ) . Blue dots with error bars show the experimental data from bilateral FOF inactivation sessions ( n = 1809 trials ) . The magenta line is the psychometric curve obtained by fitting only the lapse rate parameter to the inactivation data , while keeping all other parameters at their control values ( corresponds to magenta cross in panel B ) . The blue line shows the psychometric curve from the accumulator model fit to the inactivation data ( corresponds to peak of blue likelihood surface in panel B ) , which has a change w . r . t . control in accumulation time constant τ ( =1/λ ) , but no change in lapse rate . ( B ) Fitting the detailed click-by-click , trial-by-trial accumulator model ( Brunton et al . , 2013 ) to the inactivation data clearly distinguishes between lapse and λ . The panel shows the normalized likelihood surface , indicating quality of the model fit to the inactivation data as a function of the lapse and the λ ( =1/τ ) parameters . The black cross shows parameter values for the control data . The best fit to the inactivation data is at the peak of the blue likelihood surface ( λ = −4 . 15 , lapse = 0 . 048 ) , significantly different from control for λ , but not different from control for lapse . This best-fit lambda corresponds to λ = −0 . 241 s , a substantially leaky integrator . ( C ) Performance as a function of stimulus duration for bilateral FOF sessions ( blue , mean ± std . err . ) and the control sessions 1 day before ( black , mean ± std . err . ) . The lines are the chronometric curves generated by the accumulator model ( for inactivation data , parameter values at peak of blue likelihood surface in panel B ) . ( D ) Reverse correlation showing the relative contribution of clicks from different times to the rats' decisions for data from bilateral FOF inactivation sessions . Compare to Figure 2D . The thick dark shading shows the mean ± std . err . across trials based on the rats' choices . The thin bright lines are the reverse correlation traces generated by the accumulator model ( parameter values at peak of blue likelihood surface in panel B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 01110 . 7554/eLife . 05457 . 012Figure 4—source data 1 . MATLAB file containing resampled bilateral FOF model fits . This MATLAB file contains three variables . BF: a 300 × 9 matrix . Each row is the set of parameters which maximized the likelihood of the accumulator model fit to a resampling of the bilateral FOF data . Each column is a parameter . LL: a 300 × 1 vector with the negative log likelihood of the corresponding row of BF . parameter_names: a 9 × 1 cell with the names of the columns of BF . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 012 To further validate these results we resampled the trials ( with replacement ) and refit the model on the resampled trials 300 times . Based on this analysis , only two parameters shifted significantly from the control data: λ ( 95% C . I . = [−7 . 91 −1 . 31]; control value was 1 . 22 ) and Þ ( 95% C . I . = [0 . 10 1 . 50]; control value was 0 . 065 ) . The change in λ is both significant and substantial , and accounts for most of the change in performance . The change in decision boundary , Þ ( pronounced ‘sho’ ) is significant but very small , corresponding a horizontal shift in the psychometric curve of only one click . It is largely due to the difficulty of performing a perfectly balanced bilateral infusion . In particular , one rat , A077 , was strongly biased during the bilateral FOF inactivations ( Figure 3—figure supplement 3A ) . On average , the noise and lapse parameters also increased but due to large covariance between these parameters , it is not possible to say which of them was significantly shifted ( Figure 4—source data 1 ) . To further examine the relative contributions of the λ and Þ changes we fit two 1-parameter models . First , we fit the bilateral FOF data with a 1-parameter Þ model where only the decision boundary could change and the other 8 parameters were fixed at their best control data values . The log likelihood of the best Þ model was −1221 . 6 , substantially worse than the model in which all 9 parameters were allowed to vary ( −1102 . 5 ) . Using Bayesian or Akaike Information Criteria ( BIC or AIC; Burnham and Anderson , 2004 ) , we find that the extra parameters are indeed justified ( Table 2 ) . Second , we fit the bilateral FOF data to a 1-parameter λ model where only the accumulation time-constant could change . For this model , the best-fit value of λ was −12 . 4 and the log likelihood of the model was −1121 . 1 compared to −1102 . 5 for the best 9-parameter model , a difference of only 18 . 5 . According to BIC , the 1-parameter model is the more likely model , supporting the idea that the major effect of bilateral FOF inactivation was a change in the time constant of accumulation ( Table 2 ) . 10 . 7554/eLife . 05457 . 013Table 2Bilateral FOF model comparisonDOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 013Model# of param . Log likelihoodBICAICfull model9−1102 . 52272 . 52223†Þ model1−1221 . 62450 . 72445 . 2λ model1−1121 . 12249 . 7*2244 . 2This table shows the three models fit to the bilateral FOF data ( n = 1809 trials ) . *indicates the model with the lowest ( the most likely ) Bayesian information criterion ( BIC ) . †indicates the model with the lowest ( most informative ) Akaike information criterion ( AIC ) . In this case , the AIC and BIC select different models , suggesting a better model may be somewhere in between . That is , a model that includes the accumulator time-constant and perhaps a few additional parameters from the full model . To probe the conclusions derived from the trial-by-trial model fit , we used model-free analyses of the data ( Brunton et al . , 2013 ) . Leaky accumulation with a time constant τ ≈ −0 . 24 s would result in short trials ( with a stimulus duration less than a quarter of a second ) being essentially unimpaired , while long duration trials would be more strongly impaired . This was indeed observed in the data , with a tight correspondence between the quantitative model and experimental data ( chronometric curves , shown in Figure 4C ) . A further property of a leaky accumulator is that the more recent the clicks are , with respect to the end of the stimulus , the bigger their impact on the subject's decision . This was also observed in the data , again with a tight correspondence between quantitative model and experimental data ( reverse correlation analysis , shown in Figure 4D ) . These results indicate that the FOF is either ( a ) itself directly involved in the process of accumulating evidence , and the inactivations made the accumulator leaky; or ( b ) the FOF is a requisite component in the output pathway of an accumulator with a long time constant , that is , the FOF is part of the chain of regions that transform evidence accumulated with a time-constant longer than 0 . 24 s into an orienting decision . We also used an accumulator model to analyze the strong ipsilateral bias induced by unilateral FOF inactivations . The original model of Brunton et al . ( Figure 2—figure supplement 1 ) contains only one parameter , the decision boundary Þ , that can generate a left/right bias . We therefore extended the model with three additional parameters , each of which represented a possible imperfection that could generate a side bias . Simultaneously fitting all 12 parameters substantially increased the computational difficulty of the fitting process . In particular , efficiently fitting the original model was made possible by analytical computation of the gradient ( Brunton et al . , 2013 ) . Determining the gradient for 12 parameter model was outside the scope of this manuscript . We consequently took the strategy of first fitting the original 9-parameter model to the control data from 1 day before infusion sessions , and then , starting from those best-fitting parameter values , asking which of the bias-inducing single-parameter changes would best fit the data from the unilateral FOF inactivations ( other parameters were held fixed at the control data best-fit values ) . In other words , we asked , ‘if we changed only one parameter , which one would it be to best fit the data ? ’ Finding the maximum likelihood value was made practical by the fact that each of the four fits performed was a single-parameter fit . The four single-parameter changes we considered corresponded to hypotheses regarding possible functions of the FOF , and are conceptually illustrated in Figure 5 . ( a ) First , we considered the possibility that the FOF is part of the output pathway of the accumulator , perhaps part of computing or representing the animal's discrete choice after having categorized the accumulator value ( into ‘Go Right’ vs ‘Go Left’ categories ) ( Hanks et al . , 2015 ) , potentially in the service of preparing a motor action ( Erlich et al . , 2011 ) . Unilaterally perturbing the FOF might then bias this post-categorization representation . To implement this idea in the model , we added a parameter that biased outcomes after the R/L decision was made on each trial . Independently of the stimulus that led to the decision , we let a randomly-chosen fraction , κR , of right decisions and a randomly-chosen fraction κL of left decisions be reversed ( i . e . , R → L and L → R; see Figure 5A ) . These reversals scale the vertical endpoints of the psychometric curve towards the Went Right = 50% level . The scaling is biased when κR ≠ κL . ( b ) Next , we considered the possibility that since the FOF has been suggested as analogous to primate FEF , perturbing it might affect attentional processes , perhaps causing a lateralized sensory neglect that would bias the perceptual impact of auditory clicks from the two different sides . In other words , during unilateral inactivation , right and left clicks could have different magnitudes of their impact on the accumulating evidence ( CR and CL , instead of the single common C of Equation 1 in Figure 2—figure supplement 1 ) . We described this as a ‘unbalanced input gain’ ( see Figure 5B ) . ( c ) We next considered the possibility that the FOF plays a role in the accumulation process itself , and quantified biases in accumulation through an ‘accumulation shift’ that shifts the value of the accumulator , a , at the end of the stimulus ( see Figure 5C , equivalent to Þ in Figure 2—figure supplement 1 ) . Changes in this parameter will cause horizontal shifts in the psychometric curve . ( d ) In the fourth and final model , we considered a second possible form of lateralized sensory neglect , in the form of ‘unbalanced input noise’ ( see Figure 5D ) . In this version of the model , right and left clicks could have different signal-to-noise ratios by having different values of the sensory noise parameter ( σs , R2 and σs , L2 , instead of the single common σs2 of Figure 2—figure supplement 1 ) . For each of models ( a ) , ( b ) , and ( d ) , the original fit to the control data was constrained to be balanced . To fit the bias generated by unilateral FOF inactivation , that constraint was relaxed . For each model , we kept the corresponding ipsilateral parameter fixed , and let the contralateral parameter be free to best fit the data . The difference between the best-fitting contralateral parameter minus the ipsilateral parameter was then defined as the bias for that model . 10 . 7554/eLife . 05457 . 014Figure 5 . Conceptual illustration of four model parameters , used to quantify different sources of a lateralized bias . ( A ) Post-categorization bias: after categorizing the accumulator value into ‘Go Left’ or ‘Go Right’ decisions , a fraction , κL , of Left decisions are reversed into Right decisions , and a fraction , κR , are reversed from Right to Left . ( B ) Biased input gain , which can be thought of as a form of sensory neglect: Left and Right clicks have different impact magnitudes on the value of the accumulator . In this illustration left clicks have a much stronger impact , and decisions will consequently be biased to the left . ( C ) Accumulation shift: before categorizing the accumulator into ‘Go Left’ vs ‘Go Right’ decisions ( by comparing the accumulator's value to 0 ) , a constant is added to the value of the accumulator . ( D ) Biased sensory noise , which by differentially affecting signal-to-noise rations from the two sides , can be thought of as a form of sensory neglect distinct from biased input gain: Left and Right clicks have different magnitudes of noise in their impact . In this illustration , left clicks are more variable than right clicks , which biases decisions to the right . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 014 Of these four models , the one that best fit the experimental data was ( a ) , the post-categorization bias model in which the FOF is part of the output pathway of the accumulator ( Figure 6A , Figure 6—figure supplement 1B ) . The value of κ contralateral to the inactivation ( κC ) that best fit the data was 0 . 52 , suggesting that on over 50% of trials rats reversed their contra-choices to ipsi-choices . The next best model ( which was worse by ≈50 log-units than the post-decision model , Figure 6A ) was ( b ) , the biased input gain model . This model failed to accurately fit the data on difficult trials ( in which |#Contra − #Ipsi Clicks| is small , Figure 6—figure supplement 1C ) . The best-fit model for ( c ) , the accumulator shift model , was clearly a poor fit even when analyzed through psychometric curves , fitting particularly poorly for trials with a preponderance of contralateral clicks ( Figure 6—figure supplement 1D ) . The worst fitting model was ( d ) , the unbalanced input noise model ( Figure 6—figure supplement 1E ) . 10 . 7554/eLife . 05457 . 015Figure 6 . Unilateral FOF inactivation is best fit as a post-categorization bias . ( A ) A comparison of the likelihoods ( i . e . , best model fits ) for the four different bias mechanisms illustrated in Figure 5 . The post-categorization bias model is better than the next best model ( biased input gain ) by 50 log-units . ( B ) The 2-dimensional normalized likelihood surface for the two best single-parameter models: post-categorization bias and input gain bias . For visualization , we plot the contra-ipsi bias for the two parameters . That is , the difference between the contralateral and ipsilateral values for each parameter . The y-axis is κC − κI . The x-axis is CC − CI . By definition , in the control model ( black marker ) these biases are 0 . The peak of the magenta likelihood surface for the inactivation data is significantly different from control for post-categorization bias ( from 0 to 0 . 4588 ) but not significantly different from control for input gain bias . ( C ) Psychometric curves for control and inactivation data . The black line is the model fit to the control data ( see Figure 2A for the data points ) . The magenta circles with error bars are experimental data from unilateral FOF inactivation sessions , and indicate fraction of Contra choice trials ( mean ± binomial 95% conf . int . ) across trial groups , with different groups having different #Contra − #Ipsi clicks . The magenta line is the psychometric curve generated by the post-categorization bias model . ( D ) Performance as a function of stimulus duration for data from control sessions 1 day before ( black ) , and for data from unilateral FOF sessions ( magenta , mean ± std . err . ) . The lines are the chronometric curves generated by the corresponding model ( E ) Reverse correlation analyses showing the relative contributions of clicks throughout the stimulus in the rats' decision process . The thick dark red and green lines are the means ± std . err . across trials for contralateral and ipsilateral trials . Thin light red and green lines are the reverse correlation traces generated by the post-categorization bias model . ( F ) Psychometric curves for single-sided trials in control ( black ) , right FOF infusion ( red ) and left FOF infusion ( green ) sessions , demonstrate that even for very easy trials FOF infusions produce a vertical scaling , consistent with post-categorization bias . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 01510 . 7554/eLife . 05457 . 016Figure 6—figure supplement 1 . Psychometric and reverse correlation comparisons of data and model for unilateral FOF inactivations . Top row: For all curves , the circles with error bars indicate fraction of Contra choice trials ( mean ± se ) across trial groups , with different groups having different #Contra − #Ipsi clicks . The lines are model fits that differ from the control data in a single parameter ( see main text ) . Magenta data points in ( B–E ) are data from unilateral FOF infusions . Middle row: Reverse correlation analyses showing the relative contributions of clicks throughout the stimulus in the rats' decision process . The thick dark red and green lines are the means ± std . err . across trials for contralateral and ipsilateral trials . Thin light red and green lines are the reverse correlations predicted by the same accumulator models that were used to plot psychometric fits in corresponding column of the top row . Bottom Row: Likelihood as a function of model parameter value . This demonstrates that there was a single global maximum for each 1-parameter model . The magenta cross in each plot indicates the best-fit value of the corresponding parameter ( on the x-axis ) and the log likelihood of that model ( on the y-axis ) and corresponds to the model used to generate the psychometric curve and reverse correlations in the top and middle rows . ( A ) Data ( black circles ) from control sessions 1 day before an infusion session and a 9-parameter accumulator model fit ( line ) . ( B ) Post-categorization bias model . ( C ) Unbalanced Input model . ( D ) Accumulator Shift model . ( E ) Unbalanced input noise model . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 01610 . 7554/eLife . 05457 . 017Figure 6—figure supplement 2 . Distribution of sample from 8-parameter model of unilateral FOF inactivation . Using the Metropolis–Hastings algorithm we collected 40 , 000 samples from an 8-parameter model of the unilateral FOF inactivation . The parameters are the rows and columns of this matrix of plots . The histograms along the diagonal show the marginal distributions of individual parameters which were used to non-parametrically assess whether individual parameters were significantly different from control values . Each other plot is a scatter plot that gives an estimate of the marginal distribution for each pair of parameters . From left to right: λ and σa2 are the accumulation parameters from the original 9-parameter model . σs2 , I is the variance associated with ipsilateral clicks . The accumulator shift parameter is labeled bias here , as in the original model . Post-CatI ( i . e . , κI ) is the proportion of ipsilateral choices that are flipped to contralateral choices . σs2 , C is the variance associated with contralateral clicks . GainC is the amplitude of contralateral clicks ( i . e . , CL ) . Ipsilateral clicks are assigned a gain of 1 . Post-CatC ( i . e . , κC ) is the proportion of contralateral choices that are flipped to ipsilateral choices . Note the plot showing the 2D marginal distribution of Post-CatC vs GainC shows the same shape as the likelihood surface in Figure 6B , validating the sampling method . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 017 We verified that the post-categorization model was best by using a leave-one-session-out cross validation . For each of the 30 unilateral FOF infusion sessions , we fit the 4 models to the 29 other sessions and then evaluated the likelihood of the fit on the left-out session . Using this approach we found a significant main effect of model on likelihood/trial ( repeated-measures ANOVA F ( 3 , 87 ) = 10 . 76 , p < 10−5 ) and Bonferroni-Holm corrected post-hoc t-tests reveal that all models were significantly different from each other ( p < 0 . 005 ) except the noise and accumulator shift models ( p > 0 . 4 ) . To ask whether a combination of changes to two parameters could provide a significantly better description of the data than our single-parameter fits , we estimated the 2-dimensional likelihood surface for the two best models . This surface ( Figure 6B ) clearly demonstrated that fitting the data requires a large shift away from the balanced control value in the post-categorization bias , but not in input gain . Thus , the dominant effect of unilateral FOF inactivation could be parsimoniously explained by a post-categorization bias , consistent with the known role of the FOF in movement planning ( Erlich et al . , 2011 ) . As described above , a feature of the post-categorization model is that the psychometric curve after unilateral inactivations should be a vertical scaling of the control psychometric curve . This scaling was found in the data , with a tight correspondence between the curve generated by the quantitative model and the experimental psychometric curve ( Figure 6C ) . A similar tight correspondence between model and data were also found for the chronometric curves ( Figure 6D ) and the reverse correlation ( Figure 6E ) . Although computationally challenging , we have also explored whether higher dimensional models might reveal a different set of results . Using the Metropolis–Hastings algorithm , we estimated the best-fit posterior distribution for an 8-parameter model . This model contains the four bias parameters , as well as λ , σa2 , σS , I2 , and κI . For computational tractability , the remaining 4 parameters ( initial noise , bounds , and the two click adaptation parameters ) were fixed at the values of the best-fit control model . With 40 , 000 samples of this 8 dimensional distribution ( Figure 6—figure supplement 2 ) , we estimate that four parameters changed significantly from their control values . First , accumulator noise increased from 5 × 10−4 to 0 . 746 ( 95% C . I . = [0 . 125 12 . 123] , p = 10−4 ) , which is still a small value and would have a negligible impact on behavioral performance . Second , accumulator shift changed from 0 . 065 to 0 . 323 ( [0 . 136 0 . 890] , p = 0 . 013 ) , which would also have a minimal effect on overall reward rate . Third , κI , the changes from ipsi to contra choices decreased from 0 . 102 to 0 . 029 ( [0 . 002 0 . 071] , p = 10−4 ) . Finally , supporting our earlier analysis , κC , increased from 0 to 0 . 498 ( [0 . 176 0 . 536] , p < 10−3 ) which results in a very substantial behavioral impact , since this sets the asymptotic performance on the easiest contralateral trials to ≈50% . As in the 2D model ( Figure 6B ) the gain of contralateral clicks increased ( although not significantly ) . The likelihood of the best 8-parameter model was only marginally higher than the best 1-parameter post-categorization model ( Table 3 ) , consistent with the small shifts in the other parameters ( log likelihood of the 8-D model: −1957 . 52; log likelihood of the 1-D model: −1963 . 08 ) . Using AIC and BIC we find that the increase in likelihood is too small to justify the increase in number of parameters . 10 . 7554/eLife . 05457 . 018Table 3 . Unilateral FOF model comparisonDOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 018Model# of parametersLog likelihoodBICAICPost-categorization bias1−1963 . 13934 . 4*3928 . 2†Unbalanced input gain1−2013 . 14034 . 44028 . 2Accumulator shift1−2217 . 44443 . 04436 . 8Unbalanced input noise1−2272 . 74553 . 74547 . 48-parameter model8−1957 . 53981 . 13949 . 9This table shows the three models fit to the unilateral FOF data ( n = 3836 trials ) . *indicates the model with the lowest Bayesian information criteria ( BIC ) , that is , the most likely model . †indicates the model with the lowest Akaike information criteria ( AIC ) , that is , the most informative model . The 1-parameter post-categorization model has the lowest AIC and BIC , supporting the view that the major effect of unilateral FOF inactivation is not related to the accumulation process per se . One of the characteristics of a post-categorization bias model is that since the biasing process occurs after the accumulated evidence has been categorized into ‘Go Right’ or ‘Go Left’ , the bias is independent of whether trials are easy ( large value of |#R − #L clicks| ) or difficult ( small value of |#R − #L clicks| ) . To further probe this hypothesis , we randomly intermixed regular accumulation trials with a new set of unusually easy ‘single-sided’ trials ( Figure 6F ) . In these trials the speaker from only one side produced clicks at 100 clicks/s ( noticeably higher than the 40 Hz rate on accumulation trials ) , lasting until the ‘Go’ signal indicating the end of center port fixation . Consistent with the post-categorization bias hypothesis , the unilateral FOF inactivations produced a vertically-scaled ipsilateral bias in these single-sided trials that was similar to that seen during the accumulation trials ( Figure 6F , compare to Figure 6C data ) : that is , the bias was independent of how easy or how difficult the trials were . Given that unilateral PPC lesions in rats lead to contralateral neglect ( Crowne et al . , 1986; Reep et al . , 2004 ) , that PPC has been posited as central to rodent perceptual decision-making ( Harvey et al . , 2012; Carandini and Churchland , 2013 ) , and that neural correlates of the gradually accumulating evidence are found in PPC in our task ( Hanks et al . , 2015 ) , we predicted that unilateral PPC inactivations would cause a strong contralateral impairment ( or , in other words , in the context of our binary forced-choice task , an ipsilateral bias , similar to that seen with the FOF inactivations ) . Surprisingly , our unilateral PPC inactivations resulted in a small effect that was ≈10× smaller than the effect in the FOF . The average ipsilateral bias was 4 . 2 ± 2 . 4% ( mean ± s . e . ) ( Figure 7A; t-test t13 = 1 . 76 , p > 0 . 1 ) . Moreover , this small bias was largely due to data from the first infusion session in group 1 rats ( Figure 2—figure supplement 2A ) which led to a small but significant shift in rightward responding in right vs left infusions in the group 1 rats ( p = 0 . 024 , GLMM test ) . No consistent effects of unilateral PPC infusions were found in 10 subsequent infusion sessions with group 1 ( a total of 11 group 1 unilateral PPC infusion sessions , Figure 3—figure supplement 2 ) , even when the muscimol dose was substantially increased , to 600 ng ( Figure 7B ) . 10 . 7554/eLife . 05457 . 019Figure 7 . PPC Infusions . ( A ) As in Figure 3B , but for unilateral infusions of muscimol into the PPC , which result in a minimal impairment . In black are data from control sessions 1 day before an infusion ( n = 65 sessions , 14 rats ) . In red are data from right PPC infusions ( n = 31 sessions , 14 rats , 150 or 300 ng ) . In green are data from left PPC infusions ( n = 34 sessions , 14 rats , 150 or 300 ng ) . ( B ) As in ( A ) , but using very high doses of muscimol . Only very small effects are seen . In black are data from control sessions 1 day before an infusion ( n = 11 sessions , 9 rats ) . In red are data from right PPC infusions ( n = 3 sessions , 3 rats , 600 or 2500 ng muscimol ) . In green are data from left PPC infusions ( n = 8 sessions , 8 rats , 600 or 2500 ng ) . ( C ) Bilateral infusion of muscimol into the PPC does not produce a markedly bigger impairment . In black are data from control sessions 1 day before an infusion ( n = 8 sessions , 4 rats ) . In blue are data from bilateral PPC infusions ( n = 8 sessions , 4 rats , 150 ng per side ) . ( D ) Schematic view of the brain , duplicated from Figure 3A to remind readers of the location of FOF and PPC on the cortical surface . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 01910 . 7554/eLife . 05457 . 020Figure 7—figure supplement 1 . PPC infusions have nominal effects on the Poisson Clicks task . ( A ) The psychometric data for accumulation trials and GLMM model fits for bilateral PPC infusions in each rat ( n = 4 ) . Open circles are binned data and the small points are the predictions of the GLMM fits at sampled data points . ( A ) In black are the isoflurane control data and fits and in blue are the bilateral infusion data and fits . ( B ) The psychometric data and GLMM model fits for unilateral PPC infusions in each rat ( n = 14 ) . Open circles are binned data and the small points are the generated by the GLMM fits at sampled data points . In red are the right infusion data and fits and in green are the left infusion data and fits . The difference between right and left infusions is markedly smaller in the PPC than in the FOF . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 020 To test whether PPC inactivation could produce a quickly adapting effect ( i . e . , perhaps an effect from muscimol inactivation is observable only in the first session ) , we repeated our unilateral PPC inactivations using a second group of rats . However , no significant effect was found in any of the PPC infusion sessions in group 2 rats even on the first day of infusion ( GLMM test , p = 0 . 92; Figure 2—figure supplement 2B ) . Notably , even extremely high doses ( up to 2500 ng ) of muscimol in PPC , were ineffective at biasing the rats on accumulation trials ( Figure 7B ) . Thus , our data from group 2 suggest that the bias on the first day in group 1 occurred by chance . Our PPC coordinates for group 1 and 2 ( At 3 . 8 mm posterior and ≈3 mm lateral to Bregma ) were based on the Paxinos and Watson rat atlas , the neural correlates of accumulation found at this location ( Hanks et al . , 2015 ) and several published studies of rat PPC ( Paxinos and Watson , 2004; Nitz , 2006; Whitlock et al . , 2012 ) . Nevertheless , some authors have suggested that PPC is slightly more posterior: 4–6 mm posterior to Bregma ( Kolb and Walkey , 1987; Wilber et al . , 2014 ) . We therefore repeated the PPC experiments with a third group of rats , this time implanting cannulae at 4 . 5 mm posterior to Bregma ( Figure 3—figure supplement 1C ) . Once again , as in group 2 , there was no ipsilateral bias due to muscimol ( at 300 ng ) in the PPC ( GLMM test , p = 0 . 47 ) , nor was there a detectable effect on the first inactivation session . Performance of side LED trials at regular muscimol doses ( 150 or 300 ng ) was not significantly affected by PPC infusions ( Figure 7A; t-test t7 = 1 . 0 , p > 0 . 35 ) , and the response times on these trials were also unaffected ( repeated-measures ANOVA , F ( 1 , 6 ) = 2 . 48 , p > 0 . 15 ) . At very large doses , a significant effect on side LED trials led to a small correlation between dose and bias for side LED trials ( r = 0 . 43 , p = 0 . 032; Figure 8 yellow circles ) . Given the very large muscimol doses used , and the fact that visual cortical areas lie immediately posterior to PPC ( Paxinos and Watson , 2004 ) , this weak correlation on side LED trials may be a result of spread of muscimol to the adjacent visual cortex . 10 . 7554/eLife . 05457 . 021Figure 8 . Summary of dose-bias relationship for all unilateral infusions . Infusions into the FOF are in magenta and into the PPC are in yellow . Circles indicate bias on LED trials , squares indicate bias on accumulation trials . The magenta line is the linear fit between signed dose of FOF infusion ( +for right infusions , −for left infusions ) and performance bias ( right − left % correct ) on accumulation trials ( r = 0 . 85 , p < 10−9 ) . The yellow line is the linear fit between signed dose of PPC infusion and performance bias on accumulation trials ( r = 0 . 19 , p > 0 . 05 ) . The plotted x-location of the side LED trials are slightly offset for visualization . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 021 To summarize , unilateral inactivation of PPC did not reliably bias accumulation trials in three separately tested groups of rats . Based on the large doses used we suspected that the lack of effect was not due to a failure to inactivate PPC . There was no correlation between dose and bias magnitude in our PPC infusions on accumulation trials ( Figure 8 , yellow squares; p > 0 . 09 ) , strongly contrasting with the significant correlation between dose of muscimol infused into the FOF and bias on accumulation trials ( accumulation trials r = 0 . 85 , p < 10−9; Figure 8 , magenta squares ) . The correlation in the FOF data was specific to accumulation trials ( p > 0 . 5 for side LED trials; Figure 8 , magenta circles ) . It is possible that during unilateral inactivations , the silenced PPC may be compensated for by the PPC of the opposite hemisphere . In this case , bilateral inactivations of the PPC should produce a behavioral impairment markedly larger than any small impairment found after unilateral inactivations . To probe this hypothesis , we initially used a high dose ( 300 ng per side ) for bilateral PPC infusions in group 1 rats , but only 2 of 6 rats completed trials , with inconsistent results . The maximum dose at which subjects still performed substantial numbers of trials was 150 ng per side , double the dose per side for the bilateral FOF inactivations described above . During the bilateral PPC inactivation sessions there was a very small but significant 3 . 6% decrease in performance on accumulation trials ( Figure 7B , p < 0 . 02 vs isoflurane , GLMM test ) . This effect was not individually significant in any rat ( 0/4 ) . Critically , the effect size we found was not bigger–in fact , it was slightly smaller–than the average unilateral PPC effect , and thus does not provide support for the hypothesis of hemispheric compensation . Performance on side LED trials was not significantly different between bilateral PPC infusion and control sessions ( t-test , p > 0 . 4; Figure 7B ) . Fitting the accumulator model to the bilateral PPC data , we find that the only parameter to change was the lapse ( but the confidence intervals overlapped with the control value ) , suggesting that the effects on performance were unrelated to any specific aspect of the accumulation process . Our results from PPC contrasted with previous studies that found a strong effect of PPC inactivations in a mouse memory-guided navigation task ( Harvey et al . , 2012 ) and a strong effect of permanent unilateral PPC lesions in inducing contralateral neglect in rats ( Crowne et al . , 1986; Reep et al . , 2004 ) . This motivated us to seek a positive control task . The primate literature suggested an internally-guided decision task whose trials could be readily intermixed with our evidence accumulation task . Wilke et al . ( 2012 ) interspersed regular memory-guided saccade trials ( ‘instructed’ trials ) , in which a single saccade target was presented on each trial , with internally-guided ‘free choice’ trials , in which both an ipsilateral and a contralateral target were presented , and the monkey was rewarded regardless of its response choice . By design , subjects were free to respond as they pleased in free choice trials , and they typically displayed a bias towards one side or another in these trials . Wilke et al . found that muscimol inactivation of area LIP within PPC produced no effect on choices in the instructed memory-guided saccade trials , but produced a profound ipsilateral bias during intermixed free choice trials . Inspired by Wilke et al . 's results , we modified the task for seven of our group 2 and group 3 cannulated rats ( The six group 1 rats and one group 3 rat had been already sacrificed for histology ) . We randomly intermixed 25% free choice trials with 65% accumulation trials and 10% Side LED trials ( Figure 9A ) . Free-choice trials were indicated by a lack of auditory click stimuli , and by illumination of both side LEDs after the animals had withdrawn from the center port . We refer to sessions with interleaved accumulation , side LED , and free-choice trials as ‘free-choice’ sessions . After a few free-choice sessions with no infusions , rats performed the mix of trials reliably , and expressed a consistent bias on free choice trials but no detectable bias on accumulation trials . 10 . 7554/eLife . 05457 . 022Figure 9 . Unilateral PPC inactivations induce a strong ipsilateral bias during internally guided decisions , and can induce a strong bias on accumulation trials if the FOF is bilaterally inactivated . ( A–D ) Effect of unilateral PPC or FOF inactivations on free choice trials intermixed with regular accumulation trials . ( A ) A schematic of the three interleaved trial types: Accumulation , Side LED and Free Choice trials . Accumulation and side LED trials proceeded as described in Figure 1 . On free choice trials ( 25% of all trials ) , after withdrawal from the center port both side LEDs were illuminated and rats were rewarded for going to either port . ( B ) After a few sessions , rats quickly developed an intrinsic bias on free choice trials even while showing no bias on instructed trials ( Accumulation and side LED trials ) . When muscimol was infused into the PPC free choices were significantly biased toward the side of the infusion . This panel shows an example from a single infusion day where the side of the infusions was chosen to be opposite to their intrinsic free-choice bias , together with data from the previous and subsequent control day . ( C ) Unilateral PPC infusions generated a significant 26 ± 9% ( mean ± s . e . across rats , n = 7 ) ipsilateral bias on free choice trials compared to control sessions ( the day earlier ) . During infusion sessions there was a small 8 ± 4% ( mean ± s . e . , n = 7 rats ) contralateral bias on accumulation trials , perhaps compensatory to the free choice ipsilateral bias . There was no effect on side LED trials ( D ) . Unilateral FOF infusions generated significant ipsilateral biases on free choice and accumulation trials , but not on side LED trials . One rat ( A077 ) was extremely biased ( 90% Ipsilateral choices ) on Side LED trials during FOF inactivation , indicated as an outlier . *p < 0 . 01 . ( n = 25 session , 7 rats ) ( E ) Combined bilateral infusion of muscimol into the FOF with unilateral infusion of muscimol into the PPC results in a substantial ipsilateral bias on accumulation trials . In black are data from control sessions 1 day before an infusion ( n = 32 sessions , 4 rats ) . In red are data from bilateral FOF infusions with right PPC infusion ( n = 8 sessions , 4 rats , 75 ng per side in the FOF , 300 ng in right PPC ) . In green are data from bilateral FOF infusions with left PPC infusion ( n = 8 sessions , 4 rats , 75 ng per side in the FOF , 300 ng in left PPC ) . See Figure 9—figure supplement 1 for the individual rat results and GLMM fits . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 02210 . 7554/eLife . 05457 . 023Figure 9—figure supplement 1 . Simultaneous infusion data for each rat . The psychometric data for accumulation trials and GLMM model fits for bilateral FOF infusions in each rat ( n = 4 ) . Open circles are binned data and the small points are the predictions of the GLMM fits at sampled data points . In red are the bilateral FOF + right PPC infusion data and fits and in green are the bilateral FOF + left PPC infusion data and fits . In three of four rats the right infusions lead to significantly more rightward responses than left infusions . There is substantial variability in the overall bias of individual rats in this experiment likely due to the difficulty of performing a perfectly balanced bilateral inactivation of FOF . However , the contribution of the PPC infusion was reliable ( 3/4 rats ) despite an underlying bias . DOI: http://dx . doi . org/10 . 7554/eLife . 05457 . 023 In remarkable parallel to Wilke et al . 's results in primates , unilateral PPC inactivations ( 300 ng of muscimol ) during free-choice sessions produced a very strong and reliable ipsilateral bias on free choice trials ( Figure 9B , C; t-test t26 = 3 . 70 , p = 0 . 001 ) . The strong ipsilateral bias in free choice trials was observed even while , consistent with our previous PPC inactivations , there was no ipsilateral bias on the intermixed accumulation trials ( t-test t26 = −0 . 99 , p = 0 . 329 ) nor on the Side LED trials ( t-test t8 = 1 . 42 , p = 0 . 194; Figure 9C ) . The free-choice bias was highly reproducible: 85% ( 23/27 , sign-test , p < 0 . 001 ) individual rat PPC inactivation sessions produced an increased fraction of ipsilateral free choices when compared to free choices on immediately preceding control days ( see Figure 9B for an example of PPC infusions that were selected to ‘push’ the rats away from their innate preference seen on the day before and the day after the infusion ) . The effect on free choice trials was thus similar in its robustness and reproducibility to the effect of unilateral FOF inactivation on accumulation trials . These free choice trial inactivation results provide a clear positive control for our PPC inactivations . Moreover , they are consistent with the parietal neglect literature in both rats ( Crowne et al . , 1986; Reep et al . , 2004 ) and primates ( Mesulam , 1999 ) . Inactivation of FOF , like the PPC , also induced an ipsilateral bias on free choice trials ( t-test , t24 = 3 . 86 , p = 0 . 001 ) . Consistent with our previous experiments , FOF inactivations in free-choice sessions continued to produce an ipsilateral bias on accumulation trials ( t-test t24 = 4 . 85 , p < 0 . 001 ) but not on side LED trials ( t-test t18 = 1 . 65 , p = 0 . 117; Figure 9D ) . Are there any conditions under which silencing the PPC could affect choices in auditory click accumulation trials ? To probe whether inactivation of the FOF could reveal an effect of PPC inactivation , we bilaterally inactivated the FOF while simultaneously infusing 300 ng of muscimol unilaterally into the PPC . This combination of infusions produced a significant 15 . 1% bias ipsilateral to the side of the PPC infusion ( Figure 9E , p < 0 . 0012 , GLMM test ) . These data constitute a second positive control for our unilateral PPC inactivations . The data furthermore suggest that during auditory evidence accumulation the PPC may have a real but weak influence on choice that is normally overridden by a stronger signal from the FOF .
Our results demonstrate that the FOF is an essential part of the circuit for decisions driven by accumulating evidence ( Figure 3 ) . Unilateral FOF inactivations had a strong effect on both accumulation trials , signaled by auditory clicks , and on free choice trials , signaled by a bilateral visual stimulus ( Figure 9 ) , indicating that the effects are not specific to a single sensory modality . Critically , the model-based analyses suggested a specific location for the FOF within the functional process chain required by our accumulation of evidence task . The specific suggestion is that the FOF is not part of the accumulator but is instead part of the premotor output pathway that leads from the graded evidence accumulator to the decision-reporting motor act . This suggestion is derived from ( a ) the sharp reduction in the accumulation time constant induced by bilateral FOF inactivations ( from slightly unstable τ ≈ +0 . 8 , to very leaky τ ≈ −0 . 24 s ) , which demonstrates that the FOF is either involved in the accumulation process itself , or is part of the output pathway of the accumulator ( Figure 4 ) ; ( b ) the effects induced by unilateral inactivation of the FOF , which are captured in quantitative detail , for both accumulation trials and single-side trials , as having induced a post-categorization bias that is subsequent to the accumulation process ( Figure 6 ) ; and ( c ) the lack of an effect of FOF inactivations on side LED trials , which rules out a simple motor role for the FOF ( Figure 3 ) . A parsimonious explanation of this set of results is thus that the FOF is a requisite premotor component of the output pathway of an evidence accumulator with a long time constant ( >0 . 24 s ) . This suggestion is different from , but consistent with , the FOF's known role in short-term memory , as well as the FOF's greater importance in memory-guided ( i . e . , long time constant ) vs sensory-guided ( short time constant ) orienting decisions ( Erlich et al . , 2011 ) . In decisions driven by gradual accumulation of evidence , the gradual accumulation process occurs prior to the binary decision . An intriguing possibility suggested by our results that the decision process itself–that is , the categorization of the gradually accumulated evidence into a binary choice–might be performed in the FOF , perhaps in conjunction with the superior colliculus ( Lo and Wang , 2006 ) . It is possible that unilateral FOF inactivations would induce a hemispheric imbalance so strong that a real but nuanced role for the FOF in gradual accumulation might have been obscured by a strong post-categorization bias . We nevertheless currently favor the interpretation of the FOF's role as subsequent to , not part of , the graded accumulator . We favor this interpretation first , because of its parsimony; second , because parallel electrophysiological work from our laboratory found that the representation of the accumulated evidence in the FOF could be approximately described as the answer to the categorical question ‘if the GO signal came now , which side port should I choose ? ’; third , using AAV-CaMKII-eNpHR3 . 0 for optogenetic inactivation we found that transient unilateral silencing of the FOF was unable to cause an effect on behavior if it occurred during the evidence accumulation period , sufficiently prior to the GO signal ( Hanks et al . , 2015 ) . Those results , in combination with the pharmacological data , three control trial types , and model-based analyses reported here , all consistently support the interpretation of the FOF as a requisite component of the output pathway of an evidence accumulator with a long time constant ( >0 . 24 s ) . This view is consistent with the fact that side LED trials , which involve decisions that do not require gradual evidence accumulation or storing a motor plan in short-term memory , are not impacted by FOF inactivations . Such decisions , including perhaps decisions that require auditory evidence accumulation over only short times ( <0 . 24 s ) , may depend on pathways that bypass the FOF , perhaps involving direct connections from auditory cortex to the striatum ( Znamenskiy and Zador , 2013 ) . With pharmacological infusions , one concern is spread of inactivation to other structures . We based the volume and concentration of our infusions on previous literature to achieve inactivation volumes of 1 mm radius for our small doses and 3 mm radius for our largest doses ( Martin , 1991; Krupa et al . , 1999 ) . This spread is within the anterior-posterior bounds of FOF , but could have spread medially to cingulate cortex ( CG1 ) or laterally to M1 ( Paxinos and Watson , 2004 ) . The white matter below FOF prevent spread ventrally into the basal ganglia . In a previous study , we directly tested the effects of M1 inactivation and found them to be weaker than FOF inactivations and also they affected sensory-guided and single-sided trials equally ( Erlich et al . , 2011 ) . As such , it is unlikely that the effects we are attributing to FOF are due to M1 inactivation . We targeted our inactivations to 2 mm anterior and 1 . 25 mm lateral to Bregma . According to the Rat Atlas ( Paxinos and Watson , 2004 ) CG1 may be within the spread of the drug . However , according to a recent cell-based mapping technique the medial boundary of FOF has been underestimated ( Brecht et al . , 2004 ) , suggesting that most of spread of drug would be within FOF . Moreover , the CG1 is thought to play a role in cost-benefit decisions ( Hillman and Bilkey , 2012; Holec et al . , 2014 ) , cognitive flexibility ( Ragozzino and Rozman , 2007 ) , or emotional reactivity ( Bissière et al . , 2008 ) , not in movement planning . Therefore , the effects observed are more likely due to changes in FOF rather than an adjacent cortical region . Given the view that the PPC plays a key causal role in rodent perceptual decisions ( Harvey et al . , 2012; Carandini and Churchland , 2013 ) and that neural activity in PPC displays correlates of accumulating auditory evidence ( Hanks et al . , 2015 ) , we were surprised to find that unilateral inactivation of the PPC did not cause a side bias on accumulation , side LED , or single-sided trials . Compensation from the unperturbed hemisphere did not explain the lack of an effect , because bilateral inactivation of the PPC caused a minimal decrease in performance ( 3 . 6% ) that was not significantly different from unilateral inactivations . Thus , the PPC seems to play a far more subtle role than the FOF in choice behavior during decisions guided by auditory evidence accumulation . Unilateral PPC inactivations did cause a strong ipsilateral bias under two conditions: in internally-guided decisions ( free choice trials , signaled by a visual stimulus , Figure 9A–C ) , and when the FOF was simultaneously bilaterally inactivated in accumulation trials ( auditory stimulus , Figure 9E ) . These results suggest that in the intact brain , weak but real side choice signals from the PPC may be overridden by stronger signals from the FOF . They also suggest that the PPC's role may not be specific to a particular sensory modality . The PPC is relatively extended in the medial-lateral direction ( ≈4 mm ) but it is thin ( ≈0 . 75 mm ) in the anterior-posterior direction ( Paxinos and Watson , 2004; but see Wilber et al . , 2014 ) . The anatomical characteristics of the PPC pose two possible confounds: insufficient inactivation of the PPC in the medial-lateral direction and potential spread to adjacent regions in the anterior-posterior direction . If we assume the spread of muscimol across the cortical is isotropic ( white matter acts as a physical barrier , so we assume that the ventral spread is restricted ) , then an individual injection , due to the wide and thin shape of PPC , will either fail to inactivate all of PPC or spread to adjacent regions . Based on the literature ( Martin , 1991; Krupa et al . , 1999 ) we expect our smallest infusion to have a ≈1 mm radius and our largest to have a >3 mm radius . This technical limitation would have posed a serious challenge for our results if we observed very different effects of our small and large infusions . However , despite using a wide range of doses , we observed no significant dose–response effect with unilateral PPC infusions ( Figure 8 ) . A clear dose–response relationship ( as we found in the FOF ) would indicate that bigger infusions were silencing more and more PPC neurons required for behavior in the task . Instead , the fraction of inactivated PPC neurons had no impact on the magnitude of the behavioral effect , which is what we would expect if PPC neurons were not required for choice behavior in the task . Nonetheless , further experiments would be required to completely rule out the possibility that a small number of PPC neurons , in the most medial or lateral edge of the PPC , were spared and that these few neurons were sufficient to support intact behavioral performance on the Clicks task . Spread of muscimol into areas immediately anterior or posterior to the PPC was inevitable with our largest doses , and , in the two cases where positive effects were observed , this raises concerns about region specificity . Immediately anterior to the PPC is the somatosensory cortex ( for the trunk ) . Particularly in light of the intact motor capacities , as indicated by the intact side LED trials , the somatosensory cortex is not expected to have caused any of the observed effects . Immediately posterior to the PPC are a set of individually small visual areas , collectively referred to as secondary visual cortex ( V2; Coogan and Burkhalter , 1993; Montero , 1993; Wang and Burkhalter , 2007 ) . These visual areas are unlikely to be involved in auditory click accumulation trials , and are therefore not expected to have caused the effects we saw after simultaneous unilateral PPC and bilateral FOF inactivations ( Figure 9E ) . However , the bias we observed on free-choice trials may have been partly due to an effect in one or more of these small secondary visual areas . Nevertheless , ipsilateral biases in untrained orienting responses ( potentially analogous to the free-choice task ) due to PPC lesions have been observed in the rat even when those choices were to tactile or auditory stimuli ( Corwin et al . , 1996; Burcham et al . , 1997 , 1998 ) . We also note that our free choice results closely parallel the results from primate PPC free choice experiments ( Wilke et al . , 2012 ) , which do not suffer from this spillover concern . A minimal effect following PPC inactivation in our rat auditory click accumulation task is reminiscent of Guo et al . 's , finding of no effect after PPC inactivation in a mouse somatosensory-cued , memory-guided task ( Guo et al . , 2014 ) . But it contrasts with Harvey et al . 's finding of a severe performance impairment after PPC inactivation in a mouse visually-cued , memory-guided navigation task ( Harvey et al . , 2012 ) . Following our own preliminary reports of PPC inactivations in the Poisson Clicks task ( Erlich et al . , 2012 , 2014 ) , Raposo et al . ( 2014 ) reported an impairment after rat PPC inactivations in a visual , but not in an auditory , version of a closely related task . This contrasts with a preliminary report from Yates et al . that suggested no effect from primate PPC inactivation in a visual accumulation of evidence task ( Yates et al . , 2014 ) , as well as with multiple reports of no effect from primate PPC inactivation in visual memory-guided saccade tasks ( Chafee and Goldman-Rakic , 2000; Wardak et al . , 2002; Liu et al . , 2010; Wilke et al . , 2012 ) . The Raposo et al . results therefore suggest that rodent PPC may be unlike primate PPC in being required for accumulation of evidence for a specific modality , vision . Nevertheless , the precise location of the border between rodent PPC and visual cortex , as well as the amount of inter-animal variability in this location , remain active research questions ( Wilber et al . , 2014; see , for example , Reep et al . , 1994; Corwin and Reep , 1998 for definitions from the same authors , based on the same data , that alternately describe the border as located at −5 . 0 mm from Bregma or −4 . 4 mm from Bregma ) . The uncertainty in the location of the PPC/visual cortex border raises an alternative possibility , which is that the impairments in memory-guided visual tasks observed by Harvey et al . and Raposo et al . ( and potentially the free-choice effects seen here ) could have been due to inactivation spillover into one of the immediately adjacent small secondary visual areas , as described above , rather than inactivation of the PPC itself . This would make all the inactivation results ( Harvey et al . , 2012; Guo et al . , 2014; Raposo et al . , 2014; Yates et al . , 2014; and the results presented here . ) , including mouse , rat , and primate , fully consistent with each other . While such a reconciliation of results across multiple species may have some intellectual appeal , we emphasize that experiments that would either clearly support or rule out this possibility remain to be done . To date , the accumulation of evidence literature is mostly composed of electrophysiological experiments with primates . Based on a number of criteria , the rat PPC and FOF have been suggested as analogous to the primate PPC and FEF , respectively ( Kolb and Walkey , 1987; Reep and Corwin , 2009; Erlich et al . , 2011 ) , and accumulation of evidence signals very similar to those in primates have been found in these two rat areas ( Hanks et al . , 2015 ) . It is therefore tempting to speculate that our results in rat might also hold in primate , and to consider potential interpretations that would be consistent across mammalian model systems . There have been no published inactivation experiments in primate PPC or FEF during accumulation of evidence tasks . There have , however , been inactivation experiments in related tasks , which in general are in agreement with our findings: prefrontal perturbations strongly bias behavior while posterior parietal perturbations do not . In memory-guided saccade tasks , LIP inactivations have no effect on the choice of saccading either towards or away from the correct hemifield ( Chafee and Goldman-Rakic , 2000; Wardak et al . , 2002; Liu et al . , 2010; Wilke et al . , 2012 ) , while FEF and prefrontal inactivations reliably generate profound ipsilateral choice biases ( Sommer and Tehovnik , 1997; Dias and Segraves , 1999; Chafee and Goldman-Rakic , 2000 ) . Similarly , in a covert visual search task , LIP inactivation has no effect on error rates ( Wardak et al . , 2002 ) , while inactivation of the FEF generates significant increase in error rates for contralateral targets ( Wardak et al . , 2006 ) . Again similarly , in a memory-guided task with distractors , Suzuki and Gottlieb , ( 2013 ) found no errors after LIP inactivations while finding significant contralateral errors after prefrontal cortex inactivations . We are aware of only a few studies where LIP inactivation produces choice biases ( Wardak et al . , 2002; Balan and Gottlieb , 2009; Wilke et al . , 2012 ) . One of these studies ( Wilke et al . , 2012 ) inspired our intermixing free choice trials with accumulation trials . Consistent with a good analogy between rat PPC and primate PPC , our results in rats closely paralleled Wilke et al . 's results in primates , with our accumulation of evidence trials playing the role of their memory-guided saccade trials ( Figure 8 ) . Also consistent with our rat data and with a good analogy between rat and primate PPC , a preliminary report has suggested that unilateral primate PPC inactivations have no effect on accumulation of evidence trials , while causing an ipsilateral bias on free choice trials ( Yates et al . , 2014 ) . There has been one perturbation study in the primate FEF during an accumulation of evidence task ( Gold and Shadlen , 2000 ) . This microstimulation study concluded that ‘developing oculomotor commands may reflect the formation of the monkey's direction judgement’ ( i . e . , its decision ) . Our results in rat FOF are consistent with those of Gold and Shadlen , but go considerably further in specifically suggesting the FOF as a requisite premotor component of the output pathway of a long time constant ( >0 . 24 s ) evidence accumulator . There has been one perturbation study in the primate PPC during an accumulation of evidence task ( Hanks et al . , 2006 ) . This microstimulation study used a reaction time version of the Random Dots task , and found a pattern of results following LIP microstimulation that could be quantitatively explained in an accumulation-to-bound model if the microstimulation added a small constant offset to the value of the accumulator ( Hanks et al . , 2006 ) . Unlike the task used by Hanks et al . , our task ( and that of Yates et al . , 2014 ) was not a reaction time task , which could explain the difference in results . An alternative possibility , noted in their discussion ( Hanks et al . , 2006 ) , and which would reconcile our results with theirs , is that microstimulation may activate axon terminals or fibers of passage ( Histed et al . , 2009 , 2013; but see Tehovnik and Slocum , 2013 ) . The behavioral effects produced by microstimulation of LIP may thus have been due to activation of neurons with somata not in LIP , but in regions of the brain that project to LIP . Since muscimol does not affect axons or fibers of passage , this possibility would be consistent with our data . If the PPC plays a causal role in internally-guided decisions , but does not play a causal role in choice behavior during accumulation of evidence tasks , what role is then played by the firing rate correlates of evidence accumulation signals observed in PPC , in both rats and primates ? Accumulation signals are also correlated with confidence ( Kiani and Shadlen , 2009; Komura et al . , 2013 ) . One possibility is that , instead of being used to drive choice behavior , the accumulation signals observed in the PPC are used for computing choice confidence ( Kiani and Shadlen , 2009 ) . Confidence could be part of a process for optimizing behavior over many trials , or , in a reaction time version of the task , confidence could be part of determining when the subject chooses to commit to a decision . In the rat , our data now suggests a specific functional role for the FOF in decisions driven by accumulation of evidence: the FOF appears to be a requisite component of the output pathway of an evidence accumulator with a long time constant ( >0 . 24 s ) . Decisions with shorter processing times may involve circuits that bypass the FOF . It is possible that categorizing the graded accumulator's value into a discrete choice–the final decision itself–could occur in the FOF . In contrast , the PPC seems to play a surprisingly minimal , or subtle , role in choice behavior during decisions guided by accumulation of evidence , even while it plays an important role in internally-guided decisions . Neither region appears likely to play a major causal role in the gradual evidence accumulation process per se .
Animal use procedures were approved by the Princeton University Institutional Animal Care and Use Committee and carried out in accordance with National Institutes of Health standards . All subjects were male Long-Evans rats ( Taconic , NY ) . Rats were placed on a restricted water schedule to motivate them to work for water reward . Rats were kept on a reversed 12 hr light–dark cycle and were trained in their dark cycle . We trained 14 male Long-Evans rats on the Poisson Clicks accumulation task ( Figure 1 ) . Training took place in a behavior box with three nose ports ( left , center and right ) , and with two speakers , placed above the right and left nose ports . Each accumulation trial began with a visible light-emitting diode ( LED ) turning on in the center port . This cued the trained rat to place its nose in the center port , and keep it there until the LED was turned off . We refer to this period as the ‘nose in center’ or ‘fixation’ period . The duration of fixation was 2 s for all accumulation trials . During the fixation period a variable duration auditory stimulus ( 0 . 1–1 s , experimenter controlled ) would play , consisting of two randomly timed trains of clicks , playing simultaneously , one from the left and one from the right speaker . At the end of the auditory stimulus , the LED in the center port would extinguish , which was the signal to the rat to make their response by poking into one of the side ports . The timing of the clicks from each speaker was generated by two independent Poisson processes . The total generative click rate ( left + right rate ) was held constant at 40 clicks/s , and trial difficulty was controlled by adjusting the relative left vs right rates , as well as the duration of the stimulus . Trials where rats exited the center port during the fixation period were considered violation trials , aborted , and a new trial was started . These trials are not included in any analyses . To test whether the apparent vertical scaling of the psychometric curve after unilateral FOF inactivations was due to a lack of asymptotically easy trials , we interleaved ‘single-sided’ trials with accumulation trials ( and side LED trials in some sessions ) . A single-sided trial was much like an accumulation trial , but all the clicks came from one speaker and were played at a Poisson rate of 100 Hz . Since the 100 Hz stimuli were easily distinguished from 40 Hz stimuli and these trials did not require accumulation ( all clicks on one side ) , rats probably made their decision quickly . However , they were still required to wait until the go cue . These trials comprised ≈8% of a trials in a session . In order to control for motor effects of inactivations , in group 2 and 3 rats , we included , randomly interleaved with other trial-types , ‘side LED’ trials . On side LED trials no sounds were played during fixation , which lasted 1 s . Immediately after the end of fixation , one of the two side ports was illuminated , indicating availability of reward at the lit port ( Figure 1 ) . The right and left side LED trials , together , comprised ≈10% of the total trials . In order to find a task that was sensitive to inactivation of the PPC , we randomly interleaved ‘free-choice’ trials with the other trial-types . Free-choice trials were similar to side LED trials except at the end of fixation both side LEDs were illuminated and rats were rewarded regardless of whether they poked in the right or left nose port . These sessions took place after all of the experiments presented in Figures 3–7 , 8E . Control non-infusion sessions ( used to generate Figure 2—figure supplement 2 ) with poor performance ( <70% correct overall or fewer than eight correct trials on each side without fixation violations ) were excluded from analyses . These sessions were rare and were usually caused by problems with the hardware ( e . g . , a clogged water-reward valve or a dirty IR-photodetector ) . We collected ≈145 , 000 control trials ( range [57692 265332] ) over ≈450 ( range [291640] ) sessions from each rat ( n = 14 ) from sessions without intracranial infusions or pre-session anesthesia for a total of 2 , 057 , 074 control trials . Surgical methods were identical to those described previously ( Erlich et al . , 2011 ) . Briefly , rats were anesthetized with isoflurane and placed in a stereotax . The scalp was deflected and the skull was cleaned of tissue and blood . The stereotax was used to mark the locations of craniotomies for the FOF and PPC on the skull . Craniotomies and durotomies were performed and then the skull was coated with a thin coat of C&B Metabond ( Parkell Inc . , NY ) . Guide cannula ( Plastics One , VA ) were lowered to brain surface with dummy cannula extending 0 . 5 mm into the brain . The guide cannula were placed and secured to the skull one at a time with a small amount of Absolute Dentin ( Parkell Inc . , NY ) . After the three guide cannula were in place ( One bilateral FOF cannula and one cannula for each PPC ) dental acrylic ( Duralay , Reliance Dental Mfg . Co , IL ) was then used to cover the skull and further secure the cannula . Rats were given 5 days to recover on free water before resuming training . Group 1 rats ( n = 6 ) were implanted bilaterally in FOF ( +2 AP , ±1 . 25 ML mm from Bregma ) with 22 AWG guide cannula ( C232G-2 . 5 , Plastics One , VA ) and the medial ( 3 . 8 mm posterior , 2 . 2 mm lateral to Bregma ) and lateral ( 3 . 8 mm posterior , 3 . 4 mm lateral to Bregma ) PPC with 26 AWG guide cannula ( 6 cannula per rat total ) . Group 2 rats ( n = 4 ) were implanted in FOF and in PPC ( 3 . 8 mm posterior , ±2 . 8 mm lateral to Bregma; a total of 4 cannulae per animal ) with bilateral 22 AWG guide cannula . Group 3 rats ( n = 4 ) were implanted with bilateral 22 AWG guide cannula in FOF and in PPC ( 4 . 5 mm posterior , ±3 . 0 mm lateral to Bregma; 4 cannulae per animal ) . The tip of the guide sat at brain surface and the dummy extended 0 . 5 mm into cortex . The injector for the 22 AWG guide cannula was a 28 AWG cannula that extended 1 . 5 mm below the bottom of the guide cannula . The injector for the 26 AWG guide cannula was a 33 AWG cannula that extended 1 . 5 mm below the bottom of the guide cannula . In general , infusions were performed once a week with control training days taking place on all other days of the week in order to minimize adaptation to the effects of the muscimol and to have good stable performance in the sessions immediately before infusion sessions . On an infusion day , the rat was placed into an induction chamber with 2% isoflurane , and then transferred to a nose cone with 2% isoflurane for the infusion procedure . Caps and dummy cannula were removed and cleaned . Injectors were placed into the relevant guide cannula and extended 1 . 5 mm past the end of the guide , into cortex . We used a hamilton syringe connected via tubing filled with mineral oil to the injector to infuse 0 . 3 µl of muscimol ( of various concentrations–see ‘Results’ and Figure 3—figure supplement 2 ) into cortex . After injection , we left the injector in the brain for 4 min to allow diffusion before removal . After 4 min , cleaned dummies were placed into the guide cannula and capped and the rat was removed from isoflurane . After 30 min of recovery from isoflurane the rat was placed into a behavior box as usual . See Figure 3—figure supplement 2 for the complete list of all infusion doses , regions , and order for each rat . Previous experiments in rat cortex , performing autoradiographic estimates ( Martin , 1991 ) , as well as simultaneous muscimol inactivation and recordings ( Krupa et al . , 1999 ) , suggest that at the doses of muscimol we used , the expected area of inactivation would have an approximately ≈1 mm radius for the smallest doses and >3 mm radius for the largest doses . Since the only difference between left and right infusions ( and FOF and PPC infusions ) are the location of the infusion any differences in behavior can only be attributed to the infusion and not to handling . For bilateral infusions , we were interested in non-lateralized impairments compared to baseline performance . To rule out the possibility that the handling of rats for the infusion procedure could affect performance we did isoflurane-only sessions where rats were handled as they would be for an infusion ( taken to the infusion room , placed into an induction chamber with 2% isoflurane , dummies removed , cleaned and replaced , etc ) but given no infusion . These isoflurane-only sessions were used as a baseline to compare with the bilateral infusion sessions . All analysis and statistics were computed either in Matlab ( version 7 or better , The Mathworks , MA ) or R ( version 2 . 15 . 2 , R Foundation for Statistical Computing , Vienna , Austria ) . Sessions where there were too few trials , or otherwise had problems during the infusions were excluded . We only included the first 250 trials of a session because the effect of muscimol can decrease after a few hours , although the results are robust to including all trials . The accumulator model uses 9-parameters ( described in Figure 2—figure supplement 1 ) to transform the stimulus on each trial ( input to the model as the left and right click times ) into a probability distribution about the choice of the rat . For example , if for a given set of parameter , the model predicts that trial 1 will result in 75% chance of the rat going right , and the rat in fact went right , that trial would be assigned a likelihood of 0 . 75 . In the case that the rat went left , the trial would be assigned a likelihood of 0 . 25 . We fit the model assuming that the trials are independent . Therefore , for a model with parameters θ for all decisions D , the likelihood is given by:P ( D|θ ) =∏iP ( di|ti , R , ti , L , θ ) , the product of the likelihoods of the decisions on trial i , di , given the times of the right clicks ti , R , times of the left clicks ti , L , and the set of 9 parameters , θ . A detailed description of the procedure for fitting the accumulator model can be found in the Modeling Methods section of the supplement of Brunton et al . ( 2013 ) . For panels A , C , and D in Figure 2 we first concatenated 47 , 580 trials across sessions and rats from sessions 1 day before an infusion session . The values of the parameters that maximized P ( D|θ ) for these concatenated data are described as ‘Meta-Rat’ in Table 1 . Since the model is fit to the individual trials , we emphasize that the psychometric , chronometric and reverse correlations plot for a given model are not generated using a curve-fitting procedure for each panel . We also fit each rat individually and show the best-fit parameters for each rat in Table 1 . Even at the very easiest trials , rat behavior in our task often asymptotes at ≈90% correct . To fit this , the model includes a ‘lapse’ parameter , which represents a fraction of trials in which subjects will ignore the stimulus and choose randomly . The presence of the lapse parameter also puts a lower bound on the likelihood of any individual trial , and thus no individual trial can dominate the results and the consequent fits of the model . In Figures 3 , 4E , 6 , 8E , Figure 2—figure supplement 2A the psychometric curves were generated by concatenating trial data across sessions for each rat and fitting ( using Matlab's nlinfit ) a 4-parameter sigmoid as follows:y=y0+a1+e− ( x−x+0 ) b . For these fits , x is the click difference on each trial ( #Right Clicks − #Left Clicks ) , y is ‘P ( Went Right ) ’ , and the four parameters to be fit are: x0 , the inflection point of the sigmoid; b , the slope of the sigmoid; y0 , the minimum ‘P ( Went Right ) ’; and a + y0 is the maximum ‘P ( Went Right ) ’ . These fits are for visualization only . For chronometric analyses ( Figures 2C , 4C , 6D and Figure 2—figure supplement 2C ) , we concatenated trials for each rat across sessions and binned trials into easy , medium and hard quantiles with equal number of trials based on the relative generative left-right click rate . For each of these three difficulty levels we binned trials by stimulus duration . The detailed methods of generating the psychophysical reverse correlations ( Figures 2D , 4D , 5D and Figure 6—figure supplement 1 ) can be found in the Brunton et al . ( 2013 ) . For this analysis separation of right ( red lines ) and left ( green lines ) trials at each time in the trial indicates that there was a difference in local click rate at that time for trials in which the rat responded to the right vs to the left . If rats only used the early clicks for their decision ( a primacy strategy ) , the lines would begin separated and come together . Likewise , if they only used the clicks at the end of the stimulus ( a recency strategy ) the lines would start together and separate towards the end of the trial . To generate the accumulator model's reverse correlation , each trial was assigned as the left and right trial , according to the model's prediction for that trial . For example , if the model predicted that a trial had a 67% chance of a rightward choice , this trial would contribute 0 . 67 to the right trials and 0 . 33 to the left trials psychophysical kernel . The unilateral infusion experiments are designed specifically to test for lateralized effects . On left and right infusion days animals are handled in the exact same way: taken into the room where the infusions happen , placed under light isoflurane anesthesia , etc ( described in detail in the Infusions section of the ‘Materials and methods’ ) . The only difference is the side of the infusion . As such , this within-subject design is effective at testing biases to respond toward or away from the side of the infusion . The simplest way to estimate bias resulting from unilateral inactivations is to subtract the contralateral % correct from the ipsilateral % correct for each infusion session . To compute the overall bias we averaged the bias across sessions for each rat and then tested using a t-test whether the bias across rats was significantly different from 0 . However , this statistical test is conservative , since it collapses across all trials of differing difficulty levels and different sessions . As well , this overall bias measure was inappropriate for testing the effects within each group , since the n per group was low . To avoid false negative results ( type II error ) , as a more sensitive measure of inactivation effects we used a Generalized Linear Mixed-Model ( GLMM ) as implemented in the function ‘lmer’ in package ‘lme4’ ( Bates and Sarkar , 2007; Bates et al . , 2007 ) . For unilateral infusions we specified a mixed-effects model where the rats' choice on each trial was a logistic function of #Right − #Left Clicks ( Δ Clicks ) , infusion side and their interaction as fixed effects . The rat and an interaction of rat , infusion side , and Δ Clicks were modeled as within-subject random effects . The statistic reported in the text for unilateral infusions was the p value for the infusion side fixed effect . The plots in Figure 3—figure supplement 3 ( FOF ) and Figure 7—figure supplement 1 ( PPC ) show that the model fits for each rat are quite good , and reflect how the random effects of the model allow for each rats' data to be fit , while also finding significant fixed effects . The logistic fit sometimes misses the end point performance , which , for the FOF inactivations , was generally worse than the GLMM fit . For bilateral infusions we specified a similar model comparing bilateral infusions to isoflurane-only sessions . For these models the relevant statistic was the significance of the interaction between infusion and Δ Clicks , that is , a change in the slope of the logistic . Details of the data and code used to generate and compare the models is described in Supplementary file 1 ‘Using the lme4 package to fit generalized-linear mixed models in R’ . To test whether there were effects of unilateral infusions on response times ( RT ) on side LED trials , measured as time from go cue to side port response , we took the mean RT for each rat for left and right choices on left and right infusion days . We separated leftward and rightward responses because there can be large differences from rat to rat in the left vs right response times . We then did a repeated-measures ANOVA ( Rat X [Left vs Right] X [Ipsi vs Contra] ) to test whether there was an effect of muscimol on RTs . Since ANOVA must be balanced , only 4 rats ( out of a possible 8 group 2 and 3 rats ) performed enough side LED trials from FOF infusions for this analysis . For the PPC infusions 7 of 8 rats performed enough trials to be included . For the analyses of biases during free-choice sessions we compared the bias ( ipsilateral − contralateral % correct ) on the infusion day with the control session 1 day before and performed t-tests across sessions for each region ( PPC or FOF ) and trial-type ( free-choice , accumulation or side LED ) . Some free-choice sessions had no side LED trials which is why there are fewer sessions in the t-tests for that trial-type . Side LED trials were also not analyzed for sessions if the bias in the side LED trials during the preceding control day was greater than 20% . To fit the 9-parameter accumulator model to the bilateral FOF infusion data we concatenated all 1809 trials across all rats . We validated the error in these fits by bootstrapping . Specifically , we generated 300 sets of 1809 trials data by randomly sampling with replacement from the original data set . We then fit the 9-parameter model to each of these 300 data sets , generating 300 sets of 9 best-fit parameters . We used this distribution of fits to generate confidence intervals for each of the parameters . In order for a parameter to be considered as significantly different from the control value , the control value from the meta-rat fit had to be outside of the 95% confidence interval of the bootstrapped marginal distribution for that parameter . To better understand what aspect of the task was impaired by unilateral FOF inactivation we took advantage of the knowledge of the times of all the clicks and the rats' choices using our previous modeling work ( Figure 2—figure supplement 1 , Brunton et al . , 2013 ) to fit four constrained Accumulator Models to the unilateral FOF infusion data . For all of these analyses we relabeled the trials for left/right trials to ipsi/contra trials based on the side of the infusion . For the infusion meta-rat we combined all 3836 trials from unilateral FOF sessions . Second , to avoid computing the gradient for a 12-parameter model , we fit constrained Accumulator Models that had one parameter free and the other parameters fixed to the best-fit parameters from the control meta-rat Accumulator Model . The details of the free parameter of each model are described in the results section . We used MALTAB's built-in Metropolis–Hastings sampler ( mhsample . m in the Statistics Toolbox ) to sample from the 8-parameter distribution ( described in the ‘Results’ ) . We generated 4 separate Markov chains of 10 , 000 samples each , starting from the best-fit control parameters with a burnin of 100 . For the proposal distribution we used a uniform distribution with the range selected for each parameter ( e . g . , a smaller range for input gain than for input noise ) . We used a thin of 4 , based on test runs with known gaussian distributions . Figure 6—figure supplement 2 shows the samples generated by this sampling process . To compare models with different numbers of parameters we used two commonly used metrics , the Bayesian and Akaike information criterion ( BIC and AIC ) . The BIC is defined as , BIC = −2 × LL + k × ln ( n ) . Where LL is the maximum log likelihood of a model with k free parameters on n data points . The AIC is similarly defined , AIC = −2 × LL + 2 × k . The main difference being that AIC is not a function of the size of the data set . Since the cost of additional parameters in AIC is only 2k , the AIC favors more complicated models compared to BIC . | Imagine that you have to buy a computer before the start of the school year . You have a few options , such as a laptop or a desktop , each with its own advantages and disadvantages . A laptop is relatively light and portable , whereas a desktop has more memory and is cheaper . You will gradually accumulate evidence for and against each option , but before school starts , you have to make a decision . This gradual accumulation of evidence is an important element in many forms of decision making . It is known that the activity of many regions within the brain seem to represent accumulation of evidence , but relatively little is known about the causal role played by each region in the decision-making process . Now , by performing a series of experiments on rats , Erlich et al . have clarified the precise roles of two of these regions: the frontal orienting fields in prefrontal cortex and the posterior parietal cortex . In the experiments the rats listened to a series of clicks from two speakers , one to the left and one to the right , and then had to decide if more clicks came from the left or the right speaker . The rats normally used all of the accumulated evidence ( up to 1 second ) for their decision . When the posterior parietal cortex was silenced ( using a drug called muscimol ) , the rats continued to use all of the evidence available to them . However , when the frontal orienting fields were silenced ( again using muscimol ) , decisions were driven only by evidence accumulated over the most recent past ( just a few hundred milliseconds ) . So in the computer example , without the help of the frontal orienting fields , you would choose the laptop if the most recent piece of evidence was for the laptop , even if older evidence argued strongly for the desktop . These results show that the frontal orienting fields are necessary for making decisions based on accumulated evidence , but further experiments suggested that the accumulation process itself seems to happen elsewhere in the brain . Another set of related experiments showed that the posterior parietal cortex is involved in a different type of decision making , namely ‘free choice’ decisions in which the rat decides between two options when there is no correct answer , such as picking a cookie from a pile of identical cookies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Distinct effects of prefrontal and parietal cortex inactivations on an accumulation of evidence task in the rat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.